[PATCH v3 00/28] First class shallow clone

[PATCH v3 00/28] First class shallow clone

[Nguyễn Thái Ngọc Duy]

Compared to v2 [1], v3 grows a bit. The biggest difference is
.git/shallow is not updated by default (except when you clone from
a shallow repository). When you send something, the "safe" refs that
do not need new shallow roots are accepted at the receiver end, the
others rejected.

To accept those other refs, either use "fetch --update-shallow" or
enable receive.shallowupdate on the receiver side of a push.

Filtering "safe" refs requires walking through some commits, so
it'll be more expensive than normal full clones. This is especially
true for the receiver of a push (see 07/28 and 17/28). I envision
shallow repos are used as upstream to archive old history, so this is
not a good news. Commit cache (or pack v4) might help. We might even
be able to move some work from receive-pack to send-pack to reduce
server load..

[1] http://mid.gmane.org/1374314290-5976-1-git-send-email-pclouds@gmail.com

Nguyễn Thái Ngọc Duy (28):
  transport.h: remove send_pack prototype, already defined in send-pack.h
  send-pack: forbid pushing from a shallow repository
  clone: prevent --reference to a shallow repository
  update-server-info: do not publish shallow clones

   This part is just cleanup.

  Advertise shallow graft information on the server end
  connect.c: teach get_remote_heads to parse "shallow" lines
  shallow.c: add remove_reachable_shallow_points()
  shallow.c: add mark_new_shallow_refs()
  shallow.c: extend setup_*_shallow() to accept extra shallow points
  fetch-pack.c: move shallow update code out of fetch_pack()
  fetch-pack.h: one statement per bitfield declaration
  clone: support remote shallow repository
  fetch: support fetching from a shallow repository
  upload-pack: make sure deepening preserves shallow roots
  fetch: add --update-shallow to get refs that require updating .git/shallow

   Basic shallow fetch/clone support on git protocol

  receive-pack: reorder some code in unpack()
  Support pushing from a shallow clone
  New var GIT_SHALLOW_FILE to propagate --shallow-file to subprocesses
  connected.c: add new variant that runs with --shallow-file
  receive-pack: allow pushing with new shallow roots
  send-pack: support pushing to a shallow clone
  remote-curl: pass ref SHA-1 to fetch-pack as well
  
   Push support

  Support fetch/clone over http
  receive-pack: support pushing to a shallow clone via http
  send-pack: support pushing from a shallow clone via http

   smart-http support

  git-clone.txt: remove shallow clone limitations
  clone: use git protocol for cloning shallow repo locally
  prune: clean .git/shallow after pruning objects

   miscellaneous

 Documentation/config.txt                  |   4 +
 Documentation/fetch-options.txt           |  14 +-
 Documentation/git-clone.txt               |   7 +-
 Documentation/gitremote-helpers.txt       |   7 +
 Documentation/technical/pack-protocol.txt |   7 +-
 builtin/clone.c                           |  18 +-
 builtin/fetch-pack.c                      |  23 +-
 builtin/fetch.c                           |  15 +-
 builtin/gc.c                              |   1 +
 builtin/prune.c                           |   4 +
 builtin/receive-pack.c                    | 248 +++++++++++++++++----
 builtin/send-pack.c                       |   5 +-
 cache.h                                   |   1 +
 commit.h                                  |  19 +-
 connect.c                                 |  14 +-
 connected.c                               |  42 +++-
 connected.h                               |   2 +
 environment.c                             |   6 +
 fetch-pack.c                              | 132 ++++++++++--
 fetch-pack.h                              |  29 +--
 git.c                                     |   2 +-
 remote-curl.c                             |  33 ++-
 remote.h                                  |   5 +-
 send-pack.c                               |  20 +-
 server-info.c                             |   9 +
 shallow.c                                 | 348 +++++++++++++++++++++++++++++-
 t/t5536-fetch-shallow.sh (new +x)         | 193 +++++++++++++++++
 t/t5537-push-shallow.sh (new +x)          | 184 ++++++++++++++++
 t/t5601-clone.sh                          |   7 +
 transport-helper.c                        |   6 +
 transport.c                               |  22 +-
 transport.h                               |  16 +-
 upload-pack.c                             |   8 +-
 33 files changed, 1323 insertions(+), 128 deletions(-)
 create mode 100755 t/t5536-fetch-shallow.sh
 create mode 100755 t/t5537-push-shallow.sh

-- 
1.8.2.83.gc99314b

[PATCH v3 01/28] transport.h: remove send_pack prototype, already defined in send-pack.h

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 transport.h | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/transport.h b/transport.h
index 8f96bed..b3679bb 100644
--- a/transport.h
+++ b/transport.h
@@ -193,10 +193,4 @@ void transport_print_push_status(const char *dest, struct ref *refs,
 
 typedef void alternate_ref_fn(const struct ref *, void *);
 extern void for_each_alternate_ref(alternate_ref_fn, void *);
-
-struct send_pack_args;
-extern int send_pack(struct send_pack_args *args,
-		     int fd[], struct child_process *conn,
-		     struct ref *remote_refs,
-		     struct extra_have_objects *extra_have);
 #endif
-- 
1.8.2.83.gc99314b

[PATCH v3 02/28] send-pack: forbid pushing from a shallow repository

[Nguyễn Thái Ngọc Duy]

send-pack can send a pack with loose ends to the server.  receive-pack
before 6d4bb38 (fetch: verify we have everything we need before
updating our ref - 2011-09-01) does not detect this and keeps the pack
anyway, which corrupts the repository, at least from fsck point of
view.

send-pack will learn to safely push from a shallow repository later.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/send-pack.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index 4482f16..51121f2 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -207,6 +207,9 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 	    (send_all && args.send_mirror))
 		usage(send_pack_usage);
 
+	if (is_repository_shallow())
+		die("attempt to push from a shallow repository");
+
 	if (remote_name) {
 		remote = remote_get(remote_name);
 		if (!remote_has_url(remote, dest)) {
-- 
1.8.2.83.gc99314b

[PATCH v3 03/28] clone: prevent --reference to a shallow repository

[Nguyễn Thái Ngọc Duy]

If we borrow objects from another repository, we should also pay
attention to their $GIT_DIR/shallow (and even info/grafts). But
current alternates code does not.

Reject alternate repos that are shallow because we do not do it
right. In future we alternate code may be updated to check
$GIT_DIR/shallow properly so that this restriction could be lifted.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/clone.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/builtin/clone.c b/builtin/clone.c
index 874e0fd..900f564 100644
--- a/builtin/clone.c
+++ b/builtin/clone.c
@@ -252,6 +252,12 @@ static int add_one_reference(struct string_list_item *item, void *cb_data)
 		die(_("reference repository '%s' is not a local repository."),
 		    item->string);
 
+	if (!access(mkpath("%s/shallow", ref_git), F_OK))
+		die(_("reference repository '%s' is shallow"), item->string);
+
+	if (!access(mkpath("%s/info/grafts", ref_git), F_OK))
+		die(_("reference repository '%s' is grafted"), item->string);
+
 	strbuf_addf(&alternate, "%s/objects", ref_git);
 	add_to_alternates_file(alternate.buf);
 	strbuf_release(&alternate);
-- 
1.8.2.83.gc99314b

[PATCH v3 04/28] update-server-info: do not publish shallow clones

[Nguyễn Thái Ngọc Duy]

Dumb commit walker does not care about .git/shallow and until someone
steps up to make it happen, let's not publish shallow clones using
dumb protocols.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 server-info.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/server-info.c b/server-info.c
index 9ec744e..a8df6a5 100644
--- a/server-info.c
+++ b/server-info.c
@@ -33,6 +33,10 @@ static int update_info_refs(int force)
 	strcpy(path1 + len, "+");
 
 	safe_create_leading_directories(path0);
+	if (is_repository_shallow()) {
+		unlink(path0);
+		return error("info/refs not updated for shallow clone");
+	}
 	info_ref_fp = fopen(path1, "w");
 	if (!info_ref_fp)
 		return error("unable to update %s", path1);
@@ -217,6 +221,11 @@ static int update_info_packs(int force)
 	strcpy(name, infofile);
 	strcpy(name + namelen, "+");
 
+	if (is_repository_shallow()) {
+		unlink(infofile);
+		return error("info/packs not updated for shallow clone");
+	}
+
 	init_pack_info(infofile, force);
 
 	safe_create_leading_directories(name);
-- 
1.8.2.83.gc99314b

[PATCH v3 05/28] Advertise shallow graft information on the server end

[Nguyễn Thái Ngọc Duy]

If either receive-pack or upload-pack is called on a shallow
repository, shallow graft points will be sent after the ref
advertisement (but before the packet flush), so that the client has
the full "shape" of the server's commit graph.

This breaks the protocol for all clients trying to push to a shallow
repo, or fetch from one. Which is basically the same end result as
today's "is_repository_shallow() && die()" in receive-pack and
upload-pack. New clients will be made aware of shallow upstream and
can make use of this information.

Smart HTTP is not affected by this patch. Shallow support on
smart-http comes later separately.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/technical/pack-protocol.txt |  3 +++
 builtin/receive-pack.c                    |  4 +++-
 commit.h                                  |  1 +
 shallow.c                                 | 15 +++++++++++++++
 upload-pack.c                             |  6 ++++--
 5 files changed, 26 insertions(+), 3 deletions(-)

diff --git a/Documentation/technical/pack-protocol.txt b/Documentation/technical/pack-protocol.txt
index b898e97..eb8edd1 100644
--- a/Documentation/technical/pack-protocol.txt
+++ b/Documentation/technical/pack-protocol.txt
@@ -161,6 +161,7 @@ MUST peel the ref if it's an annotated tag.
 
 ----
   advertised-refs  =  (no-refs / list-of-refs)
+		      *shallow
 		      flush-pkt
 
   no-refs          =  PKT-LINE(zero-id SP "capabilities^{}"
@@ -174,6 +175,8 @@ MUST peel the ref if it's an annotated tag.
   other-tip        =  obj-id SP refname LF
   other-peeled     =  obj-id SP refname "^{}" LF
 
+  shallow          =  PKT-LINE("shallow" SP obj-id)
+
   capability-list  =  capability *(SP capability)
   capability       =  1*(LC_ALPHA / DIGIT / "-" / "_")
   LC_ALPHA         =  %x61-7A
diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index 67ce1ef..cc8c34f 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -178,6 +178,8 @@ static void write_head_info(void)
 	if (!sent_capabilities)
 		show_ref("capabilities^{}", null_sha1);
 
+	advertise_shallow_grafts(1);
+
 	/* EOF */
 	packet_flush(1);
 }
@@ -998,7 +1000,7 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
 	if (!enter_repo(dir, 0))
 		die("'%s' does not appear to be a git repository", dir);
 
-	if (is_repository_shallow())
+	if (is_repository_shallow() && stateless_rpc)
 		die("attempt to push into a shallow repository");
 
 	git_config(receive_pack_config, NULL);
diff --git a/commit.h b/commit.h
index bd841f4..a879526 100644
--- a/commit.h
+++ b/commit.h
@@ -205,6 +205,7 @@ extern int write_shallow_commits(struct strbuf *out, int use_pack_protocol);
 extern void setup_alternate_shallow(struct lock_file *shallow_lock,
 				    const char **alternate_shallow_file);
 extern char *setup_temporary_shallow(void);
+extern void advertise_shallow_grafts(int);
 
 int is_descendant_of(struct commit *, struct commit_list *);
 int in_merge_bases(struct commit *, struct commit *);
diff --git a/shallow.c b/shallow.c
index cdf37d6..9552512 100644
--- a/shallow.c
+++ b/shallow.c
@@ -220,3 +220,18 @@ void setup_alternate_shallow(struct lock_file *shallow_lock,
 		*alternate_shallow_file = "";
 	strbuf_release(&sb);
 }
+
+static int advertise_shallow_grafts_cb(const struct commit_graft *graft, void *cb)
+{
+	int fd = *(int*)cb;
+	if (graft->nr_parent == -1)
+		packet_write(fd, "shallow %s\n", sha1_to_hex(graft->sha1));
+	return 0;
+}
+
+void advertise_shallow_grafts(int fd)
+{
+	if (!is_repository_shallow())
+		return;
+	for_each_commit_graft(advertise_shallow_grafts_cb, &fd);
+}
diff --git a/upload-pack.c b/upload-pack.c
index c989a73..38b2a29 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -758,6 +758,7 @@ static void upload_pack(void)
 		reset_timeout();
 		head_ref_namespaced(send_ref, &symref);
 		for_each_namespaced_ref(send_ref, &symref);
+		advertise_shallow_grafts(1);
 		packet_flush(1);
 	} else {
 		head_ref_namespaced(mark_our_ref, NULL);
@@ -835,8 +836,9 @@ int main(int argc, char **argv)
 
 	if (!enter_repo(dir, strict))
 		die("'%s' does not appear to be a git repository", dir);
-	if (is_repository_shallow())
-		die("attempt to fetch/clone from a shallow repository");
+	if (is_repository_shallow() && stateless_rpc)
+		die("attempt to push into a shallow repository");
+
 	git_config(upload_pack_config, NULL);
 	upload_pack();
 	return 0;
-- 
1.8.2.83.gc99314b

[PATCH v3 06/28] connect.c: teach get_remote_heads to parse "shallow" lines

[Nguyễn Thái Ngọc Duy]

No callers pass a non-empty pointer as shallow_points at this
stage. As a result, all clients still refuse to talk to shallow
repository on the other end.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/fetch-pack.c |  2 +-
 builtin/send-pack.c  |  2 +-
 connect.c            | 12 +++++++++++-
 remote-curl.c        |  2 +-
 remote.h             |  3 ++-
 transport.c          |  7 ++++---
 6 files changed, 20 insertions(+), 8 deletions(-)

diff --git a/builtin/fetch-pack.c b/builtin/fetch-pack.c
index c8e8582..c1d918f 100644
--- a/builtin/fetch-pack.c
+++ b/builtin/fetch-pack.c
@@ -150,7 +150,7 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
 				   args.verbose ? CONNECT_VERBOSE : 0);
 	}
 
-	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL);
+	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL, NULL);
 
 	ref = fetch_pack(&args, fd, conn, ref, dest,
 			 sought, nr_sought, pack_lockfile_ptr);
diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index 51121f2..bfa9253 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -233,7 +233,7 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 
 	memset(&extra_have, 0, sizeof(extra_have));
 
-	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL, &extra_have);
+	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL, &extra_have, NULL);
 
 	transport_verify_remote_names(nr_refspecs, refspecs);
 
diff --git a/connect.c b/connect.c
index 06e88b0..d0602b0 100644
--- a/connect.c
+++ b/connect.c
@@ -122,7 +122,8 @@ static void annotate_refs_with_symref_info(struct ref *ref)
  */
 struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
 			      struct ref **list, unsigned int flags,
-			      struct extra_have_objects *extra_have)
+			      struct extra_have_objects *extra_have,
+			      struct extra_have_objects *shallow_points)
 {
 	struct ref **orig_list = list;
 	int got_at_least_one_head = 0;
@@ -148,6 +149,15 @@ struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
 		if (len > 4 && !prefixcmp(buffer, "ERR "))
 			die("remote error: %s", buffer + 4);
 
+		if (len == 48 && !prefixcmp(buffer, "shallow ")) {
+			if (get_sha1_hex(buffer + 8, old_sha1))
+				die("protocol error: expected shallow sha, got '%s'", buffer + 8);
+			if (!shallow_points)
+				die("repository on the other end cannot be shallow");
+			add_extra_have(shallow_points, old_sha1);
+			continue;
+		}
+
 		if (len < 42 || get_sha1_hex(buffer, old_sha1) || buffer[40] != ' ')
 			die("protocol error: expected sha/ref, got '%s'", buffer);
 		name = buffer + 41;
diff --git a/remote-curl.c b/remote-curl.c
index c9b891a..222210f 100644
--- a/remote-curl.c
+++ b/remote-curl.c
@@ -107,7 +107,7 @@ static struct ref *parse_git_refs(struct discovery *heads, int for_push)
 {
 	struct ref *list = NULL;
 	get_remote_heads(-1, heads->buf, heads->len, &list,
-			 for_push ? REF_NORMAL : 0, NULL);
+			 for_push ? REF_NORMAL : 0, NULL, NULL);
 	return list;
 }
 
diff --git a/remote.h b/remote.h
index 131130a..773faa9 100644
--- a/remote.h
+++ b/remote.h
@@ -143,7 +143,8 @@ struct extra_have_objects {
 };
 extern struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
 				     struct ref **list, unsigned int flags,
-				     struct extra_have_objects *);
+				     struct extra_have_objects *have,
+				     struct extra_have_objects *shallow);
 
 int resolve_remote_symref(struct ref *ref, struct ref *list);
 int ref_newer(const unsigned char *new_sha1, const unsigned char *old_sha1);
diff --git a/transport.c b/transport.c
index 7202b77..9c51767 100644
--- a/transport.c
+++ b/transport.c
@@ -511,7 +511,7 @@ static struct ref *get_refs_via_connect(struct transport *transport, int for_pus
 
 	connect_setup(transport, for_push, 0);
 	get_remote_heads(data->fd[0], NULL, 0, &refs,
-			 for_push ? REF_NORMAL : 0, &data->extra_have);
+			 for_push ? REF_NORMAL : 0, &data->extra_have, NULL);
 	data->got_remote_heads = 1;
 
 	return refs;
@@ -541,7 +541,8 @@ static int fetch_refs_via_pack(struct transport *transport,
 
 	if (!data->got_remote_heads) {
 		connect_setup(transport, 0, 0);
-		get_remote_heads(data->fd[0], NULL, 0, &refs_tmp, 0, NULL);
+		get_remote_heads(data->fd[0], NULL, 0, &refs_tmp, 0,
+				 NULL, NULL);
 		data->got_remote_heads = 1;
 	}
 
@@ -805,7 +806,7 @@ static int git_transport_push(struct transport *transport, struct ref *remote_re
 		struct ref *tmp_refs;
 		connect_setup(transport, 1, 0);
 
-		get_remote_heads(data->fd[0], NULL, 0, &tmp_refs, REF_NORMAL, NULL);
+		get_remote_heads(data->fd[0], NULL, 0, &tmp_refs, REF_NORMAL, NULL, NULL);
 		data->got_remote_heads = 1;
 	}
 
-- 
1.8.2.83.gc99314b

[PATCH v3 07/28] shallow.c: add remove_reachable_shallow_points()

[Nguyễn Thái Ngọc Duy]

When we receive a pack and the shallow points from another repository,
we may want to add more shallow points to current repo to make sure no
commits point to nowhere. However we do not want to add unnecessary
shallow points and cut our history short because the other end is a
shallow version of this repo. The output shallow points may or may not
be added to .git/shallow, depending on whether they are actually
reachable in the new pack.

This function filters such shallow points out, leaving ones that might
potentially be added. A simple has_sha1_file won't do because we may
have incomplete object islands (i.e. not connected to any refs) and
the shallow points are on these islands. In that case we want to keep
the shallow points as candidates until we figure out if the new pack
connects to such object islands.

Typical cases that use remove_reachable_shallow_points() are:

 - fetch from a repo that has longer history: in this case all remote
   shallow roots do not exist in the client repo, this function will
   be cheap as it just does a few lookup_commit_graft and
   has_sha1_file.

 - fetch from a repo that has exactly the same shallow root set
   (e.g. a clone from a shallow repo): this case may trigger
   in_merge_bases_many all the way to roots. An exception is made to
   avoid such costly path with a faith that .git/shallow does not
   usually points to unreachable commit islands.

 - push from a shallow repo that has shorter history than the
   server's: in_merge_bases_many() is unavoidable, so the longer
   history the client has the higher the server cost is. The cost may
   be reduced with the help of pack bitmaps, or commit cache, though.

 - push from a shallow clone that has exactly the same shallow root
   set: the same as the second fetch case above, .i.e. cheap by
   exception.

The function must be called before the new pack is installed, or we
won't know which objects are ours, which theirs.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 commit.h  |  3 +++
 connect.c |  2 +-
 remote.h  |  1 +
 shallow.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 50 insertions(+), 1 deletion(-)

diff --git a/commit.h b/commit.h
index a879526..98044e6 100644
--- a/commit.h
+++ b/commit.h
@@ -193,6 +193,7 @@ extern struct commit_list *get_octopus_merge_bases(struct commit_list *in);
 /* largest positive number a signed 32-bit integer can contain */
 #define INFINITE_DEPTH 0x7fffffff
 
+struct extra_have_objects;
 extern int register_shallow(const unsigned char *sha1);
 extern int unregister_shallow(const unsigned char *sha1);
 extern int for_each_commit_graft(each_commit_graft_fn, void *);
@@ -206,6 +207,8 @@ extern void setup_alternate_shallow(struct lock_file *shallow_lock,
 				    const char **alternate_shallow_file);
 extern char *setup_temporary_shallow(void);
 extern void advertise_shallow_grafts(int);
+extern void remove_reachable_shallow_points(struct extra_have_objects *out,
+					    const struct extra_have_objects *in);
 
 int is_descendant_of(struct commit *, struct commit_list *);
 int in_merge_bases(struct commit *, struct commit *);
diff --git a/connect.c b/connect.c
index d0602b0..80e4360 100644
--- a/connect.c
+++ b/connect.c
@@ -45,7 +45,7 @@ int check_ref_type(const struct ref *ref, int flags)
 	return check_ref(ref->name, strlen(ref->name), flags);
 }
 
-static void add_extra_have(struct extra_have_objects *extra, unsigned char *sha1)
+void add_extra_have(struct extra_have_objects *extra, unsigned char *sha1)
 {
 	ALLOC_GROW(extra->array, extra->nr + 1, extra->alloc);
 	hashcpy(&(extra->array[extra->nr][0]), sha1);
diff --git a/remote.h b/remote.h
index 773faa9..ff604ff 100644
--- a/remote.h
+++ b/remote.h
@@ -145,6 +145,7 @@ extern struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
 				     struct ref **list, unsigned int flags,
 				     struct extra_have_objects *have,
 				     struct extra_have_objects *shallow);
+extern void add_extra_have(struct extra_have_objects *extra, unsigned char *sha1);
 
 int resolve_remote_symref(struct ref *ref, struct ref *list);
 int ref_newer(const unsigned char *new_sha1, const unsigned char *old_sha1);
diff --git a/shallow.c b/shallow.c
index 9552512..a974d2d 100644
--- a/shallow.c
+++ b/shallow.c
@@ -2,6 +2,8 @@
 #include "commit.h"
 #include "tag.h"
 #include "pkt-line.h"
+#include "remote.h"
+#include "refs.h"
 
 static int is_shallow = -1;
 static struct stat shallow_stat;
@@ -235,3 +237,46 @@ void advertise_shallow_grafts(int fd)
 		return;
 	for_each_commit_graft(advertise_shallow_grafts_cb, &fd);
 }
+
+struct commit_array {
+	struct commit **commits;
+	int nr, alloc;
+};
+
+static int add_ref(const char *refname,
+		   const unsigned char *sha1, int flags, void *cb_data)
+{
+	struct commit_array *ca = cb_data;
+	ALLOC_GROW(ca->commits, ca->nr + 1, ca->alloc);
+	ca->commits[ca->nr++] = lookup_commit(sha1);
+	return 0;
+}
+
+void remove_reachable_shallow_points(struct extra_have_objects *out,
+				     const struct extra_have_objects *in)
+{
+	struct commit_array ca;
+	int i;
+
+	memset(&ca, 0, sizeof(ca));
+	head_ref(add_ref, &ca);
+	for_each_ref(add_ref, &ca);
+	for (i = 0; i < in->nr; i++) {
+		struct commit_graft *graft = lookup_commit_graft(in->array[i]);
+		/*
+		 * For a clone from a shallow upstream, the clone has
+		 * the same shallow roots as upstream and it will
+		 * trigger in_merge_bases_many() all the way to roots.
+		 * Avoid that costly path and assume .git/shallow is
+		 * good most of the time.
+		 */
+		if (graft && graft->nr_parent < 0)
+			continue;
+		if (has_sha1_file(in->array[i]) &&
+		    in_merge_bases_many(lookup_commit(in->array[i]),
+					ca.nr, ca.commits))
+			continue;
+		add_extra_have(out, in->array[i]);
+	}
+	free(ca.commits);
+}
-- 
1.8.2.83.gc99314b

[PATCH v3 08/28] shallow.c: add mark_new_shallow_refs()

[Nguyễn Thái Ngọc Duy]

When we receive a pack and the shallow points from another repository,
we may need to add more shallow points to current repo to make sure no
commits point to nowhere. But usually we don't want to do so because
(in future) new shallow points invalidate pack bitmaps and we need to
rebuild them again, which is not cheap.

So the default way is we allow ref updates that do not introduce new
shallow points and mark the others. If the user is fine with new
shallow point addition, we accept the marked refs.

But even so we do not blindly accept all shallow points provided. Some
of them might not point to any commits in the new pack. Some might
even do, but those might be unreachable object islands. Only shallow
points that are reachable from old and new refs can stay.

The way it's implemented is paint down from each ref, attach a bitmap
to each commit where one bit represents one ref. In order to avoid
allocating new bitmap for every commit, we try to reuse the same
bitmap for parent commits if possible. This reduces allocation and
leaks deliberately because it's hard to keep/time consuming track how
many pointers to the same buffer.

In a typical push or fetch, the new pack should not carry new shallow
roots. If the current repo does not have any commit islands refered by
the sender's shallow roots either, this function is just a few
has_sha1_file(). So quite cheap.

Once the sender diverts from that path (or the receiver detects
shallow roots attached to commit islands from remove_reachable_shallow_points),
it'll be a lot more expensive. Pack bitmaps won't help this kind of
commit traversal, but commit cache might.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 commit.h  |   4 ++
 shallow.c | 219 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 223 insertions(+)

diff --git a/commit.h b/commit.h
index 98044e6..e1fd587 100644
--- a/commit.h
+++ b/commit.h
@@ -194,6 +194,7 @@ extern struct commit_list *get_octopus_merge_bases(struct commit_list *in);
 #define INFINITE_DEPTH 0x7fffffff
 
 struct extra_have_objects;
+struct ref;
 extern int register_shallow(const unsigned char *sha1);
 extern int unregister_shallow(const unsigned char *sha1);
 extern int for_each_commit_graft(each_commit_graft_fn, void *);
@@ -209,6 +210,9 @@ extern char *setup_temporary_shallow(void);
 extern void advertise_shallow_grafts(int);
 extern void remove_reachable_shallow_points(struct extra_have_objects *out,
 					    const struct extra_have_objects *in);
+extern int mark_new_shallow_refs(const struct extra_have_objects *ref,
+				 int *ref_status, uint32_t **used,
+				 const struct extra_have_objects *shallow);
 
 int is_descendant_of(struct commit *, struct commit_list *);
 int in_merge_bases(struct commit *, struct commit *);
diff --git a/shallow.c b/shallow.c
index a974d2d..c92a1dc 100644
--- a/shallow.c
+++ b/shallow.c
@@ -4,6 +4,8 @@
 #include "pkt-line.h"
 #include "remote.h"
 #include "refs.h"
+#include "diff.h"
+#include "revision.h"
 
 static int is_shallow = -1;
 static struct stat shallow_stat;
@@ -280,3 +282,220 @@ void remove_reachable_shallow_points(struct extra_have_objects *out,
 	}
 	free(ca.commits);
 }
+
+static int paint_down(const unsigned char *sha1, int id, int nr_bits, int quick)
+{
+	int hit_bottom = 0;
+	unsigned int i, nr;
+	struct commit_list *head = NULL;
+	int bitmap_nr = (nr_bits + 31) / 32;
+	int bitmap_size = bitmap_nr * sizeof(uint32_t);
+	uint32_t *tmp = xmalloc(bitmap_size);
+	uint32_t *bitmap = xcalloc(bitmap_size, sizeof(uint32_t));
+	bitmap[id / 32] |= (1 << (id % 32));
+	commit_list_insert(lookup_commit(sha1), &head);
+	while (head) {
+		struct commit_list *p;
+		struct commit *c = head->item;
+		uint32_t *c_util = c->util;
+
+		p = head;
+		head = head->next;
+		free(p);
+
+		if (c->object.flags & (SEEN | UNINTERESTING))
+			continue;
+		else
+			c->object.flags |= SEEN;
+
+		if (c->util == NULL)
+			c->util = bitmap;
+		else {
+			/*
+			 * Deliberately leak a lot in commit->util
+			 * because there can be many pointers to the
+			 * same bitmap. Probably should allocate in a
+			 * pool and free the whole pool at the end.
+			 */
+			memcpy(tmp, c_util, bitmap_size);
+			for (i = 0; i < bitmap_nr; i++)
+				tmp[i] |= bitmap[i];
+			if (memcmp(tmp, c_util, bitmap_size)) {
+				c->util = xmalloc(bitmap_size);
+				memcpy(c->util, tmp, bitmap_size);
+			}
+		}
+
+		if (c->object.flags & BOTTOM) {
+			hit_bottom = 1;
+			if (quick) {
+				free_commit_list(head);
+				break;
+			} else
+				continue;
+		}
+
+		if (parse_commit(c))
+			die("unable to parse commit %s",
+			    sha1_to_hex(c->object.sha1));
+
+		for (p = c->parents; p; p = p->next) {
+			if (p->item->object.flags & SEEN)
+				continue;
+			if (p->item->util == NULL || p->item->util == c_util)
+				p->item->util = c->util;
+			commit_list_insert(p->item, &head);
+		}
+	}
+
+	nr = get_max_object_index();
+	for (i = 0; i < nr; i++) {
+		struct object *o = get_indexed_object(i);
+		if (o && o->type == OBJ_COMMIT) {
+			o->flags &= ~SEEN;
+		}
+	}
+
+	free(tmp);
+	return hit_bottom;
+}
+
+static int mark_uninteresting(const char *refname,
+			      const unsigned char *sha1,
+			      int flags, void *cb_data)
+{
+	struct commit *commit = lookup_commit(sha1);
+	commit->object.flags |= UNINTERESTING;
+	mark_parents_uninteresting(commit);
+	return 0;
+}
+
+struct saved_util {
+	unsigned char sha1[20];
+	void *util;
+};
+
+/*
+ * Given a set of refs and shallow roots, find out what ref can reach
+ * any of the given roots. If so mark that ref "reject_flag". If
+ * "used" is not NULL, mark all reachable roots. Return how many refs
+ * that need new shallow points.
+ */
+int mark_new_shallow_refs(const struct extra_have_objects *ref,
+			  int *ref_status, uint32_t **used,
+			  const struct extra_have_objects *shallow)
+{
+	struct saved_util *util = NULL;
+	unsigned int i, nr, ret = 0, nr_util = 0, alloc_util = 0;
+
+	/*
+	 * Quick check to see if we may need to add new shallow
+	 * roots. Go through the list of root candidates and check if
+	 * they exist (either in current repo, or in the new pack, we
+	 * can't distinguish).
+	 *
+	 * 1) If none of the new roots exist, the pack must connect to
+	 *    the main object graph, which is already guarded by
+	 *    current repo's shallow roots and we will not need to
+	 *    consider adding new shallow roots, so we can exit early.
+	 *
+	 * 2) The pack may connect to some existing object islands in
+	 *    current repo then add shallow roots to plug loose ends
+	 *    from those islands. In that case, new shallow roots must
+	 *    also exist in the repo as this stage (old objects plus
+	 *    the new pack).
+	 *
+	 * 3) The last, easiest case, is the pack contains some
+	 *    shallow roots, which may be used to tie up loose ends of
+	 *    some new refs, or redundanty (tying up loose ends of new
+	 *    object islands)
+	 */
+	for (i = 0;i < shallow->nr; i++)
+		if (has_sha1_file(shallow->array[i]))
+			break;
+	if (i == shallow->nr)
+		/*
+		 * this is the first and also the common case, where
+		 * the new pack does not carry any new shallow
+		 * points. No need to to the expensive commit traverse
+		 * dance below.
+		 */
+		return 0;
+
+	/*
+	 * Prepare the commit graph to track what refs can reach what
+	 * (new) shallow points.
+	 */
+	nr = get_max_object_index();
+	for (i = 0; i < nr; i++) {
+		struct object *o = get_indexed_object(i);
+		struct commit *c = (struct commit *)o;
+		if (!o || o->type != OBJ_COMMIT)
+			continue;
+
+		o->flags &= ~(UNINTERESTING | BOTTOM | SEEN);
+		/*
+		 * git-fetch makes use of "util" field. Save it and
+		 * restore later. For fetch/clone/push, "nr" should be
+		 * small because rev-list is delayed to pack-objects.
+		 */
+		if (c->util) {
+			ALLOC_GROW(util, nr_util+1, alloc_util);
+			hashcpy(util[nr_util].sha1, o->sha1);
+			util[nr_util].util = c->util;
+			nr_util++;
+			c->util = NULL;
+		}
+	}
+
+	/*
+	 * "--not --all" to cut short the traversal if new refs
+	 * connect to old refs. If not (e.g. force ref updates) it'll
+	 * have to go down to the current shallow roots.
+	 *
+	 * We could detect that a new commit is connected to an
+	 * existing commit by keeping new objects in a pack (i.e. the
+	 * index-pack code path) then check commit origin. If so stop
+	 * short, so we don't need to get to the bottom. But then it
+	 * will not work for case #2 because we need to go through
+	 * some of our commits before reaching new shallow roots.
+	 */
+	head_ref(mark_uninteresting, NULL);
+	for_each_ref(mark_uninteresting, NULL);
+
+	for (i = 0; i < shallow->nr; i++)
+		if (has_sha1_file(shallow->array[i])) {
+			struct commit *c = lookup_commit(shallow->array[i]);
+			c->object.flags |= BOTTOM;
+		}
+
+	for (i = 0; i < ref->nr; i++)
+		if (paint_down(ref->array[i], i, ref->nr, used == NULL)) {
+			if (ref_status)
+				ref_status[i] = 1;
+			ret++;
+		}
+
+	if (used) {
+		for (i = 0; i < shallow->nr; i++) {
+			struct commit *c = lookup_commit(shallow->array[i]);
+			used[i] = c->util;
+		}
+	}
+
+	if (nr_util) {
+		nr = get_max_object_index();
+		for (i = 0; i < nr; i++) {
+			struct object *o = get_indexed_object(i);
+			if (o && o->type == OBJ_COMMIT)
+				((struct commit *)o)->util = NULL;
+		}
+		for (i = 0; i < nr_util; i++) {
+			struct commit *c = lookup_commit(util[i].sha1);
+			c->util = util[i].util;
+		}
+		free(util);
+	}
+
+	return ret;
+}
-- 
1.8.2.83.gc99314b

[PATCH v3 09/28] shallow.c: extend setup_*_shallow() to accept extra shallow points

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 commit.h      |  8 +++++---
 fetch-pack.c  |  5 +++--
 shallow.c     | 20 +++++++++++++++-----
 upload-pack.c |  2 +-
 4 files changed, 24 insertions(+), 11 deletions(-)

diff --git a/commit.h b/commit.h
index e1fd587..3af4699 100644
--- a/commit.h
+++ b/commit.h
@@ -203,10 +203,12 @@ extern struct commit_list *get_shallow_commits(struct object_array *heads,
 		int depth, int shallow_flag, int not_shallow_flag);
 extern void check_shallow_file_for_update(void);
 extern void set_alternate_shallow_file(const char *path);
-extern int write_shallow_commits(struct strbuf *out, int use_pack_protocol);
+extern int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
+				 const struct extra_have_objects *extra);
 extern void setup_alternate_shallow(struct lock_file *shallow_lock,
-				    const char **alternate_shallow_file);
-extern char *setup_temporary_shallow(void);
+				    const char **alternate_shallow_file,
+				    const struct extra_have_objects *extra);
+extern char *setup_temporary_shallow(const struct extra_have_objects *extra);
 extern void advertise_shallow_grafts(int);
 extern void remove_reachable_shallow_points(struct extra_have_objects *out,
 					    const struct extra_have_objects *in);
diff --git a/fetch-pack.c b/fetch-pack.c
index 1042448..0e7483e 100644
--- a/fetch-pack.c
+++ b/fetch-pack.c
@@ -311,7 +311,7 @@ static int find_common(struct fetch_pack_args *args,
 	}
 
 	if (is_repository_shallow())
-		write_shallow_commits(&req_buf, 1);
+		write_shallow_commits(&req_buf, 1, NULL);
 	if (args->depth > 0)
 		packet_buf_write(&req_buf, "deepen %d", args->depth);
 	packet_buf_flush(&req_buf);
@@ -850,7 +850,8 @@ static struct ref *do_fetch_pack(struct fetch_pack_args *args,
 	if (args->stateless_rpc)
 		packet_flush(fd[1]);
 	if (args->depth > 0)
-		setup_alternate_shallow(&shallow_lock, &alternate_shallow_file);
+		setup_alternate_shallow(&shallow_lock, &alternate_shallow_file,
+					NULL);
 	else
 		alternate_shallow_file = NULL;
 	if (get_pack(args, fd, pack_lockfile))
diff --git a/shallow.c b/shallow.c
index c92a1dc..2ec0c8c 100644
--- a/shallow.c
+++ b/shallow.c
@@ -169,22 +169,31 @@ static int write_one_shallow(const struct commit_graft *graft, void *cb_data)
 	return 0;
 }
 
-int write_shallow_commits(struct strbuf *out, int use_pack_protocol)
+int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
+			  const struct extra_have_objects *extra)
 {
 	struct write_shallow_data data;
+	int i;
 	data.out = out;
 	data.use_pack_protocol = use_pack_protocol;
 	data.count = 0;
 	for_each_commit_graft(write_one_shallow, &data);
+	if (!extra)
+		return data.count;
+	for (i = 0; i < extra->nr; i++) {
+		strbuf_addstr(out, sha1_to_hex(extra->array[i]));
+		strbuf_addch(out, '\n');
+		data.count++;
+	}
 	return data.count;
 }
 
-char *setup_temporary_shallow(void)
+char *setup_temporary_shallow(const struct extra_have_objects *extra)
 {
 	struct strbuf sb = STRBUF_INIT;
 	int fd;
 
-	if (write_shallow_commits(&sb, 0)) {
+	if (write_shallow_commits(&sb, 0, extra)) {
 		struct strbuf path = STRBUF_INIT;
 		strbuf_addstr(&path, git_path("shallow_XXXXXX"));
 		fd = xmkstemp(path.buf);
@@ -203,7 +212,8 @@ char *setup_temporary_shallow(void)
 }
 
 void setup_alternate_shallow(struct lock_file *shallow_lock,
-			     const char **alternate_shallow_file)
+			     const char **alternate_shallow_file,
+			     const struct extra_have_objects *extra)
 {
 	struct strbuf sb = STRBUF_INIT;
 	int fd;
@@ -211,7 +221,7 @@ void setup_alternate_shallow(struct lock_file *shallow_lock,
 	check_shallow_file_for_update();
 	fd = hold_lock_file_for_update(shallow_lock, git_path("shallow"),
 				       LOCK_DIE_ON_ERROR);
-	if (write_shallow_commits(&sb, 0)) {
+	if (write_shallow_commits(&sb, 0, extra)) {
 		if (write_in_full(fd, sb.buf, sb.len) != sb.len)
 			die_errno("failed to write to %s",
 				  shallow_lock->filename);
diff --git a/upload-pack.c b/upload-pack.c
index 38b2a29..f082f06 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -84,7 +84,7 @@ static void create_pack_file(void)
 	char *shallow_file = NULL;
 
 	if (shallow_nr) {
-		shallow_file = setup_temporary_shallow();
+		shallow_file = setup_temporary_shallow(NULL);
 		argv[arg++] = "--shallow-file";
 		argv[arg++] = shallow_file;
 	}
-- 
1.8.2.83.gc99314b

[PATCH v3 10/28] fetch-pack.c: move shallow update code out of fetch_pack()

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 fetch-pack.c | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)

diff --git a/fetch-pack.c b/fetch-pack.c
index 0e7483e..35d097e 100644
--- a/fetch-pack.c
+++ b/fetch-pack.c
@@ -925,6 +925,18 @@ static int remove_duplicates_in_refs(struct ref **ref, int nr)
 	return dst;
 }
 
+static void update_shallow(struct fetch_pack_args *args)
+{
+	if (args->depth > 0 && alternate_shallow_file) {
+		if (*alternate_shallow_file == '\0') { /* --unshallow */
+			unlink_or_warn(git_path("shallow"));
+			rollback_lock_file(&shallow_lock);
+		} else
+			commit_lock_file(&shallow_lock);
+		return;
+	}
+}
+
 struct ref *fetch_pack(struct fetch_pack_args *args,
 		       int fd[], struct child_process *conn,
 		       const struct ref *ref,
@@ -943,15 +955,7 @@ struct ref *fetch_pack(struct fetch_pack_args *args,
 		die("no matching remote head");
 	}
 	ref_cpy = do_fetch_pack(args, fd, ref, sought, nr_sought, pack_lockfile);
-
-	if (args->depth > 0 && alternate_shallow_file) {
-		if (*alternate_shallow_file == '\0') { /* --unshallow */
-			unlink_or_warn(git_path("shallow"));
-			rollback_lock_file(&shallow_lock);
-		} else
-			commit_lock_file(&shallow_lock);
-	}
-
+	update_shallow(args);
 	reprepare_packed_git();
 	return ref_cpy;
 }
-- 
1.8.2.83.gc99314b

[PATCH v3 11/28] fetch-pack.h: one statement per bitfield declaration

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 fetch-pack.h | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

diff --git a/fetch-pack.h b/fetch-pack.h
index 461cbf3..9b08388 100644
--- a/fetch-pack.h
+++ b/fetch-pack.h
@@ -8,18 +8,18 @@ struct fetch_pack_args {
 	const char *uploadpack;
 	int unpacklimit;
 	int depth;
-	unsigned quiet:1,
-		keep_pack:1,
-		lock_pack:1,
-		use_thin_pack:1,
-		fetch_all:1,
-		stdin_refs:1,
-		verbose:1,
-		no_progress:1,
-		include_tag:1,
-		stateless_rpc:1,
-		check_self_contained_and_connected:1,
-		self_contained_and_connected:1;
+	unsigned quiet:1;
+	unsigned keep_pack:1;
+	unsigned lock_pack:1;
+	unsigned use_thin_pack:1;
+	unsigned fetch_all:1;
+	unsigned stdin_refs:1;
+	unsigned verbose:1;
+	unsigned no_progress:1;
+	unsigned include_tag:1;
+	unsigned stateless_rpc:1;
+	unsigned check_self_contained_and_connected:1;
+	unsigned self_contained_and_connected:1;
 };
 
 /*
-- 
1.8.2.83.gc99314b

[PATCH v3 12/28] clone: support remote shallow repository

[Nguyễn Thái Ngọc Duy]

The gist of cloning a remote shallow repository is simple. The remote
gives you a pack and all its shallow roots. All you need to do is set
your .git/shallow correctly.

The rule to create this .git/shallow is simple and more importantly,
cheap: if a shallow root is found in the pack, it's probably used
(i.e. reachable from some refs), so we add it. Others are dropped.

One may notice this method seems flawed by the word "probably". A
shallow point may not be reachable from any refs at all if it's
attached to an object island (a group of objects that are not
reachable by any refs).

If that object island is not complete, a new fetch request may send
more objects to connect it to some ref. At that time, because we
incorrectly installed the shallow root in this island, the user will
not see anything after that root. This is not desired.

There is a stricter method to rule out unwanted shallow roots above:
mark_new_shallow_refs(). But it's expensive and even more so for a
clone because it's proportional to the (potentially huge) number of
commits.

Given that object islands are rare (C Git never sends such islands), a
tradeoff is made to surprise the user occasionally but work faster
everyday. --strict option could be added later to enforce
mark_new_shallow_refs().

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/clone.c      |  1 +
 builtin/fetch-pack.c |  2 +-
 fetch-pack.c         | 50 +++++++++++++++++++++++++++++++++++++++++++++++---
 fetch-pack.h         |  4 ++++
 transport.c          | 11 ++++++++---
 transport.h          |  6 ++++++
 6 files changed, 67 insertions(+), 7 deletions(-)

diff --git a/builtin/clone.c b/builtin/clone.c
index 900f564..0b182ce 100644
--- a/builtin/clone.c
+++ b/builtin/clone.c
@@ -889,6 +889,7 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
 
 	remote = remote_get(option_origin);
 	transport = transport_get(remote, remote->url[0]);
+	transport->cloning = 1;
 
 	if (!transport->get_refs_list || (!is_local && !transport->fetch))
 		die(_("Don't know how to clone %s"), transport->url);
diff --git a/builtin/fetch-pack.c b/builtin/fetch-pack.c
index c1d918f..927424b 100644
--- a/builtin/fetch-pack.c
+++ b/builtin/fetch-pack.c
@@ -153,7 +153,7 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
 	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL, NULL);
 
 	ref = fetch_pack(&args, fd, conn, ref, dest,
-			 sought, nr_sought, pack_lockfile_ptr);
+			 sought, nr_sought, NULL, pack_lockfile_ptr);
 	if (pack_lockfile) {
 		printf("lock %s\n", pack_lockfile);
 		fflush(stdout);
diff --git a/fetch-pack.c b/fetch-pack.c
index 35d097e..b76581a 100644
--- a/fetch-pack.c
+++ b/fetch-pack.c
@@ -774,6 +774,7 @@ static struct ref *do_fetch_pack(struct fetch_pack_args *args,
 				 int fd[2],
 				 const struct ref *orig_ref,
 				 struct ref **sought, int nr_sought,
+				 struct extra_have_objects *shallow,
 				 char **pack_lockfile)
 {
 	struct ref *ref = copy_ref_list(orig_ref);
@@ -852,6 +853,8 @@ static struct ref *do_fetch_pack(struct fetch_pack_args *args,
 	if (args->depth > 0)
 		setup_alternate_shallow(&shallow_lock, &alternate_shallow_file,
 					NULL);
+	else if (args->cloning && shallow && shallow->nr)
+		alternate_shallow_file = setup_temporary_shallow(shallow);
 	else
 		alternate_shallow_file = NULL;
 	if (get_pack(args, fd, pack_lockfile))
@@ -925,8 +928,11 @@ static int remove_duplicates_in_refs(struct ref **ref, int nr)
 	return dst;
 }
 
-static void update_shallow(struct fetch_pack_args *args)
+static void update_shallow(struct fetch_pack_args *args,
+			   struct extra_have_objects *shallow)
 {
+	int i;
+
 	if (args->depth > 0 && alternate_shallow_file) {
 		if (*alternate_shallow_file == '\0') { /* --unshallow */
 			unlink_or_warn(git_path("shallow"));
@@ -935,6 +941,42 @@ static void update_shallow(struct fetch_pack_args *args)
 			commit_lock_file(&shallow_lock);
 		return;
 	}
+
+	if (!shallow || !shallow->nr)
+		return;
+
+	if (alternate_shallow_file) {
+		/*
+		 * The temporary shallow file is only useful for
+		 * index-pack and unpack-objects because it may
+		 * contain more roots than we want. Delete it.
+		 */
+		if (*alternate_shallow_file)
+			unlink(alternate_shallow_file);
+		free((char*)alternate_shallow_file);
+	}
+
+	if (args->cloning) {
+		/*
+		 * remote is shallow, but this is a clone, there are
+		 * no objects in repo to worry about. Accept any
+		 * shallow points that exist in the pack (iow in repo
+		 * after get_pack() and reprepare_packed_git())
+		 */
+		struct extra_have_objects extra;
+		memset(&extra, 0, sizeof(extra));
+		for (i = 0; i < shallow->nr; i++)
+			if (has_sha1_file(shallow->array[i]))
+				add_extra_have(&extra, shallow->array[i]);
+		if (extra.nr) {
+			setup_alternate_shallow(&shallow_lock,
+						&alternate_shallow_file,
+						&extra);
+			commit_lock_file(&shallow_lock);
+		}
+		free(extra.array);
+		return;
+	}
 }
 
 struct ref *fetch_pack(struct fetch_pack_args *args,
@@ -942,6 +984,7 @@ struct ref *fetch_pack(struct fetch_pack_args *args,
 		       const struct ref *ref,
 		       const char *dest,
 		       struct ref **sought, int nr_sought,
+		       struct extra_have_objects *shallow,
 		       char **pack_lockfile)
 {
 	struct ref *ref_cpy;
@@ -954,8 +997,9 @@ struct ref *fetch_pack(struct fetch_pack_args *args,
 		packet_flush(fd[1]);
 		die("no matching remote head");
 	}
-	ref_cpy = do_fetch_pack(args, fd, ref, sought, nr_sought, pack_lockfile);
-	update_shallow(args);
+	ref_cpy = do_fetch_pack(args, fd, ref, sought, nr_sought,
+				shallow, pack_lockfile);
 	reprepare_packed_git();
+	update_shallow(args, shallow);
 	return ref_cpy;
 }
diff --git a/fetch-pack.h b/fetch-pack.h
index 9b08388..cabfb60 100644
--- a/fetch-pack.h
+++ b/fetch-pack.h
@@ -4,6 +4,8 @@
 #include "string-list.h"
 #include "run-command.h"
 
+struct extra_have_objects;
+
 struct fetch_pack_args {
 	const char *uploadpack;
 	int unpacklimit;
@@ -20,6 +22,7 @@ struct fetch_pack_args {
 	unsigned stateless_rpc:1;
 	unsigned check_self_contained_and_connected:1;
 	unsigned self_contained_and_connected:1;
+	unsigned cloning:1;
 };
 
 /*
@@ -33,6 +36,7 @@ struct ref *fetch_pack(struct fetch_pack_args *args,
 		       const char *dest,
 		       struct ref **sought,
 		       int nr_sought,
+		       struct extra_have_objects *shallow,
 		       char **pack_lockfile);
 
 #endif
diff --git a/transport.c b/transport.c
index 9c51767..fa3dc16 100644
--- a/transport.c
+++ b/transport.c
@@ -455,6 +455,7 @@ struct git_transport_data {
 	int fd[2];
 	unsigned got_remote_heads : 1;
 	struct extra_have_objects extra_have;
+	struct extra_have_objects shallow;
 };
 
 static int set_git_option(struct git_transport_options *opts,
@@ -511,7 +512,9 @@ static struct ref *get_refs_via_connect(struct transport *transport, int for_pus
 
 	connect_setup(transport, for_push, 0);
 	get_remote_heads(data->fd[0], NULL, 0, &refs,
-			 for_push ? REF_NORMAL : 0, &data->extra_have, NULL);
+			 for_push ? REF_NORMAL : 0,
+			 &data->extra_have,
+			 transport->cloning ? &data->shallow : NULL);
 	data->got_remote_heads = 1;
 
 	return refs;
@@ -538,17 +541,19 @@ static int fetch_refs_via_pack(struct transport *transport,
 	args.depth = data->options.depth;
 	args.check_self_contained_and_connected =
 		data->options.check_self_contained_and_connected;
+	args.cloning = transport->cloning;
 
 	if (!data->got_remote_heads) {
 		connect_setup(transport, 0, 0);
 		get_remote_heads(data->fd[0], NULL, 0, &refs_tmp, 0,
-				 NULL, NULL);
+				 NULL,
+				 transport->cloning ? &data->shallow : NULL);
 		data->got_remote_heads = 1;
 	}
 
 	refs = fetch_pack(&args, data->fd, data->conn,
 			  refs_tmp ? refs_tmp : transport->remote_refs,
-			  dest, to_fetch, nr_heads,
+			  dest, to_fetch, nr_heads, &data->shallow,
 			  &transport->pack_lockfile);
 	close(data->fd[0]);
 	close(data->fd[1]);
diff --git a/transport.h b/transport.h
index b3679bb..59842d4 100644
--- a/transport.h
+++ b/transport.h
@@ -35,6 +35,12 @@ struct transport {
 	 */
 	unsigned cannot_reuse : 1;
 
+	/*
+	 * A hint from caller that it will be performing a clone, not
+	 * normal fetch. IOW the repository is guaranteed empty.
+	 */
+	unsigned cloning : 1;
+
 	/**
 	 * Returns 0 if successful, positive if the option is not
 	 * recognized or is inapplicable, and negative if the option
-- 
1.8.2.83.gc99314b

[PATCH v3 13/28] fetch: support fetching from a shallow repository

[Nguyễn Thái Ngọc Duy]

This patch just put together pieces from the previous patches:

 - Before getting the new pack, we need to remove all new reachable
   shallow roots. The remaining roots may or may not be added to
   .git/shallow.

 - After getting the pack, walk all new refs until they connect to
   current refs, or hit the bottom of current repo, or hit new
   shallow roots.

Those refs that hit new shallow roots are rejected because by default
we do not allow to update .git/shallow (the only exception so far is
cloning from a shallow repo, which is more like creating .git/shallow
than updating it)

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/fetch.c                   |   9 +++
 fetch-pack.c                      |  35 ++++++++++-
 remote.h                          |   1 +
 t/t5536-fetch-shallow.sh (new +x) | 128 ++++++++++++++++++++++++++++++++++++++
 transport.c                       |  11 +++-
 5 files changed, 179 insertions(+), 5 deletions(-)
 create mode 100755 t/t5536-fetch-shallow.sh

diff --git a/builtin/fetch.c b/builtin/fetch.c
index bd7a101..7b41a7e 100644
--- a/builtin/fetch.c
+++ b/builtin/fetch.c
@@ -405,6 +405,8 @@ static int iterate_ref_map(void *cb_data, unsigned char sha1[20])
 	struct ref **rm = cb_data;
 	struct ref *ref = *rm;
 
+	while (ref && ref->status == REF_STATUS_REJECT_SHALLOW)
+		ref = ref->next;
 	if (!ref)
 		return -1; /* end of the list */
 	*rm = ref->next;
@@ -451,6 +453,13 @@ static int store_updated_refs(const char *raw_url, const char *remote_name,
 			struct ref *ref = NULL;
 			const char *merge_status_marker = "";
 
+			if (rm->status == REF_STATUS_REJECT_SHALLOW) {
+				if (want_status == FETCH_HEAD_MERGE)
+					warning(_("reject %s because shallow roots are not allowed to be updated"),
+						rm->peer_ref ? rm->peer_ref->name : rm->name);
+				continue;
+			}
+
 			commit = lookup_commit_reference_gently(rm->old_sha1, 1);
 			if (!commit)
 				rm->fetch_head_status = FETCH_HEAD_NOT_FOR_MERGE;
diff --git a/fetch-pack.c b/fetch-pack.c
index b76581a..64fa5d2 100644
--- a/fetch-pack.c
+++ b/fetch-pack.c
@@ -855,7 +855,17 @@ static struct ref *do_fetch_pack(struct fetch_pack_args *args,
 					NULL);
 	else if (args->cloning && shallow && shallow->nr)
 		alternate_shallow_file = setup_temporary_shallow(shallow);
-	else
+	else if (!args->cloning && shallow && shallow->nr) {
+		struct extra_have_objects extra;
+		memset(&extra, 0, sizeof(extra));
+		remove_reachable_shallow_points(&extra, shallow);
+		if (extra.nr) {
+			alternate_shallow_file = setup_temporary_shallow(&extra);
+			free(shallow->array);
+			*shallow = extra;
+		} else
+			alternate_shallow_file = NULL;
+	} else
 		alternate_shallow_file = NULL;
 	if (get_pack(args, fd, pack_lockfile))
 		die("git fetch-pack: fetch failed.");
@@ -929,8 +939,11 @@ static int remove_duplicates_in_refs(struct ref **ref, int nr)
 }
 
 static void update_shallow(struct fetch_pack_args *args,
+			   struct ref **sought, int nr_sought,
 			   struct extra_have_objects *shallow)
 {
+	struct extra_have_objects ref;
+	int *status;
 	int i;
 
 	if (args->depth > 0 && alternate_shallow_file) {
@@ -977,6 +990,24 @@ static void update_shallow(struct fetch_pack_args *args,
 		free(extra.array);
 		return;
 	}
+
+	memset(&ref, 0, sizeof(ref));
+	for (i = 0; i < nr_sought; i++)
+		add_extra_have(&ref, sought[i]->old_sha1);
+
+	status = xcalloc(nr_sought, sizeof(*status));
+
+	/*
+	 * remote is also shallow, check what ref is safe to update
+	 * without updating .git/shallow
+	 */
+	if (mark_new_shallow_refs(&ref, status, NULL, shallow)) {
+		for (i = 0; i < nr_sought; i++)
+			if (status[i])
+				sought[i]->status = REF_STATUS_REJECT_SHALLOW;
+	}
+	free(status);
+	free(ref.array);
 }
 
 struct ref *fetch_pack(struct fetch_pack_args *args,
@@ -1000,6 +1031,6 @@ struct ref *fetch_pack(struct fetch_pack_args *args,
 	ref_cpy = do_fetch_pack(args, fd, ref, sought, nr_sought,
 				shallow, pack_lockfile);
 	reprepare_packed_git();
-	update_shallow(args, shallow);
+	update_shallow(args, sought, nr_sought, shallow);
 	return ref_cpy;
 }
diff --git a/remote.h b/remote.h
index ff604ff..e519c26 100644
--- a/remote.h
+++ b/remote.h
@@ -109,6 +109,7 @@ struct ref {
 		REF_STATUS_REJECT_FETCH_FIRST,
 		REF_STATUS_REJECT_NEEDS_FORCE,
 		REF_STATUS_REJECT_STALE,
+		REF_STATUS_REJECT_SHALLOW,
 		REF_STATUS_UPTODATE,
 		REF_STATUS_REMOTE_REJECT,
 		REF_STATUS_EXPECTING_REPORT
diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
new file mode 100755
index 0000000..41de74d
--- /dev/null
+++ b/t/t5536-fetch-shallow.sh
@@ -0,0 +1,128 @@
+#!/bin/sh
+
+test_description='fetch/clone from a shallow clone'
+
+. ./test-lib.sh
+
+commit() {
+	echo "$1" >tracked &&
+	git add tracked &&
+	git commit -m "$1"
+}
+
+test_expect_success 'setup' '
+	commit 1 &&
+	commit 2 &&
+	commit 3 &&
+	commit 4 &&
+	git config --global transfer.fsckObjects true
+'
+
+test_expect_success 'setup shallow clone' '
+	git clone --no-local --depth=2 .git shallow &&
+	git --git-dir=shallow/.git log --format=%s >actual &&
+	cat <<EOF >expect &&
+4
+3
+EOF
+	test_cmp expect actual
+'
+
+test_expect_success 'clone from shallow clone' '
+	git clone --no-local shallow shallow2 &&
+	(
+	cd shallow2 &&
+	git fsck &&
+	git log --format=%s >actual &&
+	cat <<EOF >expect &&
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
+test_expect_success 'fetch from shallow clone' '
+	(
+	cd shallow &&
+	commit 5
+	) &&
+	(
+	cd shallow2 &&
+	git fetch &&
+	git fsck &&
+	git log --format=%s origin/master >actual &&
+	cat <<EOF >expect &&
+5
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
+test_expect_success 'fetch --depth from shallow clone' '
+	(
+	cd shallow &&
+	commit 6
+	) &&
+	(
+	cd shallow2 &&
+	git fetch --depth=2 &&
+	git fsck &&
+	git log --format=%s origin/master >actual &&
+	cat <<EOF >expect &&
+6
+5
+EOF
+	test_cmp expect actual
+	)
+'
+
+test_expect_success 'fetch something upstream has but hidden by clients shallow boundaries' '
+	# the blob "1" is available in .git but hidden by the
+	# shallow2/.git/shallow and it should be resent
+	! git --git-dir=shallow2/.git cat-file blob `echo 1|git hash-object --stdin` >/dev/null &&
+	echo 1 > 1.t &&
+	git add 1.t &&
+	git commit -m add-1-back &&
+	(
+	cd shallow2 &&
+	git fetch ../.git +refs/heads/master:refs/remotes/top/master &&
+	git fsck &&
+	git log --format=%s top/master >actual &&
+	cat <<EOF >expect &&
+add-1-back
+4
+3
+EOF
+	test_cmp expect actual
+	) &&
+	git --git-dir=shallow2/.git cat-file blob `echo 1|git hash-object --stdin` >/dev/null
+
+'
+
+test_expect_success 'fetch that requires changes in .git/shallow is filtered' '
+	(
+	cd shallow &&
+	git checkout --orphan no-shallow &&
+	commit no-shallow
+	) &&
+	git init notshallow &&
+	(
+	cd notshallow &&
+	git fetch ../shallow/.git refs/heads/*:refs/remotes/shallow/*&&
+	git for-each-ref --format="%(refname)" >actual.refs &&
+	cat <<EOF >expect.refs &&
+refs/remotes/shallow/no-shallow
+EOF
+	test_cmp expect.refs actual.refs &&
+	git log --format=%s shallow/no-shallow >actual &&
+	cat <<EOF >expect &&
+no-shallow
+EOF
+	test_cmp expect actual
+	)
+'
+
+test_done
diff --git a/transport.c b/transport.c
index fa3dc16..d6d14eb 100644
--- a/transport.c
+++ b/transport.c
@@ -514,7 +514,7 @@ static struct ref *get_refs_via_connect(struct transport *transport, int for_pus
 	get_remote_heads(data->fd[0], NULL, 0, &refs,
 			 for_push ? REF_NORMAL : 0,
 			 &data->extra_have,
-			 transport->cloning ? &data->shallow : NULL);
+			 &data->shallow);
 	data->got_remote_heads = 1;
 
 	return refs;
@@ -546,8 +546,7 @@ static int fetch_refs_via_pack(struct transport *transport,
 	if (!data->got_remote_heads) {
 		connect_setup(transport, 0, 0);
 		get_remote_heads(data->fd[0], NULL, 0, &refs_tmp, 0,
-				 NULL,
-				 transport->cloning ? &data->shallow : NULL);
+				 NULL, &data->shallow);
 		data->got_remote_heads = 1;
 	}
 
@@ -719,6 +718,10 @@ static int print_one_push_status(struct ref *ref, const char *dest, int count, i
 		print_ref_status('!', "[rejected]", ref, ref->peer_ref,
 						 "stale info", porcelain);
 		break;
+	case REF_STATUS_REJECT_SHALLOW:
+		print_ref_status('!', "[rejected]", ref, ref->peer_ref,
+						 "new shallow roots not allowed", porcelain);
+		break;
 	case REF_STATUS_REMOTE_REJECT:
 		print_ref_status('!', "[remote rejected]", ref,
 						 ref->deletion ? NULL : ref->peer_ref,
@@ -814,6 +817,8 @@ static int git_transport_push(struct transport *transport, struct ref *remote_re
 		get_remote_heads(data->fd[0], NULL, 0, &tmp_refs, REF_NORMAL, NULL, NULL);
 		data->got_remote_heads = 1;
 	}
+	if (data->shallow.nr)
+		die("pushing to a shallow repository is not supported");
 
 	memset(&args, 0, sizeof(args));
 	args.send_mirror = !!(flags & TRANSPORT_PUSH_MIRROR);
-- 
1.8.2.83.gc99314b

[PATCH v3 14/28] upload-pack: make sure deepening preserves shallow roots

[Nguyễn Thái Ngọc Duy]

When "fetch --depth=N" where N exceeds the longest chain of history in
the source repo, usually we just send an "unshallow" line to the
client so full history is obtained.

When the source repo is shallow we need to make sure to "unshallow"
the current shallow point _and_ "shallow" again when the commit
reaches its shallow bottom in the source repo.

This should fix both cases: large <N> and --unshallow.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/fetch-options.txt |  8 ++++++--
 shallow.c                       |  6 +++++-
 t/t5536-fetch-shallow.sh        | 16 ++++++++++++++++
 upload-pack.c                   |  2 +-
 4 files changed, 28 insertions(+), 4 deletions(-)

diff --git a/Documentation/fetch-options.txt b/Documentation/fetch-options.txt
index ba1fe49..a83d2b4 100644
--- a/Documentation/fetch-options.txt
+++ b/Documentation/fetch-options.txt
@@ -14,8 +14,12 @@
 	branch history. Tags for the deepened commits are not fetched.
 
 --unshallow::
-	Convert a shallow repository to a complete one, removing all
-	the limitations imposed by shallow repositories.
+	If the source repository is complete, convert a shallow
+	repository to a complete one, removing all the limitations
+	imposed by shallow repositories.
++
+If the source repository is shallow, fetch as much as possible so that
+the current repository has the same history as the source repository.
 
 ifndef::git-pull[]
 --dry-run::
diff --git a/shallow.c b/shallow.c
index 2ec0c8c..1baf6b5 100644
--- a/shallow.c
+++ b/shallow.c
@@ -73,6 +73,7 @@ struct commit_list *get_shallow_commits(struct object_array *heads, int depth,
 	struct commit_list *result = NULL;
 	struct object_array stack = OBJECT_ARRAY_INIT;
 	struct commit *commit = NULL;
+	struct commit_graft *graft;
 
 	while (commit || i < heads->nr || stack.nr) {
 		struct commit_list *p;
@@ -97,7 +98,10 @@ struct commit_list *get_shallow_commits(struct object_array *heads, int depth,
 		if (parse_commit(commit))
 			die("invalid commit");
 		cur_depth++;
-		if (cur_depth >= depth) {
+		if ((depth != INFINITE_DEPTH && cur_depth >= depth) ||
+		    (is_repository_shallow() && !commit->parents &&
+		     (graft = lookup_commit_graft(commit->object.sha1)) != NULL &&
+		     graft->nr_parent < 0)) {
 			commit_list_insert(commit, &result);
 			commit->object.flags |= shallow_flag;
 			commit = NULL;
diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
index 41de74d..e011ead 100755
--- a/t/t5536-fetch-shallow.sh
+++ b/t/t5536-fetch-shallow.sh
@@ -79,6 +79,22 @@ EOF
 	)
 '
 
+test_expect_success 'fetch --unshallow from shallow clone' '
+	(
+	cd shallow2 &&
+	git fetch --unshallow &&
+	git fsck &&
+	git log --format=%s origin/master >actual &&
+	cat <<EOF >expect &&
+6
+5
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
 test_expect_success 'fetch something upstream has but hidden by clients shallow boundaries' '
 	# the blob "1" is available in .git but hidden by the
 	# shallow2/.git/shallow and it should be resent
diff --git a/upload-pack.c b/upload-pack.c
index f082f06..28269c7 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -619,7 +619,7 @@ static void receive_needs(void)
 	if (depth > 0) {
 		struct commit_list *result = NULL, *backup = NULL;
 		int i;
-		if (depth == INFINITE_DEPTH)
+		if (depth == INFINITE_DEPTH && !is_repository_shallow())
 			for (i = 0; i < shallows.nr; i++) {
 				struct object *object = shallows.objects[i].item;
 				object->flags |= NOT_SHALLOW;
-- 
1.8.2.83.gc99314b

[PATCH v3 15/28] fetch: add --update-shallow to get refs that require updating .git/shallow

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/fetch-options.txt |  6 ++++++
 builtin/fetch.c                 |  6 +++++-
 fetch-pack.c                    | 26 ++++++++++++++++++++++++++
 fetch-pack.h                    |  1 +
 t/t5536-fetch-shallow.sh        | 22 ++++++++++++++++++++++
 transport.c                     |  4 ++++
 transport.h                     |  4 ++++
 7 files changed, 68 insertions(+), 1 deletion(-)

diff --git a/Documentation/fetch-options.txt b/Documentation/fetch-options.txt
index a83d2b4..54043e3 100644
--- a/Documentation/fetch-options.txt
+++ b/Documentation/fetch-options.txt
@@ -21,6 +21,12 @@
 If the source repository is shallow, fetch as much as possible so that
 the current repository has the same history as the source repository.
 
+--update-shallow::
+	By default when fetching from a shallow repository,
+	`git fetch` refuses refs that require updating
+	.git/shallow. This option updates .git/shallow and accept such
+	refs.
+
 ifndef::git-pull[]
 --dry-run::
 	Show what would be done, without making any changes.
diff --git a/builtin/fetch.c b/builtin/fetch.c
index 7b41a7e..d2e4fc0 100644
--- a/builtin/fetch.c
+++ b/builtin/fetch.c
@@ -36,7 +36,7 @@ static int prune = -1; /* unspecified */
 
 static int all, append, dry_run, force, keep, multiple, update_head_ok, verbosity;
 static int progress = -1, recurse_submodules = RECURSE_SUBMODULES_DEFAULT;
-static int tags = TAGS_DEFAULT, unshallow;
+static int tags = TAGS_DEFAULT, unshallow, update_shallow;
 static const char *depth;
 static const char *upload_pack;
 static struct strbuf default_rla = STRBUF_INIT;
@@ -104,6 +104,8 @@ static struct option builtin_fetch_options[] = {
 	{ OPTION_STRING, 0, "recurse-submodules-default",
 		   &recurse_submodules_default, NULL,
 		   N_("default mode for recursion"), PARSE_OPT_HIDDEN },
+	OPT_BOOL(0, "update-shallow", &update_shallow,
+		 N_("accept refs that update .git/shallow")),
 	OPT_END()
 };
 
@@ -768,6 +770,8 @@ static struct transport *prepare_transport(struct remote *remote)
 		set_option(transport, TRANS_OPT_KEEP, "yes");
 	if (depth)
 		set_option(transport, TRANS_OPT_DEPTH, depth);
+	if (update_shallow)
+		set_option(transport, TRANS_OPT_UPDATE_SHALLOW, "yes");
 	return transport;
 }
 
diff --git a/fetch-pack.c b/fetch-pack.c
index 64fa5d2..82aa5db 100644
--- a/fetch-pack.c
+++ b/fetch-pack.c
@@ -995,6 +995,32 @@ static void update_shallow(struct fetch_pack_args *args,
 	for (i = 0; i < nr_sought; i++)
 		add_extra_have(&ref, sought[i]->old_sha1);
 
+	if (args->update_shallow) {
+		/*
+		 * remote is also shallow, .git/shallow may be updated
+		 * so all refs can be accepted. Make sure we only add
+		 * shallow roots that are actually reachable from new
+		 * refs.
+		 */
+		uint32_t **used = xmalloc(sizeof(*used) * shallow->nr);
+		if (mark_new_shallow_refs(&ref, NULL, used, shallow)) {
+			struct extra_have_objects extra;
+			memset(&extra, 0, sizeof(extra));
+			for (i = 0; i < shallow->nr; i++)
+				if (used[i])
+					add_extra_have(&extra,
+						       shallow->array[i]);
+			setup_alternate_shallow(&shallow_lock,
+						&alternate_shallow_file,
+						&extra);
+			commit_lock_file(&shallow_lock);
+			free(extra.array);
+		}
+		free(used);
+		free(ref.array);
+		return;
+	}
+
 	status = xcalloc(nr_sought, sizeof(*status));
 
 	/*
diff --git a/fetch-pack.h b/fetch-pack.h
index cabfb60..5cfb77b 100644
--- a/fetch-pack.h
+++ b/fetch-pack.h
@@ -23,6 +23,7 @@ struct fetch_pack_args {
 	unsigned check_self_contained_and_connected:1;
 	unsigned self_contained_and_connected:1;
 	unsigned cloning:1;
+	unsigned update_shallow:1;
 };
 
 /*
diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
index e011ead..95b6313 100755
--- a/t/t5536-fetch-shallow.sh
+++ b/t/t5536-fetch-shallow.sh
@@ -141,4 +141,26 @@ EOF
 	)
 '
 
+test_expect_success 'fetch --update-shallow' '
+	(
+	cd notshallow &&
+	git fetch --update-shallow ../shallow/.git refs/heads/*:refs/remotes/shallow/* &&
+	git fsck &&
+	git for-each-ref --format="%(refname)" |sort >actual.refs &&
+	cat <<EOF >expect.refs &&
+refs/remotes/shallow/master
+refs/remotes/shallow/no-shallow
+EOF
+	test_cmp expect.refs actual.refs &&
+	git log --format=%s shallow/master >actual &&
+	cat <<EOF >expect &&
+6
+5
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
 test_done
diff --git a/transport.c b/transport.c
index d6d14eb..c0be6b1 100644
--- a/transport.c
+++ b/transport.c
@@ -476,6 +476,9 @@ static int set_git_option(struct git_transport_options *opts,
 	} else if (!strcmp(name, TRANS_OPT_KEEP)) {
 		opts->keep = !!value;
 		return 0;
+	} else if (!strcmp(name, TRANS_OPT_UPDATE_SHALLOW)) {
+		opts->update_shallow = !!value;
+		return 0;
 	} else if (!strcmp(name, TRANS_OPT_DEPTH)) {
 		if (!value)
 			opts->depth = 0;
@@ -542,6 +545,7 @@ static int fetch_refs_via_pack(struct transport *transport,
 	args.check_self_contained_and_connected =
 		data->options.check_self_contained_and_connected;
 	args.cloning = transport->cloning;
+	args.update_shallow = data->options.update_shallow;
 
 	if (!data->got_remote_heads) {
 		connect_setup(transport, 0, 0);
diff --git a/transport.h b/transport.h
index 59842d4..02ea248 100644
--- a/transport.h
+++ b/transport.h
@@ -11,6 +11,7 @@ struct git_transport_options {
 	unsigned followtags : 1;
 	unsigned check_self_contained_and_connected : 1;
 	unsigned self_contained_and_connected : 1;
+	unsigned update_shallow : 1;
 	int depth;
 	const char *uploadpack;
 	const char *receivepack;
@@ -152,6 +153,9 @@ struct transport *transport_get(struct remote *, const char *);
 /* Aggressively fetch annotated tags if possible */
 #define TRANS_OPT_FOLLOWTAGS "followtags"
 
+/* Accept refs that may update .git/shallow without --depth */
+#define TRANS_OPT_UPDATE_SHALLOW "updateshallow"
+
 /**
  * Returns 0 if the option was used, non-zero otherwise. Prints a
  * message to stderr if the option is not used.
-- 
1.8.2.83.gc99314b

[PATCH v3 16/28] receive-pack: reorder some code in unpack()

[Nguyễn Thái Ngọc Duy]

This is the preparation for adding --shallow-file to both
unpack-objects and index-pack. To sum up:

 - status/code, ip/child, unpacker/keeper and i (now ac) are moved out
   to function top level

 - successful flow now ends at the end of the function

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/receive-pack.c | 74 +++++++++++++++++++++++---------------------------
 1 file changed, 34 insertions(+), 40 deletions(-)

diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index cc8c34f..22e162d 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -822,8 +822,11 @@ static const char *pack_lockfile;
 static const char *unpack(int err_fd)
 {
 	struct pack_header hdr;
+	const char *argv[7];
 	const char *hdr_err;
+	int status, ac = 0;
 	char hdr_arg[38];
+	struct child_process child;
 	int fsck_objects = (receive_fsck_objects >= 0
 			    ? receive_fsck_objects
 			    : transfer_fsck_objects >= 0
@@ -840,63 +843,54 @@ static const char *unpack(int err_fd)
 			"--pack_header=%"PRIu32",%"PRIu32,
 			ntohl(hdr.hdr_version), ntohl(hdr.hdr_entries));
 
+	memset(&child, 0, sizeof(child));
 	if (ntohl(hdr.hdr_entries) < unpack_limit) {
-		int code, i = 0;
-		struct child_process child;
-		const char *unpacker[5];
-		unpacker[i++] = "unpack-objects";
+		argv[ac++] = "unpack-objects";
 		if (quiet)
-			unpacker[i++] = "-q";
+			argv[ac++] = "-q";
 		if (fsck_objects)
-			unpacker[i++] = "--strict";
-		unpacker[i++] = hdr_arg;
-		unpacker[i++] = NULL;
-		memset(&child, 0, sizeof(child));
-		child.argv = unpacker;
+			argv[ac++] = "--strict";
+		argv[ac++] = hdr_arg;
+		argv[ac++] = NULL;
+		child.argv = argv;
 		child.no_stdout = 1;
 		child.err = err_fd;
 		child.git_cmd = 1;
-		code = run_command(&child);
-		if (!code)
-			return NULL;
-		return "unpack-objects abnormal exit";
+		status = run_command(&child);
+		if (status)
+			return "unpack-objects abnormal exit";
 	} else {
-		const char *keeper[7];
-		int s, status, i = 0;
+		int s;
 		char keep_arg[256];
-		struct child_process ip;
 
 		s = sprintf(keep_arg, "--keep=receive-pack %"PRIuMAX" on ", (uintmax_t) getpid());
 		if (gethostname(keep_arg + s, sizeof(keep_arg) - s))
 			strcpy(keep_arg + s, "localhost");
 
-		keeper[i++] = "index-pack";
-		keeper[i++] = "--stdin";
+		argv[ac++] = "index-pack";
+		argv[ac++] = "--stdin";
 		if (fsck_objects)
-			keeper[i++] = "--strict";
+			argv[ac++] = "--strict";
 		if (fix_thin)
-			keeper[i++] = "--fix-thin";
-		keeper[i++] = hdr_arg;
-		keeper[i++] = keep_arg;
-		keeper[i++] = NULL;
-		memset(&ip, 0, sizeof(ip));
-		ip.argv = keeper;
-		ip.out = -1;
-		ip.err = err_fd;
-		ip.git_cmd = 1;
-		status = start_command(&ip);
-		if (status) {
+			argv[ac++] = "--fix-thin";
+		argv[ac++] = hdr_arg;
+		argv[ac++] = keep_arg;
+		argv[ac++] = NULL;
+		child.argv = argv;
+		child.out = -1;
+		child.err = err_fd;
+		child.git_cmd = 1;
+		status = start_command(&child);
+		if (status)
 			return "index-pack fork failed";
-		}
-		pack_lockfile = index_pack_lockfile(ip.out);
-		close(ip.out);
-		status = finish_command(&ip);
-		if (!status) {
-			reprepare_packed_git();
-			return NULL;
-		}
-		return "index-pack abnormal exit";
+		pack_lockfile = index_pack_lockfile(child.out);
+		close(child.out);
+		status = finish_command(&child);
+		if (status)
+			return "index-pack abnormal exit";
+		reprepare_packed_git();
 	}
+	return NULL;
 }
 
 static const char *unpack_with_sideband(void)
-- 
1.8.2.83.gc99314b

[PATCH v3 17/28] Support pushing from a shallow clone

[Nguyễn Thái Ngọc Duy]

Pushing from a shallow clone using today's send-pack and receive-pack
may work, if the transferred pack does not end up at any graft
points. If it does, recent receive-pack that does connectivity check
will reject the push. If receive-pack is old and does not have the
connectivity check, the upstream repo becomes corrupt.

The pack protocol is updated and send-pack now sends all shallow
grafts before it sends the commands, if the repo is shallow. This
protocol extension will break current receive-pack, which is intended,
mostly to stop corrupting the upstream repo.

Changes on the receiver are similar to what has been done in
fetch-pack, i.e. filter out refs that require new shallow roots then
go along as usual.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/technical/pack-protocol.txt |  4 +-
 builtin/receive-pack.c                    | 68 +++++++++++++++++++++++++++++-
 builtin/send-pack.c                       |  2 +-
 send-pack.c                               |  3 ++
 t/t5537-push-shallow.sh (new +x)          | 70 +++++++++++++++++++++++++++++++
 5 files changed, 143 insertions(+), 4 deletions(-)
 create mode 100755 t/t5537-push-shallow.sh

diff --git a/Documentation/technical/pack-protocol.txt b/Documentation/technical/pack-protocol.txt
index eb8edd1..c73b62f 100644
--- a/Documentation/technical/pack-protocol.txt
+++ b/Documentation/technical/pack-protocol.txt
@@ -464,7 +464,9 @@ contain all the objects that the server will need to complete the new
 references.
 
 ----
-  update-request    =  command-list [pack-file]
+  update-request    =  *shallow command-list [pack-file]
+
+  shallow           =  PKT-LINE("shallow" SP obj-id)
 
   command-list      =  PKT-LINE(command NUL capability-list LF)
 		       *PKT-LINE(command LF)
diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index 22e162d..254feff 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -43,6 +43,9 @@ static int fix_thin = 1;
 static const char *head_name;
 static void *head_name_to_free;
 static int sent_capabilities;
+static int shallow_push;
+static const char* alternate_shallow_file;
+static struct extra_have_objects shallow;
 
 static enum deny_action parse_deny_action(const char *var, const char *value)
 {
@@ -189,6 +192,7 @@ struct command {
 	const char *error_string;
 	unsigned int skip_update:1,
 		     did_not_exist:1;
+	int index;
 	unsigned char old_sha1[20];
 	unsigned char new_sha1[20];
 	char ref_name[FLEX_ARRAY]; /* more */
@@ -687,7 +691,7 @@ static int iterate_receive_command_list(void *cb_data, unsigned char sha1[20])
 	struct command *cmd = *cmd_list;
 
 	while (cmd) {
-		if (!is_null_sha1(cmd->new_sha1)) {
+		if (!is_null_sha1(cmd->new_sha1) && !cmd->skip_update) {
 			hashcpy(sha1, cmd->new_sha1);
 			*cmd_list = cmd->next;
 			return 0;
@@ -712,8 +716,25 @@ static void reject_updates_to_hidden(struct command *commands)
 	}
 }
 
+static void filter_shallow_refs(struct command *commands,
+				const int *ref_status, int nr_ref)
+{
+	struct command *cmd;
+	for (cmd = commands; cmd; cmd = cmd->next)
+		if (!is_null_sha1(cmd->new_sha1) && ref_status[cmd->index]) {
+			cmd->error_string = "shallow update not allowed";
+			cmd->skip_update = 1;
+		}
+	if (*alternate_shallow_file) {
+		unlink(alternate_shallow_file);
+		alternate_shallow_file = NULL;
+	}
+}
+
 static void execute_commands(struct command *commands, const char *unpacker_error)
 {
+	int *ref_status = NULL;
+	struct extra_have_objects ref;
 	struct command *cmd;
 	unsigned char sha1[20];
 
@@ -723,6 +744,20 @@ static void execute_commands(struct command *commands, const char *unpacker_erro
 		return;
 	}
 
+	memset(&ref, 0, sizeof(ref));
+	if (shallow_push) {
+		int i;
+		for (i = 0, cmd = commands; cmd; i++, cmd = cmd->next)
+			if (!is_null_sha1(cmd->new_sha1)) {
+				add_extra_have(&ref, cmd->new_sha1);
+				cmd->index = i;
+			}
+
+		ref_status = xmalloc(sizeof(*ref_status) * ref.nr);
+		if (mark_new_shallow_refs(&ref, ref_status, NULL, &shallow))
+			filter_shallow_refs(commands, ref_status, ref.nr);
+	}
+
 	cmd = commands;
 	if (check_everything_connected(iterate_receive_command_list,
 				       0, &cmd))
@@ -752,6 +787,8 @@ static void execute_commands(struct command *commands, const char *unpacker_erro
 
 		cmd->error_string = update(cmd);
 	}
+	free(ref.array);
+	free(ref_status);
 }
 
 static struct command *read_head_info(void)
@@ -768,6 +805,15 @@ static struct command *read_head_info(void)
 		line = packet_read_line(0, &len);
 		if (!line)
 			break;
+
+		if (len == 48 && !prefixcmp(line, "shallow ")) {
+			if (get_sha1_hex(line + 8, old_sha1))
+				die("protocol error: expected shallow sha, got '%s'", line + 8);
+			add_extra_have(&shallow, old_sha1);
+			shallow_push = 1;
+			continue;
+		}
+
 		if (len < 83 ||
 		    line[40] != ' ' ||
 		    line[81] != ' ' ||
@@ -822,7 +868,7 @@ static const char *pack_lockfile;
 static const char *unpack(int err_fd)
 {
 	struct pack_header hdr;
-	const char *argv[7];
+	const char *argv[9];
 	const char *hdr_err;
 	int status, ac = 0;
 	char hdr_arg[38];
@@ -843,6 +889,24 @@ static const char *unpack(int err_fd)
 			"--pack_header=%"PRIu32",%"PRIu32,
 			ntohl(hdr.hdr_version), ntohl(hdr.hdr_entries));
 
+	if (shallow_push) {
+		struct extra_have_objects extra;
+		memset(&extra, 0, sizeof(extra));
+		remove_reachable_shallow_points(&extra, &shallow);
+		if (extra.nr) {
+			alternate_shallow_file = setup_temporary_shallow(&extra);
+			free(shallow.array);
+			shallow = extra;
+		} else {
+			alternate_shallow_file = NULL;
+			shallow_push = 0;
+		}
+		if (alternate_shallow_file && *alternate_shallow_file) {
+			argv[ac++] = "--shallow-file";
+			argv[ac++] = alternate_shallow_file;
+		}
+	}
+
 	memset(&child, 0, sizeof(child));
 	if (ntohl(hdr.hdr_entries) < unpack_limit) {
 		argv[ac++] = "unpack-objects";
diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index bfa9253..5db1311 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -207,7 +207,7 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 	    (send_all && args.send_mirror))
 		usage(send_pack_usage);
 
-	if (is_repository_shallow())
+	if (is_repository_shallow() && args.stateless_rpc)
 		die("attempt to push from a shallow repository");
 
 	if (remote_name) {
diff --git a/send-pack.c b/send-pack.c
index fab62e3..8b5571c 100644
--- a/send-pack.c
+++ b/send-pack.c
@@ -213,6 +213,9 @@ int send_pack(struct send_pack_args *args,
 		return 0;
 	}
 
+	if (!args->dry_run)
+		advertise_shallow_grafts(out);
+
 	/*
 	 * Finally, tell the other end!
 	 */
diff --git a/t/t5537-push-shallow.sh b/t/t5537-push-shallow.sh
new file mode 100755
index 0000000..650c31a
--- /dev/null
+++ b/t/t5537-push-shallow.sh
@@ -0,0 +1,70 @@
+#!/bin/sh
+
+test_description='push from/to a shallow clone'
+
+. ./test-lib.sh
+
+commit() {
+	echo "$1" >tracked &&
+	git add tracked &&
+	git commit -m "$1"
+}
+
+test_expect_success 'setup' '
+	git config --global transfer.fsckObjects true &&
+	commit 1 &&
+	commit 2 &&
+	commit 3 &&
+	commit 4 &&
+	(
+	git init full-abc &&
+	cd full-abc &&
+	commit a &&
+	commit b &&
+	commit c
+	) &&
+	git clone --no-local --depth=2 .git shallow &&
+	git --git-dir=shallow/.git log --format=%s >actual &&
+	cat <<EOF >expect &&
+4
+3
+EOF
+	test_cmp expect actual &&
+	git clone --no-local --depth=2 full-abc/.git shallow2 &&
+	git --git-dir=shallow2/.git log --format=%s >actual &&
+	cat <<EOF >expect &&
+c
+b
+EOF
+	test_cmp expect actual
+'
+
+test_expect_success 'push from shallow clone' '
+	(
+	cd shallow &&
+	commit 5 &&
+	git push ../.git +master:refs/remotes/shallow/master
+	) &&
+	git log --format=%s shallow/master >actual &&
+	git fsck &&
+	cat <<EOF >expect &&
+5
+4
+3
+2
+1
+EOF
+	test_cmp expect actual
+'
+
+test_expect_success 'push from shallow clone, with grafted roots' '
+	(
+	cd shallow2 &&
+	test_must_fail git push ../.git +master:refs/remotes/shallow2/master 2>err &&
+	grep "shallow2/master.*shallow update not allowed" err
+	) &&
+	test_must_fail git rev-parse shallow2/master &&
+	git fsck
+'
+
+test_done
-- 
1.8.2.83.gc99314b

[PATCH v3 18/28] New var GIT_SHALLOW_FILE to propagate --shallow-file to subprocesses

[Nguyễn Thái Ngọc Duy]

This may be needed when a hook is run after a new shallow pack is
received, but .git/shallow is not settled yet. A temporary shallow
file to plug all loose ends should be used instead. GIT_SHALLOW_FILE
is overriden by --shallow-file.

--shallow-file does not work in this case because the hook may spawn
many git subprocesses and the launch commands do not have
--shallow-file as it's a recent addition.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 cache.h       | 1 +
 commit.h      | 2 +-
 environment.c | 6 ++++++
 git.c         | 2 +-
 shallow.c     | 4 +++-
 5 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/cache.h b/cache.h
index ce377e1..185d692 100644
--- a/cache.h
+++ b/cache.h
@@ -354,6 +354,7 @@ static inline enum object_type object_type(unsigned int mode)
 #define DB_ENVIRONMENT "GIT_OBJECT_DIRECTORY"
 #define INDEX_ENVIRONMENT "GIT_INDEX_FILE"
 #define GRAFT_ENVIRONMENT "GIT_GRAFT_FILE"
+#define GIT_SHALLOW_FILE_ENVIRONMENT "GIT_SHALLOW_FILE"
 #define TEMPLATE_DIR_ENVIRONMENT "GIT_TEMPLATE_DIR"
 #define CONFIG_ENVIRONMENT "GIT_CONFIG"
 #define CONFIG_DATA_ENVIRONMENT "GIT_CONFIG_PARAMETERS"
diff --git a/commit.h b/commit.h
index 3af4699..0ff70fa 100644
--- a/commit.h
+++ b/commit.h
@@ -202,7 +202,7 @@ extern int is_repository_shallow(void);
 extern struct commit_list *get_shallow_commits(struct object_array *heads,
 		int depth, int shallow_flag, int not_shallow_flag);
 extern void check_shallow_file_for_update(void);
-extern void set_alternate_shallow_file(const char *path);
+extern void set_alternate_shallow_file(const char *path, int override);
 extern int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
 				 const struct extra_have_objects *extra);
 extern void setup_alternate_shallow(struct lock_file *shallow_lock,
diff --git a/environment.c b/environment.c
index 0a15349..b73b39d 100644
--- a/environment.c
+++ b/environment.c
@@ -10,6 +10,7 @@
 #include "cache.h"
 #include "refs.h"
 #include "fmt-merge-msg.h"
+#include "commit.h"
 
 int trust_executable_bit = 1;
 int trust_ctime = 1;
@@ -97,6 +98,7 @@ const char * const local_repo_env[] = {
 	INDEX_ENVIRONMENT,
 	NO_REPLACE_OBJECTS_ENVIRONMENT,
 	GIT_PREFIX_ENVIRONMENT,
+	GIT_SHALLOW_FILE_ENVIRONMENT,
 	NULL
 };
 
@@ -124,6 +126,7 @@ static char *expand_namespace(const char *raw_namespace)
 static void setup_git_env(void)
 {
 	const char *gitfile;
+	const char *shallow_file;
 
 	git_dir = getenv(GIT_DIR_ENVIRONMENT);
 	if (!git_dir)
@@ -147,6 +150,9 @@ static void setup_git_env(void)
 		read_replace_refs = 0;
 	namespace = expand_namespace(getenv(GIT_NAMESPACE_ENVIRONMENT));
 	namespace_len = strlen(namespace);
+	shallow_file = getenv(GIT_SHALLOW_FILE_ENVIRONMENT);
+	if (shallow_file)
+		set_alternate_shallow_file(shallow_file, 0);
 }
 
 int is_bare_repository(void)
diff --git a/git.c b/git.c
index cb5208d..179c4f6 100644
--- a/git.c
+++ b/git.c
@@ -162,7 +162,7 @@ static int handle_options(const char ***argv, int *argc, int *envchanged)
 		} else if (!strcmp(cmd, "--shallow-file")) {
 			(*argv)++;
 			(*argc)--;
-			set_alternate_shallow_file((*argv)[0]);
+			set_alternate_shallow_file((*argv)[0], 1);
 			if (envchanged)
 				*envchanged = 1;
 		} else if (!strcmp(cmd, "-C")) {
diff --git a/shallow.c b/shallow.c
index 1baf6b5..abb04db 100644
--- a/shallow.c
+++ b/shallow.c
@@ -11,10 +11,12 @@ static int is_shallow = -1;
 static struct stat shallow_stat;
 static char *alternate_shallow_file;
 
-void set_alternate_shallow_file(const char *path)
+void set_alternate_shallow_file(const char *path, int override)
 {
 	if (is_shallow != -1)
 		die("BUG: is_repository_shallow must not be called before set_alternate_shallow_file");
+	if (alternate_shallow_file && !override)
+		return;
 	free(alternate_shallow_file);
 	alternate_shallow_file = path ? xstrdup(path) : NULL;
 }
-- 
1.8.2.83.gc99314b

[PATCH v3 19/28] connected.c: add new variant that runs with --shallow-file

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 connected.c | 42 ++++++++++++++++++++++++++++++++++--------
 connected.h |  2 ++
 2 files changed, 36 insertions(+), 8 deletions(-)

diff --git a/connected.c b/connected.c
index fae8d64..427389d 100644
--- a/connected.c
+++ b/connected.c
@@ -19,17 +19,17 @@ int check_everything_connected(sha1_iterate_fn fn, int quiet, void *cb_data)
  *
  * Returns 0 if everything is connected, non-zero otherwise.
  */
-int check_everything_connected_with_transport(sha1_iterate_fn fn,
-					      int quiet,
-					      void *cb_data,
-					      struct transport *transport)
+static int check_everything_connected_real(sha1_iterate_fn fn,
+					   int quiet,
+					   void *cb_data,
+					   struct transport *transport,
+					   const char *shallow_file)
 {
 	struct child_process rev_list;
-	const char *argv[] = {"rev-list", "--objects",
-			      "--stdin", "--not", "--all", NULL, NULL};
+	const char *argv[9];
 	char commit[41];
 	unsigned char sha1[20];
-	int err = 0;
+	int err = 0, ac = 0;
 	struct packed_git *new_pack = NULL;
 
 	if (fn(cb_data, sha1))
@@ -47,8 +47,18 @@ int check_everything_connected_with_transport(sha1_iterate_fn fn,
 		strbuf_release(&idx_file);
 	}
 
+	if (shallow_file) {
+		argv[ac++] = "--shallow-file";
+		argv[ac++] = shallow_file;
+	}
+	argv[ac++] = "rev-list";
+	argv[ac++] = "--objects";
+	argv[ac++] = "--stdin";
+	argv[ac++] = "--not";
+	argv[ac++] = "--all";
 	if (quiet)
-		argv[5] = "--quiet";
+		argv[ac++] = "--quiet";
+	argv[ac] = NULL;
 
 	memset(&rev_list, 0, sizeof(rev_list));
 	rev_list.argv = argv;
@@ -92,3 +102,19 @@ int check_everything_connected_with_transport(sha1_iterate_fn fn,
 	sigchain_pop(SIGPIPE);
 	return finish_command(&rev_list) || err;
 }
+
+int check_everything_connected_with_transport(sha1_iterate_fn fn,
+					      int quiet,
+					      void *cb_data,
+					      struct transport *transport)
+{
+	return check_everything_connected_real(fn, quiet, cb_data,
+					       transport, NULL);
+}
+
+int check_shallow_connected(sha1_iterate_fn fn, int quiet, void *cb_data,
+			    const char *shallow_file)
+{
+	return check_everything_connected_real(fn, quiet, cb_data,
+					       NULL, shallow_file);
+}
diff --git a/connected.h b/connected.h
index 0b060b7..071d408 100644
--- a/connected.h
+++ b/connected.h
@@ -18,6 +18,8 @@ typedef int (*sha1_iterate_fn)(void *, unsigned char [20]);
  * Return 0 if Ok, non zero otherwise (i.e. some missing objects)
  */
 extern int check_everything_connected(sha1_iterate_fn, int quiet, void *cb_data);
+extern int check_shallow_connected(sha1_iterate_fn, int quiet, void *cb_data,
+				   const char *shallow_file);
 extern int check_everything_connected_with_transport(sha1_iterate_fn, int quiet,
 						     void *cb_data,
 						     struct transport *transport);
-- 
1.8.2.83.gc99314b

[PATCH v3 20/28] receive-pack: allow pushing with new shallow roots

[Nguyễn Thái Ngọc Duy]

This is more expensive than the current mode and potentially
invalidates caches like pack bitmaps. It's only enabled if
receive.shallowupdate is on.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/config.txt |   4 ++
 builtin/receive-pack.c   | 111 ++++++++++++++++++++++++++++++++++++++++++++---
 t/t5537-push-shallow.sh  |  16 +++++++
 3 files changed, 125 insertions(+), 6 deletions(-)

diff --git a/Documentation/config.txt b/Documentation/config.txt
index ab26963..1a0bd0d 100644
--- a/Documentation/config.txt
+++ b/Documentation/config.txt
@@ -2026,6 +2026,10 @@ receive.updateserverinfo::
 	If set to true, git-receive-pack will run git-update-server-info
 	after receiving data from git-push and updating refs.
 
+receive.shallowupdate::
+	If set to true, .git/shallow can be updated when new refs
+	require new shallow roots. Otherwise those refs are rejected.
+
 remote.pushdefault::
 	The remote to push to by default.  Overrides
 	`branch.<name>.remote` for all branches, and is overridden by
diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index 254feff..366ecde 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -43,7 +43,7 @@ static int fix_thin = 1;
 static const char *head_name;
 static void *head_name_to_free;
 static int sent_capabilities;
-static int shallow_push;
+static int shallow_push, shallow_update;
 static const char* alternate_shallow_file;
 static struct extra_have_objects shallow;
 
@@ -124,6 +124,11 @@ static int receive_pack_config(const char *var, const char *value, void *cb)
 		return 0;
 	}
 
+	if (strcmp(var, "receive.shallowupdate") == 0) {
+		shallow_update = git_config_bool(var, value);
+		return 0;
+	}
+
 	return git_default_config(var, value, cb);
 }
 
@@ -191,6 +196,7 @@ struct command {
 	struct command *next;
 	const char *error_string;
 	unsigned int skip_update:1,
+		     checked_connectivity:1,
 		     did_not_exist:1;
 	int index;
 	unsigned char old_sha1[20];
@@ -424,7 +430,44 @@ static void refuse_unconfigured_deny_delete_current(void)
 		rp_error("%s", refuse_unconfigured_deny_delete_current_msg[i]);
 }
 
-static const char *update(struct command *cmd)
+static int command_singleton_iterator(void *cb_data, unsigned char sha1[20]);
+static int update_ref_shallow(struct command *cmd, uint32_t **used_shallow)
+{
+	static struct lock_file shallow_lock;
+	struct extra_have_objects extra;
+	const char *alt_file;
+	uint32_t mask = 1 << (cmd->index % 32);
+	int i;
+
+	memset(&extra, 0, sizeof(extra));
+	for (i = 0; i < shallow.nr; i++)
+		if (used_shallow[i] &&
+		    used_shallow[i][cmd->index / 32] & mask)
+			add_extra_have(&extra, shallow.array[i]);
+
+	setup_alternate_shallow(&shallow_lock, &alt_file, &extra);
+	if (check_shallow_connected(command_singleton_iterator,
+				     0, cmd, alt_file)) {
+		rollback_lock_file(&shallow_lock);
+		free(extra.array);
+		return -1;
+	}
+
+	commit_lock_file(&shallow_lock);
+
+	/*
+	 * Make sure setup_alternate_shallow() for the next ref does
+	 * not lose these new roots..
+	 */
+	for (i = 0; i < extra.nr; i++)
+		register_shallow(extra.array[i]);
+
+	cmd->checked_connectivity = 1;
+	free(extra.array);
+	return 0;
+}
+
+static const char *update(struct command *cmd, uint32_t **used_shallow)
 {
 	const char *name = cmd->ref_name;
 	struct strbuf namespaced_name_buf = STRBUF_INIT;
@@ -532,6 +575,10 @@ static const char *update(struct command *cmd)
 		return NULL; /* good */
 	}
 	else {
+		if (shallow_push && !cmd->checked_connectivity &&
+		    update_ref_shallow(cmd, used_shallow))
+			return "shallow error";
+
 		lock = lock_any_ref_for_update(namespaced_name, old_sha1,
 					       0, NULL);
 		if (!lock) {
@@ -678,6 +725,8 @@ static void set_connectivity_errors(struct command *commands)
 
 	for (cmd = commands; cmd; cmd = cmd->next) {
 		struct command *singleton = cmd;
+		if (shallow_push && !cmd->checked_connectivity)
+			continue;
 		if (!check_everything_connected(command_singleton_iterator,
 						0, &singleton))
 			continue;
@@ -691,7 +740,8 @@ static int iterate_receive_command_list(void *cb_data, unsigned char sha1[20])
 	struct command *cmd = *cmd_list;
 
 	while (cmd) {
-		if (!is_null_sha1(cmd->new_sha1) && !cmd->skip_update) {
+		if (!is_null_sha1(cmd->new_sha1) && !cmd->skip_update &&
+		    (!shallow_push || cmd->checked_connectivity)) {
 			hashcpy(sha1, cmd->new_sha1);
 			*cmd_list = cmd->next;
 			return 0;
@@ -733,9 +783,10 @@ static void filter_shallow_refs(struct command *commands,
 
 static void execute_commands(struct command *commands, const char *unpacker_error)
 {
-	int *ref_status = NULL;
+	int *ref_status = NULL, checked_connectivity;
 	struct extra_have_objects ref;
 	struct command *cmd;
+	uint32_t **used_shallow = NULL;
 	unsigned char sha1[20];
 
 	if (unpacker_error) {
@@ -754,8 +805,39 @@ static void execute_commands(struct command *commands, const char *unpacker_erro
 			}
 
 		ref_status = xmalloc(sizeof(*ref_status) * ref.nr);
-		if (mark_new_shallow_refs(&ref, ref_status, NULL, &shallow))
+		if (shallow_update)
+			used_shallow = xmalloc(sizeof(*used_shallow) * shallow.nr);
+		if (!mark_new_shallow_refs(&ref, ref_status, used_shallow,
+					   &shallow))
+			shallow_push = 0;
+		else if (!shallow_update) {
 			filter_shallow_refs(commands, ref_status, ref.nr);
+			shallow_push = 0;
+		}
+	}
+
+	/*
+	 * If shallow_push is still on at this point, we know some of
+	 * the new refs may require extra shallow roots. But we don't
+	 * know yet which ref will be accepted because hooks may
+	 * reject some of them. So we can't add all new shallow roots
+	 * in. Go through ref by ref and only add relevant shallow
+	 * roots right before write_ref_sha1.
+	 */
+	if (shallow_push) {
+		int i;
+		/* keep hooks happy */
+		setenv(GIT_SHALLOW_FILE_ENVIRONMENT, alternate_shallow_file, 1);
+		for (i = 0, cmd = commands; i < ref.nr; i++, cmd = cmd->next) {
+			/*
+			 * marker for iterate_receive_command_list.
+			 * All safe refs are run through the next
+			 * check_everything_connected() the rest one
+			 * by one in update()
+			 */
+			cmd->checked_connectivity = !ref_status[i];
+			cmd->index = i;
+		}
 	}
 
 	cmd = commands;
@@ -778,6 +860,7 @@ static void execute_commands(struct command *commands, const char *unpacker_erro
 	free(head_name_to_free);
 	head_name = head_name_to_free = resolve_refdup("HEAD", sha1, 0, NULL);
 
+	checked_connectivity = 1;
 	for (cmd = commands; cmd; cmd = cmd->next) {
 		if (cmd->error_string)
 			continue;
@@ -785,10 +868,26 @@ static void execute_commands(struct command *commands, const char *unpacker_erro
 		if (cmd->skip_update)
 			continue;
 
-		cmd->error_string = update(cmd);
+		cmd->error_string = update(cmd, used_shallow);
+		if (shallow_push && !cmd->error_string &&
+		    !cmd->checked_connectivity) {
+			error("BUG: connectivity check has not been run on ref %s",
+			      cmd->ref_name);
+			checked_connectivity = 0;
+		}
+	}
+
+	if (shallow_push) {
+		if (!checked_connectivity)
+			error("BUG: run 'git fsck' for safety.\n"
+			      "If there are errors, try to remove "
+			      "the reported refs above");
+		if (*alternate_shallow_file)
+			unlink(alternate_shallow_file);
 	}
 	free(ref.array);
 	free(ref_status);
+	free(used_shallow);
 }
 
 static struct command *read_head_info(void)
diff --git a/t/t5537-push-shallow.sh b/t/t5537-push-shallow.sh
index 650c31a..0084a31 100755
--- a/t/t5537-push-shallow.sh
+++ b/t/t5537-push-shallow.sh
@@ -67,4 +67,20 @@ test_expect_success 'push from shallow clone, with grafted roots' '
 	git fsck
 '
 
+test_expect_success 'add new shallow root with receive.updateshallow on' '
+	git config receive.shallowupdate true &&
+	(
+	cd shallow2 &&
+	GIT_TRACE=2 git push ../.git +master:refs/remotes/shallow2/master
+	) &&
+	git log --format=%s shallow2/master >actual &&
+	git fsck &&
+	cat <<EOF >expect &&
+c
+b
+EOF
+	test_cmp expect actual &&
+	git config receive.shallowupdate false
+'
+
 test_done
-- 
1.8.2.83.gc99314b

[PATCH v3 21/28] send-pack: support pushing to a shallow clone

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/send-pack.c     |  5 ++++-
 t/t5537-push-shallow.sh | 38 ++++++++++++++++++++++++++++++++++++++
 transport.c             |  5 ++---
 3 files changed, 44 insertions(+), 4 deletions(-)

diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index 5db1311..0031165 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -100,6 +100,7 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 	int fd[2];
 	struct child_process *conn;
 	struct extra_have_objects extra_have;
+	struct extra_have_objects shallow;
 	struct ref *remote_refs, *local_refs;
 	int ret;
 	int helper_status = 0;
@@ -232,8 +233,10 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 	}
 
 	memset(&extra_have, 0, sizeof(extra_have));
+	memset(&shallow, 0, sizeof(shallow));
 
-	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL, &extra_have, NULL);
+	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL,
+			 &extra_have, &shallow);
 
 	transport_verify_remote_names(nr_refspecs, refspecs);
 
diff --git a/t/t5537-push-shallow.sh b/t/t5537-push-shallow.sh
index 0084a31..ccb41b6 100755
--- a/t/t5537-push-shallow.sh
+++ b/t/t5537-push-shallow.sh
@@ -83,4 +83,42 @@ EOF
 	git config receive.shallowupdate false
 '
 
+test_expect_success 'push from shallow to shallow' '
+	(
+	cd shallow &&
+	git --git-dir=../shallow2/.git config receive.shallowupdate true &&
+	git push ../shallow2/.git +master:refs/remotes/shallow/master &&
+	git --git-dir=../shallow2/.git config receive.shallowupdate false
+	) &&
+	(
+	cd shallow2 &&
+	git log --format=%s shallow/master >actual &&
+	git fsck &&
+	cat <<EOF >expect &&
+5
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
+test_expect_success 'push from full to shallow' '
+	! git --git-dir=shallow2/.git cat-file blob `echo 1|git hash-object --stdin` &&
+	commit 1 &&
+	git push shallow2/.git +master:refs/remotes/top/master &&
+	(
+	cd shallow2 &&
+	git log --format=%s top/master >actual &&
+	git fsck &&
+	cat <<EOF >expect &&
+1
+4
+3
+EOF
+	test_cmp expect actual &&
+	git cat-file blob `echo 1|git hash-object --stdin` >/dev/null
+	)
+'
+
 test_done
diff --git a/transport.c b/transport.c
index c0be6b1..6a3fe9b 100644
--- a/transport.c
+++ b/transport.c
@@ -818,11 +818,10 @@ static int git_transport_push(struct transport *transport, struct ref *remote_re
 		struct ref *tmp_refs;
 		connect_setup(transport, 1, 0);
 
-		get_remote_heads(data->fd[0], NULL, 0, &tmp_refs, REF_NORMAL, NULL, NULL);
+		get_remote_heads(data->fd[0], NULL, 0, &tmp_refs, REF_NORMAL,
+				 NULL, &data->shallow);
 		data->got_remote_heads = 1;
 	}
-	if (data->shallow.nr)
-		die("pushing to a shallow repository is not supported");
 
 	memset(&args, 0, sizeof(args));
 	args.send_mirror = !!(flags & TRANSPORT_PUSH_MIRROR);
-- 
1.8.2.83.gc99314b

[PATCH v3 22/28] remote-curl: pass ref SHA-1 to fetch-pack as well

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/fetch-pack.c | 7 +++++++
 remote-curl.c        | 3 ++-
 2 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/builtin/fetch-pack.c b/builtin/fetch-pack.c
index 927424b..aa6e596 100644
--- a/builtin/fetch-pack.c
+++ b/builtin/fetch-pack.c
@@ -13,6 +13,13 @@ static void add_sought_entry_mem(struct ref ***sought, int *nr, int *alloc,
 				 const char *name, int namelen)
 {
 	struct ref *ref = xcalloc(1, sizeof(*ref) + namelen + 1);
+	unsigned char sha1[20];
+
+	if (namelen > 41 && name[40] == ' ' && !get_sha1_hex(name, sha1)) {
+		hashcpy(ref->old_sha1, sha1);
+		name += 41;
+		namelen -= 41;
+	}
 
 	memcpy(ref->name, name, namelen);
 	ref->name[namelen] = '\0';
diff --git a/remote-curl.c b/remote-curl.c
index 222210f..25d6730 100644
--- a/remote-curl.c
+++ b/remote-curl.c
@@ -719,7 +719,8 @@ static int fetch_git(struct discovery *heads,
 		struct ref *ref = to_fetch[i];
 		if (!ref->name || !*ref->name)
 			die("cannot fetch by sha1 over smart http");
-		packet_buf_write(&preamble, "%s\n", ref->name);
+		packet_buf_write(&preamble, "%s %s\n",
+				 sha1_to_hex(ref->old_sha1), ref->name);
 	}
 	packet_buf_flush(&preamble);
 
-- 
1.8.2.83.gc99314b

[PATCH v3 23/28] Support fetch/clone over http

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/gitremote-helpers.txt |  7 +++++++
 builtin/fetch-pack.c                | 16 +++++++++++++---
 remote-curl.c                       | 30 ++++++++++++++++++++++++++++--
 t/t5536-fetch-shallow.sh            | 27 +++++++++++++++++++++++++++
 transport-helper.c                  |  6 ++++++
 upload-pack.c                       |  2 --
 6 files changed, 81 insertions(+), 7 deletions(-)

diff --git a/Documentation/gitremote-helpers.txt b/Documentation/gitremote-helpers.txt
index f1f4ca9..c2908db 100644
--- a/Documentation/gitremote-helpers.txt
+++ b/Documentation/gitremote-helpers.txt
@@ -437,6 +437,13 @@ set by Git if the remote helper has the 'option' capability.
 'option check-connectivity' \{'true'|'false'\}::
 	Request the helper to check connectivity of a clone.
 
+'option cloning \{'true'|'false'\}::
+	Notify the helper this is a clone request (i.e. the current
+	repository is guaranteed empty).
+
+'option update-shallow \{'true'|'false'\}::
+	Allow to extend .git/shallow if the new refs require it.
+
 SEE ALSO
 --------
 linkgit:git-remote[1]
diff --git a/builtin/fetch-pack.c b/builtin/fetch-pack.c
index aa6e596..64d1c0e 100644
--- a/builtin/fetch-pack.c
+++ b/builtin/fetch-pack.c
@@ -46,6 +46,7 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
 	char **pack_lockfile_ptr = NULL;
 	struct child_process *conn;
 	struct fetch_pack_args args;
+	struct extra_have_objects shallow;
 
 	packet_trace_identity("fetch-pack");
 
@@ -113,6 +114,14 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
 			args.check_self_contained_and_connected = 1;
 			continue;
 		}
+		if (!strcmp("--cloning", arg)) {
+			args.cloning = 1;
+			continue;
+		}
+		if (!strcmp("--update-shallow", arg)) {
+			args.update_shallow = 1;
+			continue;
+		}
 		usage(fetch_pack_usage);
 	}
 
@@ -157,10 +166,11 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
 				   args.verbose ? CONNECT_VERBOSE : 0);
 	}
 
-	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL, NULL);
+	memset(&shallow, 0, sizeof(shallow));
+	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL, &shallow);
 
-	ref = fetch_pack(&args, fd, conn, ref, dest,
-			 sought, nr_sought, NULL, pack_lockfile_ptr);
+	ref = fetch_pack(&args, fd, conn, ref, dest, sought, nr_sought,
+			 &shallow, pack_lockfile_ptr);
 	if (pack_lockfile) {
 		printf("lock %s\n", pack_lockfile);
 		fflush(stdout);
diff --git a/remote-curl.c b/remote-curl.c
index 25d6730..0ae06b3 100644
--- a/remote-curl.c
+++ b/remote-curl.c
@@ -20,6 +20,8 @@ struct options {
 	unsigned long depth;
 	unsigned progress : 1,
 		check_self_contained_and_connected : 1,
+		cloning : 1,
+		update_shallow : 1,
 		followtags : 1,
 		dry_run : 1,
 		thin : 1;
@@ -88,6 +90,24 @@ static int set_option(const char *name, const char *value)
 		strbuf_release(&val);
 		return 0;
 	}
+	else if (!strcmp(name, "cloning")) {
+		if (!strcmp(value, "true"))
+			options.cloning = 1;
+		else if (!strcmp(value, "false"))
+			options.cloning = 0;
+		else
+			return -1;
+		return 0;
+	}
+	else if (!strcmp(name, "update-shallow")) {
+		if (!strcmp(value, "true"))
+			options.update_shallow = 1;
+		else if (!strcmp(value, "false"))
+			options.update_shallow = 0;
+		else
+			return -1;
+		return 0;
+	}
 	else {
 		return 1 /* unsupported */;
 	}
@@ -99,6 +119,7 @@ struct discovery {
 	char *buf;
 	size_t len;
 	struct ref *refs;
+	struct extra_have_objects shallow;
 	unsigned proto_git : 1;
 };
 static struct discovery *last_discovery;
@@ -107,7 +128,7 @@ static struct ref *parse_git_refs(struct discovery *heads, int for_push)
 {
 	struct ref *list = NULL;
 	get_remote_heads(-1, heads->buf, heads->len, &list,
-			 for_push ? REF_NORMAL : 0, NULL, NULL);
+			 for_push ? REF_NORMAL : 0, NULL, &heads->shallow);
 	return list;
 }
 
@@ -168,6 +189,7 @@ static void free_discovery(struct discovery *d)
 	if (d) {
 		if (d == last_discovery)
 			last_discovery = NULL;
+		free(d->shallow.array);
 		free(d->buf_alloc);
 		free_refs(d->refs);
 		free(d);
@@ -688,7 +710,7 @@ static int fetch_git(struct discovery *heads,
 	struct strbuf preamble = STRBUF_INIT;
 	char *depth_arg = NULL;
 	int argc = 0, i, err;
-	const char *argv[16];
+	const char *argv[17];
 
 	argv[argc++] = "fetch-pack";
 	argv[argc++] = "--stateless-rpc";
@@ -704,6 +726,10 @@ static int fetch_git(struct discovery *heads,
 	}
 	if (options.check_self_contained_and_connected)
 		argv[argc++] = "--check-self-contained-and-connected";
+	if (options.cloning)
+		argv[argc++] = "--cloning";
+	if (options.update_shallow)
+		argv[argc++] = "--update-shallow";
 	if (!options.progress)
 		argv[argc++] = "--no-progress";
 	if (options.depth) {
diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
index 95b6313..bc96436 100755
--- a/t/t5536-fetch-shallow.sh
+++ b/t/t5536-fetch-shallow.sh
@@ -163,4 +163,31 @@ EOF
 	)
 '
 
+if test -n "$NO_CURL" -o -z "$GIT_TEST_HTTPD"; then
+	say 'skipping remaining tests, git built without http support'
+	test_done
+fi
+
+LIB_HTTPD_PORT=${LIB_HTTPD_PORT-'5536'}
+. "$TEST_DIRECTORY"/lib-httpd.sh
+start_httpd
+
+test_expect_success 'clone http repository' '
+	git clone --bare --no-local shallow "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	git clone $HTTPD_URL/smart/repo.git clone &&
+	(
+	cd clone &&
+	git fsck &&
+	git log --format=%s origin/master >actual &&
+	cat <<EOF >expect &&
+6
+5
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
+stop_httpd
 test_done
diff --git a/transport-helper.c b/transport-helper.c
index 673b7c2..e2b4203 100644
--- a/transport-helper.c
+++ b/transport-helper.c
@@ -360,6 +360,12 @@ static int fetch_with_fetch(struct transport *transport,
 	    data->transport_options.check_self_contained_and_connected)
 		set_helper_option(transport, "check-connectivity", "true");
 
+	if (transport->cloning)
+		set_helper_option(transport, "cloning", "true");
+
+	if (data->transport_options.update_shallow)
+		set_helper_option(transport, "update-shallow", "true");
+
 	for (i = 0; i < nr_heads; i++) {
 		const struct ref *posn = to_fetch[i];
 		if (posn->status & REF_STATUS_UPTODATE)
diff --git a/upload-pack.c b/upload-pack.c
index 28269c7..2d02297 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -836,8 +836,6 @@ int main(int argc, char **argv)
 
 	if (!enter_repo(dir, strict))
 		die("'%s' does not appear to be a git repository", dir);
-	if (is_repository_shallow() && stateless_rpc)
-		die("attempt to push into a shallow repository");
 
 	git_config(upload_pack_config, NULL);
 	upload_pack();
-- 
1.8.2.83.gc99314b

[PATCH v3 24/28] receive-pack: support pushing to a shallow clone via http

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/receive-pack.c  |  3 ---
 t/t5537-push-shallow.sh | 35 +++++++++++++++++++++++++++++++++++
 2 files changed, 35 insertions(+), 3 deletions(-)

diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index 366ecde..ec681ba 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -1157,9 +1157,6 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
 	if (!enter_repo(dir, 0))
 		die("'%s' does not appear to be a git repository", dir);
 
-	if (is_repository_shallow() && stateless_rpc)
-		die("attempt to push into a shallow repository");
-
 	git_config(receive_pack_config, NULL);
 
 	if (0 <= transfer_unpack_limit)
diff --git a/t/t5537-push-shallow.sh b/t/t5537-push-shallow.sh
index ccb41b6..d252a78 100755
--- a/t/t5537-push-shallow.sh
+++ b/t/t5537-push-shallow.sh
@@ -16,6 +16,7 @@ test_expect_success 'setup' '
 	commit 2 &&
 	commit 3 &&
 	commit 4 &&
+	git clone . full &&
 	(
 	git init full-abc &&
 	cd full-abc &&
@@ -121,4 +122,38 @@ EOF
 	)
 '
 
+if test -n "$NO_CURL" -o -z "$GIT_TEST_HTTPD"; then
+	say 'skipping remaining tests, git built without http support'
+	test_done
+fi
+
+LIB_HTTPD_PORT=${LIB_HTTPD_PORT-'5537'}
+. "$TEST_DIRECTORY"/lib-httpd.sh
+start_httpd
+
+test_expect_success 'push to shallow repo via http' '
+	git clone --bare --no-local shallow "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	(
+	cd "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	git config http.receivepack true
+	) &&
+	(
+	cd full &&
+	commit 9 &&
+	git push $HTTPD_URL/smart/repo.git +master:refs/remotes/top/master
+	) &&
+	(
+	cd "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	git fsck &&
+	git log --format=%s top/master >actual &&
+	cat <<EOF >expect &&
+9
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
+stop_httpd
 test_done
-- 
1.8.2.83.gc99314b

[PATCH v3 25/28] send-pack: support pushing from a shallow clone via http

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/send-pack.c     |  3 ---
 send-pack.c             | 19 +++++++++++++++++--
 t/t5537-push-shallow.sh | 25 +++++++++++++++++++++++++
 3 files changed, 42 insertions(+), 5 deletions(-)

diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index 0031165..966b45e 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -208,9 +208,6 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 	    (send_all && args.send_mirror))
 		usage(send_pack_usage);
 
-	if (is_repository_shallow() && args.stateless_rpc)
-		die("attempt to push from a shallow repository");
-
 	if (remote_name) {
 		remote = remote_get(remote_name);
 		if (!remote_has_url(remote, dest)) {
diff --git a/send-pack.c b/send-pack.c
index 8b5571c..13167ea 100644
--- a/send-pack.c
+++ b/send-pack.c
@@ -174,6 +174,21 @@ static int sideband_demux(int in, int out, void *data)
 	return ret;
 }
 
+static int advertise_shallow_grafts_cb(const struct commit_graft *graft, void *cb)
+{
+	struct strbuf *sb = cb;
+	if (graft->nr_parent == -1)
+		packet_buf_write(sb, "shallow %s\n", sha1_to_hex(graft->sha1));
+	return 0;
+}
+
+void advertise_shallow_grafts_buf(struct strbuf *sb)
+{
+	if (!is_repository_shallow())
+		return;
+	for_each_commit_graft(advertise_shallow_grafts_cb, sb);
+}
+
 int send_pack(struct send_pack_args *args,
 	      int fd[], struct child_process *conn,
 	      struct ref *remote_refs,
@@ -214,7 +229,7 @@ int send_pack(struct send_pack_args *args,
 	}
 
 	if (!args->dry_run)
-		advertise_shallow_grafts(out);
+		advertise_shallow_grafts_buf(&req_buf);
 
 	/*
 	 * Finally, tell the other end!
@@ -275,7 +290,7 @@ int send_pack(struct send_pack_args *args,
 	}
 
 	if (args->stateless_rpc) {
-		if (!args->dry_run && cmds_sent) {
+		if (!args->dry_run && (cmds_sent || is_repository_shallow())) {
 			packet_buf_flush(&req_buf);
 			send_sideband(out, -1, req_buf.buf, req_buf.len, LARGE_PACKET_MAX);
 		}
diff --git a/t/t5537-push-shallow.sh b/t/t5537-push-shallow.sh
index d252a78..f8200f7 100755
--- a/t/t5537-push-shallow.sh
+++ b/t/t5537-push-shallow.sh
@@ -155,5 +155,30 @@ EOF
 	)
 '
 
+test_expect_success 'push from shallow repo via http' '
+	mv "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" shallow-upstream.git &&
+	git clone --bare --no-local full "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	(
+	cd "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	git config http.receivepack true
+	) &&
+	commit 10 &&
+	git push $HTTPD_URL/smart/repo.git +master:refs/remotes/top/master &&
+	(
+	cd "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	git fsck &&
+	git log --format=%s top/master >actual &&
+	cat <<EOF >expect &&
+10
+1
+4
+3
+2
+1
+EOF
+	test_cmp expect actual
+	)
+'
+
 stop_httpd
 test_done
-- 
1.8.2.83.gc99314b

[PATCH v3 26/28] git-clone.txt: remove shallow clone limitations

[Nguyễn Thái Ngọc Duy]

Now that git supports push/pull from/to a shallow clone, these
limitations are not true anymore.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/git-clone.txt | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/Documentation/git-clone.txt b/Documentation/git-clone.txt
index 450f158..4987857 100644
--- a/Documentation/git-clone.txt
+++ b/Documentation/git-clone.txt
@@ -181,12 +181,7 @@ objects from the source repository into a pack in the cloned repository.
 
 --depth <depth>::
 	Create a 'shallow' clone with a history truncated to the
-	specified number of revisions.  A shallow repository has a
-	number of limitations (you cannot clone or fetch from
-	it, nor push from nor into it), but is adequate if you
-	are only interested in the recent history of a large project
-	with a long history, and would want to send in fixes
-	as patches.
+	specified number of revisions.
 
 --[no-]single-branch::
 	Clone only the history leading to the tip of a single branch,
-- 
1.8.2.83.gc99314b

[PATCH v3 27/28] clone: use git protocol for cloning shallow repo locally

[Nguyễn Thái Ngọc Duy]

clone_local() does not handle $SRC/shallow. It could be made so, but
it's simpler to use fetch-pack/upload-pack instead.

This used by be caught by the check in upload-pack, which is triggered
by transport_get_remote_refs(), even in local clone case. The check is
now gone and check_everything_connected() should catch the result
incomplete repo. But check_everything_connected() will soon be skipped
in local clone case, opening a door to corrupt repo. This patch should
close that door.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/clone.c  | 11 +++++++++--
 t/t5601-clone.sh |  7 +++++++
 2 files changed, 16 insertions(+), 2 deletions(-)

diff --git a/builtin/clone.c b/builtin/clone.c
index 0b182ce..71ee68b 100644
--- a/builtin/clone.c
+++ b/builtin/clone.c
@@ -797,8 +797,15 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
 	else
 		repo = repo_name;
 	is_local = option_local != 0 && path && !is_bundle;
-	if (is_local && option_depth)
-		warning(_("--depth is ignored in local clones; use file:// instead."));
+	if (is_local) {
+		if (option_depth)
+			warning(_("--depth is ignored in local clones; use file:// instead."));
+		if (!access(mkpath("%s/shallow", path), F_OK)) {
+			if (option_local > 0)
+				warning(_("source repository is shallow, ignoring --local"));
+			is_local = 0;
+		}
+	}
 	if (option_local > 0 && !is_local)
 		warning(_("--local is ignored"));
 
diff --git a/t/t5601-clone.sh b/t/t5601-clone.sh
index 1d1c875..c226cff 100755
--- a/t/t5601-clone.sh
+++ b/t/t5601-clone.sh
@@ -340,4 +340,11 @@ test_expect_success 'clone from a repository with two identical branches' '
 
 '
 
+test_expect_success 'shallow clone locally' '
+	git clone --depth=1 --no-local src ssrrcc &&
+	git clone ssrrcc ddsstt &&
+	test_cmp ssrrcc/.git/shallow ddsstt/.git/shallow &&
+	( cd ddsstt && git fsck )
+'
+
 test_done
-- 
1.8.2.83.gc99314b

[PATCH v3 28/28] prune: clean .git/shallow after pruning objects

[Nguyễn Thái Ngọc Duy]

This patch teaches "prune" to remove shallow roots that are no longer
reachable from any refs (e.g. when the relevant refs are removed).

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/gc.c    |  1 +
 builtin/prune.c |  4 ++++
 commit.h        |  1 +
 shallow.c       | 43 +++++++++++++++++++++++++++++++++++++++++--
 4 files changed, 47 insertions(+), 2 deletions(-)

diff --git a/builtin/gc.c b/builtin/gc.c
index c14190f..cec8ecd 100644
--- a/builtin/gc.c
+++ b/builtin/gc.c
@@ -16,6 +16,7 @@
 #include "run-command.h"
 #include "sigchain.h"
 #include "argv-array.h"
+#include "commit.h"
 
 #define FAILED_RUN "failed to run %s"
 
diff --git a/builtin/prune.c b/builtin/prune.c
index 6366917..023aea8 100644
--- a/builtin/prune.c
+++ b/builtin/prune.c
@@ -170,5 +170,9 @@ int cmd_prune(int argc, const char **argv, const char *prefix)
 	s = mkpathdup("%s/pack", get_object_directory());
 	remove_temporary_files(s);
 	free(s);
+
+	if (!show_only && is_repository_shallow())
+		prune_shallow();
+
 	return 0;
 }
diff --git a/commit.h b/commit.h
index 0ff70fa..ff93a59 100644
--- a/commit.h
+++ b/commit.h
@@ -215,6 +215,7 @@ extern void remove_reachable_shallow_points(struct extra_have_objects *out,
 extern int mark_new_shallow_refs(const struct extra_have_objects *ref,
 				 int *ref_status, uint32_t **used,
 				 const struct extra_have_objects *shallow);
+extern void prune_shallow(void);
 
 int is_descendant_of(struct commit *, struct commit_list *);
 int in_merge_bases(struct commit *, struct commit *);
diff --git a/shallow.c b/shallow.c
index abb04db..0eda40f 100644
--- a/shallow.c
+++ b/shallow.c
@@ -157,6 +157,7 @@ struct write_shallow_data {
 	struct strbuf *out;
 	int use_pack_protocol;
 	int count;
+	int seen_only;
 };
 
 static int write_one_shallow(const struct commit_graft *graft, void *cb_data)
@@ -165,6 +166,11 @@ static int write_one_shallow(const struct commit_graft *graft, void *cb_data)
 	const char *hex = sha1_to_hex(graft->sha1);
 	if (graft->nr_parent != -1)
 		return 0;
+	if (data->seen_only) {
+		struct commit *c = lookup_commit(graft->sha1);
+		if (!c || !(c->object.flags & SEEN))
+			return 0;
+	}
 	data->count++;
 	if (data->use_pack_protocol)
 		packet_buf_write(data->out, "shallow %s", hex);
@@ -175,14 +181,16 @@ static int write_one_shallow(const struct commit_graft *graft, void *cb_data)
 	return 0;
 }
 
-int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
-			  const struct extra_have_objects *extra)
+static int write_shallow_commits_1(struct strbuf *out, int use_pack_protocol,
+				   const struct extra_have_objects *extra,
+				   int seen_only)
 {
 	struct write_shallow_data data;
 	int i;
 	data.out = out;
 	data.use_pack_protocol = use_pack_protocol;
 	data.count = 0;
+	data.seen_only = seen_only;
 	for_each_commit_graft(write_one_shallow, &data);
 	if (!extra)
 		return data.count;
@@ -194,6 +202,12 @@ int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
 	return data.count;
 }
 
+int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
+			  const struct extra_have_objects *extra)
+{
+	return write_shallow_commits_1(out, use_pack_protocol, extra, 0);
+}
+
 char *setup_temporary_shallow(const struct extra_have_objects *extra)
 {
 	struct strbuf sb = STRBUF_INIT;
@@ -515,3 +529,28 @@ int mark_new_shallow_refs(const struct extra_have_objects *ref,
 
 	return ret;
 }
+
+/*
+ * mark_reachable_objects() should have been run prior to this and all
+ * reachable commits marked as "SEEN".
+ */
+void prune_shallow(void)
+{
+	static struct lock_file shallow_lock;
+	struct strbuf sb = STRBUF_INIT;
+	int fd;
+
+	check_shallow_file_for_update();
+	fd = hold_lock_file_for_update(&shallow_lock, git_path("shallow"),
+				       LOCK_DIE_ON_ERROR);
+	if (write_shallow_commits_1(&sb, 0, NULL, 1)) {
+		if (write_in_full(fd, sb.buf, sb.len) != sb.len)
+			die_errno("failed to write to %s",
+				  shallow_lock.filename);
+		commit_lock_file(&shallow_lock);
+	} else {
+		unlink(git_path("shallow"));
+		rollback_lock_file(&shallow_lock);
+	}
+	strbuf_release(&sb);
+}
-- 
1.8.2.83.gc99314b

Re: [PATCH v3 04/28] update-server-info: do not publish shallow clones

[Junio C Hamano]

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> Dumb commit walker does not care about .git/shallow and until someone
> steps up to make it happen, let's not publish shallow clones using
> dumb protocols.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  server-info.c | 9 +++++++++
>  1 file changed, 9 insertions(+)

I doubt that pros-and-cons is in this patch's favor.  Without this
patch, if a fetch requests commits just on the surface in this
shallow repository, the walker would happily get the necessary
objects just fine.  If the request has to dig deeper and cross the
shallow boundary, the walker will get a failure and error out.

With this patch, both will error out.  So overall, the patch did not
make anything safer (e.g. preventing from introducing new corruption
on the recipient's end) while breaking a case where it worked just
fine, no?

Or am I missing something?  Not that dumb walkers would matter very
much these days, ...

> diff --git a/server-info.c b/server-info.c
> index 9ec744e..a8df6a5 100644
> --- a/server-info.c
> +++ b/server-info.c
> @@ -33,6 +33,10 @@ static int update_info_refs(int force)
>  	strcpy(path1 + len, "+");
>  
>  	safe_create_leading_directories(path0);
> +	if (is_repository_shallow()) {
> +		unlink(path0);
> +		return error("info/refs not updated for shallow clone");
> +	}
>  	info_ref_fp = fopen(path1, "w");
>  	if (!info_ref_fp)
>  		return error("unable to update %s", path1);
> @@ -217,6 +221,11 @@ static int update_info_packs(int force)
>  	strcpy(name, infofile);
>  	strcpy(name + namelen, "+");
>  
> +	if (is_repository_shallow()) {
> +		unlink(infofile);
> +		return error("info/packs not updated for shallow clone");
> +	}
> +
>  	init_pack_info(infofile, force);
>  
>  	safe_create_leading_directories(name);

Re: [PATCH v3 06/28] connect.c: teach get_remote_heads to parse "shallow" lines

[Junio C Hamano]

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> No callers pass a non-empty pointer as shallow_points at this
> stage. As a result, all clients still refuse to talk to shallow
> repository on the other end.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  builtin/fetch-pack.c |  2 +-
>  builtin/send-pack.c  |  2 +-
>  connect.c            | 12 +++++++++++-
>  remote-curl.c        |  2 +-
>  remote.h             |  3 ++-
>  transport.c          |  7 ++++---
>  6 files changed, 20 insertions(+), 8 deletions(-)
>
> diff --git a/builtin/fetch-pack.c b/builtin/fetch-pack.c
> index c8e8582..c1d918f 100644
> --- a/builtin/fetch-pack.c
> +++ b/builtin/fetch-pack.c
> @@ -150,7 +150,7 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
>  				   args.verbose ? CONNECT_VERBOSE : 0);
>  	}
>  
> -	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL);
> +	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL, NULL);
>  
>  	ref = fetch_pack(&args, fd, conn, ref, dest,
>  			 sought, nr_sought, pack_lockfile_ptr);
> diff --git a/builtin/send-pack.c b/builtin/send-pack.c
> index 51121f2..bfa9253 100644
> --- a/builtin/send-pack.c
> +++ b/builtin/send-pack.c
> @@ -233,7 +233,7 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
>  
>  	memset(&extra_have, 0, sizeof(extra_have));
>  
> -	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL, &extra_have);
> +	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL, &extra_have, NULL);
>  
>  	transport_verify_remote_names(nr_refspecs, refspecs);
>  
> diff --git a/connect.c b/connect.c
> index 06e88b0..d0602b0 100644
> --- a/connect.c
> +++ b/connect.c
> @@ -122,7 +122,8 @@ static void annotate_refs_with_symref_info(struct ref *ref)
>   */
>  struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
>  			      struct ref **list, unsigned int flags,
> -			      struct extra_have_objects *extra_have)
> +			      struct extra_have_objects *extra_have,
> +			      struct extra_have_objects *shallow_points)

The _shape_ of the information you would want to keep track of for
the shallow cut-off points may exactly be the same as that is for
extra-have endspoints, but it still feels wrong to throw these
shallow cut-off points into a structure called "extra have". After
all, these are objects the other end does not have, which is the
direct opposite of extra-have.

Perhaps a preparatory patch needs to rename the structure type to
object_name_list or something.  And then we can make the variable
names, not typenames, responsible for signalling what they mean,
i.e.

	get_remote_heads(...
        	struct list_of_objects *extra_have,
                struct list_of_objects *shallow_points);

when we introduce the new parameter.

Re: [PATCH v3 07/28] shallow.c: add remove_reachable_shallow_points()

[Junio C Hamano]

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> When we receive a pack and the shallow points from another repository,
> we may want to add more shallow points to current repo to make sure no
> commits point to nowhere. However we do not want to add unnecessary
> shallow points and cut our history short because the other end is a
> shallow version of this repo. The output shallow points may or may not
> be added to .git/shallow, depending on whether they are actually
> reachable in the new pack.
>
> This function filters such shallow points out, leaving ones that might
> potentially be added. A simple has_sha1_file won't do because we may
> have incomplete object islands (i.e. not connected to any refs) and
> the shallow points are on these islands. In that case we want to keep
> the shallow points as candidates until we figure out if the new pack
> connects to such object islands.
>
> Typical cases that use remove_reachable_shallow_points() are:
>
>  - fetch from a repo that has longer history: in this case all remote
>    shallow roots do not exist in the client repo, this function will
>    be cheap as it just does a few lookup_commit_graft and
>    has_sha1_file.

It is unclear why.  If you fetched from a repository more complete
than you, you may deepen your history which may allow you to unplug
the shallow points you originally had, and remote "shallow root" (by
the way, lets find a single good phrase to represent this thing and
stick to it) may want to replace them, no?

>  - fetch from a repo that has exactly the same shallow root set
>    (e.g. a clone from a shallow repo): this case may trigger
>    in_merge_bases_many all the way to roots. An exception is made to
>    avoid such costly path with a faith that .git/shallow does not
>    usually points to unreachable commit islands.

... and when the faith is broken, you will end up with a broken
repository???

>  - push from a shallow repo that has shorter history than the
>    server's: in_merge_bases_many() is unavoidable, so the longer
>    history the client has the higher the server cost is. The cost may
>    be reduced with the help of pack bitmaps, or commit cache, though.
>
>  - push from a shallow clone that has exactly the same shallow root
>    set: the same as the second fetch case above, .i.e. cheap by
>    exception.
>
> The function must be called before the new pack is installed, or we
> won't know which objects are ours, which theirs.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  commit.h  |  3 +++
>  connect.c |  2 +-
>  remote.h  |  1 +
>  shallow.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
>  4 files changed, 50 insertions(+), 1 deletion(-)
>
> diff --git a/commit.h b/commit.h
> index a879526..98044e6 100644
> --- a/commit.h
> +++ b/commit.h
> @@ -193,6 +193,7 @@ extern struct commit_list *get_octopus_merge_bases(struct commit_list *in);
>  /* largest positive number a signed 32-bit integer can contain */
>  #define INFINITE_DEPTH 0x7fffffff
>  
> +struct extra_have_objects;
>  extern int register_shallow(const unsigned char *sha1);
>  extern int unregister_shallow(const unsigned char *sha1);
>  extern int for_each_commit_graft(each_commit_graft_fn, void *);
> @@ -206,6 +207,8 @@ extern void setup_alternate_shallow(struct lock_file *shallow_lock,
>  				    const char **alternate_shallow_file);
>  extern char *setup_temporary_shallow(void);
>  extern void advertise_shallow_grafts(int);
> +extern void remove_reachable_shallow_points(struct extra_have_objects *out,
> +					    const struct extra_have_objects *in);
>  
>  int is_descendant_of(struct commit *, struct commit_list *);
>  int in_merge_bases(struct commit *, struct commit *);
> diff --git a/connect.c b/connect.c
> index d0602b0..80e4360 100644
> --- a/connect.c
> +++ b/connect.c
> @@ -45,7 +45,7 @@ int check_ref_type(const struct ref *ref, int flags)
>  	return check_ref(ref->name, strlen(ref->name), flags);
>  }
>  
> -static void add_extra_have(struct extra_have_objects *extra, unsigned char *sha1)
> +void add_extra_have(struct extra_have_objects *extra, unsigned char *sha1)
>  {
>  	ALLOC_GROW(extra->array, extra->nr + 1, extra->alloc);
>  	hashcpy(&(extra->array[extra->nr][0]), sha1);
> diff --git a/remote.h b/remote.h
> index 773faa9..ff604ff 100644
> --- a/remote.h
> +++ b/remote.h
> @@ -145,6 +145,7 @@ extern struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
>  				     struct ref **list, unsigned int flags,
>  				     struct extra_have_objects *have,
>  				     struct extra_have_objects *shallow);
> +extern void add_extra_have(struct extra_have_objects *extra, unsigned char *sha1);
>  
>  int resolve_remote_symref(struct ref *ref, struct ref *list);
>  int ref_newer(const unsigned char *new_sha1, const unsigned char *old_sha1);
> diff --git a/shallow.c b/shallow.c
> index 9552512..a974d2d 100644
> --- a/shallow.c
> +++ b/shallow.c
> @@ -2,6 +2,8 @@
>  #include "commit.h"
>  #include "tag.h"
>  #include "pkt-line.h"
> +#include "remote.h"
> +#include "refs.h"
>  
>  static int is_shallow = -1;
>  static struct stat shallow_stat;
> @@ -235,3 +237,46 @@ void advertise_shallow_grafts(int fd)
>  		return;
>  	for_each_commit_graft(advertise_shallow_grafts_cb, &fd);
>  }
> +
> +struct commit_array {
> +	struct commit **commits;
> +	int nr, alloc;
> +};
> +
> +static int add_ref(const char *refname,
> +		   const unsigned char *sha1, int flags, void *cb_data)
> +{
> +	struct commit_array *ca = cb_data;
> +	ALLOC_GROW(ca->commits, ca->nr + 1, ca->alloc);
> +	ca->commits[ca->nr++] = lookup_commit(sha1);
> +	return 0;
> +}

Can't a ref point at a non-commit-ish?  Is the code prepared to deal
with such an entry (possibly a NULL pointer) in the commit_array
struct?

> +void remove_reachable_shallow_points(struct extra_have_objects *out,
> +				     const struct extra_have_objects *in)
> +{
> +	struct commit_array ca;
> +	int i;
> +
> +	memset(&ca, 0, sizeof(ca));
> +	head_ref(add_ref, &ca);
> +	for_each_ref(add_ref, &ca);
> +	for (i = 0; i < in->nr; i++) {
> +		struct commit_graft *graft = lookup_commit_graft(in->array[i]);
> +		/*
> +		 * For a clone from a shallow upstream, the clone has
> +		 * the same shallow roots as upstream and it will
> +		 * trigger in_merge_bases_many() all the way to roots.
> +		 * Avoid that costly path and assume .git/shallow is
> +		 * good most of the time.
> +		 */
> +		if (graft && graft->nr_parent < 0)
> +			continue;
> +		if (has_sha1_file(in->array[i]) &&
> +		    in_merge_bases_many(lookup_commit(in->array[i]),
> +					ca.nr, ca.commits))
> +			continue;
> +		add_extra_have(out, in->array[i]);
> +	}
> +	free(ca.commits);
> +}

Re: [PATCH v3 08/28] shallow.c: add mark_new_shallow_refs()

[Junio C Hamano]

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> When we receive a pack and the shallow points from another repository,
> we may need to add more shallow points to current repo to make sure no
> commits point to nowhere. But usually we don't want to do so because
> (in future) new shallow points invalidate pack bitmaps and we need to
> rebuild them again, which is not cheap.
>
> So the default way is we allow ref updates that do not introduce new
> shallow points and mark the others. If the user is fine with new
> shallow point addition, we accept the marked refs.
>
> But even so we do not blindly accept all shallow points provided. Some
> of them might not point to any commits in the new pack. Some might
> even do, but those might be unreachable object islands. Only shallow
> points that are reachable from old and new refs can stay.
>
> The way it's implemented is paint down from each ref, attach a bitmap
> to each commit where one bit represents one ref. In order to avoid
> allocating new bitmap for every commit, we try to reuse the same
> bitmap for parent commits if possible. This reduces allocation and
> leaks deliberately because it's hard to keep/time consuming track how
> many pointers to the same buffer.
>
> In a typical push or fetch, the new pack should not carry new shallow
> roots. If the current repo does not have any commit islands refered by
> the sender's shallow roots either, this function is just a few
> has_sha1_file(). So quite cheap.
>
> Once the sender diverts from that path (or the receiver detects
> shallow roots attached to commit islands from remove_reachable_shallow_points),
> it'll be a lot more expensive. Pack bitmaps won't help this kind of
> commit traversal, but commit cache might.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  commit.h  |   4 ++
>  shallow.c | 219 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 223 insertions(+)

Hmph.  the use of ->util field in this patch feels that it was
something commit-slab data structure was invented to solve.


> diff --git a/shallow.c b/shallow.c
> index a974d2d..c92a1dc 100644
> --- a/shallow.c
> +++ b/shallow.c
> @@ -4,6 +4,8 @@
>  #include "pkt-line.h"
>  #include "remote.h"
>  #include "refs.h"
> +#include "diff.h"
> +#include "revision.h"
>  
>  static int is_shallow = -1;
>  static struct stat shallow_stat;
> @@ -280,3 +282,220 @@ void remove_reachable_shallow_points(struct extra_have_objects *out,
>  	}
>  	free(ca.commits);
>  }
> +
> +static int paint_down(const unsigned char *sha1, int id, int nr_bits, int quick)
> +{
> +	int hit_bottom = 0;
> +	unsigned int i, nr;
> +	struct commit_list *head = NULL;
> +	int bitmap_nr = (nr_bits + 31) / 32;
> +	int bitmap_size = bitmap_nr * sizeof(uint32_t);
> +	uint32_t *tmp = xmalloc(bitmap_size);
> +	uint32_t *bitmap = xcalloc(bitmap_size, sizeof(uint32_t));
> +	bitmap[id / 32] |= (1 << (id % 32));
> +	commit_list_insert(lookup_commit(sha1), &head);
> +	while (head) {
> +		struct commit_list *p;
> +		struct commit *c = head->item;
> +		uint32_t *c_util = c->util;
> +
> +		p = head;
> +		head = head->next;
> +		free(p);
> +
> +		if (c->object.flags & (SEEN | UNINTERESTING))
> +			continue;
> +		else
> +			c->object.flags |= SEEN;
> +
> +		if (c->util == NULL)
> +			c->util = bitmap;
> +		else {
> +			/*
> +			 * Deliberately leak a lot in commit->util
> +			 * because there can be many pointers to the
> +			 * same bitmap. Probably should allocate in a
> +			 * pool and free the whole pool at the end.
> +			 */

... or perhaps make the bitmap into

	struct {
	        int refcnt;
        	uint32_t bits[FLEX_ARRAY];
	}

and refcnt them?


> +			memcpy(tmp, c_util, bitmap_size);
> +			for (i = 0; i < bitmap_nr; i++)
> +				tmp[i] |= bitmap[i];
> +			if (memcmp(tmp, c_util, bitmap_size)) {
> +				c->util = xmalloc(bitmap_size);
> +				memcpy(c->util, tmp, bitmap_size);
> +			}
> +		}
> +
> +		if (c->object.flags & BOTTOM) {
> +			hit_bottom = 1;
> +			if (quick) {
> +				free_commit_list(head);
> +				break;
> +			} else
> +				continue;
> +		}
> +
> +		if (parse_commit(c))
> +			die("unable to parse commit %s",
> +			    sha1_to_hex(c->object.sha1));
> +
> +		for (p = c->parents; p; p = p->next) {
> +			if (p->item->object.flags & SEEN)
> +				continue;
> +			if (p->item->util == NULL || p->item->util == c_util)
> +				p->item->util = c->util;
> +			commit_list_insert(p->item, &head);
> +		}
> +	}
> +
> +	nr = get_max_object_index();
> +	for (i = 0; i < nr; i++) {
> +		struct object *o = get_indexed_object(i);
> +		if (o && o->type == OBJ_COMMIT) {
> +			o->flags &= ~SEEN;
> +		}
> +	}
> +
> +	free(tmp);
> +	return hit_bottom;
> +}
> +
> +static int mark_uninteresting(const char *refname,
> +			      const unsigned char *sha1,
> +			      int flags, void *cb_data)
> +{
> +	struct commit *commit = lookup_commit(sha1);
> +	commit->object.flags |= UNINTERESTING;
> +	mark_parents_uninteresting(commit);
> +	return 0;
> +}
> +
> +struct saved_util {
> +	unsigned char sha1[20];
> +	void *util;
> +};
> +
> +/*
> + * Given a set of refs and shallow roots, find out what ref can reach
> + * any of the given roots. If so mark that ref "reject_flag". If
> + * "used" is not NULL, mark all reachable roots. Return how many refs
> + * that need new shallow points.
> + */
> +int mark_new_shallow_refs(const struct extra_have_objects *ref,
> +			  int *ref_status, uint32_t **used,
> +			  const struct extra_have_objects *shallow)
> +{
> +	struct saved_util *util = NULL;
> +	unsigned int i, nr, ret = 0, nr_util = 0, alloc_util = 0;
> +
> +	/*
> +	 * Quick check to see if we may need to add new shallow
> +	 * roots. Go through the list of root candidates and check if
> +	 * they exist (either in current repo, or in the new pack, we
> +	 * can't distinguish).
> +	 *
> +	 * 1) If none of the new roots exist, the pack must connect to
> +	 *    the main object graph, which is already guarded by
> +	 *    current repo's shallow roots and we will not need to
> +	 *    consider adding new shallow roots, so we can exit early.
> +	 *
> +	 * 2) The pack may connect to some existing object islands in
> +	 *    current repo then add shallow roots to plug loose ends
> +	 *    from those islands. In that case, new shallow roots must
> +	 *    also exist in the repo as this stage (old objects plus
> +	 *    the new pack).
> +	 *
> +	 * 3) The last, easiest case, is the pack contains some
> +	 *    shallow roots, which may be used to tie up loose ends of
> +	 *    some new refs, or redundanty (tying up loose ends of new
> +	 *    object islands)
> +	 */
> +	for (i = 0;i < shallow->nr; i++)
> +		if (has_sha1_file(shallow->array[i]))
> +			break;
> +	if (i == shallow->nr)
> +		/*
> +		 * this is the first and also the common case, where
> +		 * the new pack does not carry any new shallow
> +		 * points. No need to to the expensive commit traverse
> +		 * dance below.
> +		 */
> +		return 0;

I am Confused.

The loop only made sure that all the elements in the array[] is
still missing (or, ... is this function supposed to be called before
installing a new pack???  It is unclear.  But if new objects were
unpacked while receiving, then there is no "not install the new
objects and check" possible, so I'd assume this is called after
receiving and registering a new pack to the object store).

But then, can it be that you had N-1 "shallow points" originally,
the pack has a reference to a new missing commit, and the array has
N "shallow points" in total?  Or is the caller expected to call this
function with shallow pointing at a pre-transfer shallow points?

> +	/*
> +	 * Prepare the commit graph to track what refs can reach what
> +	 * (new) shallow points.
> +	 */
> +	nr = get_max_object_index();

Hmph. At this point (again, there is no description on where in the
overall sequence of events this function is designed to be called,
so it is impossible to review the logic), is it expected that we
have seen all the objects under the sun and marked them in a
specific way?

> +	for (i = 0; i < nr; i++) {
> +		struct object *o = get_indexed_object(i);
> +		struct commit *c = (struct commit *)o;
> +		if (!o || o->type != OBJ_COMMIT)
> +			continue;
> +
> +		o->flags &= ~(UNINTERESTING | BOTTOM | SEEN);
> +		/*
> +		 * git-fetch makes use of "util" field. Save it and
> +		 * restore later. For fetch/clone/push, "nr" should be
> +		 * small because rev-list is delayed to pack-objects.
> +		 */
> +		if (c->util) {
> +			ALLOC_GROW(util, nr_util+1, alloc_util);
> +			hashcpy(util[nr_util].sha1, o->sha1);
> +			util[nr_util].util = c->util;
> +			nr_util++;
> +			c->util = NULL;
> +		}
> +	}
> +
> +	/*
> +	 * "--not --all" to cut short the traversal if new refs
> +	 * connect to old refs. If not (e.g. force ref updates) it'll
> +	 * have to go down to the current shallow roots.
> +	 *
> +	 * We could detect that a new commit is connected to an
> +	 * existing commit by keeping new objects in a pack (i.e. the
> +	 * index-pack code path) then check commit origin. If so stop
> +	 * short, so we don't need to get to the bottom. But then it
> +	 * will not work for case #2 because we need to go through
> +	 * some of our commits before reaching new shallow roots.
> +	 */
> +	head_ref(mark_uninteresting, NULL);
> +	for_each_ref(mark_uninteresting, NULL);
> +
> +	for (i = 0; i < shallow->nr; i++)
> +		if (has_sha1_file(shallow->array[i])) {
> +			struct commit *c = lookup_commit(shallow->array[i]);
> +			c->object.flags |= BOTTOM;
> +		}
> +
> +	for (i = 0; i < ref->nr; i++)
> +		if (paint_down(ref->array[i], i, ref->nr, used == NULL)) {
> +			if (ref_status)
> +				ref_status[i] = 1;
> +			ret++;
> +		}
> +
> +	if (used) {
> +		for (i = 0; i < shallow->nr; i++) {
> +			struct commit *c = lookup_commit(shallow->array[i]);
> +			used[i] = c->util;
> +		}
> +	}
> +
> +	if (nr_util) {
> +		nr = get_max_object_index();
> +		for (i = 0; i < nr; i++) {
> +			struct object *o = get_indexed_object(i);
> +			if (o && o->type == OBJ_COMMIT)
> +				((struct commit *)o)->util = NULL;
> +		}
> +		for (i = 0; i < nr_util; i++) {
> +			struct commit *c = lookup_commit(util[i].sha1);
> +			c->util = util[i].util;
> +		}
> +		free(util);
> +	}
> +
> +	return ret;
> +}

Re: [PATCH v3 09/28] shallow.c: extend setup_*_shallow() to accept extra shallow points

[Junio C Hamano]

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  commit.h      |  8 +++++---
>  fetch-pack.c  |  5 +++--
>  shallow.c     | 20 +++++++++++++++-----
>  upload-pack.c |  2 +-
>  4 files changed, 24 insertions(+), 11 deletions(-)
>
> diff --git a/commit.h b/commit.h
> index e1fd587..3af4699 100644
> --- a/commit.h
> +++ b/commit.h
> @@ -203,10 +203,12 @@ extern struct commit_list *get_shallow_commits(struct object_array *heads,
>  		int depth, int shallow_flag, int not_shallow_flag);
>  extern void check_shallow_file_for_update(void);
>  extern void set_alternate_shallow_file(const char *path);
> -extern int write_shallow_commits(struct strbuf *out, int use_pack_protocol);
> +extern int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
> +				 const struct extra_have_objects *extra);

Confusing.

Sounds as if you got the extra ".have" and storing them in the
.git/shallow file (which of course would not make much sense), which
is not what is going on.

Also it is unclear how the sanity check the previous step seems to
make and the new list of shallow commit names this patch adds to
existing functions are designed to interact. I think the whole
design needs a better explanation of the flow at the higher level.

Re: [PATCH v3 06/28] connect.c: teach get_remote_heads to parse "shallow" lines

[Junio C Hamano]

Junio C Hamano <gitster@pobox.com> writes:

> Perhaps a preparatory patch needs to rename the structure type to
> object_name_list or something.  And then we can make the variable
> names, not typenames, responsible for signalling what they mean,
> i.e.
>
> 	get_remote_heads(...
>         	struct list_of_objects *extra_have,
>                 struct list_of_objects *shallow_points);
>
> when we introduce the new parameter.

Yuck, and these are not "list-of-objects", either.  They are
"list-of-object-names".

Re: [PATCH v3 03/28] clone: prevent --reference to a shallow repository

[Eric Sunshine]

On Sun, Nov 24, 2013 at 10:55 PM, Nguyễn Thái Ngọc Duy
<pclouds@gmail.com> wrote:
> If we borrow objects from another repository, we should also pay
> attention to their $GIT_DIR/shallow (and even info/grafts). But
> current alternates code does not.
>
> Reject alternate repos that are shallow because we do not do it
> right. In future we alternate code may be updated to check

s/we/the/

> $GIT_DIR/shallow properly so that this restriction could be lifted.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  builtin/clone.c | 6 ++++++
>  1 file changed, 6 insertions(+)
>
> diff --git a/builtin/clone.c b/builtin/clone.c
> index 874e0fd..900f564 100644
> --- a/builtin/clone.c
> +++ b/builtin/clone.c
> @@ -252,6 +252,12 @@ static int add_one_reference(struct string_list_item *item, void *cb_data)
>                 die(_("reference repository '%s' is not a local repository."),
>                     item->string);
>
> +       if (!access(mkpath("%s/shallow", ref_git), F_OK))
> +               die(_("reference repository '%s' is shallow"), item->string);
> +
> +       if (!access(mkpath("%s/info/grafts", ref_git), F_OK))
> +               die(_("reference repository '%s' is grafted"), item->string);
> +
>         strbuf_addf(&alternate, "%s/objects", ref_git);
>         add_to_alternates_file(alternate.buf);
>         strbuf_release(&alternate);
> --
> 1.8.2.83.gc99314b

Re: [PATCH v3 04/28] update-server-info: do not publish shallow clones

[Duy Nguyen]

On Tue, Nov 26, 2013 at 3:08 AM, Junio C Hamano <gitster@pobox.com> wrote:
> Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:
>
>> Dumb commit walker does not care about .git/shallow and until someone
>> steps up to make it happen, let's not publish shallow clones using
>> dumb protocols.
>>
>> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
>> ---
>>  server-info.c | 9 +++++++++
>>  1 file changed, 9 insertions(+)
>
> I doubt that pros-and-cons is in this patch's favor.  Without this
> patch, if a fetch requests commits just on the surface in this
> shallow repository, the walker would happily get the necessary
> objects just fine.  If the request has to dig deeper and cross the
> shallow boundary, the walker will get a failure and error out.
>
> With this patch, both will error out.  So overall, the patch did not
> make anything safer (e.g. preventing from introducing new corruption
> on the recipient's end) while breaking a case where it worked just
> fine, no?
>
> Or am I missing something?  Not that dumb walkers would matter very
> much these days, ...

No you're not. If it may fail, in my opinion it should fail early than
walk all the way and fail. But yeah dropping the patch is fine too. I
don't care too much about dumb walkers.
-- 
Duy

Re: [PATCH v3 08/28] shallow.c: add mark_new_shallow_refs()

[Duy Nguyen]

On Tue, Nov 26, 2013 at 5:20 AM, Junio C Hamano <gitster@pobox.com> wrote:
> Hmph.  the use of ->util field in this patch feels that it was
> something commit-slab data structure was invented to solve.

Good stuff! Thanks.

>> +             if (c->util == NULL)
>> +                     c->util = bitmap;
>> +             else {
>> +                     /*
>> +                      * Deliberately leak a lot in commit->util
>> +                      * because there can be many pointers to the
>> +                      * same bitmap. Probably should allocate in a
>> +                      * pool and free the whole pool at the end.
>> +                      */
>
> ... or perhaps make the bitmap into
>
>         struct {
>                 int refcnt;
>                 uint32_t bits[FLEX_ARRAY];
>         }
>
> and refcnt them?

I still prefer memory pools so I just need to do a few free() than
walking through all the commits again and refcnt-- or free() them.

>> +     /*
>> +      * Quick check to see if we may need to add new shallow
>> +      * roots. Go through the list of root candidates and check if
>> +      * they exist (either in current repo, or in the new pack, we
>> +      * can't distinguish).
>> +      *
>> +      * 1) If none of the new roots exist, the pack must connect to
>> +      *    the main object graph, which is already guarded by
>> +      *    current repo's shallow roots and we will not need to
>> +      *    consider adding new shallow roots, so we can exit early.
>> +      *
>> +      * 2) The pack may connect to some existing object islands in
>> +      *    current repo then add shallow roots to plug loose ends
>> +      *    from those islands. In that case, new shallow roots must
>> +      *    also exist in the repo as this stage (old objects plus
>> +      *    the new pack).
>> +      *
>> +      * 3) The last, easiest case, is the pack contains some
>> +      *    shallow roots, which may be used to tie up loose ends of
>> +      *    some new refs, or redundanty (tying up loose ends of new
>> +      *    object islands)
>> +      */
>> +     for (i = 0;i < shallow->nr; i++)
>> +             if (has_sha1_file(shallow->array[i]))
>> +                     break;
>> +     if (i == shallow->nr)
>> +             /*
>> +              * this is the first and also the common case, where
>> +              * the new pack does not carry any new shallow
>> +              * points. No need to to the expensive commit traverse
>> +              * dance below.
>> +              */
>> +             return 0;
>
> I am Confused.
>
> The loop only made sure that all the elements in the array[] is
> still missing (or, ... is this function supposed to be called before
> installing a new pack???  It is unclear.  But if new objects were
> unpacked while receiving, then there is no "not install the new
> objects and check" possible, so I'd assume this is called after
> receiving and registering a new pack to the object store).

Yes it's called after installing the new packs (or after unpacking).
I'll mention about this.

> But then, can it be that you had N-1 "shallow points" originally,
> the pack has a reference to a new missing commit, and the array has
> N "shallow points" in total?  Or is the caller expected to call this
> function with shallow pointing at a pre-transfer shallow points?

Another thing I did not mention is all shallow commits we have are
already filtered out by remove_reachable_shallow_points. By the time
you get here, the array only contains the shallow commits we don't
have that may or may not be referenced by a something (oh yeah I did
not handle tags!) in the new pack. So if all of them are missing (i.e.
the new pack does not reference to any of them), they're useless and
can be thrown away.

Sorry to break the patches this way and lose the overall call flow.
It's just too big to put all into one patch. 13/28 is the one that put
the pieces together but basically

 1. receive the remote's .git/shallow
 2. call remote_reachable_shallow_points() to exclude our shallow commits
 3. get the pack and install it (or unpack it)
 4. call this function to determine what new ref needs new shallow
commits from the result of #2

>> +     /*
>> +      * Prepare the commit graph to track what refs can reach what
>> +      * (new) shallow points.
>> +      */
>> +     nr = get_max_object_index();
>
> Hmph. At this point (again, there is no description on where in the
> overall sequence of events this function is designed to be called,
> so it is impossible to review the logic), is it expected that we
> have seen all the objects under the sun and marked them in a
> specific way?

No. At this point we just make sure to have clean flags in commit
objects. paint_down() is the one that walks through and marks them as
seen.
-- 
Duy

Re: [PATCH v3 17/28] Support pushing from a shallow clone

[Eric Sunshine]

On Sun, Nov 24, 2013 at 10:55 PM, Nguyễn Thái Ngọc Duy
<pclouds@gmail.com> wrote:
> Pushing from a shallow clone using today's send-pack and receive-pack
> may work, if the transferred pack does not end up at any graft
> points. If it does, recent receive-pack that does connectivity check
> will reject the push. If receive-pack is old and does not have the
> connectivity check, the upstream repo becomes corrupt.
>
> The pack protocol is updated and send-pack now sends all shallow
> grafts before it sends the commands, if the repo is shallow. This
> protocol extension will break current receive-pack, which is intended,
> mostly to stop corrupting the upstream repo.
>
> Changes on the receiver are similar to what has been done in
> fetch-pack, i.e. filter out refs that require new shallow roots then
> go along as usual.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
> diff --git a/Documentation/technical/pack-protocol.txt b/Documentation/technical/pack-protocol.txt
> index eb8edd1..c73b62f 100644
> --- a/Documentation/technical/pack-protocol.txt
> +++ b/Documentation/technical/pack-protocol.txt
> @@ -43,6 +43,9 @@ static int fix_thin = 1;
>  static const char *head_name;
>  static void *head_name_to_free;
>  static int sent_capabilities;
> +static int shallow_push;
> +static const char* alternate_shallow_file;

s/char\* /char */

> +static struct extra_have_objects shallow;
>
>  static enum deny_action parse_deny_action(const char *var, const char *value)
>  {

Re: [PATCH v3 08/28] shallow.c: add mark_new_shallow_refs()

[Junio C Hamano]

Duy Nguyen <pclouds@gmail.com> writes:

> On Tue, Nov 26, 2013 at 5:20 AM, Junio C Hamano <gitster@pobox.com> wrote:
>> Hmph.  the use of ->util field in this patch feels that it was
>> something commit-slab data structure was invented to solve.
>
> Good stuff! Thanks.
>
>>> +             if (c->util == NULL)
>>> +                     c->util = bitmap;
>>> +             else {
>>> +                     /*
>>> +                      * Deliberately leak a lot in commit->util
>>> +                      * because there can be many pointers to the
>>> +                      * same bitmap. Probably should allocate in a
>>> +                      * pool and free the whole pool at the end.
>>> +                      */
>>
>> ... or perhaps make the bitmap into
>>
>>         struct {
>>                 int refcnt;
>>                 uint32_t bits[FLEX_ARRAY];
>>         }
>>
>> and refcnt them?
>
> I still prefer memory pools so I just need to do a few free() than
> walking through all the commits again and refcnt-- or free() them.

Fair enough.

> Sorry to break the patches this way and lose the overall call flow.
> It's just too big to put all into one patch. 13/28 is the one that put
> the pieces together but basically
>
>  1. receive the remote's .git/shallow
>  2. call remote_reachable_shallow_points() to exclude our shallow commits
>  3. get the pack and install it (or unpack it)
>  4. call this function to determine what new ref needs new shallow
> commits from the result of #2

Thanks for a roadmap.  Will find time to re-read the thing with it.

Re: [PATCH v3 27/28] clone: use git protocol for cloning shallow repo locally

[Eric Sunshine]

On Sun, Nov 24, 2013 at 10:55 PM, Nguyễn Thái Ngọc Duy
<pclouds@gmail.com> wrote:
> clone_local() does not handle $SRC/shallow. It could be made so, but
> it's simpler to use fetch-pack/upload-pack instead.
>
> This used by be caught by the check in upload-pack, which is triggered

s/used by/used to/

> by transport_get_remote_refs(), even in local clone case. The check is
> now gone and check_everything_connected() should catch the result
> incomplete repo. But check_everything_connected() will soon be skipped
> in local clone case, opening a door to corrupt repo. This patch should
> close that door.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>

Re: [PATCH v3 15/28] fetch: add --update-shallow to get refs that require updating .git/shallow

[Eric Sunshine]

On Sun, Nov 24, 2013 at 10:55 PM, Nguyễn Thái Ngọc Duy
<pclouds@gmail.com> wrote:
> diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
> index e011ead..95b6313 100755
> --- a/t/t5536-fetch-shallow.sh
> +++ b/t/t5536-fetch-shallow.sh
> @@ -141,4 +141,26 @@ EOF
>         )
>  '
>
> +test_expect_success 'fetch --update-shallow' '
> +       (
> +       cd notshallow &&
> +       git fetch --update-shallow ../shallow/.git refs/heads/*:refs/remotes/shallow/* &&
> +       git fsck &&
> +       git for-each-ref --format="%(refname)" |sort >actual.refs &&

Exit status of git-for-each-ref could be lost down the pipe.

> +       cat <<EOF >expect.refs &&
> +refs/remotes/shallow/master
> +refs/remotes/shallow/no-shallow
> +EOF
> +       test_cmp expect.refs actual.refs &&
> +       git log --format=%s shallow/master >actual &&
> +       cat <<EOF >expect &&
> +6
> +5
> +4
> +3
> +EOF
> +       test_cmp expect actual
> +       )
> +'
> +
>  test_done

Re: [PATCH v3 13/28] fetch: support fetching from a shallow repository

[Eric Sunshine]

On Sun, Nov 24, 2013 at 10:55 PM, Nguyễn Thái Ngọc Duy
<pclouds@gmail.com> wrote:
> diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
> new file mode 100755
> index 0000000..41de74d
> --- /dev/null
> +++ b/t/t5536-fetch-shallow.sh
> @@ -0,0 +1,128 @@
> +#!/bin/sh
> +
> +test_description='fetch/clone from a shallow clone'
> +
> +. ./test-lib.sh
> +
> +commit() {
> +       echo "$1" >tracked &&
> +       git add tracked &&
> +       git commit -m "$1"
> +}
> +
> +test_expect_success 'fetch something upstream has but hidden by clients shallow boundaries' '
> +       # the blob "1" is available in .git but hidden by the
> +       # shallow2/.git/shallow and it should be resent
> +       ! git --git-dir=shallow2/.git cat-file blob `echo 1|git hash-object --stdin` >/dev/null &&
> +       echo 1 > 1.t &&

s/> />/

> +       git add 1.t &&
> +       git commit -m add-1-back &&
> +       (
> +       cd shallow2 &&
> +       git fetch ../.git +refs/heads/master:refs/remotes/top/master &&
> +       git fsck &&
> +       git log --format=%s top/master >actual &&
> +       cat <<EOF >expect &&
> +add-1-back
> +4
> +3
> +EOF
> +       test_cmp expect actual
> +       ) &&
> +       git --git-dir=shallow2/.git cat-file blob `echo 1|git hash-object --stdin` >/dev/null
> +
> +'

Re: [PATCH v3 15/28] fetch: add --update-shallow to get refs that require updating .git/shallow

[Duy Nguyen]

On Wed, Nov 27, 2013 at 8:53 AM, Eric Sunshine <sunshine@sunshineco.com> wrote:
> On Sun, Nov 24, 2013 at 10:55 PM, Nguyễn Thái Ngọc Duy
> <pclouds@gmail.com> wrote:
>> diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
>> index e011ead..95b6313 100755
>> --- a/t/t5536-fetch-shallow.sh
>> +++ b/t/t5536-fetch-shallow.sh
>> @@ -141,4 +141,26 @@ EOF
>>         )
>>  '
>>
>> +test_expect_success 'fetch --update-shallow' '
>> +       (
>> +       cd notshallow &&
>> +       git fetch --update-shallow ../shallow/.git refs/heads/*:refs/remotes/shallow/* &&
>> +       git fsck &&
>> +       git for-each-ref --format="%(refname)" |sort >actual.refs &&
>
> Exit status of git-for-each-ref could be lost down the pipe.

I think it's ok. If for-each-ref is broken or something the final
output in actual.refs will be different anyway.
-- 
Duy

Re: [PATCH v3 06/28] connect.c: teach get_remote_heads to parse "shallow" lines

[Duy Nguyen]

On Tue, Nov 26, 2013 at 5:42 AM, Junio C Hamano <gitster@pobox.com> wrote:
> Junio C Hamano <gitster@pobox.com> writes:
>
>> Perhaps a preparatory patch needs to rename the structure type to
>> object_name_list or something.  And then we can make the variable
>> names, not typenames, responsible for signalling what they mean,
>> i.e.
>>
>>       get_remote_heads(...
>>               struct list_of_objects *extra_have,
>>                 struct list_of_objects *shallow_points);
>>
>> when we introduce the new parameter.
>
> Yuck, and these are not "list-of-objects", either.  They are
> "list-of-object-names".

I was thinking of renaming it to sha1_array and what do you know, we
do have 'struct sha1_array' with exact same (or better) functionality
that extra_have_objects provides. Will replace extra_have_objects with
current sha1_array.
-- 
Duy

Re: [PATCH v3 15/28] fetch: add --update-shallow to get refs that require updating .git/shallow

[Junio C Hamano]

Duy Nguyen <pclouds@gmail.com> writes:

> On Wed, Nov 27, 2013 at 8:53 AM, Eric Sunshine <sunshine@sunshineco.com> wrote:
>> On Sun, Nov 24, 2013 at 10:55 PM, Nguyễn Thái Ngọc Duy
>> <pclouds@gmail.com> wrote:
>>> diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
>>> index e011ead..95b6313 100755
>>> --- a/t/t5536-fetch-shallow.sh
>>> +++ b/t/t5536-fetch-shallow.sh
>>> @@ -141,4 +141,26 @@ EOF
>>>         )
>>>  '
>>>
>>> +test_expect_success 'fetch --update-shallow' '
>>> +       (
>>> +       cd notshallow &&
>>> +       git fetch --update-shallow ../shallow/.git refs/heads/*:refs/remotes/shallow/* &&
>>> +       git fsck &&
>>> +       git for-each-ref --format="%(refname)" |sort >actual.refs &&
>>
>> Exit status of git-for-each-ref could be lost down the pipe.
>
> I think it's ok. If for-each-ref is broken or something the final
> output in actual.refs will be different anyway.

You can tell for-each-ref to sort its output in a way you specify,
no?

Re: [PATCH v3 16/28] receive-pack: reorder some code in unpack()

[Junio C Hamano]

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> This is the preparation for adding --shallow-file to both
> unpack-objects and index-pack. To sum up:
>
>  - status/code, ip/child, unpacker/keeper and i (now ac) are moved out
>    to function top level
>
>  - successful flow now ends at the end of the function
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>

Sounds sensible; for a new piece of code, we probably should see if
argv_array API is more appropriate, and I think it is for this case.

Thanks.

> ---
>  builtin/receive-pack.c | 74 +++++++++++++++++++++++---------------------------
>  1 file changed, 34 insertions(+), 40 deletions(-)
>
> diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
> index cc8c34f..22e162d 100644
> --- a/builtin/receive-pack.c
> +++ b/builtin/receive-pack.c
> @@ -822,8 +822,11 @@ static const char *pack_lockfile;
>  static const char *unpack(int err_fd)
>  {
>  	struct pack_header hdr;
> +	const char *argv[7];
>  	const char *hdr_err;
> +	int status, ac = 0;
>  	char hdr_arg[38];
> +	struct child_process child;
>  	int fsck_objects = (receive_fsck_objects >= 0
>  			    ? receive_fsck_objects
>  			    : transfer_fsck_objects >= 0
> @@ -840,63 +843,54 @@ static const char *unpack(int err_fd)
>  			"--pack_header=%"PRIu32",%"PRIu32,
>  			ntohl(hdr.hdr_version), ntohl(hdr.hdr_entries));
>  
> +	memset(&child, 0, sizeof(child));
>  	if (ntohl(hdr.hdr_entries) < unpack_limit) {
> -		int code, i = 0;
> -		struct child_process child;
> -		const char *unpacker[5];
> -		unpacker[i++] = "unpack-objects";
> +		argv[ac++] = "unpack-objects";
>  		if (quiet)
> -			unpacker[i++] = "-q";
> +			argv[ac++] = "-q";
>  		if (fsck_objects)
> -			unpacker[i++] = "--strict";
> -		unpacker[i++] = hdr_arg;
> -		unpacker[i++] = NULL;
> -		memset(&child, 0, sizeof(child));
> -		child.argv = unpacker;
> +			argv[ac++] = "--strict";
> +		argv[ac++] = hdr_arg;
> +		argv[ac++] = NULL;
> +		child.argv = argv;
>  		child.no_stdout = 1;
>  		child.err = err_fd;
>  		child.git_cmd = 1;
> -		code = run_command(&child);
> -		if (!code)
> -			return NULL;
> -		return "unpack-objects abnormal exit";
> +		status = run_command(&child);
> +		if (status)
> +			return "unpack-objects abnormal exit";
>  	} else {
> -		const char *keeper[7];
> -		int s, status, i = 0;
> +		int s;
>  		char keep_arg[256];
> -		struct child_process ip;
>  
>  		s = sprintf(keep_arg, "--keep=receive-pack %"PRIuMAX" on ", (uintmax_t) getpid());
>  		if (gethostname(keep_arg + s, sizeof(keep_arg) - s))
>  			strcpy(keep_arg + s, "localhost");
>  
> -		keeper[i++] = "index-pack";
> -		keeper[i++] = "--stdin";
> +		argv[ac++] = "index-pack";
> +		argv[ac++] = "--stdin";
>  		if (fsck_objects)
> -			keeper[i++] = "--strict";
> +			argv[ac++] = "--strict";
>  		if (fix_thin)
> -			keeper[i++] = "--fix-thin";
> -		keeper[i++] = hdr_arg;
> -		keeper[i++] = keep_arg;
> -		keeper[i++] = NULL;
> -		memset(&ip, 0, sizeof(ip));
> -		ip.argv = keeper;
> -		ip.out = -1;
> -		ip.err = err_fd;
> -		ip.git_cmd = 1;
> -		status = start_command(&ip);
> -		if (status) {
> +			argv[ac++] = "--fix-thin";
> +		argv[ac++] = hdr_arg;
> +		argv[ac++] = keep_arg;
> +		argv[ac++] = NULL;
> +		child.argv = argv;
> +		child.out = -1;
> +		child.err = err_fd;
> +		child.git_cmd = 1;
> +		status = start_command(&child);
> +		if (status)
>  			return "index-pack fork failed";
> -		}
> -		pack_lockfile = index_pack_lockfile(ip.out);
> -		close(ip.out);
> -		status = finish_command(&ip);
> -		if (!status) {
> -			reprepare_packed_git();
> -			return NULL;
> -		}
> -		return "index-pack abnormal exit";
> +		pack_lockfile = index_pack_lockfile(child.out);
> +		close(child.out);
> +		status = finish_command(&child);
> +		if (status)
> +			return "index-pack abnormal exit";
> +		reprepare_packed_git();
>  	}
> +	return NULL;
>  }
>  
>  static const char *unpack_with_sideband(void)

[PATCH v4 00/28] First class shallow clone

[Nguyễn Thái Ngọc Duy]

This reroll should fix all comments I have received in v3. I reordered
the shallow checks a bit so in common case it should be as cheap as
a normal fetch or push. See 08/28 and 20/28 for the big picture. I'm not
entirely happy with the hook issue in 20/28, but it looks good enough
for me. There are a few XXXes for further improvement, but I'll keep
them until this lands.

To recap, this series allows you to clone from a shallow repo, push to
or fetch from any shallow repo. Normally it will reject new refs
that introduce new shallow boundaries to your repository, so if you're
in a full clone, it will always stay a full clone. Use "fetch
--update-shallow" or set receive.shallowupdate to accept those refs.

Nguyễn Thái Ngọc Duy (28):
  transport.h: remove send_pack prototype, already defined in send-pack.h
  Replace struct extra_have_objects with struct sha1_array
  send-pack: forbid pushing from a shallow repository
  clone: prevent --reference to a shallow repository
  Make the sender advertise shallow commits to the receiver
  connect.c: teach get_remote_heads to parse "shallow" lines
  shallow.c: extend setup_*_shallow() to accept extra shallow commits
  shallow.c: the 8 steps to select new commits for .git/shallow
  shallow.c: steps 6 and 7 to select new commits for .git/shallow
  fetch-pack.c: move shallow update code out of fetch_pack()
  fetch-pack.h: one statement per bitfield declaration
  clone: support remote shallow repository
  fetch: support fetching from a shallow repository
  upload-pack: make sure deepening preserves shallow roots
  fetch: add --update-shallow to accept refs that update .git/shallow
  receive-pack: reorder some code in unpack()
  receive/send-pack: support pushing from a shallow clone
  New var GIT_SHALLOW_FILE to propagate --shallow-file to subprocesses
  connected.c: add new variant that runs with --shallow-file
  receive-pack: allow pushes that update .git/shallow
  send-pack: support pushing to a shallow clone
  remote-curl: pass ref SHA-1 to fetch-pack as well
  Support shallow fetch/clone over smart-http
  receive-pack: support pushing to a shallow clone via http
  send-pack: support pushing from a shallow clone via http
  clone: use git protocol for cloning shallow repo locally
  prune: clean .git/shallow after pruning objects
  git-clone.txt: remove shallow clone limitations

 Documentation/config.txt                  |   4 +
 Documentation/fetch-options.txt           |  14 +-
 Documentation/git-clone.txt               |   7 +-
 Documentation/git-prune.txt               |   2 +
 Documentation/gitremote-helpers.txt       |   7 +
 Documentation/technical/pack-protocol.txt |   7 +-
 builtin/clone.c                           |  18 +-
 builtin/fetch-pack.c                      |  23 +-
 builtin/fetch.c                           |  15 +-
 builtin/gc.c                              |   1 +
 builtin/prune.c                           |   4 +
 builtin/receive-pack.c                    | 314 +++++++++++++++----
 builtin/send-pack.c                       |   9 +-
 cache.h                                   |   3 +
 commit.h                                  |  39 ++-
 connect.c                                 |  22 +-
 connected.c                               |  42 ++-
 connected.h                               |   2 +
 environment.c                             |   6 +
 fetch-pack.c                              | 132 +++++++-
 fetch-pack.h                              |  29 +-
 git.c                                     |   2 +-
 remote-curl.c                             |  34 ++-
 remote.h                                  |   9 +-
 send-pack.c                               |  27 +-
 send-pack.h                               |   2 +-
 shallow.c                                 | 486 +++++++++++++++++++++++++++++-
 t/t5304-prune.sh                          |  10 +
 t/t5536-fetch-shallow.sh (new +x)         | 203 +++++++++++++
 t/t5537-push-shallow.sh (new +x)          | 183 +++++++++++
 t/t5601-clone.sh                          |   7 +
 trace.c                                   |   2 +-
 transport-helper.c                        |   6 +
 transport.c                               |  25 +-
 transport.h                               |  16 +-
 upload-pack.c                             |   8 +-
 36 files changed, 1555 insertions(+), 165 deletions(-)
 create mode 100755 t/t5536-fetch-shallow.sh
 create mode 100755 t/t5537-push-shallow.sh

-- 
1.8.5.1.25.g8667982

[PATCH v4 01/28] transport.h: remove send_pack prototype, already defined in send-pack.h

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 transport.h | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/transport.h b/transport.h
index 8f96bed..b3679bb 100644
--- a/transport.h
+++ b/transport.h
@@ -193,10 +193,4 @@ void transport_print_push_status(const char *dest, struct ref *refs,
 
 typedef void alternate_ref_fn(const struct ref *, void *);
 extern void for_each_alternate_ref(alternate_ref_fn, void *);
-
-struct send_pack_args;
-extern int send_pack(struct send_pack_args *args,
-		     int fd[], struct child_process *conn,
-		     struct ref *remote_refs,
-		     struct extra_have_objects *extra_have);
 #endif
-- 
1.8.5.1.25.g8667982

[PATCH v4 02/28] Replace struct extra_have_objects with struct sha1_array

[Nguyễn Thái Ngọc Duy]

The latter can do everything the former can and is used in many more
places.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/send-pack.c |  5 ++---
 connect.c           | 12 +++---------
 remote.h            |  7 ++-----
 send-pack.c         |  7 ++++---
 send-pack.h         |  2 +-
 transport.c         |  3 ++-
 6 files changed, 14 insertions(+), 22 deletions(-)

diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index 4482f16..faaa603 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -10,6 +10,7 @@
 #include "quote.h"
 #include "transport.h"
 #include "version.h"
+#include "sha1-array.h"
 
 static const char send_pack_usage[] =
 "git send-pack [--all | --mirror] [--dry-run] [--force] [--receive-pack=<git-receive-pack>] [--verbose] [--thin] [<host>:]<directory> [<ref>...]\n"
@@ -99,7 +100,7 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 	const char *dest = NULL;
 	int fd[2];
 	struct child_process *conn;
-	struct extra_have_objects extra_have;
+	struct sha1_array extra_have = SHA1_ARRAY_INIT;
 	struct ref *remote_refs, *local_refs;
 	int ret;
 	int helper_status = 0;
@@ -228,8 +229,6 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 			args.verbose ? CONNECT_VERBOSE : 0);
 	}
 
-	memset(&extra_have, 0, sizeof(extra_have));
-
 	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL, &extra_have);
 
 	transport_verify_remote_names(nr_refspecs, refspecs);
diff --git a/connect.c b/connect.c
index 06e88b0..48eec41 100644
--- a/connect.c
+++ b/connect.c
@@ -8,6 +8,7 @@
 #include "connect.h"
 #include "url.h"
 #include "string-list.h"
+#include "sha1-array.h"
 
 static char *server_capabilities;
 static const char *parse_feature_value(const char *, const char *, int *);
@@ -45,13 +46,6 @@ int check_ref_type(const struct ref *ref, int flags)
 	return check_ref(ref->name, strlen(ref->name), flags);
 }
 
-static void add_extra_have(struct extra_have_objects *extra, unsigned char *sha1)
-{
-	ALLOC_GROW(extra->array, extra->nr + 1, extra->alloc);
-	hashcpy(&(extra->array[extra->nr][0]), sha1);
-	extra->nr++;
-}
-
 static void die_initial_contact(int got_at_least_one_head)
 {
 	if (got_at_least_one_head)
@@ -122,7 +116,7 @@ static void annotate_refs_with_symref_info(struct ref *ref)
  */
 struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
 			      struct ref **list, unsigned int flags,
-			      struct extra_have_objects *extra_have)
+			      struct sha1_array *extra_have)
 {
 	struct ref **orig_list = list;
 	int got_at_least_one_head = 0;
@@ -160,7 +154,7 @@ struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
 
 		if (extra_have &&
 		    name_len == 5 && !memcmp(".have", name, 5)) {
-			add_extra_have(extra_have, old_sha1);
+			sha1_array_append(extra_have, old_sha1);
 			continue;
 		}
 
diff --git a/remote.h b/remote.h
index 131130a..984519b 100644
--- a/remote.h
+++ b/remote.h
@@ -137,13 +137,10 @@ int check_ref_type(const struct ref *ref, int flags);
  */
 void free_refs(struct ref *ref);
 
-struct extra_have_objects {
-	int nr, alloc;
-	unsigned char (*array)[20];
-};
+struct sha1_array;
 extern struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
 				     struct ref **list, unsigned int flags,
-				     struct extra_have_objects *);
+				     struct sha1_array *extra_have);
 
 int resolve_remote_symref(struct ref *ref, struct ref *list);
 int ref_newer(const unsigned char *new_sha1, const unsigned char *old_sha1);
diff --git a/send-pack.c b/send-pack.c
index fab62e3..14005fa 100644
--- a/send-pack.c
+++ b/send-pack.c
@@ -10,6 +10,7 @@
 #include "quote.h"
 #include "transport.h"
 #include "version.h"
+#include "sha1-array.h"
 
 static int feed_object(const unsigned char *sha1, int fd, int negative)
 {
@@ -28,7 +29,7 @@ static int feed_object(const unsigned char *sha1, int fd, int negative)
 /*
  * Make a pack stream and spit it out into file descriptor fd
  */
-static int pack_objects(int fd, struct ref *refs, struct extra_have_objects *extra, struct send_pack_args *args)
+static int pack_objects(int fd, struct ref *refs, struct sha1_array *extra, struct send_pack_args *args)
 {
 	/*
 	 * The child becomes pack-objects --revs; we feed
@@ -71,7 +72,7 @@ static int pack_objects(int fd, struct ref *refs, struct extra_have_objects *ext
 	 * parameters by writing to the pipe.
 	 */
 	for (i = 0; i < extra->nr; i++)
-		if (!feed_object(extra->array[i], po.in, 1))
+		if (!feed_object(extra->sha1[i], po.in, 1))
 			break;
 
 	while (refs) {
@@ -177,7 +178,7 @@ static int sideband_demux(int in, int out, void *data)
 int send_pack(struct send_pack_args *args,
 	      int fd[], struct child_process *conn,
 	      struct ref *remote_refs,
-	      struct extra_have_objects *extra_have)
+	      struct sha1_array *extra_have)
 {
 	int in = fd[0];
 	int out = fd[1];
diff --git a/send-pack.h b/send-pack.h
index 05d7ab1..8e84392 100644
--- a/send-pack.h
+++ b/send-pack.h
@@ -16,6 +16,6 @@ struct send_pack_args {
 
 int send_pack(struct send_pack_args *args,
 	      int fd[], struct child_process *conn,
-	      struct ref *remote_refs, struct extra_have_objects *extra_have);
+	      struct ref *remote_refs, struct sha1_array *extra_have);
 
 #endif
diff --git a/transport.c b/transport.c
index 7202b77..12e46ad 100644
--- a/transport.c
+++ b/transport.c
@@ -14,6 +14,7 @@
 #include "url.h"
 #include "submodule.h"
 #include "string-list.h"
+#include "sha1-array.h"
 
 /* rsync support */
 
@@ -454,7 +455,7 @@ struct git_transport_data {
 	struct child_process *conn;
 	int fd[2];
 	unsigned got_remote_heads : 1;
-	struct extra_have_objects extra_have;
+	struct sha1_array extra_have;
 };
 
 static int set_git_option(struct git_transport_options *opts,
-- 
1.8.5.1.25.g8667982

[PATCH v4 03/28] send-pack: forbid pushing from a shallow repository

[Nguyễn Thái Ngọc Duy]

send-pack can send a pack with loose ends to the server.  receive-pack
before 6d4bb38 (fetch: verify we have everything we need before
updating our ref - 2011-09-01) does not detect this and keeps the pack
anyway, which corrupts the repository, at least from fsck point of
view.

send-pack will learn to safely push from a shallow repository later.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/send-pack.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index faaa603..961df04 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -208,6 +208,9 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 	    (send_all && args.send_mirror))
 		usage(send_pack_usage);
 
+	if (is_repository_shallow())
+		die("attempt to push from a shallow repository");
+
 	if (remote_name) {
 		remote = remote_get(remote_name);
 		if (!remote_has_url(remote, dest)) {
-- 
1.8.5.1.25.g8667982

[PATCH v4 04/28] clone: prevent --reference to a shallow repository

[Nguyễn Thái Ngọc Duy]

If we borrow objects from another repository, we should also pay
attention to their $GIT_DIR/shallow (and even info/grafts). But
current alternates code does not.

Reject alternate repos that are shallow because we do not do it
right. In future the alternate code may be updated to check
$GIT_DIR/shallow properly so that this restriction could be lifted.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/clone.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/builtin/clone.c b/builtin/clone.c
index 874e0fd..900f564 100644
--- a/builtin/clone.c
+++ b/builtin/clone.c
@@ -252,6 +252,12 @@ static int add_one_reference(struct string_list_item *item, void *cb_data)
 		die(_("reference repository '%s' is not a local repository."),
 		    item->string);
 
+	if (!access(mkpath("%s/shallow", ref_git), F_OK))
+		die(_("reference repository '%s' is shallow"), item->string);
+
+	if (!access(mkpath("%s/info/grafts", ref_git), F_OK))
+		die(_("reference repository '%s' is grafted"), item->string);
+
 	strbuf_addf(&alternate, "%s/objects", ref_git);
 	add_to_alternates_file(alternate.buf);
 	strbuf_release(&alternate);
-- 
1.8.5.1.25.g8667982

[PATCH v4 05/28] Make the sender advertise shallow commits to the receiver

[Nguyễn Thái Ngọc Duy]

If either receive-pack or upload-pack is called on a shallow
repository, shallow commits (*) will be sent after the ref
advertisement (but before the packet flush), so that the receiver has
the full "shape" of the sender's commit graph. This will be needed for
the receiver to update its .git/shallow if necessary.

This breaks the protocol for all clients trying to push to a shallow
repo, or fetch from one. Which is basically the same end result as
today's "is_repository_shallow() && die()" in receive-pack and
upload-pack. New clients will be made aware of shallow upstream and
can make use of this information.

The sender must send all shallow commits that are sent in the
following pack. It may send more shallow commits than necessary.

upload-pack for example may choose to advertise no shallow commits if
it knows in advance that the pack it's going to send contains no
shallow commits. But upload-pack is the server, so we choose the
cheaper way, send full .git/shallow and let the client deal with it.

Smart HTTP is not affected by this patch. Shallow support on
smart-http comes later separately.

(*) A shallow commit is a commit that terminates the revision
    walker. It is usually put in .git/shallow in order to keep the
    revision walker from going out of bound because there is no
    guarantee that objects behind this commit is available.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/technical/pack-protocol.txt |  3 +++
 builtin/receive-pack.c                    |  4 +++-
 commit.h                                  |  1 +
 shallow.c                                 | 15 +++++++++++++++
 upload-pack.c                             |  6 ++++--
 5 files changed, 26 insertions(+), 3 deletions(-)

diff --git a/Documentation/technical/pack-protocol.txt b/Documentation/technical/pack-protocol.txt
index b898e97..eb8edd1 100644
--- a/Documentation/technical/pack-protocol.txt
+++ b/Documentation/technical/pack-protocol.txt
@@ -161,6 +161,7 @@ MUST peel the ref if it's an annotated tag.
 
 ----
   advertised-refs  =  (no-refs / list-of-refs)
+		      *shallow
 		      flush-pkt
 
   no-refs          =  PKT-LINE(zero-id SP "capabilities^{}"
@@ -174,6 +175,8 @@ MUST peel the ref if it's an annotated tag.
   other-tip        =  obj-id SP refname LF
   other-peeled     =  obj-id SP refname "^{}" LF
 
+  shallow          =  PKT-LINE("shallow" SP obj-id)
+
   capability-list  =  capability *(SP capability)
   capability       =  1*(LC_ALPHA / DIGIT / "-" / "_")
   LC_ALPHA         =  %x61-7A
diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index 67ce1ef..cc8c34f 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -178,6 +178,8 @@ static void write_head_info(void)
 	if (!sent_capabilities)
 		show_ref("capabilities^{}", null_sha1);
 
+	advertise_shallow_grafts(1);
+
 	/* EOF */
 	packet_flush(1);
 }
@@ -998,7 +1000,7 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
 	if (!enter_repo(dir, 0))
 		die("'%s' does not appear to be a git repository", dir);
 
-	if (is_repository_shallow())
+	if (is_repository_shallow() && stateless_rpc)
 		die("attempt to push into a shallow repository");
 
 	git_config(receive_pack_config, NULL);
diff --git a/commit.h b/commit.h
index bd841f4..a879526 100644
--- a/commit.h
+++ b/commit.h
@@ -205,6 +205,7 @@ extern int write_shallow_commits(struct strbuf *out, int use_pack_protocol);
 extern void setup_alternate_shallow(struct lock_file *shallow_lock,
 				    const char **alternate_shallow_file);
 extern char *setup_temporary_shallow(void);
+extern void advertise_shallow_grafts(int);
 
 int is_descendant_of(struct commit *, struct commit_list *);
 int in_merge_bases(struct commit *, struct commit *);
diff --git a/shallow.c b/shallow.c
index cdf37d6..9552512 100644
--- a/shallow.c
+++ b/shallow.c
@@ -220,3 +220,18 @@ void setup_alternate_shallow(struct lock_file *shallow_lock,
 		*alternate_shallow_file = "";
 	strbuf_release(&sb);
 }
+
+static int advertise_shallow_grafts_cb(const struct commit_graft *graft, void *cb)
+{
+	int fd = *(int*)cb;
+	if (graft->nr_parent == -1)
+		packet_write(fd, "shallow %s\n", sha1_to_hex(graft->sha1));
+	return 0;
+}
+
+void advertise_shallow_grafts(int fd)
+{
+	if (!is_repository_shallow())
+		return;
+	for_each_commit_graft(advertise_shallow_grafts_cb, &fd);
+}
diff --git a/upload-pack.c b/upload-pack.c
index c989a73..38b2a29 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -758,6 +758,7 @@ static void upload_pack(void)
 		reset_timeout();
 		head_ref_namespaced(send_ref, &symref);
 		for_each_namespaced_ref(send_ref, &symref);
+		advertise_shallow_grafts(1);
 		packet_flush(1);
 	} else {
 		head_ref_namespaced(mark_our_ref, NULL);
@@ -835,8 +836,9 @@ int main(int argc, char **argv)
 
 	if (!enter_repo(dir, strict))
 		die("'%s' does not appear to be a git repository", dir);
-	if (is_repository_shallow())
-		die("attempt to fetch/clone from a shallow repository");
+	if (is_repository_shallow() && stateless_rpc)
+		die("attempt to push into a shallow repository");
+
 	git_config(upload_pack_config, NULL);
 	upload_pack();
 	return 0;
-- 
1.8.5.1.25.g8667982

[PATCH v4 06/28] connect.c: teach get_remote_heads to parse "shallow" lines

[Nguyễn Thái Ngọc Duy]

No callers pass a non-empty pointer as shallow_points at this
stage. As a result, all clients still refuse to talk to shallow
repository on the other end.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/fetch-pack.c |  2 +-
 builtin/send-pack.c  |  2 +-
 connect.c            | 12 +++++++++++-
 remote-curl.c        |  2 +-
 remote.h             |  3 ++-
 transport.c          |  7 ++++---
 6 files changed, 20 insertions(+), 8 deletions(-)

diff --git a/builtin/fetch-pack.c b/builtin/fetch-pack.c
index c8e8582..c1d918f 100644
--- a/builtin/fetch-pack.c
+++ b/builtin/fetch-pack.c
@@ -150,7 +150,7 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
 				   args.verbose ? CONNECT_VERBOSE : 0);
 	}
 
-	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL);
+	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL, NULL);
 
 	ref = fetch_pack(&args, fd, conn, ref, dest,
 			 sought, nr_sought, pack_lockfile_ptr);
diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index 961df04..62cc4d3 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -232,7 +232,7 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 			args.verbose ? CONNECT_VERBOSE : 0);
 	}
 
-	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL, &extra_have);
+	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL, &extra_have, NULL);
 
 	transport_verify_remote_names(nr_refspecs, refspecs);
 
diff --git a/connect.c b/connect.c
index 48eec41..efadd3c 100644
--- a/connect.c
+++ b/connect.c
@@ -116,7 +116,8 @@ static void annotate_refs_with_symref_info(struct ref *ref)
  */
 struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
 			      struct ref **list, unsigned int flags,
-			      struct sha1_array *extra_have)
+			      struct sha1_array *extra_have,
+			      struct sha1_array *shallow_points)
 {
 	struct ref **orig_list = list;
 	int got_at_least_one_head = 0;
@@ -142,6 +143,15 @@ struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
 		if (len > 4 && !prefixcmp(buffer, "ERR "))
 			die("remote error: %s", buffer + 4);
 
+		if (len == 48 && !prefixcmp(buffer, "shallow ")) {
+			if (get_sha1_hex(buffer + 8, old_sha1))
+				die("protocol error: expected shallow sha-1, got '%s'", buffer + 8);
+			if (!shallow_points)
+				die("repository on the other end cannot be shallow");
+			sha1_array_append(shallow_points, old_sha1);
+			continue;
+		}
+
 		if (len < 42 || get_sha1_hex(buffer, old_sha1) || buffer[40] != ' ')
 			die("protocol error: expected sha/ref, got '%s'", buffer);
 		name = buffer + 41;
diff --git a/remote-curl.c b/remote-curl.c
index c9b891a..222210f 100644
--- a/remote-curl.c
+++ b/remote-curl.c
@@ -107,7 +107,7 @@ static struct ref *parse_git_refs(struct discovery *heads, int for_push)
 {
 	struct ref *list = NULL;
 	get_remote_heads(-1, heads->buf, heads->len, &list,
-			 for_push ? REF_NORMAL : 0, NULL);
+			 for_push ? REF_NORMAL : 0, NULL, NULL);
 	return list;
 }
 
diff --git a/remote.h b/remote.h
index 984519b..5d217d5 100644
--- a/remote.h
+++ b/remote.h
@@ -140,7 +140,8 @@ void free_refs(struct ref *ref);
 struct sha1_array;
 extern struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
 				     struct ref **list, unsigned int flags,
-				     struct sha1_array *extra_have);
+				     struct sha1_array *extra_have,
+				     struct sha1_array *shallow);
 
 int resolve_remote_symref(struct ref *ref, struct ref *list);
 int ref_newer(const unsigned char *new_sha1, const unsigned char *old_sha1);
diff --git a/transport.c b/transport.c
index 12e46ad..90453df 100644
--- a/transport.c
+++ b/transport.c
@@ -512,7 +512,7 @@ static struct ref *get_refs_via_connect(struct transport *transport, int for_pus
 
 	connect_setup(transport, for_push, 0);
 	get_remote_heads(data->fd[0], NULL, 0, &refs,
-			 for_push ? REF_NORMAL : 0, &data->extra_have);
+			 for_push ? REF_NORMAL : 0, &data->extra_have, NULL);
 	data->got_remote_heads = 1;
 
 	return refs;
@@ -542,7 +542,8 @@ static int fetch_refs_via_pack(struct transport *transport,
 
 	if (!data->got_remote_heads) {
 		connect_setup(transport, 0, 0);
-		get_remote_heads(data->fd[0], NULL, 0, &refs_tmp, 0, NULL);
+		get_remote_heads(data->fd[0], NULL, 0, &refs_tmp, 0,
+				 NULL, NULL);
 		data->got_remote_heads = 1;
 	}
 
@@ -806,7 +807,7 @@ static int git_transport_push(struct transport *transport, struct ref *remote_re
 		struct ref *tmp_refs;
 		connect_setup(transport, 1, 0);
 
-		get_remote_heads(data->fd[0], NULL, 0, &tmp_refs, REF_NORMAL, NULL);
+		get_remote_heads(data->fd[0], NULL, 0, &tmp_refs, REF_NORMAL, NULL, NULL);
 		data->got_remote_heads = 1;
 	}
 
-- 
1.8.5.1.25.g8667982

[PATCH v4 07/28] shallow.c: extend setup_*_shallow() to accept extra shallow commits

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 commit.h      |  8 +++++---
 fetch-pack.c  |  5 +++--
 shallow.c     | 20 +++++++++++++++-----
 upload-pack.c |  2 +-
 4 files changed, 24 insertions(+), 11 deletions(-)

diff --git a/commit.h b/commit.h
index a879526..1faf717 100644
--- a/commit.h
+++ b/commit.h
@@ -201,10 +201,12 @@ extern struct commit_list *get_shallow_commits(struct object_array *heads,
 		int depth, int shallow_flag, int not_shallow_flag);
 extern void check_shallow_file_for_update(void);
 extern void set_alternate_shallow_file(const char *path);
-extern int write_shallow_commits(struct strbuf *out, int use_pack_protocol);
+extern int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
+				 const struct sha1_array *extra);
 extern void setup_alternate_shallow(struct lock_file *shallow_lock,
-				    const char **alternate_shallow_file);
-extern char *setup_temporary_shallow(void);
+				    const char **alternate_shallow_file,
+				    const struct sha1_array *extra);
+extern char *setup_temporary_shallow(const struct sha1_array *extra);
 extern void advertise_shallow_grafts(int);
 
 int is_descendant_of(struct commit *, struct commit_list *);
diff --git a/fetch-pack.c b/fetch-pack.c
index 1042448..0e7483e 100644
--- a/fetch-pack.c
+++ b/fetch-pack.c
@@ -311,7 +311,7 @@ static int find_common(struct fetch_pack_args *args,
 	}
 
 	if (is_repository_shallow())
-		write_shallow_commits(&req_buf, 1);
+		write_shallow_commits(&req_buf, 1, NULL);
 	if (args->depth > 0)
 		packet_buf_write(&req_buf, "deepen %d", args->depth);
 	packet_buf_flush(&req_buf);
@@ -850,7 +850,8 @@ static struct ref *do_fetch_pack(struct fetch_pack_args *args,
 	if (args->stateless_rpc)
 		packet_flush(fd[1]);
 	if (args->depth > 0)
-		setup_alternate_shallow(&shallow_lock, &alternate_shallow_file);
+		setup_alternate_shallow(&shallow_lock, &alternate_shallow_file,
+					NULL);
 	else
 		alternate_shallow_file = NULL;
 	if (get_pack(args, fd, pack_lockfile))
diff --git a/shallow.c b/shallow.c
index 9552512..f9d1633 100644
--- a/shallow.c
+++ b/shallow.c
@@ -165,22 +165,31 @@ static int write_one_shallow(const struct commit_graft *graft, void *cb_data)
 	return 0;
 }
 
-int write_shallow_commits(struct strbuf *out, int use_pack_protocol)
+int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
+			  const struct sha1_array *extra)
 {
 	struct write_shallow_data data;
+	int i;
 	data.out = out;
 	data.use_pack_protocol = use_pack_protocol;
 	data.count = 0;
 	for_each_commit_graft(write_one_shallow, &data);
+	if (!extra)
+		return data.count;
+	for (i = 0; i < extra->nr; i++) {
+		strbuf_addstr(out, sha1_to_hex(extra->sha1[i]));
+		strbuf_addch(out, '\n');
+		data.count++;
+	}
 	return data.count;
 }
 
-char *setup_temporary_shallow(void)
+char *setup_temporary_shallow(const struct sha1_array *extra)
 {
 	struct strbuf sb = STRBUF_INIT;
 	int fd;
 
-	if (write_shallow_commits(&sb, 0)) {
+	if (write_shallow_commits(&sb, 0, extra)) {
 		struct strbuf path = STRBUF_INIT;
 		strbuf_addstr(&path, git_path("shallow_XXXXXX"));
 		fd = xmkstemp(path.buf);
@@ -199,7 +208,8 @@ char *setup_temporary_shallow(void)
 }
 
 void setup_alternate_shallow(struct lock_file *shallow_lock,
-			     const char **alternate_shallow_file)
+			     const char **alternate_shallow_file,
+			     const struct sha1_array *extra)
 {
 	struct strbuf sb = STRBUF_INIT;
 	int fd;
@@ -207,7 +217,7 @@ void setup_alternate_shallow(struct lock_file *shallow_lock,
 	check_shallow_file_for_update();
 	fd = hold_lock_file_for_update(shallow_lock, git_path("shallow"),
 				       LOCK_DIE_ON_ERROR);
-	if (write_shallow_commits(&sb, 0)) {
+	if (write_shallow_commits(&sb, 0, extra)) {
 		if (write_in_full(fd, sb.buf, sb.len) != sb.len)
 			die_errno("failed to write to %s",
 				  shallow_lock->filename);
diff --git a/upload-pack.c b/upload-pack.c
index 38b2a29..f082f06 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -84,7 +84,7 @@ static void create_pack_file(void)
 	char *shallow_file = NULL;
 
 	if (shallow_nr) {
-		shallow_file = setup_temporary_shallow();
+		shallow_file = setup_temporary_shallow(NULL);
 		argv[arg++] = "--shallow-file";
 		argv[arg++] = shallow_file;
 	}
-- 
1.8.5.1.25.g8667982

[PATCH v4 08/28] shallow.c: the 8 steps to select new commits for .git/shallow

[Nguyễn Thái Ngọc Duy]

Suppose a fetch or push is requested between two shallow repositories
(with no history deepening or shortening). A pack that contains
necessary objects is transferred over together with .git/shallow of
the sender. The receiver has to determine whether it needs to update
.git/shallow if new refs needs new shallow comits.

The rule here is avoid updating .git/shallow by default. But we don't
want to waste the received pack. If the pack contains two refs, one
needs new shallow commits installed in .git/shallow and one does not,
we keep the latter and reject/warn about the former.

Even if .git/shallow update is allowed, we only add shallow commits
strictly necessary for the former ref (remember the sender can send
more shallow commits than necessary) and pay attention not to
accidentally cut the receiver history short (no history shortening is
asked for)

So the steps to figure out what ref need what new shallow commits are:

1. Split the sender shallow commit list into "ours" and "theirs" list
   by has_sha1_file. Those that exist in current repo in "ours", the
   remaining in "theirs".

2. Check the receiver .git/shallow, remove from "ours" the ones that
   also exist in .git/shallow.

3. Fetch the new pack. Either install or unpack it.

4. Do has_sha1_file on "theirs" list again. Drop the ones that fail
   has_sha1_file. Obviously the new pack does not need them.

5. If the pack is kept, remove from "ours" the ones that do not exist
   in the new pack.

6. Walk the new refs to answer the question "what shallow commits,
   both ours and theirs, are required in .git/shallow in order to add
   this ref?". Shallow commits not associated to any refs are removed
   from their respective list.

7. (*) Check reachability (from the current refs) of all remaining
   commits in "ours". Those reachable are removed. We do not want to
   cut any part of our (reachable) history. We only check up
   commits. True reachability test is done by
   check_everything_connected() at the end as usual.

8. Combine the final "ours" and "theirs" and add them all to
   .git/shallow. Install new refs. The case where some hook rejects
   some refs on a push is explained in more detail in the push
   patches.

Of these steps, #6 and #7 are expensive. Both require walking through
some commits, or in the worst case all commits. And we rather avoid
them in at least common case, where the transferred pack does not
contain any shallow commits that the sender advertises. Let's look at
each scenario:

1) the sender has longer history than the receiver

   All shallow commits from the sender will be put into "theirs" list
   at step 1 because none of them exists in current repo. In the
   common case, "theirs" becomes empty at step 4 and exit early.

2) the sender has shorter history than the receiver

   All shallow commits from the sender are likely in "ours" list at
   step 1. In the common case, if the new pack is kept, we could empty
   "ours" and exit early at step 5.

   If the pack is not kept, we hit the expensive step 6 then exit
   after "ours" is emptied. There'll be only a handful of objects to
   walk in fast-forward case. If it's forced update, we may need to
   walk to the bottom.

3) the sender has same .git/shallow as the receiver

   This is similar to case 2 except that "ours" should be emptied at
   step 2 and exit early.

A fetch after "clone --depth=X" is case 1. A fetch after "clone" (from
a shallow repo) is case 3. Luckily they're cheap for the common case.

A push from "clone --depth=X" falls into case 2, which is expensive.
Some more work may be done at the sender/client side to avoid more
work on the server side: if the transferred pack does not contain any
shallow commits, send-pack should not send any shallow commits to the
receive-pack, effectively turning it into a normal push and avoid all
steps.

This patch implements all steps except #3, already handled by
fetch-pack and receive-pack, #6 and #7, which has their own patch due
to their size.

(*) in previous versions step 7 was put before step 3. I reorder it so
    that the common case that keeps the pack does not need to walk
    commits at all. In future if we implement faster commit
    reachability check (maybe with the help of pack bitmaps or commit
    cache), step 7 could become cheap and be moved up before 6 again.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 cache.h   |  2 ++
 commit.h  | 15 +++++++++++++
 shallow.c | 72 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 trace.c   |  2 +-
 4 files changed, 90 insertions(+), 1 deletion(-)

diff --git a/cache.h b/cache.h
index ce377e1..55dd4e3 100644
--- a/cache.h
+++ b/cache.h
@@ -1236,6 +1236,8 @@ __attribute__((format (printf, 2, 3)))
 extern void trace_argv_printf(const char **argv, const char *format, ...);
 extern void trace_repo_setup(const char *prefix);
 extern int trace_want(const char *key);
+__attribute__((format (printf, 2, 3)))
+extern void trace_printf_key(const char *key, const char *fmt, ...);
 extern void trace_strbuf(const char *key, const struct strbuf *buf);
 
 void packet_trace_identity(const char *prog);
diff --git a/commit.h b/commit.h
index 1faf717..9ead93b 100644
--- a/commit.h
+++ b/commit.h
@@ -193,6 +193,8 @@ extern struct commit_list *get_octopus_merge_bases(struct commit_list *in);
 /* largest positive number a signed 32-bit integer can contain */
 #define INFINITE_DEPTH 0x7fffffff
 
+struct sha1_array;
+struct ref;
 extern int register_shallow(const unsigned char *sha1);
 extern int unregister_shallow(const unsigned char *sha1);
 extern int for_each_commit_graft(each_commit_graft_fn, void *);
@@ -209,6 +211,19 @@ extern void setup_alternate_shallow(struct lock_file *shallow_lock,
 extern char *setup_temporary_shallow(const struct sha1_array *extra);
 extern void advertise_shallow_grafts(int);
 
+struct shallow_info {
+	struct sha1_array *shallow;
+	int *ours, nr_ours;
+	int *theirs, nr_theirs;
+	struct sha1_array *ref;
+};
+
+extern void prepare_shallow_info(struct shallow_info *, struct sha1_array *);
+extern void clear_shallow_info(struct shallow_info *);
+extern void remove_nonexistent_theirs_shallow(struct shallow_info *);
+extern void remove_nonexistent_ours_in_pack(struct shallow_info *,
+					    struct packed_git *);
+
 int is_descendant_of(struct commit *, struct commit_list *);
 int in_merge_bases(struct commit *, struct commit *);
 int in_merge_bases_many(struct commit *, int, struct commit **);
diff --git a/shallow.c b/shallow.c
index f9d1633..a6547ca 100644
--- a/shallow.c
+++ b/shallow.c
@@ -2,6 +2,12 @@
 #include "commit.h"
 #include "tag.h"
 #include "pkt-line.h"
+#include "remote.h"
+#include "refs.h"
+#include "sha1-array.h"
+#include "diff.h"
+#include "revision.h"
+#include "commit-slab.h"
 
 static int is_shallow = -1;
 static struct stat shallow_stat;
@@ -245,3 +251,69 @@ void advertise_shallow_grafts(int fd)
 		return;
 	for_each_commit_graft(advertise_shallow_grafts_cb, &fd);
 }
+
+#define TRACE_KEY "GIT_TRACE_SHALLOW"
+
+/*
+ * Step 1, split sender shallow commits into "ours" and "theirs"
+ * Step 2, clean "ours" based on .git/shallow
+ */
+void prepare_shallow_info(struct shallow_info *info, struct sha1_array *sa)
+{
+	int i;
+	trace_printf_key(TRACE_KEY, "shallow: prepare_shallow_info\n");
+	memset(info, 0, sizeof(*info));
+	info->shallow = sa;
+	if (!sa)
+		return;
+	info->ours = xmalloc(sizeof(*info->ours) * sa->nr);
+	info->theirs = xmalloc(sizeof(*info->theirs) * sa->nr);
+	for (i = 0; i < sa->nr; i++) {
+		if (has_sha1_file(sa->sha1[i])) {
+			struct commit_graft *graft;
+			graft = lookup_commit_graft(sa->sha1[i]);
+			if (graft && graft->nr_parent < 0)
+				continue;
+			info->ours[info->nr_ours++] = i;
+		} else
+			info->theirs[info->nr_theirs++] = i;
+	}
+}
+
+void clear_shallow_info(struct shallow_info *info)
+{
+	free(info->ours);
+	free(info->theirs);
+}
+
+/* Step 4, remove non-existent ones in "theirs" after getting the pack */
+
+void remove_nonexistent_theirs_shallow(struct shallow_info *info)
+{
+	unsigned char (*sha1)[20] = info->shallow->sha1;
+	int i, dst;
+	trace_printf_key(TRACE_KEY, "shallow: remove_nonexistent_theirs_shallow\n");
+	for (i = dst = 0; i < info->nr_theirs; i++) {
+		if (i != dst)
+			info->theirs[dst] = info->theirs[i];
+		if (has_sha1_file(sha1[info->theirs[i]]))
+			dst++;
+	}
+	info->nr_theirs = dst;
+}
+
+/* Step 5, remove non-existent ones in "ours" in the pack */
+void remove_nonexistent_ours_in_pack(struct shallow_info *info,
+				     struct packed_git *p)
+{
+	unsigned char (*sha1)[20] = info->shallow->sha1;
+	int i, dst;
+	trace_printf_key(TRACE_KEY, "shallow: remove_nonexistent_ours_in_pack\n");
+	for (i = dst = 0; i < info->nr_ours; i++) {
+		if (i != dst)
+			info->ours[dst] = info->ours[i];
+		if (find_pack_entry_one(sha1[info->ours[i]], p))
+			dst++;
+	}
+	info->nr_ours = dst;
+}
diff --git a/trace.c b/trace.c
index 3d744d1..08180a9 100644
--- a/trace.c
+++ b/trace.c
@@ -76,7 +76,7 @@ static void trace_vprintf(const char *key, const char *fmt, va_list ap)
 }
 
 __attribute__((format (printf, 2, 3)))
-static void trace_printf_key(const char *key, const char *fmt, ...)
+void trace_printf_key(const char *key, const char *fmt, ...)
 {
 	va_list ap;
 	va_start(ap, fmt);
-- 
1.8.5.1.25.g8667982

[PATCH v4 09/28] shallow.c: steps 6 and 7 to select new commits for .git/shallow

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 commit.h  |   3 +
 shallow.c | 295 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 298 insertions(+)

diff --git a/commit.h b/commit.h
index 9ead93b..69bca3e 100644
--- a/commit.h
+++ b/commit.h
@@ -223,6 +223,9 @@ extern void clear_shallow_info(struct shallow_info *);
 extern void remove_nonexistent_theirs_shallow(struct shallow_info *);
 extern void remove_nonexistent_ours_in_pack(struct shallow_info *,
 					    struct packed_git *);
+extern void assign_shallow_commits_to_refs(struct shallow_info *info,
+					   uint32_t **used,
+					   int *ref_status);
 
 int is_descendant_of(struct commit *, struct commit_list *);
 int in_merge_bases(struct commit *, struct commit *);
diff --git a/shallow.c b/shallow.c
index a6547ca..975d699 100644
--- a/shallow.c
+++ b/shallow.c
@@ -317,3 +317,298 @@ void remove_nonexistent_ours_in_pack(struct shallow_info *info,
 	}
 	info->nr_ours = dst;
 }
+
+define_commit_slab(ref_bitmap, uint32_t *);
+
+struct paint_info {
+	struct ref_bitmap ref_bitmap;
+	unsigned nr_bits;
+	char **slab;
+	char *free, *end;
+	unsigned slab_count;
+};
+
+static uint32_t *paint_alloc(struct paint_info *info)
+{
+	unsigned nr = (info->nr_bits + 31) / 32;
+	unsigned size = nr * sizeof(uint32_t);
+	void *p;
+	if (!info->slab_count || info->free + size > info->end) {
+		info->slab_count++;
+		info->slab = xrealloc(info->slab,
+				      info->slab_count * sizeof(*info->slab));
+		info->free = xmalloc(COMMIT_SLAB_SIZE);
+		info->slab[info->slab_count - 1] = info->free;
+		info->end = info->free + COMMIT_SLAB_SIZE;
+	}
+	p = info->free;
+	info->free += size;
+	return p;
+}
+
+/*
+ * Given a commit SHA-1, walk down to parents until either SEEN,
+ * UNINTERESTING or BOTTOM is hit. Set the id-th bit in ref_bitmap for
+ * all walked commits.
+ */
+static void paint_down(struct paint_info *info, const unsigned char *sha1,
+		       int id)
+{
+	unsigned int i, nr;
+	struct commit_list *head = NULL;
+	int bitmap_nr = (info->nr_bits + 31) / 32;
+	int bitmap_size = bitmap_nr * sizeof(uint32_t);
+	uint32_t *tmp = xmalloc(bitmap_size); /* to be freed before return */
+	uint32_t *bitmap = paint_alloc(info);
+	struct commit *c = lookup_commit_reference_gently(sha1, 1);
+	if (!c)
+		return;
+	memset(bitmap, 0, bitmap_size);
+	bitmap[id / 32] |= (1 << (id % 32));
+	commit_list_insert(c, &head);
+	while (head) {
+		struct commit_list *p;
+		struct commit *c = head->item;
+		uint32_t **refs = ref_bitmap_at(&info->ref_bitmap, c);
+
+		p = head;
+		head = head->next;
+		free(p);
+
+		/* XXX check "UNINTERESTING" from pack bitmaps if available */
+		if (c->object.flags & (SEEN | UNINTERESTING))
+			continue;
+		else
+			c->object.flags |= SEEN;
+
+		if (*refs == NULL)
+			*refs = bitmap;
+		else {
+			memcpy(tmp, *refs, bitmap_size);
+			for (i = 0; i < bitmap_nr; i++)
+				tmp[i] |= bitmap[i];
+			if (memcmp(tmp, *refs, bitmap_size)) {
+				*refs = paint_alloc(info);
+				memcpy(*refs, tmp, bitmap_size);
+			}
+		}
+
+		if (c->object.flags & BOTTOM)
+			continue;
+
+		if (parse_commit(c))
+			die("unable to parse commit %s",
+			    sha1_to_hex(c->object.sha1));
+
+		for (p = c->parents; p; p = p->next) {
+			uint32_t **p_refs = ref_bitmap_at(&info->ref_bitmap,
+							  p->item);
+			if (p->item->object.flags & SEEN)
+				continue;
+			if (*p_refs == NULL || *p_refs == *refs)
+				*p_refs = *refs;
+			commit_list_insert(p->item, &head);
+		}
+	}
+
+	nr = get_max_object_index();
+	for (i = 0; i < nr; i++) {
+		struct object *o = get_indexed_object(i);
+		if (o && o->type == OBJ_COMMIT) {
+			o->flags &= ~SEEN;
+		}
+	}
+
+	free(tmp);
+}
+
+static int mark_uninteresting(const char *refname,
+			      const unsigned char *sha1,
+			      int flags, void *cb_data)
+{
+	struct commit *commit = lookup_commit_reference_gently(sha1, 1);
+	if (!commit)
+		return 0;
+	commit->object.flags |= UNINTERESTING;
+	mark_parents_uninteresting(commit);
+	return 0;
+}
+
+static void post_assign_shallow(struct shallow_info *info,
+				struct ref_bitmap *ref_bitmap,
+				int *ref_status);
+/*
+ * Step 6(+7), associate shallow commits with new refs
+ *
+ * info->ref must be initialized before calling this function.
+ *
+ * If used is not NULL, it's an array of info->shallow->nr
+ * bitmaps. The n-th bit set in the m-th bitmap if ref[n] needs the
+ * m-th shallow commit from info->shallow.
+ *
+ * If used is NULL, "ours" and "theirs" are updated. And if ref_status
+ * is not NULL it's an array of ref->nr ints. ref_status[i] is true if
+ * the ref needs some shallow commits from either info->ours or
+ * info->theirs.
+ */
+void assign_shallow_commits_to_refs(struct shallow_info *info,
+				    uint32_t **used, int *ref_status)
+{
+	unsigned char (*sha1)[20] = info->shallow->sha1;
+	struct sha1_array *ref = info->ref;
+	unsigned int i, nr;
+	int *shallow, nr_shallow = 0;
+	struct paint_info pi;
+
+	trace_printf_key(TRACE_KEY, "shallow: assign_shallow_commits_to_refs\n");
+	shallow = xmalloc(sizeof(*shallow) * (info->nr_ours + info->nr_theirs));
+	for (i = 0; i < info->nr_ours; i++)
+		shallow[nr_shallow++] = info->ours[i];
+	for (i = 0; i < info->nr_theirs; i++)
+		shallow[nr_shallow++] = info->theirs[i];
+
+	/*
+	 * Prepare the commit graph to track what refs can reach what
+	 * (new) shallow commits.
+	 */
+	nr = get_max_object_index();
+	for (i = 0; i < nr; i++) {
+		struct object *o = get_indexed_object(i);
+		if (!o || o->type != OBJ_COMMIT)
+			continue;
+
+		o->flags &= ~(UNINTERESTING | BOTTOM | SEEN);
+	}
+
+	memset(&pi, 0, sizeof(pi));
+	init_ref_bitmap(&pi.ref_bitmap);
+	pi.nr_bits = ref->nr;
+
+	/*
+	 * "--not --all" to cut short the traversal if new refs
+	 * connect to old refs. If not (e.g. force ref updates) it'll
+	 * have to go down to the current shallow commits.
+	 */
+	head_ref(mark_uninteresting, NULL);
+	for_each_ref(mark_uninteresting, NULL);
+
+	/* Mark potential bottoms so we won't go out of bound */
+	for (i = 0; i < nr_shallow; i++) {
+		struct commit *c = lookup_commit(sha1[shallow[i]]);
+		c->object.flags |= BOTTOM;
+	}
+
+	for (i = 0; i < ref->nr; i++)
+		paint_down(&pi, ref->sha1[i], i);
+
+	if (used) {
+		int bitmap_size = ((pi.nr_bits + 31) / 32) * sizeof(uint32_t);
+		memset(used, 0, sizeof(*used) * info->shallow->nr);
+		for (i = 0; i < nr_shallow; i++) {
+			const struct commit *c = lookup_commit(sha1[shallow[i]]);
+			uint32_t **map = ref_bitmap_at(&pi.ref_bitmap, c);
+			if (*map)
+				used[shallow[i]] = xmemdupz(*map, bitmap_size);
+		}
+		/*
+		 * unreachable shallow commits are not removed from
+		 * "ours" and "theirs". The user is supposed to run
+		 * step 7 on every ref separately and not trust "ours"
+		 * and "theirs" any more.
+		 */
+	} else
+		post_assign_shallow(info, &pi.ref_bitmap, ref_status);
+
+	clear_ref_bitmap(&pi.ref_bitmap);
+	for (i = 0; i < pi.slab_count; i++)
+		free(pi.slab[i]);
+	free(pi.slab);
+	free(shallow);
+}
+
+struct commit_array {
+	struct commit **commits;
+	int nr, alloc;
+};
+
+static int add_ref(const char *refname,
+		   const unsigned char *sha1, int flags, void *cb_data)
+{
+	struct commit_array *ca = cb_data;
+	ALLOC_GROW(ca->commits, ca->nr + 1, ca->alloc);
+	ca->commits[ca->nr] = lookup_commit_reference_gently(sha1, 1);
+	if (ca->commits[ca->nr])
+		ca->nr++;
+	return 0;
+}
+
+static void update_refstatus(int *ref_status, int nr, uint32_t *bitmap)
+{
+	int i;
+	if (!ref_status)
+		return;
+	for (i = 0; i < nr; i++)
+		if (bitmap[i / 32] & (1 << (i % 32)))
+			ref_status[i]++;
+}
+
+/*
+ * Step 7, reachability test on "ours" at commit level
+ */
+static void post_assign_shallow(struct shallow_info *info,
+				struct ref_bitmap *ref_bitmap,
+				int *ref_status)
+{
+	unsigned char (*sha1)[20] = info->shallow->sha1;
+	struct commit *c;
+	uint32_t **bitmap;
+	int dst, i, j;
+	int bitmap_nr = (info->ref->nr + 31) / 32;
+	struct commit_array ca;
+
+	trace_printf_key(TRACE_KEY, "shallow: post_assign_shallow\n");
+	if (ref_status)
+		memset(ref_status, 0, sizeof(*ref_status) * info->ref->nr);
+
+	/* Remove unreachable shallow commits from "theirs" */
+	for (i = dst = 0; i < info->nr_theirs; i++) {
+		if (i != dst)
+			info->theirs[dst] = info->theirs[i];
+		c = lookup_commit(sha1[info->theirs[i]]);
+		bitmap = ref_bitmap_at(ref_bitmap, c);
+		if (!*bitmap)
+			continue;
+		for (j = 0; j < bitmap_nr; j++)
+			if (bitmap[0][j]) {
+				update_refstatus(ref_status, info->ref->nr, *bitmap);
+				dst++;
+				break;
+			}
+	}
+	info->nr_theirs = dst;
+
+	memset(&ca, 0, sizeof(ca));
+	head_ref(add_ref, &ca);
+	for_each_ref(add_ref, &ca);
+
+	/* Remove unreachable shallow commits from "ours" */
+	for (i = dst = 0; i < info->nr_ours; i++) {
+		if (i != dst)
+			info->ours[dst] = info->ours[i];
+		c = lookup_commit(sha1[info->ours[i]]);
+		bitmap = ref_bitmap_at(ref_bitmap, c);
+		if (!*bitmap)
+			continue;
+		for (j = 0; j < bitmap_nr; j++)
+			if (bitmap[0][j] &&
+			    /* Step 7, reachability test at commit level */
+			    !in_merge_bases_many(c, ca.nr, ca.commits)) {
+				update_refstatus(ref_status, info->ref->nr, *bitmap);
+				dst++;
+				break;
+			}
+	}
+	info->nr_ours = dst;
+
+	free(ca.commits);
+}
-- 
1.8.5.1.25.g8667982

[PATCH v4 10/28] fetch-pack.c: move shallow update code out of fetch_pack()

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 fetch-pack.c | 22 +++++++++++++---------
 1 file changed, 13 insertions(+), 9 deletions(-)

diff --git a/fetch-pack.c b/fetch-pack.c
index 0e7483e..35d097e 100644
--- a/fetch-pack.c
+++ b/fetch-pack.c
@@ -925,6 +925,18 @@ static int remove_duplicates_in_refs(struct ref **ref, int nr)
 	return dst;
 }
 
+static void update_shallow(struct fetch_pack_args *args)
+{
+	if (args->depth > 0 && alternate_shallow_file) {
+		if (*alternate_shallow_file == '\0') { /* --unshallow */
+			unlink_or_warn(git_path("shallow"));
+			rollback_lock_file(&shallow_lock);
+		} else
+			commit_lock_file(&shallow_lock);
+		return;
+	}
+}
+
 struct ref *fetch_pack(struct fetch_pack_args *args,
 		       int fd[], struct child_process *conn,
 		       const struct ref *ref,
@@ -943,15 +955,7 @@ struct ref *fetch_pack(struct fetch_pack_args *args,
 		die("no matching remote head");
 	}
 	ref_cpy = do_fetch_pack(args, fd, ref, sought, nr_sought, pack_lockfile);
-
-	if (args->depth > 0 && alternate_shallow_file) {
-		if (*alternate_shallow_file == '\0') { /* --unshallow */
-			unlink_or_warn(git_path("shallow"));
-			rollback_lock_file(&shallow_lock);
-		} else
-			commit_lock_file(&shallow_lock);
-	}
-
+	update_shallow(args);
 	reprepare_packed_git();
 	return ref_cpy;
 }
-- 
1.8.5.1.25.g8667982

[PATCH v4 11/28] fetch-pack.h: one statement per bitfield declaration

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 fetch-pack.h | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

diff --git a/fetch-pack.h b/fetch-pack.h
index 461cbf3..9b08388 100644
--- a/fetch-pack.h
+++ b/fetch-pack.h
@@ -8,18 +8,18 @@ struct fetch_pack_args {
 	const char *uploadpack;
 	int unpacklimit;
 	int depth;
-	unsigned quiet:1,
-		keep_pack:1,
-		lock_pack:1,
-		use_thin_pack:1,
-		fetch_all:1,
-		stdin_refs:1,
-		verbose:1,
-		no_progress:1,
-		include_tag:1,
-		stateless_rpc:1,
-		check_self_contained_and_connected:1,
-		self_contained_and_connected:1;
+	unsigned quiet:1;
+	unsigned keep_pack:1;
+	unsigned lock_pack:1;
+	unsigned use_thin_pack:1;
+	unsigned fetch_all:1;
+	unsigned stdin_refs:1;
+	unsigned verbose:1;
+	unsigned no_progress:1;
+	unsigned include_tag:1;
+	unsigned stateless_rpc:1;
+	unsigned check_self_contained_and_connected:1;
+	unsigned self_contained_and_connected:1;
 };
 
 /*
-- 
1.8.5.1.25.g8667982

[PATCH v4 12/28] clone: support remote shallow repository

[Nguyễn Thái Ngọc Duy]

Cloning from a shallow repository does not follow the "8 steps for new
.git/shallow" because if it does we need to get through step 6 for all
refs. That means commit walking down to the bottom.

Instead the rule to create .git/shallow is simpler and, more
importantly, cheap: if a shallow commit is found in the pack, it's
probably used (i.e. reachable from some refs), so we add it. Others
are dropped.

One may notice this method seems flawed by the word "probably". A
shallow commit may not be reachable from any refs at all if it's
attached to an object island (a group of objects that are not
reachable by any refs).

If that object island is not complete, a new fetch request may send
more objects to connect it to some ref. At that time, because we
incorrectly installed the shallow commit in this island, the user will
not see anything after that commit (fsck is still ok). This is not
desired.

Given that object islands are rare (C Git never sends such islands for
security reasons) and do not really harm the repository integrity, a
tradeoff is made to surprise the user occasionally but work faster
everyday.

A new option --strict could be added later that follows exactly the 8
steps. "git prune" can also learn to remove dangling objects _and_ the
shallow commits that are attached to them from .git/shallow.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/clone.c      |  1 +
 builtin/fetch-pack.c |  2 +-
 fetch-pack.c         | 54 +++++++++++++++++++++++++++++++++++++++++++++++++---
 fetch-pack.h         |  4 ++++
 transport.c          | 11 ++++++++---
 transport.h          |  6 ++++++
 6 files changed, 71 insertions(+), 7 deletions(-)

diff --git a/builtin/clone.c b/builtin/clone.c
index 900f564..0b182ce 100644
--- a/builtin/clone.c
+++ b/builtin/clone.c
@@ -889,6 +889,7 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
 
 	remote = remote_get(option_origin);
 	transport = transport_get(remote, remote->url[0]);
+	transport->cloning = 1;
 
 	if (!transport->get_refs_list || (!is_local && !transport->fetch))
 		die(_("Don't know how to clone %s"), transport->url);
diff --git a/builtin/fetch-pack.c b/builtin/fetch-pack.c
index c1d918f..927424b 100644
--- a/builtin/fetch-pack.c
+++ b/builtin/fetch-pack.c
@@ -153,7 +153,7 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
 	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL, NULL);
 
 	ref = fetch_pack(&args, fd, conn, ref, dest,
-			 sought, nr_sought, pack_lockfile_ptr);
+			 sought, nr_sought, NULL, pack_lockfile_ptr);
 	if (pack_lockfile) {
 		printf("lock %s\n", pack_lockfile);
 		fflush(stdout);
diff --git a/fetch-pack.c b/fetch-pack.c
index 35d097e..e18a9fd 100644
--- a/fetch-pack.c
+++ b/fetch-pack.c
@@ -13,6 +13,7 @@
 #include "transport.h"
 #include "version.h"
 #include "prio-queue.h"
+#include "sha1-array.h"
 
 static int transfer_unpack_limit = -1;
 static int fetch_unpack_limit = -1;
@@ -774,6 +775,7 @@ static struct ref *do_fetch_pack(struct fetch_pack_args *args,
 				 int fd[2],
 				 const struct ref *orig_ref,
 				 struct ref **sought, int nr_sought,
+				 struct shallow_info *si,
 				 char **pack_lockfile)
 {
 	struct ref *ref = copy_ref_list(orig_ref);
@@ -852,6 +854,8 @@ static struct ref *do_fetch_pack(struct fetch_pack_args *args,
 	if (args->depth > 0)
 		setup_alternate_shallow(&shallow_lock, &alternate_shallow_file,
 					NULL);
+	else if (args->cloning && si->shallow && si->shallow->nr)
+		alternate_shallow_file = setup_temporary_shallow(si->shallow);
 	else
 		alternate_shallow_file = NULL;
 	if (get_pack(args, fd, pack_lockfile))
@@ -925,8 +929,11 @@ static int remove_duplicates_in_refs(struct ref **ref, int nr)
 	return dst;
 }
 
-static void update_shallow(struct fetch_pack_args *args)
+static void update_shallow(struct fetch_pack_args *args,
+			   struct shallow_info *si)
 {
+	int i;
+
 	if (args->depth > 0 && alternate_shallow_file) {
 		if (*alternate_shallow_file == '\0') { /* --unshallow */
 			unlink_or_warn(git_path("shallow"));
@@ -935,6 +942,42 @@ static void update_shallow(struct fetch_pack_args *args)
 			commit_lock_file(&shallow_lock);
 		return;
 	}
+
+	if (!si->shallow || !si->shallow->nr)
+		return;
+
+	if (alternate_shallow_file) {
+		/*
+		 * The temporary shallow file is only useful for
+		 * index-pack and unpack-objects because it may
+		 * contain more roots than we want. Delete it.
+		 */
+		if (*alternate_shallow_file)
+			unlink(alternate_shallow_file);
+		free((char*)alternate_shallow_file);
+	}
+
+	if (args->cloning) {
+		/*
+		 * remote is shallow, but this is a clone, there are
+		 * no objects in repo to worry about. Accept any
+		 * shallow points that exist in the pack (iow in repo
+		 * after get_pack() and reprepare_packed_git())
+		 */
+		struct sha1_array extra = SHA1_ARRAY_INIT;
+		unsigned char (*sha1)[20] = si->shallow->sha1;
+		for (i = 0; i < si->shallow->nr; i++)
+			if (has_sha1_file(sha1[i]))
+				sha1_array_append(&extra, sha1[i]);
+		if (extra.nr) {
+			setup_alternate_shallow(&shallow_lock,
+						&alternate_shallow_file,
+						&extra);
+			commit_lock_file(&shallow_lock);
+		}
+		sha1_array_clear(&extra);
+		return;
+	}
 }
 
 struct ref *fetch_pack(struct fetch_pack_args *args,
@@ -942,9 +985,11 @@ struct ref *fetch_pack(struct fetch_pack_args *args,
 		       const struct ref *ref,
 		       const char *dest,
 		       struct ref **sought, int nr_sought,
+		       struct sha1_array *shallow,
 		       char **pack_lockfile)
 {
 	struct ref *ref_cpy;
+	struct shallow_info si;
 
 	fetch_pack_setup();
 	if (nr_sought)
@@ -954,8 +999,11 @@ struct ref *fetch_pack(struct fetch_pack_args *args,
 		packet_flush(fd[1]);
 		die("no matching remote head");
 	}
-	ref_cpy = do_fetch_pack(args, fd, ref, sought, nr_sought, pack_lockfile);
-	update_shallow(args);
+	prepare_shallow_info(&si, shallow);
+	ref_cpy = do_fetch_pack(args, fd, ref, sought, nr_sought,
+				&si, pack_lockfile);
 	reprepare_packed_git();
+	update_shallow(args, &si);
+	clear_shallow_info(&si);
 	return ref_cpy;
 }
diff --git a/fetch-pack.h b/fetch-pack.h
index 9b08388..ce59537 100644
--- a/fetch-pack.h
+++ b/fetch-pack.h
@@ -4,6 +4,8 @@
 #include "string-list.h"
 #include "run-command.h"
 
+struct sha1_array;
+
 struct fetch_pack_args {
 	const char *uploadpack;
 	int unpacklimit;
@@ -20,6 +22,7 @@ struct fetch_pack_args {
 	unsigned stateless_rpc:1;
 	unsigned check_self_contained_and_connected:1;
 	unsigned self_contained_and_connected:1;
+	unsigned cloning:1;
 };
 
 /*
@@ -33,6 +36,7 @@ struct ref *fetch_pack(struct fetch_pack_args *args,
 		       const char *dest,
 		       struct ref **sought,
 		       int nr_sought,
+		       struct sha1_array *shallow,
 		       char **pack_lockfile);
 
 #endif
diff --git a/transport.c b/transport.c
index 90453df..91c4667 100644
--- a/transport.c
+++ b/transport.c
@@ -456,6 +456,7 @@ struct git_transport_data {
 	int fd[2];
 	unsigned got_remote_heads : 1;
 	struct sha1_array extra_have;
+	struct sha1_array shallow;
 };
 
 static int set_git_option(struct git_transport_options *opts,
@@ -512,7 +513,9 @@ static struct ref *get_refs_via_connect(struct transport *transport, int for_pus
 
 	connect_setup(transport, for_push, 0);
 	get_remote_heads(data->fd[0], NULL, 0, &refs,
-			 for_push ? REF_NORMAL : 0, &data->extra_have, NULL);
+			 for_push ? REF_NORMAL : 0,
+			 &data->extra_have,
+			 transport->cloning ? &data->shallow : NULL);
 	data->got_remote_heads = 1;
 
 	return refs;
@@ -539,17 +542,19 @@ static int fetch_refs_via_pack(struct transport *transport,
 	args.depth = data->options.depth;
 	args.check_self_contained_and_connected =
 		data->options.check_self_contained_and_connected;
+	args.cloning = transport->cloning;
 
 	if (!data->got_remote_heads) {
 		connect_setup(transport, 0, 0);
 		get_remote_heads(data->fd[0], NULL, 0, &refs_tmp, 0,
-				 NULL, NULL);
+				 NULL,
+				 transport->cloning ? &data->shallow : NULL);
 		data->got_remote_heads = 1;
 	}
 
 	refs = fetch_pack(&args, data->fd, data->conn,
 			  refs_tmp ? refs_tmp : transport->remote_refs,
-			  dest, to_fetch, nr_heads,
+			  dest, to_fetch, nr_heads, &data->shallow,
 			  &transport->pack_lockfile);
 	close(data->fd[0]);
 	close(data->fd[1]);
diff --git a/transport.h b/transport.h
index b3679bb..59842d4 100644
--- a/transport.h
+++ b/transport.h
@@ -35,6 +35,12 @@ struct transport {
 	 */
 	unsigned cannot_reuse : 1;
 
+	/*
+	 * A hint from caller that it will be performing a clone, not
+	 * normal fetch. IOW the repository is guaranteed empty.
+	 */
+	unsigned cloning : 1;
+
 	/**
 	 * Returns 0 if successful, positive if the option is not
 	 * recognized or is inapplicable, and negative if the option
-- 
1.8.5.1.25.g8667982

[PATCH v4 13/28] fetch: support fetching from a shallow repository

[Nguyễn Thái Ngọc Duy]

This patch just put together pieces from the 8 steps patch. We stop at
step 7 and reject refs that require new shallow commits.

Note that, by rejecting refs that require new shallow commits, we
leave dangling objects in the repo, which become "object islands" by
the next "git fetch" of the same source.

If the first fetch our "ours" set is zero and we do practically
nothing at step 7, "ours" is full at the next fetch and we may need to
walk through commits for reachability test. Room for improvement.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/fetch.c                   |   9 +++
 fetch-pack.c                      |  32 +++++++++-
 remote.h                          |   1 +
 t/t5536-fetch-shallow.sh (new +x) | 128 ++++++++++++++++++++++++++++++++++++++
 transport.c                       |  11 +++-
 5 files changed, 176 insertions(+), 5 deletions(-)
 create mode 100755 t/t5536-fetch-shallow.sh

diff --git a/builtin/fetch.c b/builtin/fetch.c
index bd7a101..7b41a7e 100644
--- a/builtin/fetch.c
+++ b/builtin/fetch.c
@@ -405,6 +405,8 @@ static int iterate_ref_map(void *cb_data, unsigned char sha1[20])
 	struct ref **rm = cb_data;
 	struct ref *ref = *rm;
 
+	while (ref && ref->status == REF_STATUS_REJECT_SHALLOW)
+		ref = ref->next;
 	if (!ref)
 		return -1; /* end of the list */
 	*rm = ref->next;
@@ -451,6 +453,13 @@ static int store_updated_refs(const char *raw_url, const char *remote_name,
 			struct ref *ref = NULL;
 			const char *merge_status_marker = "";
 
+			if (rm->status == REF_STATUS_REJECT_SHALLOW) {
+				if (want_status == FETCH_HEAD_MERGE)
+					warning(_("reject %s because shallow roots are not allowed to be updated"),
+						rm->peer_ref ? rm->peer_ref->name : rm->name);
+				continue;
+			}
+
 			commit = lookup_commit_reference_gently(rm->old_sha1, 1);
 			if (!commit)
 				rm->fetch_head_status = FETCH_HEAD_NOT_FOR_MERGE;
diff --git a/fetch-pack.c b/fetch-pack.c
index e18a9fd..7c11aff 100644
--- a/fetch-pack.c
+++ b/fetch-pack.c
@@ -854,7 +854,7 @@ static struct ref *do_fetch_pack(struct fetch_pack_args *args,
 	if (args->depth > 0)
 		setup_alternate_shallow(&shallow_lock, &alternate_shallow_file,
 					NULL);
-	else if (args->cloning && si->shallow && si->shallow->nr)
+	else if (si->nr_ours || si->nr_theirs)
 		alternate_shallow_file = setup_temporary_shallow(si->shallow);
 	else
 		alternate_shallow_file = NULL;
@@ -930,8 +930,11 @@ static int remove_duplicates_in_refs(struct ref **ref, int nr)
 }
 
 static void update_shallow(struct fetch_pack_args *args,
+			   struct ref **sought, int nr_sought,
 			   struct shallow_info *si)
 {
+	struct sha1_array ref = SHA1_ARRAY_INIT;
+	int *status;
 	int i;
 
 	if (args->depth > 0 && alternate_shallow_file) {
@@ -978,6 +981,31 @@ static void update_shallow(struct fetch_pack_args *args,
 		sha1_array_clear(&extra);
 		return;
 	}
+
+	if (!si->nr_ours && !si->nr_theirs)
+		return;
+
+	remove_nonexistent_theirs_shallow(si);
+	/* XXX remove_nonexistent_ours_in_pack() */
+	if (!si->nr_ours && !si->nr_theirs)
+		return;
+	for (i = 0; i < nr_sought; i++)
+		sha1_array_append(&ref, sought[i]->old_sha1);
+	si->ref = &ref;
+
+	/*
+	 * remote is also shallow, check what ref is safe to update
+	 * without updating .git/shallow
+	 */
+	status = xcalloc(nr_sought, sizeof(*status));
+	assign_shallow_commits_to_refs(si, NULL, status);
+	if (si->nr_ours || si->nr_theirs) {
+		for (i = 0; i < nr_sought; i++)
+			if (status[i])
+				sought[i]->status = REF_STATUS_REJECT_SHALLOW;
+	}
+	free(status);
+	sha1_array_clear(&ref);
 }
 
 struct ref *fetch_pack(struct fetch_pack_args *args,
@@ -1003,7 +1031,7 @@ struct ref *fetch_pack(struct fetch_pack_args *args,
 	ref_cpy = do_fetch_pack(args, fd, ref, sought, nr_sought,
 				&si, pack_lockfile);
 	reprepare_packed_git();
-	update_shallow(args, &si);
+	update_shallow(args, sought, nr_sought, &si);
 	clear_shallow_info(&si);
 	return ref_cpy;
 }
diff --git a/remote.h b/remote.h
index 5d217d5..3498091 100644
--- a/remote.h
+++ b/remote.h
@@ -109,6 +109,7 @@ struct ref {
 		REF_STATUS_REJECT_FETCH_FIRST,
 		REF_STATUS_REJECT_NEEDS_FORCE,
 		REF_STATUS_REJECT_STALE,
+		REF_STATUS_REJECT_SHALLOW,
 		REF_STATUS_UPTODATE,
 		REF_STATUS_REMOTE_REJECT,
 		REF_STATUS_EXPECTING_REPORT
diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
new file mode 100755
index 0000000..d211052
--- /dev/null
+++ b/t/t5536-fetch-shallow.sh
@@ -0,0 +1,128 @@
+#!/bin/sh
+
+test_description='fetch/clone from a shallow clone'
+
+. ./test-lib.sh
+
+commit() {
+	echo "$1" >tracked &&
+	git add tracked &&
+	git commit -m "$1"
+}
+
+test_expect_success 'setup' '
+	commit 1 &&
+	commit 2 &&
+	commit 3 &&
+	commit 4 &&
+	git config --global transfer.fsckObjects true
+'
+
+test_expect_success 'setup shallow clone' '
+	git clone --no-local --depth=2 .git shallow &&
+	git --git-dir=shallow/.git log --format=%s >actual &&
+	cat <<EOF >expect &&
+4
+3
+EOF
+	test_cmp expect actual
+'
+
+test_expect_success 'clone from shallow clone' '
+	git clone --no-local shallow shallow2 &&
+	(
+	cd shallow2 &&
+	git fsck &&
+	git log --format=%s >actual &&
+	cat <<EOF >expect &&
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
+test_expect_success 'fetch from shallow clone' '
+	(
+	cd shallow &&
+	commit 5
+	) &&
+	(
+	cd shallow2 &&
+	git fetch &&
+	git fsck &&
+	git log --format=%s origin/master >actual &&
+	cat <<EOF >expect &&
+5
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
+test_expect_success 'fetch --depth from shallow clone' '
+	(
+	cd shallow &&
+	commit 6
+	) &&
+	(
+	cd shallow2 &&
+	git fetch --depth=2 &&
+	git fsck &&
+	git log --format=%s origin/master >actual &&
+	cat <<EOF >expect &&
+6
+5
+EOF
+	test_cmp expect actual
+	)
+'
+
+test_expect_success 'fetch something upstream has but hidden by clients shallow boundaries' '
+	# the blob "1" is available in .git but hidden by the
+	# shallow2/.git/shallow and it should be resent
+	! git --git-dir=shallow2/.git cat-file blob `echo 1|git hash-object --stdin` >/dev/null &&
+	echo 1 >1.t &&
+	git add 1.t &&
+	git commit -m add-1-back &&
+	(
+	cd shallow2 &&
+	git fetch ../.git +refs/heads/master:refs/remotes/top/master &&
+	git fsck &&
+	git log --format=%s top/master >actual &&
+	cat <<EOF >expect &&
+add-1-back
+4
+3
+EOF
+	test_cmp expect actual
+	) &&
+	git --git-dir=shallow2/.git cat-file blob `echo 1|git hash-object --stdin` >/dev/null
+
+'
+
+test_expect_success 'fetch that requires changes in .git/shallow is filtered' '
+	(
+	cd shallow &&
+	git checkout --orphan no-shallow &&
+	commit no-shallow
+	) &&
+	git init notshallow &&
+	(
+	cd notshallow &&
+	git fetch ../shallow/.git refs/heads/*:refs/remotes/shallow/*&&
+	git for-each-ref --format="%(refname)" >actual.refs &&
+	cat <<EOF >expect.refs &&
+refs/remotes/shallow/no-shallow
+EOF
+	test_cmp expect.refs actual.refs &&
+	git log --format=%s shallow/no-shallow >actual &&
+	cat <<EOF >expect &&
+no-shallow
+EOF
+	test_cmp expect actual
+	)
+'
+
+test_done
diff --git a/transport.c b/transport.c
index 91c4667..491360b 100644
--- a/transport.c
+++ b/transport.c
@@ -515,7 +515,7 @@ static struct ref *get_refs_via_connect(struct transport *transport, int for_pus
 	get_remote_heads(data->fd[0], NULL, 0, &refs,
 			 for_push ? REF_NORMAL : 0,
 			 &data->extra_have,
-			 transport->cloning ? &data->shallow : NULL);
+			 &data->shallow);
 	data->got_remote_heads = 1;
 
 	return refs;
@@ -547,8 +547,7 @@ static int fetch_refs_via_pack(struct transport *transport,
 	if (!data->got_remote_heads) {
 		connect_setup(transport, 0, 0);
 		get_remote_heads(data->fd[0], NULL, 0, &refs_tmp, 0,
-				 NULL,
-				 transport->cloning ? &data->shallow : NULL);
+				 NULL, &data->shallow);
 		data->got_remote_heads = 1;
 	}
 
@@ -720,6 +719,10 @@ static int print_one_push_status(struct ref *ref, const char *dest, int count, i
 		print_ref_status('!', "[rejected]", ref, ref->peer_ref,
 						 "stale info", porcelain);
 		break;
+	case REF_STATUS_REJECT_SHALLOW:
+		print_ref_status('!', "[rejected]", ref, ref->peer_ref,
+						 "new shallow roots not allowed", porcelain);
+		break;
 	case REF_STATUS_REMOTE_REJECT:
 		print_ref_status('!', "[remote rejected]", ref,
 						 ref->deletion ? NULL : ref->peer_ref,
@@ -815,6 +818,8 @@ static int git_transport_push(struct transport *transport, struct ref *remote_re
 		get_remote_heads(data->fd[0], NULL, 0, &tmp_refs, REF_NORMAL, NULL, NULL);
 		data->got_remote_heads = 1;
 	}
+	if (data->shallow.nr)
+		die("pushing to a shallow repository is not supported");
 
 	memset(&args, 0, sizeof(args));
 	args.send_mirror = !!(flags & TRANSPORT_PUSH_MIRROR);
-- 
1.8.5.1.25.g8667982

[PATCH v4 14/28] upload-pack: make sure deepening preserves shallow roots

[Nguyễn Thái Ngọc Duy]

When "fetch --depth=N" where N exceeds the longest chain of history in
the source repo, usually we just send an "unshallow" line to the
client so full history is obtained.

When the source repo is shallow we need to make sure to "unshallow"
the current shallow point _and_ "shallow" again when the commit
reaches its shallow bottom in the source repo.

This should fix both cases: large <N> and --unshallow.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/fetch-options.txt |  8 ++++++--
 shallow.c                       |  6 +++++-
 t/t5536-fetch-shallow.sh        | 16 ++++++++++++++++
 upload-pack.c                   |  2 +-
 4 files changed, 28 insertions(+), 4 deletions(-)

diff --git a/Documentation/fetch-options.txt b/Documentation/fetch-options.txt
index ba1fe49..a83d2b4 100644
--- a/Documentation/fetch-options.txt
+++ b/Documentation/fetch-options.txt
@@ -14,8 +14,12 @@
 	branch history. Tags for the deepened commits are not fetched.
 
 --unshallow::
-	Convert a shallow repository to a complete one, removing all
-	the limitations imposed by shallow repositories.
+	If the source repository is complete, convert a shallow
+	repository to a complete one, removing all the limitations
+	imposed by shallow repositories.
++
+If the source repository is shallow, fetch as much as possible so that
+the current repository has the same history as the source repository.
 
 ifndef::git-pull[]
 --dry-run::
diff --git a/shallow.c b/shallow.c
index 975d699..ae46b10 100644
--- a/shallow.c
+++ b/shallow.c
@@ -75,6 +75,7 @@ struct commit_list *get_shallow_commits(struct object_array *heads, int depth,
 	struct commit_list *result = NULL;
 	struct object_array stack = OBJECT_ARRAY_INIT;
 	struct commit *commit = NULL;
+	struct commit_graft *graft;
 
 	while (commit || i < heads->nr || stack.nr) {
 		struct commit_list *p;
@@ -99,7 +100,10 @@ struct commit_list *get_shallow_commits(struct object_array *heads, int depth,
 		if (parse_commit(commit))
 			die("invalid commit");
 		cur_depth++;
-		if (cur_depth >= depth) {
+		if ((depth != INFINITE_DEPTH && cur_depth >= depth) ||
+		    (is_repository_shallow() && !commit->parents &&
+		     (graft = lookup_commit_graft(commit->object.sha1)) != NULL &&
+		     graft->nr_parent < 0)) {
 			commit_list_insert(commit, &result);
 			commit->object.flags |= shallow_flag;
 			commit = NULL;
diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
index d211052..022cb2c 100755
--- a/t/t5536-fetch-shallow.sh
+++ b/t/t5536-fetch-shallow.sh
@@ -79,6 +79,22 @@ EOF
 	)
 '
 
+test_expect_success 'fetch --unshallow from shallow clone' '
+	(
+	cd shallow2 &&
+	git fetch --unshallow &&
+	git fsck &&
+	git log --format=%s origin/master >actual &&
+	cat <<EOF >expect &&
+6
+5
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
 test_expect_success 'fetch something upstream has but hidden by clients shallow boundaries' '
 	# the blob "1" is available in .git but hidden by the
 	# shallow2/.git/shallow and it should be resent
diff --git a/upload-pack.c b/upload-pack.c
index f082f06..28269c7 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -619,7 +619,7 @@ static void receive_needs(void)
 	if (depth > 0) {
 		struct commit_list *result = NULL, *backup = NULL;
 		int i;
-		if (depth == INFINITE_DEPTH)
+		if (depth == INFINITE_DEPTH && !is_repository_shallow())
 			for (i = 0; i < shallows.nr; i++) {
 				struct object *object = shallows.objects[i].item;
 				object->flags |= NOT_SHALLOW;
-- 
1.8.5.1.25.g8667982

[PATCH v4 15/28] fetch: add --update-shallow to accept refs that update .git/shallow

[Nguyễn Thái Ngọc Duy]

The same steps are done as in when --update-shallow is not given. The
only difference is we now add all shallow commits in "ours" and
"theirs" to .git/shallow (aka "step 8").

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/fetch-options.txt |  6 ++++++
 builtin/fetch.c                 |  6 +++++-
 fetch-pack.c                    | 27 +++++++++++++++++++++++++++
 fetch-pack.h                    |  1 +
 t/t5536-fetch-shallow.sh        | 32 ++++++++++++++++++++++++++++++++
 transport.c                     |  4 ++++
 transport.h                     |  4 ++++
 7 files changed, 79 insertions(+), 1 deletion(-)

diff --git a/Documentation/fetch-options.txt b/Documentation/fetch-options.txt
index a83d2b4..54043e3 100644
--- a/Documentation/fetch-options.txt
+++ b/Documentation/fetch-options.txt
@@ -21,6 +21,12 @@
 If the source repository is shallow, fetch as much as possible so that
 the current repository has the same history as the source repository.
 
+--update-shallow::
+	By default when fetching from a shallow repository,
+	`git fetch` refuses refs that require updating
+	.git/shallow. This option updates .git/shallow and accept such
+	refs.
+
 ifndef::git-pull[]
 --dry-run::
 	Show what would be done, without making any changes.
diff --git a/builtin/fetch.c b/builtin/fetch.c
index 7b41a7e..d2e4fc0 100644
--- a/builtin/fetch.c
+++ b/builtin/fetch.c
@@ -36,7 +36,7 @@ static int prune = -1; /* unspecified */
 
 static int all, append, dry_run, force, keep, multiple, update_head_ok, verbosity;
 static int progress = -1, recurse_submodules = RECURSE_SUBMODULES_DEFAULT;
-static int tags = TAGS_DEFAULT, unshallow;
+static int tags = TAGS_DEFAULT, unshallow, update_shallow;
 static const char *depth;
 static const char *upload_pack;
 static struct strbuf default_rla = STRBUF_INIT;
@@ -104,6 +104,8 @@ static struct option builtin_fetch_options[] = {
 	{ OPTION_STRING, 0, "recurse-submodules-default",
 		   &recurse_submodules_default, NULL,
 		   N_("default mode for recursion"), PARSE_OPT_HIDDEN },
+	OPT_BOOL(0, "update-shallow", &update_shallow,
+		 N_("accept refs that update .git/shallow")),
 	OPT_END()
 };
 
@@ -768,6 +770,8 @@ static struct transport *prepare_transport(struct remote *remote)
 		set_option(transport, TRANS_OPT_KEEP, "yes");
 	if (depth)
 		set_option(transport, TRANS_OPT_DEPTH, depth);
+	if (update_shallow)
+		set_option(transport, TRANS_OPT_UPDATE_SHALLOW, "yes");
 	return transport;
 }
 
diff --git a/fetch-pack.c b/fetch-pack.c
index 7c11aff..cd219b8 100644
--- a/fetch-pack.c
+++ b/fetch-pack.c
@@ -993,6 +993,33 @@ static void update_shallow(struct fetch_pack_args *args,
 		sha1_array_append(&ref, sought[i]->old_sha1);
 	si->ref = &ref;
 
+	if (args->update_shallow) {
+		/*
+		 * remote is also shallow, .git/shallow may be updated
+		 * so all refs can be accepted. Make sure we only add
+		 * shallow roots that are actually reachable from new
+		 * refs.
+		 */
+		struct sha1_array extra = SHA1_ARRAY_INIT;
+		unsigned char (*sha1)[20] = si->shallow->sha1;
+		assign_shallow_commits_to_refs(si, NULL, NULL);
+		if (!si->nr_ours && !si->nr_theirs) {
+			sha1_array_clear(&ref);
+			return;
+		}
+		for (i = 0; i < si->nr_ours; i++)
+			sha1_array_append(&extra, sha1[si->ours[i]]);
+		for (i = 0; i < si->nr_theirs; i++)
+			sha1_array_append(&extra, sha1[si->theirs[i]]);
+		setup_alternate_shallow(&shallow_lock,
+					&alternate_shallow_file,
+					&extra);
+		commit_lock_file(&shallow_lock);
+		sha1_array_clear(&extra);
+		sha1_array_clear(&ref);
+		return;
+	}
+
 	/*
 	 * remote is also shallow, check what ref is safe to update
 	 * without updating .git/shallow
diff --git a/fetch-pack.h b/fetch-pack.h
index ce59537..ada02f5 100644
--- a/fetch-pack.h
+++ b/fetch-pack.h
@@ -23,6 +23,7 @@ struct fetch_pack_args {
 	unsigned check_self_contained_and_connected:1;
 	unsigned self_contained_and_connected:1;
 	unsigned cloning:1;
+	unsigned update_shallow:1;
 };
 
 /*
diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
index 022cb2c..3ae9092 100755
--- a/t/t5536-fetch-shallow.sh
+++ b/t/t5536-fetch-shallow.sh
@@ -141,4 +141,36 @@ EOF
 	)
 '
 
+test_expect_success 'fetch --update-shallow' '
+	(
+	cd shallow &&
+	git checkout master &&
+	commit 7 &&
+	git tag -m foo heavy-tag HEAD^ &&
+	git tag light-tag HEAD^:tracked
+	) &&
+	(
+	cd notshallow &&
+	git fetch --update-shallow ../shallow/.git refs/heads/*:refs/remotes/shallow/* &&
+	git fsck &&
+	git for-each-ref --sort=refname --format="%(refname)" >actual.refs &&
+	cat <<EOF >expect.refs &&
+refs/remotes/shallow/master
+refs/remotes/shallow/no-shallow
+refs/tags/heavy-tag
+refs/tags/light-tag
+EOF
+	test_cmp expect.refs actual.refs &&
+	git log --format=%s shallow/master >actual &&
+	cat <<EOF >expect &&
+7
+6
+5
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
 test_done
diff --git a/transport.c b/transport.c
index 491360b..a09fdb6 100644
--- a/transport.c
+++ b/transport.c
@@ -477,6 +477,9 @@ static int set_git_option(struct git_transport_options *opts,
 	} else if (!strcmp(name, TRANS_OPT_KEEP)) {
 		opts->keep = !!value;
 		return 0;
+	} else if (!strcmp(name, TRANS_OPT_UPDATE_SHALLOW)) {
+		opts->update_shallow = !!value;
+		return 0;
 	} else if (!strcmp(name, TRANS_OPT_DEPTH)) {
 		if (!value)
 			opts->depth = 0;
@@ -543,6 +546,7 @@ static int fetch_refs_via_pack(struct transport *transport,
 	args.check_self_contained_and_connected =
 		data->options.check_self_contained_and_connected;
 	args.cloning = transport->cloning;
+	args.update_shallow = data->options.update_shallow;
 
 	if (!data->got_remote_heads) {
 		connect_setup(transport, 0, 0);
diff --git a/transport.h b/transport.h
index 59842d4..02ea248 100644
--- a/transport.h
+++ b/transport.h
@@ -11,6 +11,7 @@ struct git_transport_options {
 	unsigned followtags : 1;
 	unsigned check_self_contained_and_connected : 1;
 	unsigned self_contained_and_connected : 1;
+	unsigned update_shallow : 1;
 	int depth;
 	const char *uploadpack;
 	const char *receivepack;
@@ -152,6 +153,9 @@ struct transport *transport_get(struct remote *, const char *);
 /* Aggressively fetch annotated tags if possible */
 #define TRANS_OPT_FOLLOWTAGS "followtags"
 
+/* Accept refs that may update .git/shallow without --depth */
+#define TRANS_OPT_UPDATE_SHALLOW "updateshallow"
+
 /**
  * Returns 0 if the option was used, non-zero otherwise. Prints a
  * message to stderr if the option is not used.
-- 
1.8.5.1.25.g8667982

[PATCH v4 16/28] receive-pack: reorder some code in unpack()

[Nguyễn Thái Ngọc Duy]

This is the preparation for adding --shallow-file to both
unpack-objects and index-pack. To sum up:

 - struct argv_array used instead of const char **

 - status/code, ip/child, unpacker/keeper are moved out to function
   top level

 - successful flow now ends at the end of the function

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/receive-pack.c | 70 ++++++++++++++++++++++----------------------------
 1 file changed, 30 insertions(+), 40 deletions(-)

diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index cc8c34f..8927ddf 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -13,6 +13,7 @@
 #include "string-list.h"
 #include "sha1-array.h"
 #include "connected.h"
+#include "argv-array.h"
 #include "version.h"
 
 static const char receive_pack_usage[] = "git receive-pack <git-dir>";
@@ -822,8 +823,11 @@ static const char *pack_lockfile;
 static const char *unpack(int err_fd)
 {
 	struct pack_header hdr;
+	struct argv_array av = ARGV_ARRAY_INIT;
 	const char *hdr_err;
+	int status;
 	char hdr_arg[38];
+	struct child_process child;
 	int fsck_objects = (receive_fsck_objects >= 0
 			    ? receive_fsck_objects
 			    : transfer_fsck_objects >= 0
@@ -840,63 +844,49 @@ static const char *unpack(int err_fd)
 			"--pack_header=%"PRIu32",%"PRIu32,
 			ntohl(hdr.hdr_version), ntohl(hdr.hdr_entries));
 
+	memset(&child, 0, sizeof(child));
 	if (ntohl(hdr.hdr_entries) < unpack_limit) {
-		int code, i = 0;
-		struct child_process child;
-		const char *unpacker[5];
-		unpacker[i++] = "unpack-objects";
+		argv_array_pushl(&av, "unpack-objects", hdr_arg, NULL);
 		if (quiet)
-			unpacker[i++] = "-q";
+			argv_array_push(&av, "-q");
 		if (fsck_objects)
-			unpacker[i++] = "--strict";
-		unpacker[i++] = hdr_arg;
-		unpacker[i++] = NULL;
-		memset(&child, 0, sizeof(child));
-		child.argv = unpacker;
+			argv_array_push(&av, "--strict");
+		child.argv = av.argv;
 		child.no_stdout = 1;
 		child.err = err_fd;
 		child.git_cmd = 1;
-		code = run_command(&child);
-		if (!code)
-			return NULL;
-		return "unpack-objects abnormal exit";
+		status = run_command(&child);
+		if (status)
+			return "unpack-objects abnormal exit";
 	} else {
-		const char *keeper[7];
-		int s, status, i = 0;
+		int s;
 		char keep_arg[256];
-		struct child_process ip;
 
 		s = sprintf(keep_arg, "--keep=receive-pack %"PRIuMAX" on ", (uintmax_t) getpid());
 		if (gethostname(keep_arg + s, sizeof(keep_arg) - s))
 			strcpy(keep_arg + s, "localhost");
 
-		keeper[i++] = "index-pack";
-		keeper[i++] = "--stdin";
+		argv_array_pushl(&av, "index-pack",
+				 "--stdin", hdr_arg, keep_arg, NULL);
 		if (fsck_objects)
-			keeper[i++] = "--strict";
+			argv_array_push(&av, "--strict");
 		if (fix_thin)
-			keeper[i++] = "--fix-thin";
-		keeper[i++] = hdr_arg;
-		keeper[i++] = keep_arg;
-		keeper[i++] = NULL;
-		memset(&ip, 0, sizeof(ip));
-		ip.argv = keeper;
-		ip.out = -1;
-		ip.err = err_fd;
-		ip.git_cmd = 1;
-		status = start_command(&ip);
-		if (status) {
+			argv_array_push(&av, "--fix-thin");
+		child.argv = av.argv;
+		child.out = -1;
+		child.err = err_fd;
+		child.git_cmd = 1;
+		status = start_command(&child);
+		if (status)
 			return "index-pack fork failed";
-		}
-		pack_lockfile = index_pack_lockfile(ip.out);
-		close(ip.out);
-		status = finish_command(&ip);
-		if (!status) {
-			reprepare_packed_git();
-			return NULL;
-		}
-		return "index-pack abnormal exit";
+		pack_lockfile = index_pack_lockfile(child.out);
+		close(child.out);
+		status = finish_command(&child);
+		if (status)
+			return "index-pack abnormal exit";
+		reprepare_packed_git();
 	}
+	return NULL;
 }
 
 static const char *unpack_with_sideband(void)
-- 
1.8.5.1.25.g8667982

[PATCH v4 17/28] receive/send-pack: support pushing from a shallow clone

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/technical/pack-protocol.txt |  4 +-
 builtin/receive-pack.c                    | 78 +++++++++++++++++++++++++++----
 builtin/send-pack.c                       |  2 +-
 send-pack.c                               |  3 ++
 t/t5537-push-shallow.sh (new +x)          | 70 +++++++++++++++++++++++++++
 5 files changed, 146 insertions(+), 11 deletions(-)
 create mode 100755 t/t5537-push-shallow.sh

diff --git a/Documentation/technical/pack-protocol.txt b/Documentation/technical/pack-protocol.txt
index eb8edd1..c73b62f 100644
--- a/Documentation/technical/pack-protocol.txt
+++ b/Documentation/technical/pack-protocol.txt
@@ -464,7 +464,9 @@ contain all the objects that the server will need to complete the new
 references.
 
 ----
-  update-request    =  command-list [pack-file]
+  update-request    =  *shallow command-list [pack-file]
+
+  shallow           =  PKT-LINE("shallow" SP obj-id)
 
   command-list      =  PKT-LINE(command NUL capability-list LF)
 		       *PKT-LINE(command LF)
diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index 8927ddf..b9de2e8 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -44,6 +44,7 @@ static int fix_thin = 1;
 static const char *head_name;
 static void *head_name_to_free;
 static int sent_capabilities;
+static const char *alt_shallow_file;
 
 static enum deny_action parse_deny_action(const char *var, const char *value)
 {
@@ -190,6 +191,7 @@ struct command {
 	const char *error_string;
 	unsigned int skip_update:1,
 		     did_not_exist:1;
+	int index;
 	unsigned char old_sha1[20];
 	unsigned char new_sha1[20];
 	char ref_name[FLEX_ARRAY]; /* more */
@@ -688,7 +690,7 @@ static int iterate_receive_command_list(void *cb_data, unsigned char sha1[20])
 	struct command *cmd = *cmd_list;
 
 	while (cmd) {
-		if (!is_null_sha1(cmd->new_sha1)) {
+		if (!is_null_sha1(cmd->new_sha1) && !cmd->skip_update) {
 			hashcpy(sha1, cmd->new_sha1);
 			*cmd_list = cmd->next;
 			return 0;
@@ -755,7 +757,7 @@ static void execute_commands(struct command *commands, const char *unpacker_erro
 	}
 }
 
-static struct command *read_head_info(void)
+static struct command *read_head_info(struct sha1_array *shallow)
 {
 	struct command *commands = NULL;
 	struct command **p = &commands;
@@ -769,6 +771,14 @@ static struct command *read_head_info(void)
 		line = packet_read_line(0, &len);
 		if (!line)
 			break;
+
+		if (len == 48 && !prefixcmp(line, "shallow ")) {
+			if (get_sha1_hex(line + 8, old_sha1))
+				die("protocol error: expected shallow sha, got '%s'", line + 8);
+			sha1_array_append(shallow, old_sha1);
+			continue;
+		}
+
 		if (len < 83 ||
 		    line[40] != ' ' ||
 		    line[81] != ' ' ||
@@ -820,7 +830,7 @@ static const char *parse_pack_header(struct pack_header *hdr)
 
 static const char *pack_lockfile;
 
-static const char *unpack(int err_fd)
+static const char *unpack(int err_fd, struct shallow_info *si)
 {
 	struct pack_header hdr;
 	struct argv_array av = ARGV_ARRAY_INIT;
@@ -844,6 +854,11 @@ static const char *unpack(int err_fd)
 			"--pack_header=%"PRIu32",%"PRIu32,
 			ntohl(hdr.hdr_version), ntohl(hdr.hdr_entries));
 
+	if (si->nr_ours || si->nr_theirs) {
+		alt_shallow_file = setup_temporary_shallow(si->shallow);
+		argv_array_pushl(&av, "--shallow-file", alt_shallow_file, NULL);
+	}
+
 	memset(&child, 0, sizeof(child));
 	if (ntohl(hdr.hdr_entries) < unpack_limit) {
 		argv_array_pushl(&av, "unpack-objects", hdr_arg, NULL);
@@ -889,13 +904,13 @@ static const char *unpack(int err_fd)
 	return NULL;
 }
 
-static const char *unpack_with_sideband(void)
+static const char *unpack_with_sideband(struct shallow_info *si)
 {
 	struct async muxer;
 	const char *ret;
 
 	if (!use_sideband)
-		return unpack(0);
+		return unpack(0, si);
 
 	memset(&muxer, 0, sizeof(muxer));
 	muxer.proc = copy_to_sideband;
@@ -903,12 +918,48 @@ static const char *unpack_with_sideband(void)
 	if (start_async(&muxer))
 		return NULL;
 
-	ret = unpack(muxer.in);
+	ret = unpack(muxer.in, si);
 
 	finish_async(&muxer);
 	return ret;
 }
 
+static void update_shallow_info(struct command *commands,
+				struct shallow_info *si,
+				struct sha1_array *ref)
+{
+	struct command *cmd;
+	int *ref_status;
+	remove_nonexistent_theirs_shallow(si);
+	/* XXX remove_nonexistent_ours_in_pack() */
+	if (!si->nr_ours && !si->nr_theirs)
+		return;
+
+	for (cmd = commands; cmd; cmd = cmd->next) {
+		if (is_null_sha1(cmd->new_sha1))
+			continue;
+		sha1_array_append(ref, cmd->new_sha1);
+		cmd->index = ref->nr - 1;
+	}
+	si->ref = ref;
+
+	ref_status = xmalloc(sizeof(*ref_status) * ref->nr);
+	assign_shallow_commits_to_refs(si, NULL, ref_status);
+	for (cmd = commands; cmd; cmd = cmd->next) {
+		if (is_null_sha1(cmd->new_sha1))
+			continue;
+		if (ref_status[cmd->index]) {
+			cmd->error_string = "shallow update not allowed";
+			cmd->skip_update = 1;
+		}
+	}
+	if (alt_shallow_file && *alt_shallow_file) {
+		unlink(alt_shallow_file);
+		alt_shallow_file = NULL;
+	}
+	free(ref_status);
+}
+
 static void report(struct command *commands, const char *unpack_status)
 {
 	struct command *cmd;
@@ -950,6 +1001,9 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
 	int i;
 	char *dir = NULL;
 	struct command *commands;
+	struct sha1_array shallow = SHA1_ARRAY_INIT;
+	struct sha1_array ref = SHA1_ARRAY_INIT;
+	struct shallow_info si;
 
 	packet_trace_identity("receive-pack");
 
@@ -1006,11 +1060,14 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
 	if (advertise_refs)
 		return 0;
 
-	if ((commands = read_head_info()) != NULL) {
+	if ((commands = read_head_info(&shallow)) != NULL) {
 		const char *unpack_status = NULL;
 
-		if (!delete_only(commands))
-			unpack_status = unpack_with_sideband();
+		prepare_shallow_info(&si, &shallow);
+		if (!delete_only(commands)) {
+			unpack_status = unpack_with_sideband(&si);
+			update_shallow_info(commands, &si, &ref);
+		}
 		execute_commands(commands, unpack_status);
 		if (pack_lockfile)
 			unlink_or_warn(pack_lockfile);
@@ -1027,8 +1084,11 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
 		}
 		if (auto_update_server_info)
 			update_server_info(0);
+		clear_shallow_info(&si);
 	}
 	if (use_sideband)
 		packet_flush(1);
+	sha1_array_clear(&shallow);
+	sha1_array_clear(&ref);
 	return 0;
 }
diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index 62cc4d3..ea2ab28 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -208,7 +208,7 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 	    (send_all && args.send_mirror))
 		usage(send_pack_usage);
 
-	if (is_repository_shallow())
+	if (is_repository_shallow() && args.stateless_rpc)
 		die("attempt to push from a shallow repository");
 
 	if (remote_name) {
diff --git a/send-pack.c b/send-pack.c
index 14005fa..cd536b4 100644
--- a/send-pack.c
+++ b/send-pack.c
@@ -214,6 +214,9 @@ int send_pack(struct send_pack_args *args,
 		return 0;
 	}
 
+	if (!args->dry_run)
+		advertise_shallow_grafts(out);
+
 	/*
 	 * Finally, tell the other end!
 	 */
diff --git a/t/t5537-push-shallow.sh b/t/t5537-push-shallow.sh
new file mode 100755
index 0000000..650c31a
--- /dev/null
+++ b/t/t5537-push-shallow.sh
@@ -0,0 +1,70 @@
+#!/bin/sh
+
+test_description='push from/to a shallow clone'
+
+. ./test-lib.sh
+
+commit() {
+	echo "$1" >tracked &&
+	git add tracked &&
+	git commit -m "$1"
+}
+
+test_expect_success 'setup' '
+	git config --global transfer.fsckObjects true &&
+	commit 1 &&
+	commit 2 &&
+	commit 3 &&
+	commit 4 &&
+	(
+	git init full-abc &&
+	cd full-abc &&
+	commit a &&
+	commit b &&
+	commit c
+	) &&
+	git clone --no-local --depth=2 .git shallow &&
+	git --git-dir=shallow/.git log --format=%s >actual &&
+	cat <<EOF >expect &&
+4
+3
+EOF
+	test_cmp expect actual &&
+	git clone --no-local --depth=2 full-abc/.git shallow2 &&
+	git --git-dir=shallow2/.git log --format=%s >actual &&
+	cat <<EOF >expect &&
+c
+b
+EOF
+	test_cmp expect actual
+'
+
+test_expect_success 'push from shallow clone' '
+	(
+	cd shallow &&
+	commit 5 &&
+	git push ../.git +master:refs/remotes/shallow/master
+	) &&
+	git log --format=%s shallow/master >actual &&
+	git fsck &&
+	cat <<EOF >expect &&
+5
+4
+3
+2
+1
+EOF
+	test_cmp expect actual
+'
+
+test_expect_success 'push from shallow clone, with grafted roots' '
+	(
+	cd shallow2 &&
+	test_must_fail git push ../.git +master:refs/remotes/shallow2/master 2>err &&
+	grep "shallow2/master.*shallow update not allowed" err
+	) &&
+	test_must_fail git rev-parse shallow2/master &&
+	git fsck
+'
+
+test_done
-- 
1.8.5.1.25.g8667982

[PATCH v4 18/28] New var GIT_SHALLOW_FILE to propagate --shallow-file to subprocesses

[Nguyễn Thái Ngọc Duy]

This may be needed when a hook is run after a new shallow pack is
received, but .git/shallow is not settled yet. A temporary shallow
file to plug all loose ends should be used instead. GIT_SHALLOW_FILE
is overriden by --shallow-file.

--shallow-file does not work in this case because the hook may spawn
many git subprocesses and the launch commands do not have
--shallow-file as it's a recent addition.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 cache.h       | 1 +
 commit.h      | 2 +-
 environment.c | 6 ++++++
 git.c         | 2 +-
 shallow.c     | 4 +++-
 5 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/cache.h b/cache.h
index 55dd4e3..8b13287 100644
--- a/cache.h
+++ b/cache.h
@@ -354,6 +354,7 @@ static inline enum object_type object_type(unsigned int mode)
 #define DB_ENVIRONMENT "GIT_OBJECT_DIRECTORY"
 #define INDEX_ENVIRONMENT "GIT_INDEX_FILE"
 #define GRAFT_ENVIRONMENT "GIT_GRAFT_FILE"
+#define GIT_SHALLOW_FILE_ENVIRONMENT "GIT_SHALLOW_FILE"
 #define TEMPLATE_DIR_ENVIRONMENT "GIT_TEMPLATE_DIR"
 #define CONFIG_ENVIRONMENT "GIT_CONFIG"
 #define CONFIG_DATA_ENVIRONMENT "GIT_CONFIG_PARAMETERS"
diff --git a/commit.h b/commit.h
index 69bca3e..79649ef 100644
--- a/commit.h
+++ b/commit.h
@@ -202,7 +202,7 @@ extern int is_repository_shallow(void);
 extern struct commit_list *get_shallow_commits(struct object_array *heads,
 		int depth, int shallow_flag, int not_shallow_flag);
 extern void check_shallow_file_for_update(void);
-extern void set_alternate_shallow_file(const char *path);
+extern void set_alternate_shallow_file(const char *path, int override);
 extern int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
 				 const struct sha1_array *extra);
 extern void setup_alternate_shallow(struct lock_file *shallow_lock,
diff --git a/environment.c b/environment.c
index 0a15349..b73b39d 100644
--- a/environment.c
+++ b/environment.c
@@ -10,6 +10,7 @@
 #include "cache.h"
 #include "refs.h"
 #include "fmt-merge-msg.h"
+#include "commit.h"
 
 int trust_executable_bit = 1;
 int trust_ctime = 1;
@@ -97,6 +98,7 @@ const char * const local_repo_env[] = {
 	INDEX_ENVIRONMENT,
 	NO_REPLACE_OBJECTS_ENVIRONMENT,
 	GIT_PREFIX_ENVIRONMENT,
+	GIT_SHALLOW_FILE_ENVIRONMENT,
 	NULL
 };
 
@@ -124,6 +126,7 @@ static char *expand_namespace(const char *raw_namespace)
 static void setup_git_env(void)
 {
 	const char *gitfile;
+	const char *shallow_file;
 
 	git_dir = getenv(GIT_DIR_ENVIRONMENT);
 	if (!git_dir)
@@ -147,6 +150,9 @@ static void setup_git_env(void)
 		read_replace_refs = 0;
 	namespace = expand_namespace(getenv(GIT_NAMESPACE_ENVIRONMENT));
 	namespace_len = strlen(namespace);
+	shallow_file = getenv(GIT_SHALLOW_FILE_ENVIRONMENT);
+	if (shallow_file)
+		set_alternate_shallow_file(shallow_file, 0);
 }
 
 int is_bare_repository(void)
diff --git a/git.c b/git.c
index cb5208d..179c4f6 100644
--- a/git.c
+++ b/git.c
@@ -162,7 +162,7 @@ static int handle_options(const char ***argv, int *argc, int *envchanged)
 		} else if (!strcmp(cmd, "--shallow-file")) {
 			(*argv)++;
 			(*argc)--;
-			set_alternate_shallow_file((*argv)[0]);
+			set_alternate_shallow_file((*argv)[0], 1);
 			if (envchanged)
 				*envchanged = 1;
 		} else if (!strcmp(cmd, "-C")) {
diff --git a/shallow.c b/shallow.c
index ae46b10..5710690 100644
--- a/shallow.c
+++ b/shallow.c
@@ -13,10 +13,12 @@ static int is_shallow = -1;
 static struct stat shallow_stat;
 static char *alternate_shallow_file;
 
-void set_alternate_shallow_file(const char *path)
+void set_alternate_shallow_file(const char *path, int override)
 {
 	if (is_shallow != -1)
 		die("BUG: is_repository_shallow must not be called before set_alternate_shallow_file");
+	if (alternate_shallow_file && !override)
+		return;
 	free(alternate_shallow_file);
 	alternate_shallow_file = path ? xstrdup(path) : NULL;
 }
-- 
1.8.5.1.25.g8667982

[PATCH v4 19/28] connected.c: add new variant that runs with --shallow-file

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 connected.c | 42 ++++++++++++++++++++++++++++++++++--------
 connected.h |  2 ++
 2 files changed, 36 insertions(+), 8 deletions(-)

diff --git a/connected.c b/connected.c
index fae8d64..427389d 100644
--- a/connected.c
+++ b/connected.c
@@ -19,17 +19,17 @@ int check_everything_connected(sha1_iterate_fn fn, int quiet, void *cb_data)
  *
  * Returns 0 if everything is connected, non-zero otherwise.
  */
-int check_everything_connected_with_transport(sha1_iterate_fn fn,
-					      int quiet,
-					      void *cb_data,
-					      struct transport *transport)
+static int check_everything_connected_real(sha1_iterate_fn fn,
+					   int quiet,
+					   void *cb_data,
+					   struct transport *transport,
+					   const char *shallow_file)
 {
 	struct child_process rev_list;
-	const char *argv[] = {"rev-list", "--objects",
-			      "--stdin", "--not", "--all", NULL, NULL};
+	const char *argv[9];
 	char commit[41];
 	unsigned char sha1[20];
-	int err = 0;
+	int err = 0, ac = 0;
 	struct packed_git *new_pack = NULL;
 
 	if (fn(cb_data, sha1))
@@ -47,8 +47,18 @@ int check_everything_connected_with_transport(sha1_iterate_fn fn,
 		strbuf_release(&idx_file);
 	}
 
+	if (shallow_file) {
+		argv[ac++] = "--shallow-file";
+		argv[ac++] = shallow_file;
+	}
+	argv[ac++] = "rev-list";
+	argv[ac++] = "--objects";
+	argv[ac++] = "--stdin";
+	argv[ac++] = "--not";
+	argv[ac++] = "--all";
 	if (quiet)
-		argv[5] = "--quiet";
+		argv[ac++] = "--quiet";
+	argv[ac] = NULL;
 
 	memset(&rev_list, 0, sizeof(rev_list));
 	rev_list.argv = argv;
@@ -92,3 +102,19 @@ int check_everything_connected_with_transport(sha1_iterate_fn fn,
 	sigchain_pop(SIGPIPE);
 	return finish_command(&rev_list) || err;
 }
+
+int check_everything_connected_with_transport(sha1_iterate_fn fn,
+					      int quiet,
+					      void *cb_data,
+					      struct transport *transport)
+{
+	return check_everything_connected_real(fn, quiet, cb_data,
+					       transport, NULL);
+}
+
+int check_shallow_connected(sha1_iterate_fn fn, int quiet, void *cb_data,
+			    const char *shallow_file)
+{
+	return check_everything_connected_real(fn, quiet, cb_data,
+					       NULL, shallow_file);
+}
diff --git a/connected.h b/connected.h
index 0b060b7..071d408 100644
--- a/connected.h
+++ b/connected.h
@@ -18,6 +18,8 @@ typedef int (*sha1_iterate_fn)(void *, unsigned char [20]);
  * Return 0 if Ok, non zero otherwise (i.e. some missing objects)
  */
 extern int check_everything_connected(sha1_iterate_fn, int quiet, void *cb_data);
+extern int check_shallow_connected(sha1_iterate_fn, int quiet, void *cb_data,
+				   const char *shallow_file);
 extern int check_everything_connected_with_transport(sha1_iterate_fn, int quiet,
 						     void *cb_data,
 						     struct transport *transport);
-- 
1.8.5.1.25.g8667982

[PATCH v4 20/28] receive-pack: allow pushes that update .git/shallow

[Nguyễn Thái Ngọc Duy]

The basic 8 steps to update .git/shallow does not fully apply here
because the user may choose to accept just a few refs (while fetch
always accepts all refs). The steps are modified a bit.

1-6. same as before. After calling assign_shallow_commits_to_refs at
   step 6, each shallow commit has a bitmap that marks all refs that
   require it.

7. mark all "ours" shallow commits that are reachable from any
   refs. We will need to do the original step 7 on them later.

8. go over all shallow commit bitmaps, mark refs that require new
   shallow commits.

9. setup a strict temporary shallow file to plug all the holes, even
   if it may cut some of our history short. This file is used by all
   hooks. The hooks could use --shallow-file=$GIT_DIR/shallow to
   overcome this and reach everything in current repo.

10. go over the new refs one by one. For each ref, do the reachability
   test if it needs a shallow commit on the list from step 7. Remove
   it if it's reachable from our refs. Gather all required shallow
   commits, run check_everything_connected() with the new ref, then
   install them to .git/shallow.

This mode is disabled by default and can be turned on with
receive.shallowupdate

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/config.txt |   4 ++
 builtin/receive-pack.c   | 163 +++++++++++++++++++++++++++++++++++++++++++----
 commit.h                 |   9 +++
 shallow.c                |  23 +++++++
 t/t5537-push-shallow.sh  |  15 +++++
 5 files changed, 201 insertions(+), 13 deletions(-)

diff --git a/Documentation/config.txt b/Documentation/config.txt
index ab26963..1a0bd0d 100644
--- a/Documentation/config.txt
+++ b/Documentation/config.txt
@@ -2026,6 +2026,10 @@ receive.updateserverinfo::
 	If set to true, git-receive-pack will run git-update-server-info
 	after receiving data from git-push and updating refs.
 
+receive.shallowupdate::
+	If set to true, .git/shallow can be updated when new refs
+	require new shallow roots. Otherwise those refs are rejected.
+
 remote.pushdefault::
 	The remote to push to by default.  Overrides
 	`branch.<name>.remote` for all branches, and is overridden by
diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index b9de2e8..5c85bb4 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -44,6 +44,7 @@ static int fix_thin = 1;
 static const char *head_name;
 static void *head_name_to_free;
 static int sent_capabilities;
+static int shallow_update;
 static const char *alt_shallow_file;
 
 static enum deny_action parse_deny_action(const char *var, const char *value)
@@ -123,6 +124,11 @@ static int receive_pack_config(const char *var, const char *value, void *cb)
 		return 0;
 	}
 
+	if (strcmp(var, "receive.shallowupdate") == 0) {
+		shallow_update = git_config_bool(var, value);
+		return 0;
+	}
+
 	return git_default_config(var, value, cb);
 }
 
@@ -423,7 +429,46 @@ static void refuse_unconfigured_deny_delete_current(void)
 		rp_error("%s", refuse_unconfigured_deny_delete_current_msg[i]);
 }
 
-static const char *update(struct command *cmd)
+static int command_singleton_iterator(void *cb_data, unsigned char sha1[20]);
+static int update_shallow_ref(struct command *cmd, struct shallow_info *si)
+{
+	static struct lock_file shallow_lock;
+	struct sha1_array extra = SHA1_ARRAY_INIT;
+	const char *alt_file;
+	uint32_t mask = 1 << (cmd->index % 32);
+	int i;
+
+	trace_printf_key("GIT_TRACE_SHALLOW",
+			 "shallow: update_shallow_ref %s\n", cmd->ref_name);
+	for (i = 0; i < si->shallow->nr; i++)
+		if (si->used_shallow[i] &&
+		    (si->used_shallow[i][cmd->index / 32] & mask) &&
+		    !delayed_reachability_test(si, i))
+			sha1_array_append(&extra, si->shallow->sha1[i]);
+
+	setup_alternate_shallow(&shallow_lock, &alt_file, &extra);
+	if (check_shallow_connected(command_singleton_iterator,
+				    0, cmd, alt_file)) {
+		rollback_lock_file(&shallow_lock);
+		sha1_array_clear(&extra);
+		return -1;
+	}
+
+	commit_lock_file(&shallow_lock);
+
+	/*
+	 * Make sure setup_alternate_shallow() for the next ref does
+	 * not lose these new roots..
+	 */
+	for (i = 0; i < extra.nr; i++)
+		register_shallow(extra.sha1[i]);
+
+	si->shallow_ref[cmd->index] = 0;
+	sha1_array_clear(&extra);
+	return 0;
+}
+
+static const char *update(struct command *cmd, struct shallow_info *si)
 {
 	const char *name = cmd->ref_name;
 	struct strbuf namespaced_name_buf = STRBUF_INIT;
@@ -531,6 +576,10 @@ static const char *update(struct command *cmd)
 		return NULL; /* good */
 	}
 	else {
+		if (shallow_update && si->shallow_ref[cmd->index] &&
+		    update_shallow_ref(cmd, si))
+			return "shallow error";
+
 		lock = lock_any_ref_for_update(namespaced_name, old_sha1,
 					       0, NULL);
 		if (!lock) {
@@ -671,12 +720,16 @@ static int command_singleton_iterator(void *cb_data, unsigned char sha1[20])
 	return 0;
 }
 
-static void set_connectivity_errors(struct command *commands)
+static void set_connectivity_errors(struct command *commands,
+				    struct shallow_info *si)
 {
 	struct command *cmd;
 
 	for (cmd = commands; cmd; cmd = cmd->next) {
 		struct command *singleton = cmd;
+		if (shallow_update && si->shallow_ref[cmd->index])
+			/* to be checked in update_shallow_ref() */
+			continue;
 		if (!check_everything_connected(command_singleton_iterator,
 						0, &singleton))
 			continue;
@@ -684,18 +737,26 @@ static void set_connectivity_errors(struct command *commands)
 	}
 }
 
+struct iterate_data {
+	struct command *cmds;
+	struct shallow_info *si;
+};
+
 static int iterate_receive_command_list(void *cb_data, unsigned char sha1[20])
 {
-	struct command **cmd_list = cb_data;
+	struct iterate_data *data = cb_data;
+	struct command **cmd_list = &data->cmds;
 	struct command *cmd = *cmd_list;
 
-	while (cmd) {
+	for (; cmd; cmd = cmd->next) {
+		if (shallow_update && data->si->shallow_ref[cmd->index])
+			/* to be checked in update_shallow_ref() */
+			continue;
 		if (!is_null_sha1(cmd->new_sha1) && !cmd->skip_update) {
 			hashcpy(sha1, cmd->new_sha1);
 			*cmd_list = cmd->next;
 			return 0;
 		}
-		cmd = cmd->next;
 	}
 	*cmd_list = NULL;
 	return -1; /* end of list */
@@ -715,10 +776,14 @@ static void reject_updates_to_hidden(struct command *commands)
 	}
 }
 
-static void execute_commands(struct command *commands, const char *unpacker_error)
+static void execute_commands(struct command *commands,
+			     const char *unpacker_error,
+			     struct shallow_info *si)
 {
+	int checked_connectivity;
 	struct command *cmd;
 	unsigned char sha1[20];
+	struct iterate_data data;
 
 	if (unpacker_error) {
 		for (cmd = commands; cmd; cmd = cmd->next)
@@ -726,10 +791,10 @@ static void execute_commands(struct command *commands, const char *unpacker_erro
 		return;
 	}
 
-	cmd = commands;
-	if (check_everything_connected(iterate_receive_command_list,
-				       0, &cmd))
-		set_connectivity_errors(commands);
+	data.cmds = commands;
+	data.si = si;
+	if (check_everything_connected(iterate_receive_command_list, 0, &data))
+		set_connectivity_errors(commands, si);
 
 	reject_updates_to_hidden(commands);
 
@@ -746,6 +811,7 @@ static void execute_commands(struct command *commands, const char *unpacker_erro
 	free(head_name_to_free);
 	head_name = head_name_to_free = resolve_refdup("HEAD", sha1, 0, NULL);
 
+	checked_connectivity = 1;
 	for (cmd = commands; cmd; cmd = cmd->next) {
 		if (cmd->error_string)
 			continue;
@@ -753,7 +819,22 @@ static void execute_commands(struct command *commands, const char *unpacker_erro
 		if (cmd->skip_update)
 			continue;
 
-		cmd->error_string = update(cmd);
+		cmd->error_string = update(cmd, si);
+		if (shallow_update && !cmd->error_string &&
+		    si->shallow_ref[cmd->index]) {
+			error("BUG: connectivity check has not been run on ref %s",
+			      cmd->ref_name);
+			checked_connectivity = 0;
+		}
+	}
+
+	if (shallow_update) {
+		if (!checked_connectivity)
+			error("BUG: run 'git fsck' for safety.\n"
+			      "If there are errors, try to remove "
+			      "the reported refs above");
+		if (alt_shallow_file && *alt_shallow_file)
+			unlink(alt_shallow_file);
 	}
 }
 
@@ -924,6 +1005,53 @@ static const char *unpack_with_sideband(struct shallow_info *si)
 	return ret;
 }
 
+static void prepare_shallow_update(struct command *commands,
+				   struct shallow_info *si)
+{
+	int i, j, k, bitmap_size = (si->ref->nr + 31) / 32;
+
+	si->used_shallow = xmalloc(sizeof(*si->used_shallow) *
+				   si->shallow->nr);
+	assign_shallow_commits_to_refs(si, si->used_shallow, NULL);
+
+	si->need_reachability_test =
+		xcalloc(si->shallow->nr, sizeof(*si->need_reachability_test));
+	si->reachable =
+		xcalloc(si->shallow->nr, sizeof(*si->reachable));
+	si->shallow_ref = xcalloc(si->ref->nr, sizeof(*si->shallow_ref));
+
+	for (i = 0; i < si->nr_ours; i++)
+		si->need_reachability_test[si->ours[i]] = 1;
+
+	for (i = 0; i < si->shallow->nr; i++) {
+		if (!si->used_shallow[i])
+			continue;
+		for (j = 0; j < bitmap_size; j++) {
+			if (!si->used_shallow[i][j])
+				continue;
+			si->need_reachability_test[i]++;
+			for (k = 0; k < 32; k++)
+				if (si->used_shallow[i][j] & (1 << k))
+					si->shallow_ref[j * 32 + k]++;
+		}
+
+		/*
+		 * true for those associated with some refs and belong
+		 * in "ours" list aka "step 7 not done yet"
+		 */
+		si->need_reachability_test[i] =
+			si->need_reachability_test[i] > 1;
+	}
+
+	/*
+	 * keep hooks happy by forcing a temporary shallow file via
+	 * env variable because we can't add --shallow-file to every
+	 * command. check_everything_connected() will be done with
+	 * true .git/shallow though.
+	 */
+	setenv(GIT_SHALLOW_FILE_ENVIRONMENT, alt_shallow_file, 1);
+}
+
 static void update_shallow_info(struct command *commands,
 				struct shallow_info *si,
 				struct sha1_array *ref)
@@ -932,8 +1060,10 @@ static void update_shallow_info(struct command *commands,
 	int *ref_status;
 	remove_nonexistent_theirs_shallow(si);
 	/* XXX remove_nonexistent_ours_in_pack() */
-	if (!si->nr_ours && !si->nr_theirs)
+	if (!si->nr_ours && !si->nr_theirs) {
+		shallow_update = 0;
 		return;
+	}
 
 	for (cmd = commands; cmd; cmd = cmd->next) {
 		if (is_null_sha1(cmd->new_sha1))
@@ -943,6 +1073,11 @@ static void update_shallow_info(struct command *commands,
 	}
 	si->ref = ref;
 
+	if (shallow_update) {
+		prepare_shallow_update(commands, si);
+		return;
+	}
+
 	ref_status = xmalloc(sizeof(*ref_status) * ref->nr);
 	assign_shallow_commits_to_refs(si, NULL, ref_status);
 	for (cmd = commands; cmd; cmd = cmd->next) {
@@ -1064,11 +1199,13 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
 		const char *unpack_status = NULL;
 
 		prepare_shallow_info(&si, &shallow);
+		if (!si.nr_ours && !si.nr_theirs)
+			shallow_update = 0;
 		if (!delete_only(commands)) {
 			unpack_status = unpack_with_sideband(&si);
 			update_shallow_info(commands, &si, &ref);
 		}
-		execute_commands(commands, unpack_status);
+		execute_commands(commands, unpack_status, &si);
 		if (pack_lockfile)
 			unlink_or_warn(pack_lockfile);
 		if (report_status)
diff --git a/commit.h b/commit.h
index 79649ef..a1f2d49 100644
--- a/commit.h
+++ b/commit.h
@@ -216,6 +216,14 @@ struct shallow_info {
 	int *ours, nr_ours;
 	int *theirs, nr_theirs;
 	struct sha1_array *ref;
+
+	/* for receive-pack */
+	uint32_t **used_shallow;
+	int *need_reachability_test;
+	int *reachable;
+	int *shallow_ref;
+	struct commit **commits;
+	int nr_commits;
 };
 
 extern void prepare_shallow_info(struct shallow_info *, struct sha1_array *);
@@ -226,6 +234,7 @@ extern void remove_nonexistent_ours_in_pack(struct shallow_info *,
 extern void assign_shallow_commits_to_refs(struct shallow_info *info,
 					   uint32_t **used,
 					   int *ref_status);
+extern int delayed_reachability_test(struct shallow_info *si, int c);
 
 int is_descendant_of(struct commit *, struct commit_list *);
 int in_merge_bases(struct commit *, struct commit *);
diff --git a/shallow.c b/shallow.c
index 5710690..e611639 100644
--- a/shallow.c
+++ b/shallow.c
@@ -618,3 +618,26 @@ static void post_assign_shallow(struct shallow_info *info,
 
 	free(ca.commits);
 }
+
+/* (Delayed) step 7, reachability test at commit level */
+int delayed_reachability_test(struct shallow_info *si, int c)
+{
+	if (si->need_reachability_test[c]) {
+		struct commit *commit = lookup_commit(si->shallow->sha1[c]);
+
+		if (!si->commits) {
+			struct commit_array ca;
+			memset(&ca, 0, sizeof(ca));
+			head_ref(add_ref, &ca);
+			for_each_ref(add_ref, &ca);
+			si->commits = ca.commits;
+			si->nr_commits = ca.nr;
+		}
+
+		si->reachable[c] =in_merge_bases_many(commit,
+						       si->nr_commits,
+						       si->commits);
+		si->need_reachability_test[c] = 0;
+	}
+	return si->reachable[c];
+}
diff --git a/t/t5537-push-shallow.sh b/t/t5537-push-shallow.sh
index 650c31a..ff5eb5b 100755
--- a/t/t5537-push-shallow.sh
+++ b/t/t5537-push-shallow.sh
@@ -67,4 +67,19 @@ test_expect_success 'push from shallow clone, with grafted roots' '
 	git fsck
 '
 
+test_expect_success 'add new shallow root with receive.updateshallow on' '
+	test_config receive.shallowupdate true &&
+	(
+	cd shallow2 &&
+	git push ../.git +master:refs/remotes/shallow2/master
+	) &&
+	git log --format=%s shallow2/master >actual &&
+	git fsck &&
+	cat <<EOF >expect &&
+c
+b
+EOF
+	test_cmp expect actual
+'
+
 test_done
-- 
1.8.5.1.25.g8667982

[PATCH v4 21/28] send-pack: support pushing to a shallow clone

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/send-pack.c     |  4 +++-
 t/t5537-push-shallow.sh | 38 ++++++++++++++++++++++++++++++++++++++
 transport.c             |  5 ++---
 3 files changed, 43 insertions(+), 4 deletions(-)

diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index ea2ab28..664dd20 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -101,6 +101,7 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 	int fd[2];
 	struct child_process *conn;
 	struct sha1_array extra_have = SHA1_ARRAY_INIT;
+	struct sha1_array shallow = SHA1_ARRAY_INIT;
 	struct ref *remote_refs, *local_refs;
 	int ret;
 	int helper_status = 0;
@@ -232,7 +233,8 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 			args.verbose ? CONNECT_VERBOSE : 0);
 	}
 
-	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL, &extra_have, NULL);
+	get_remote_heads(fd[0], NULL, 0, &remote_refs, REF_NORMAL,
+			 &extra_have, &shallow);
 
 	transport_verify_remote_names(nr_refspecs, refspecs);
 
diff --git a/t/t5537-push-shallow.sh b/t/t5537-push-shallow.sh
index ff5eb5b..f5c74e6 100755
--- a/t/t5537-push-shallow.sh
+++ b/t/t5537-push-shallow.sh
@@ -82,4 +82,42 @@ EOF
 	test_cmp expect actual
 '
 
+test_expect_success 'push from shallow to shallow' '
+	(
+	cd shallow &&
+	git --git-dir=../shallow2/.git config receive.shallowupdate true &&
+	git push ../shallow2/.git +master:refs/remotes/shallow/master &&
+	git --git-dir=../shallow2/.git config receive.shallowupdate false
+	) &&
+	(
+	cd shallow2 &&
+	git log --format=%s shallow/master >actual &&
+	git fsck &&
+	cat <<EOF >expect &&
+5
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
+test_expect_success 'push from full to shallow' '
+	! git --git-dir=shallow2/.git cat-file blob `echo 1|git hash-object --stdin` &&
+	commit 1 &&
+	git push shallow2/.git +master:refs/remotes/top/master &&
+	(
+	cd shallow2 &&
+	git log --format=%s top/master >actual &&
+	git fsck &&
+	cat <<EOF >expect &&
+1
+4
+3
+EOF
+	test_cmp expect actual &&
+	git cat-file blob `echo 1|git hash-object --stdin` >/dev/null
+	)
+'
+
 test_done
diff --git a/transport.c b/transport.c
index a09fdb6..d596abb 100644
--- a/transport.c
+++ b/transport.c
@@ -819,11 +819,10 @@ static int git_transport_push(struct transport *transport, struct ref *remote_re
 		struct ref *tmp_refs;
 		connect_setup(transport, 1, 0);
 
-		get_remote_heads(data->fd[0], NULL, 0, &tmp_refs, REF_NORMAL, NULL, NULL);
+		get_remote_heads(data->fd[0], NULL, 0, &tmp_refs, REF_NORMAL,
+				 NULL, &data->shallow);
 		data->got_remote_heads = 1;
 	}
-	if (data->shallow.nr)
-		die("pushing to a shallow repository is not supported");
 
 	memset(&args, 0, sizeof(args));
 	args.send_mirror = !!(flags & TRANSPORT_PUSH_MIRROR);
-- 
1.8.5.1.25.g8667982

[PATCH v4 22/28] remote-curl: pass ref SHA-1 to fetch-pack as well

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/fetch-pack.c | 7 +++++++
 remote-curl.c        | 3 ++-
 2 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/builtin/fetch-pack.c b/builtin/fetch-pack.c
index 927424b..aa6e596 100644
--- a/builtin/fetch-pack.c
+++ b/builtin/fetch-pack.c
@@ -13,6 +13,13 @@ static void add_sought_entry_mem(struct ref ***sought, int *nr, int *alloc,
 				 const char *name, int namelen)
 {
 	struct ref *ref = xcalloc(1, sizeof(*ref) + namelen + 1);
+	unsigned char sha1[20];
+
+	if (namelen > 41 && name[40] == ' ' && !get_sha1_hex(name, sha1)) {
+		hashcpy(ref->old_sha1, sha1);
+		name += 41;
+		namelen -= 41;
+	}
 
 	memcpy(ref->name, name, namelen);
 	ref->name[namelen] = '\0';
diff --git a/remote-curl.c b/remote-curl.c
index 222210f..25d6730 100644
--- a/remote-curl.c
+++ b/remote-curl.c
@@ -719,7 +719,8 @@ static int fetch_git(struct discovery *heads,
 		struct ref *ref = to_fetch[i];
 		if (!ref->name || !*ref->name)
 			die("cannot fetch by sha1 over smart http");
-		packet_buf_write(&preamble, "%s\n", ref->name);
+		packet_buf_write(&preamble, "%s %s\n",
+				 sha1_to_hex(ref->old_sha1), ref->name);
 	}
 	packet_buf_flush(&preamble);
 
-- 
1.8.5.1.25.g8667982

[PATCH v4 23/28] Support shallow fetch/clone over smart-http

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/gitremote-helpers.txt |  7 +++++++
 builtin/fetch-pack.c                | 16 +++++++++++++---
 remote-curl.c                       | 31 +++++++++++++++++++++++++++++--
 t/t5536-fetch-shallow.sh            | 27 +++++++++++++++++++++++++++
 transport-helper.c                  |  6 ++++++
 upload-pack.c                       |  2 --
 6 files changed, 82 insertions(+), 7 deletions(-)

diff --git a/Documentation/gitremote-helpers.txt b/Documentation/gitremote-helpers.txt
index f1f4ca9..c2908db 100644
--- a/Documentation/gitremote-helpers.txt
+++ b/Documentation/gitremote-helpers.txt
@@ -437,6 +437,13 @@ set by Git if the remote helper has the 'option' capability.
 'option check-connectivity' \{'true'|'false'\}::
 	Request the helper to check connectivity of a clone.
 
+'option cloning \{'true'|'false'\}::
+	Notify the helper this is a clone request (i.e. the current
+	repository is guaranteed empty).
+
+'option update-shallow \{'true'|'false'\}::
+	Allow to extend .git/shallow if the new refs require it.
+
 SEE ALSO
 --------
 linkgit:git-remote[1]
diff --git a/builtin/fetch-pack.c b/builtin/fetch-pack.c
index aa6e596..81fae38 100644
--- a/builtin/fetch-pack.c
+++ b/builtin/fetch-pack.c
@@ -3,6 +3,7 @@
 #include "fetch-pack.h"
 #include "remote.h"
 #include "connect.h"
+#include "sha1-array.h"
 
 static const char fetch_pack_usage[] =
 "git fetch-pack [--all] [--stdin] [--quiet|-q] [--keep|-k] [--thin] "
@@ -46,6 +47,7 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
 	char **pack_lockfile_ptr = NULL;
 	struct child_process *conn;
 	struct fetch_pack_args args;
+	struct sha1_array shallow = SHA1_ARRAY_INIT;
 
 	packet_trace_identity("fetch-pack");
 
@@ -113,6 +115,14 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
 			args.check_self_contained_and_connected = 1;
 			continue;
 		}
+		if (!strcmp("--cloning", arg)) {
+			args.cloning = 1;
+			continue;
+		}
+		if (!strcmp("--update-shallow", arg)) {
+			args.update_shallow = 1;
+			continue;
+		}
 		usage(fetch_pack_usage);
 	}
 
@@ -157,10 +167,10 @@ int cmd_fetch_pack(int argc, const char **argv, const char *prefix)
 				   args.verbose ? CONNECT_VERBOSE : 0);
 	}
 
-	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL, NULL);
+	get_remote_heads(fd[0], NULL, 0, &ref, 0, NULL, &shallow);
 
-	ref = fetch_pack(&args, fd, conn, ref, dest,
-			 sought, nr_sought, NULL, pack_lockfile_ptr);
+	ref = fetch_pack(&args, fd, conn, ref, dest, sought, nr_sought,
+			 &shallow, pack_lockfile_ptr);
 	if (pack_lockfile) {
 		printf("lock %s\n", pack_lockfile);
 		fflush(stdout);
diff --git a/remote-curl.c b/remote-curl.c
index 25d6730..3d97e3d 100644
--- a/remote-curl.c
+++ b/remote-curl.c
@@ -10,6 +10,7 @@
 #include "sideband.h"
 #include "argv-array.h"
 #include "credential.h"
+#include "sha1-array.h"
 
 static struct remote *remote;
 /* always ends with a trailing slash */
@@ -20,6 +21,8 @@ struct options {
 	unsigned long depth;
 	unsigned progress : 1,
 		check_self_contained_and_connected : 1,
+		cloning : 1,
+		update_shallow : 1,
 		followtags : 1,
 		dry_run : 1,
 		thin : 1;
@@ -88,6 +91,24 @@ static int set_option(const char *name, const char *value)
 		strbuf_release(&val);
 		return 0;
 	}
+	else if (!strcmp(name, "cloning")) {
+		if (!strcmp(value, "true"))
+			options.cloning = 1;
+		else if (!strcmp(value, "false"))
+			options.cloning = 0;
+		else
+			return -1;
+		return 0;
+	}
+	else if (!strcmp(name, "update-shallow")) {
+		if (!strcmp(value, "true"))
+			options.update_shallow = 1;
+		else if (!strcmp(value, "false"))
+			options.update_shallow = 0;
+		else
+			return -1;
+		return 0;
+	}
 	else {
 		return 1 /* unsupported */;
 	}
@@ -99,6 +120,7 @@ struct discovery {
 	char *buf;
 	size_t len;
 	struct ref *refs;
+	struct sha1_array shallow;
 	unsigned proto_git : 1;
 };
 static struct discovery *last_discovery;
@@ -107,7 +129,7 @@ static struct ref *parse_git_refs(struct discovery *heads, int for_push)
 {
 	struct ref *list = NULL;
 	get_remote_heads(-1, heads->buf, heads->len, &list,
-			 for_push ? REF_NORMAL : 0, NULL, NULL);
+			 for_push ? REF_NORMAL : 0, NULL, &heads->shallow);
 	return list;
 }
 
@@ -168,6 +190,7 @@ static void free_discovery(struct discovery *d)
 	if (d) {
 		if (d == last_discovery)
 			last_discovery = NULL;
+		free(d->shallow.sha1);
 		free(d->buf_alloc);
 		free_refs(d->refs);
 		free(d);
@@ -688,7 +711,7 @@ static int fetch_git(struct discovery *heads,
 	struct strbuf preamble = STRBUF_INIT;
 	char *depth_arg = NULL;
 	int argc = 0, i, err;
-	const char *argv[16];
+	const char *argv[17];
 
 	argv[argc++] = "fetch-pack";
 	argv[argc++] = "--stateless-rpc";
@@ -704,6 +727,10 @@ static int fetch_git(struct discovery *heads,
 	}
 	if (options.check_self_contained_and_connected)
 		argv[argc++] = "--check-self-contained-and-connected";
+	if (options.cloning)
+		argv[argc++] = "--cloning";
+	if (options.update_shallow)
+		argv[argc++] = "--update-shallow";
 	if (!options.progress)
 		argv[argc++] = "--no-progress";
 	if (options.depth) {
diff --git a/t/t5536-fetch-shallow.sh b/t/t5536-fetch-shallow.sh
index 3ae9092..79ce472 100755
--- a/t/t5536-fetch-shallow.sh
+++ b/t/t5536-fetch-shallow.sh
@@ -173,4 +173,31 @@ EOF
 	)
 '
 
+if test -n "$NO_CURL" -o -z "$GIT_TEST_HTTPD"; then
+	say 'skipping remaining tests, git built without http support'
+	test_done
+fi
+
+LIB_HTTPD_PORT=${LIB_HTTPD_PORT-'5536'}
+. "$TEST_DIRECTORY"/lib-httpd.sh
+start_httpd
+
+test_expect_success 'clone http repository' '
+	git clone --bare --no-local shallow "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	git clone $HTTPD_URL/smart/repo.git clone &&
+	(
+	cd clone &&
+	git fsck &&
+	git log --format=%s origin/master >actual &&
+	cat <<EOF >expect &&
+6
+5
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
+stop_httpd
 test_done
diff --git a/transport-helper.c b/transport-helper.c
index 673b7c2..e2b4203 100644
--- a/transport-helper.c
+++ b/transport-helper.c
@@ -360,6 +360,12 @@ static int fetch_with_fetch(struct transport *transport,
 	    data->transport_options.check_self_contained_and_connected)
 		set_helper_option(transport, "check-connectivity", "true");
 
+	if (transport->cloning)
+		set_helper_option(transport, "cloning", "true");
+
+	if (data->transport_options.update_shallow)
+		set_helper_option(transport, "update-shallow", "true");
+
 	for (i = 0; i < nr_heads; i++) {
 		const struct ref *posn = to_fetch[i];
 		if (posn->status & REF_STATUS_UPTODATE)
diff --git a/upload-pack.c b/upload-pack.c
index 28269c7..2d02297 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -836,8 +836,6 @@ int main(int argc, char **argv)
 
 	if (!enter_repo(dir, strict))
 		die("'%s' does not appear to be a git repository", dir);
-	if (is_repository_shallow() && stateless_rpc)
-		die("attempt to push into a shallow repository");
 
 	git_config(upload_pack_config, NULL);
 	upload_pack();
-- 
1.8.5.1.25.g8667982

[PATCH v4 24/28] receive-pack: support pushing to a shallow clone via http

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/receive-pack.c  |  3 ---
 t/t5537-push-shallow.sh | 35 +++++++++++++++++++++++++++++++++++
 2 files changed, 35 insertions(+), 3 deletions(-)

diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c
index 5c85bb4..78fe8ee 100644
--- a/builtin/receive-pack.c
+++ b/builtin/receive-pack.c
@@ -1179,9 +1179,6 @@ int cmd_receive_pack(int argc, const char **argv, const char *prefix)
 	if (!enter_repo(dir, 0))
 		die("'%s' does not appear to be a git repository", dir);
 
-	if (is_repository_shallow() && stateless_rpc)
-		die("attempt to push into a shallow repository");
-
 	git_config(receive_pack_config, NULL);
 
 	if (0 <= transfer_unpack_limit)
diff --git a/t/t5537-push-shallow.sh b/t/t5537-push-shallow.sh
index f5c74e6..866621a 100755
--- a/t/t5537-push-shallow.sh
+++ b/t/t5537-push-shallow.sh
@@ -16,6 +16,7 @@ test_expect_success 'setup' '
 	commit 2 &&
 	commit 3 &&
 	commit 4 &&
+	git clone . full &&
 	(
 	git init full-abc &&
 	cd full-abc &&
@@ -120,4 +121,38 @@ EOF
 	)
 '
 
+if test -n "$NO_CURL" -o -z "$GIT_TEST_HTTPD"; then
+	say 'skipping remaining tests, git built without http support'
+	test_done
+fi
+
+LIB_HTTPD_PORT=${LIB_HTTPD_PORT-'5537'}
+. "$TEST_DIRECTORY"/lib-httpd.sh
+start_httpd
+
+test_expect_success 'push to shallow repo via http' '
+	git clone --bare --no-local shallow "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	(
+	cd "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	git config http.receivepack true
+	) &&
+	(
+	cd full &&
+	commit 9 &&
+	git push $HTTPD_URL/smart/repo.git +master:refs/remotes/top/master
+	) &&
+	(
+	cd "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	git fsck &&
+	git log --format=%s top/master >actual &&
+	cat <<EOF >expect &&
+9
+4
+3
+EOF
+	test_cmp expect actual
+	)
+'
+
+stop_httpd
 test_done
-- 
1.8.5.1.25.g8667982

[PATCH v4 25/28] send-pack: support pushing from a shallow clone via http

[Nguyễn Thái Ngọc Duy]

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/send-pack.c     |  3 ---
 send-pack.c             | 19 +++++++++++++++++--
 t/t5537-push-shallow.sh | 25 +++++++++++++++++++++++++
 3 files changed, 42 insertions(+), 5 deletions(-)

diff --git a/builtin/send-pack.c b/builtin/send-pack.c
index 664dd20..cc25744 100644
--- a/builtin/send-pack.c
+++ b/builtin/send-pack.c
@@ -209,9 +209,6 @@ int cmd_send_pack(int argc, const char **argv, const char *prefix)
 	    (send_all && args.send_mirror))
 		usage(send_pack_usage);
 
-	if (is_repository_shallow() && args.stateless_rpc)
-		die("attempt to push from a shallow repository");
-
 	if (remote_name) {
 		remote = remote_get(remote_name);
 		if (!remote_has_url(remote, dest)) {
diff --git a/send-pack.c b/send-pack.c
index cd536b4..848d15e 100644
--- a/send-pack.c
+++ b/send-pack.c
@@ -175,6 +175,21 @@ static int sideband_demux(int in, int out, void *data)
 	return ret;
 }
 
+static int advertise_shallow_grafts_cb(const struct commit_graft *graft, void *cb)
+{
+	struct strbuf *sb = cb;
+	if (graft->nr_parent == -1)
+		packet_buf_write(sb, "shallow %s\n", sha1_to_hex(graft->sha1));
+	return 0;
+}
+
+void advertise_shallow_grafts_buf(struct strbuf *sb)
+{
+	if (!is_repository_shallow())
+		return;
+	for_each_commit_graft(advertise_shallow_grafts_cb, sb);
+}
+
 int send_pack(struct send_pack_args *args,
 	      int fd[], struct child_process *conn,
 	      struct ref *remote_refs,
@@ -215,7 +230,7 @@ int send_pack(struct send_pack_args *args,
 	}
 
 	if (!args->dry_run)
-		advertise_shallow_grafts(out);
+		advertise_shallow_grafts_buf(&req_buf);
 
 	/*
 	 * Finally, tell the other end!
@@ -276,7 +291,7 @@ int send_pack(struct send_pack_args *args,
 	}
 
 	if (args->stateless_rpc) {
-		if (!args->dry_run && cmds_sent) {
+		if (!args->dry_run && (cmds_sent || is_repository_shallow())) {
 			packet_buf_flush(&req_buf);
 			send_sideband(out, -1, req_buf.buf, req_buf.len, LARGE_PACKET_MAX);
 		}
diff --git a/t/t5537-push-shallow.sh b/t/t5537-push-shallow.sh
index 866621a..0a6e40f 100755
--- a/t/t5537-push-shallow.sh
+++ b/t/t5537-push-shallow.sh
@@ -154,5 +154,30 @@ EOF
 	)
 '
 
+test_expect_success 'push from shallow repo via http' '
+	mv "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" shallow-upstream.git &&
+	git clone --bare --no-local full "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	(
+	cd "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	git config http.receivepack true
+	) &&
+	commit 10 &&
+	git push $HTTPD_URL/smart/repo.git +master:refs/remotes/top/master &&
+	(
+	cd "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
+	git fsck &&
+	git log --format=%s top/master >actual &&
+	cat <<EOF >expect &&
+10
+1
+4
+3
+2
+1
+EOF
+	test_cmp expect actual
+	)
+'
+
 stop_httpd
 test_done
-- 
1.8.5.1.25.g8667982

[PATCH v4 26/28] clone: use git protocol for cloning shallow repo locally

[Nguyễn Thái Ngọc Duy]

clone_local() does not handle $SRC/shallow. It could be made so, but
it's simpler to use fetch-pack/upload-pack instead.

This used to be caught by the check in upload-pack, which is triggered
by transport_get_remote_refs(), even in local clone case. The check is
now gone and check_everything_connected() should catch the result
incomplete repo. But check_everything_connected() will soon be skipped
in local clone case, opening a door to corrupt repo. This patch should
close that door.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 builtin/clone.c  | 11 +++++++++--
 t/t5601-clone.sh |  7 +++++++
 2 files changed, 16 insertions(+), 2 deletions(-)

diff --git a/builtin/clone.c b/builtin/clone.c
index 0b182ce..71ee68b 100644
--- a/builtin/clone.c
+++ b/builtin/clone.c
@@ -797,8 +797,15 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
 	else
 		repo = repo_name;
 	is_local = option_local != 0 && path && !is_bundle;
-	if (is_local && option_depth)
-		warning(_("--depth is ignored in local clones; use file:// instead."));
+	if (is_local) {
+		if (option_depth)
+			warning(_("--depth is ignored in local clones; use file:// instead."));
+		if (!access(mkpath("%s/shallow", path), F_OK)) {
+			if (option_local > 0)
+				warning(_("source repository is shallow, ignoring --local"));
+			is_local = 0;
+		}
+	}
 	if (option_local > 0 && !is_local)
 		warning(_("--local is ignored"));
 
diff --git a/t/t5601-clone.sh b/t/t5601-clone.sh
index 1d1c875..c226cff 100755
--- a/t/t5601-clone.sh
+++ b/t/t5601-clone.sh
@@ -340,4 +340,11 @@ test_expect_success 'clone from a repository with two identical branches' '
 
 '
 
+test_expect_success 'shallow clone locally' '
+	git clone --depth=1 --no-local src ssrrcc &&
+	git clone ssrrcc ddsstt &&
+	test_cmp ssrrcc/.git/shallow ddsstt/.git/shallow &&
+	( cd ddsstt && git fsck )
+'
+
 test_done
-- 
1.8.5.1.25.g8667982

[PATCH v4 27/28] prune: clean .git/shallow after pruning objects

[Nguyễn Thái Ngọc Duy]

This patch teaches "prune" to remove shallow roots that are no longer
reachable from any refs (e.g. when the relevant refs are removed).

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/git-prune.txt |  2 ++
 builtin/gc.c                |  1 +
 builtin/prune.c             |  4 ++++
 commit.h                    |  1 +
 shallow.c                   | 55 +++++++++++++++++++++++++++++++++++++++++++--
 t/t5304-prune.sh            | 10 +++++++++
 6 files changed, 71 insertions(+), 2 deletions(-)

diff --git a/Documentation/git-prune.txt b/Documentation/git-prune.txt
index bf82410..058ac0d 100644
--- a/Documentation/git-prune.txt
+++ b/Documentation/git-prune.txt
@@ -24,6 +24,8 @@ objects unreachable from any of these head objects from the object database.
 In addition, it
 prunes the unpacked objects that are also found in packs by
 running 'git prune-packed'.
+It also removes entries from .git/shallow that are not reachable by
+any ref.
 
 Note that unreachable, packed objects will remain.  If this is
 not desired, see linkgit:git-repack[1].
diff --git a/builtin/gc.c b/builtin/gc.c
index c14190f..cec8ecd 100644
--- a/builtin/gc.c
+++ b/builtin/gc.c
@@ -16,6 +16,7 @@
 #include "run-command.h"
 #include "sigchain.h"
 #include "argv-array.h"
+#include "commit.h"
 
 #define FAILED_RUN "failed to run %s"
 
diff --git a/builtin/prune.c b/builtin/prune.c
index 6366917..2214040 100644
--- a/builtin/prune.c
+++ b/builtin/prune.c
@@ -170,5 +170,9 @@ int cmd_prune(int argc, const char **argv, const char *prefix)
 	s = mkpathdup("%s/pack", get_object_directory());
 	remove_temporary_files(s);
 	free(s);
+
+	if (is_repository_shallow())
+		prune_shallow(show_only);
+
 	return 0;
 }
diff --git a/commit.h b/commit.h
index a1f2d49..affe210 100644
--- a/commit.h
+++ b/commit.h
@@ -235,6 +235,7 @@ extern void assign_shallow_commits_to_refs(struct shallow_info *info,
 					   uint32_t **used,
 					   int *ref_status);
 extern int delayed_reachability_test(struct shallow_info *si, int c);
+extern void prune_shallow(int show_only);
 
 int is_descendant_of(struct commit *, struct commit_list *);
 int in_merge_bases(struct commit *, struct commit *);
diff --git a/shallow.c b/shallow.c
index e611639..f2ab72c 100644
--- a/shallow.c
+++ b/shallow.c
@@ -155,10 +155,14 @@ void check_shallow_file_for_update(void)
 		die("shallow file was changed during fetch");
 }
 
+#define SEEN_ONLY 1
+#define VERBOSE   2
+
 struct write_shallow_data {
 	struct strbuf *out;
 	int use_pack_protocol;
 	int count;
+	unsigned flags;
 };
 
 static int write_one_shallow(const struct commit_graft *graft, void *cb_data)
@@ -167,6 +171,15 @@ static int write_one_shallow(const struct commit_graft *graft, void *cb_data)
 	const char *hex = sha1_to_hex(graft->sha1);
 	if (graft->nr_parent != -1)
 		return 0;
+	if (data->flags & SEEN_ONLY) {
+		struct commit *c = lookup_commit(graft->sha1);
+		if (!c || !(c->object.flags & SEEN)) {
+			if (data->flags & VERBOSE)
+				printf("Removing %s from .git/shallow\n",
+				       sha1_to_hex(c->object.sha1));
+			return 0;
+		}
+	}
 	data->count++;
 	if (data->use_pack_protocol)
 		packet_buf_write(data->out, "shallow %s", hex);
@@ -177,14 +190,16 @@ static int write_one_shallow(const struct commit_graft *graft, void *cb_data)
 	return 0;
 }
 
-int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
-			  const struct sha1_array *extra)
+static int write_shallow_commits_1(struct strbuf *out, int use_pack_protocol,
+				   const struct sha1_array *extra,
+				   unsigned flags)
 {
 	struct write_shallow_data data;
 	int i;
 	data.out = out;
 	data.use_pack_protocol = use_pack_protocol;
 	data.count = 0;
+	data.flags = flags;
 	for_each_commit_graft(write_one_shallow, &data);
 	if (!extra)
 		return data.count;
@@ -196,6 +211,12 @@ int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
 	return data.count;
 }
 
+int write_shallow_commits(struct strbuf *out, int use_pack_protocol,
+			  const struct sha1_array *extra)
+{
+	return write_shallow_commits_1(out, use_pack_protocol, extra, 0);
+}
+
 char *setup_temporary_shallow(const struct sha1_array *extra)
 {
 	struct strbuf sb = STRBUF_INIT;
@@ -258,6 +279,36 @@ void advertise_shallow_grafts(int fd)
 	for_each_commit_graft(advertise_shallow_grafts_cb, &fd);
 }
 
+/*
+ * mark_reachable_objects() should have been run prior to this and all
+ * reachable commits marked as "SEEN".
+ */
+void prune_shallow(int show_only)
+{
+	static struct lock_file shallow_lock;
+	struct strbuf sb = STRBUF_INIT;
+	int fd;
+
+	if (show_only) {
+		write_shallow_commits_1(&sb, 0, NULL, SEEN_ONLY | VERBOSE);
+		strbuf_release(&sb);
+		return;
+	}
+	check_shallow_file_for_update();
+	fd = hold_lock_file_for_update(&shallow_lock, git_path("shallow"),
+				       LOCK_DIE_ON_ERROR);
+	if (write_shallow_commits_1(&sb, 0, NULL, SEEN_ONLY)) {
+		if (write_in_full(fd, sb.buf, sb.len) != sb.len)
+			die_errno("failed to write to %s",
+				  shallow_lock.filename);
+		commit_lock_file(&shallow_lock);
+	} else {
+		unlink(git_path("shallow"));
+		rollback_lock_file(&shallow_lock);
+	}
+	strbuf_release(&sb);
+}
+
 #define TRACE_KEY "GIT_TRACE_SHALLOW"
 
 /*
diff --git a/t/t5304-prune.sh b/t/t5304-prune.sh
index e4bb3a1..66c9a41 100755
--- a/t/t5304-prune.sh
+++ b/t/t5304-prune.sh
@@ -221,4 +221,14 @@ EOF
 	test_cmp expected actual
 '
 
+test_expect_success 'prune .git/shallow' '
+	SHA1=`echo hi|git commit-tree HEAD^{tree}` &&
+	echo $SHA1 >.git/shallow &&
+	git prune --dry-run >out &&
+	grep $SHA1 .git/shallow &&
+	grep $SHA1 out &&
+	git prune &&
+	! test -f .git/shallow
+'
+
 test_done
-- 
1.8.5.1.25.g8667982

[PATCH v4 28/28] git-clone.txt: remove shallow clone limitations

[Nguyễn Thái Ngọc Duy]

Now that git supports data transfer from or to a shallow clone, these
limitations are not true anymore.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 Documentation/git-clone.txt | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/Documentation/git-clone.txt b/Documentation/git-clone.txt
index 450f158..4987857 100644
--- a/Documentation/git-clone.txt
+++ b/Documentation/git-clone.txt
@@ -181,12 +181,7 @@ objects from the source repository into a pack in the cloned repository.
 
 --depth <depth>::
 	Create a 'shallow' clone with a history truncated to the
-	specified number of revisions.  A shallow repository has a
-	number of limitations (you cannot clone or fetch from
-	it, nor push from nor into it), but is adequate if you
-	are only interested in the recent history of a large project
-	with a long history, and would want to send in fixes
-	as patches.
+	specified number of revisions.
 
 --[no-]single-branch::
 	Clone only the history leading to the tip of a single branch,
-- 
1.8.5.1.25.g8667982

Re: [PATCH v4 23/28] Support shallow fetch/clone over smart-http

[Jeff King]

On Thu, Dec 05, 2013 at 08:02:50PM +0700, Nguyễn Thái Ngọc Duy wrote:

> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  Documentation/gitremote-helpers.txt |  7 +++++++
>  builtin/fetch-pack.c                | 16 +++++++++++++---
>  remote-curl.c                       | 31 +++++++++++++++++++++++++++++--
>  t/t5536-fetch-shallow.sh            | 27 +++++++++++++++++++++++++++

I'm getting test failures in 'next' with GIT_TEST_HTTPD set, and they
are bisectable to this patch (actually, the moral equivalent of it, as
it looks like the commit message was tweaked along with the test number
when it was applied). The failures look like this:

  $ GIT_TEST_HTTPD=1 ./t5537-fetch-shallow.sh -v -i
  [...]
  ok 9 - fetch --update-shallow
  
  expecting success: 
          git clone --bare --no-local shallow "$HTTPD_DOCUMENT_ROOT_PATH/repo.git" &&
          git clone $HTTPD_URL/smart/repo.git clone &&
          (
          cd clone &&
          git fsck &&
          git log --format=%s origin/master >actual &&
          cat <<EOF >expect &&
  6
  5
  4
  3
  EOF
          test_cmp expect actual
          )
  
  Cloning into bare repository '/home/peff/compile/git/t/trash directory.t5537-fetch-shallow/httpd/www/repo.git'...
  remote: Counting objects: 19, done.        
  remote: Compressing objects: 100% (7/7), done.        
  remote: Total 19 (delta 0), reused 6 (delta 0)        
  Receiving objects: 100% (19/19), done.
  Checking connectivity... done.
  Cloning into 'clone'...
  remote: Counting objects: 19, done.        
  remote: Compressing objects: 100% (7/7), done.        
  remote: Total 19 (delta 0), reused 19 (delta 0)        
  Unpacking objects: 100% (19/19), done.
  Checking connectivity... done.
  Checking object directories: 100% (256/256), done.
  --- expect      2014-01-08 11:20:20.178546452 +0000
  +++ actual      2014-01-08 11:20:20.178546452 +0000
  @@ -1,3 +1,4 @@
  +7
   6
   5
   4
  not ok 10 - clone http repository


If you do end up tweaking the test, you may also want to fix:

> +LIB_HTTPD_PORT=${LIB_HTTPD_PORT-'5536'}
> +. "$TEST_DIRECTORY"/lib-httpd.sh

Since the test number got switched, it would be nice to tweak the port
number to match it, in case the real t5536 ever starts using lib-httpd.

-Peff

[PATCH] t5537: fix incorrect expectation in test case 10

[Nguyễn Thái Ngọc Duy]

Commit 48d25ca adds a new commit "7" to the repo that the next test case
in commit 1609488 clones from. But the next test case does not expect
this commit. For these tests, it's the bottom that's important, not
the top. Fix the expected commit list.

While at it, fix the default http port number to 5537. Otherwise when
t5536 learns to test httpd, running test in parallel may fail.

References:

48d25ca fetch: add --update-shallow to accept... - 2013-12-05
1609488 smart-http: support shallow fetch/clone - 2013-12-05

Noticed-by: Jeff King <peff@peff.net>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 I obviously made a mistake with patch reordering or something. And
 embarassing because I did not run tests with GIT_TEST_HTTPD again
 before submission.

 t/t5537-fetch-shallow.sh | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/t/t5537-fetch-shallow.sh b/t/t5537-fetch-shallow.sh
index 79ce472..b0fa738 100755
--- a/t/t5537-fetch-shallow.sh
+++ b/t/t5537-fetch-shallow.sh
@@ -178,7 +178,7 @@ if test -n "$NO_CURL" -o -z "$GIT_TEST_HTTPD"; then
 	test_done
 fi
 
-LIB_HTTPD_PORT=${LIB_HTTPD_PORT-'5536'}
+LIB_HTTPD_PORT=${LIB_HTTPD_PORT-'5537'}
 . "$TEST_DIRECTORY"/lib-httpd.sh
 start_httpd
 
@@ -190,6 +190,7 @@ test_expect_success 'clone http repository' '
 	git fsck &&
 	git log --format=%s origin/master >actual &&
 	cat <<EOF >expect &&
+7
 6
 5
 4
-- 
1.8.5.2.240.g8478abd

Re: [PATCH] t5537: fix incorrect expectation in test case 10

[Jeff King]

On Wed, Jan 08, 2014 at 07:13:19PM +0700, Nguyễn Thái Ngọc Duy wrote:

> Commit 48d25ca adds a new commit "7" to the repo that the next test case
> in commit 1609488 clones from. But the next test case does not expect
> this commit. For these tests, it's the bottom that's important, not
> the top. Fix the expected commit list.

Given the test output, I had a feeling it was something like this but
didn't dive in (and figured it would be obvious to you).

Patch looks sane. Thanks for a quick turnaround.

-Peff