[PATCH v1 0/2] add additional config settings for merge

[PATCH v1 0/2] add additional config settings for merge

[Ben Peart]

This enables the user to set a couple of additional options for merge.

1. merge.aggressive - this is to try to resolve a few more trivial
   merge cases.  It is documented in read-tree and is not something you
   can pass into merge itself.

2. merge.renames - this is to save git from having to go through the entire
   3 trees to see if there were any renames that happened.

For the work item repro that I have been using this drops the merge time
from ~1 hour to ~5 minutes and the unmerged entries goes down from
~40,000 to 1.

Helped-by: Kevin Willford <kewillf@microsoft.com>
Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Ben Peart <Ben.Peart@microsoft.com>

Base Ref: master
Web-Diff: https://github.com/benpeart/git/commit/a3d157f5be
Checkout: git fetch https://github.com/benpeart/git merge-options-v1 && git checkout a3d157f5be

Ben Peart (2):
  merge: Add merge.renames config setting
  merge: Add merge.aggressive config setting

 Documentation/merge-config.txt | 9 +++++++++
 merge-recursive.c              | 2 ++
 2 files changed, 11 insertions(+)


base-commit: 0b0cc9f86731f894cff8dd25299a9b38c254569e
-- 
2.17.0.windows.1

[PATCH v1 1/2] merge: Add merge.renames config setting

[Ben Peart]

Add the ability to control rename detection for merge via a config setting.

Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Ben Peart <benpeart@microsoft.com>
---
 Documentation/merge-config.txt | 5 +++++
 merge-recursive.c              | 1 +
 2 files changed, 6 insertions(+)

diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
index 12b6bbf591..656f909eb3 100644
--- a/Documentation/merge-config.txt
+++ b/Documentation/merge-config.txt
@@ -37,6 +37,11 @@ merge.renameLimit::
 	during a merge; if not specified, defaults to the value of
 	diff.renameLimit.
 
+merge.renames::
+	Whether and how Git detects renames.  If set to "false",
+	rename detection is disabled. If set to "true", basic rename
+	detection is enabled. This is the default.
+
 merge.renormalize::
 	Tell Git that canonical representation of files in the
 	repository has changed over time (e.g. earlier commits record
diff --git a/merge-recursive.c b/merge-recursive.c
index 9c05eb7f70..cd5367e890 100644
--- a/merge-recursive.c
+++ b/merge-recursive.c
@@ -3256,6 +3256,7 @@ static void merge_recursive_config(struct merge_options *o)
 	git_config_get_int("merge.verbosity", &o->verbosity);
 	git_config_get_int("diff.renamelimit", &o->diff_rename_limit);
 	git_config_get_int("merge.renamelimit", &o->merge_rename_limit);
+	git_config_get_bool("merge.renames", &o->detect_rename);
 	git_config(git_xmerge_config, NULL);
 }
 
-- 
2.17.0.windows.1

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Elijah Newren]

On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
> --- a/Documentation/merge-config.txt
> +++ b/Documentation/merge-config.txt
> @@ -37,6 +37,11 @@ merge.renameLimit::
>         during a merge; if not specified, defaults to the value of
>         diff.renameLimit.
>
> +merge.renames::
> +       Whether and how Git detects renames.  If set to "false",
> +       rename detection is disabled. If set to "true", basic rename
> +       detection is enabled. This is the default.

One can already control o->detect_rename via the -Xno-renames and
-Xfind-renames options.  I think the documentation should mention that
"false" is the same as passing -Xno-renames, and "true" is the same as
passing -Xfind-renames.  However, find-renames does take similarity
threshold as a parameter, so there's a question whether this option
should provide some way to do the same.  I'm not sure the answer to
that; it may be that we'd want a separate config option for that, and
we can wait to add it until someone actually wants it.

>  merge.renormalize::
>         Tell Git that canonical representation of files in the
>         repository has changed over time (e.g. earlier commits record
> diff --git a/merge-recursive.c b/merge-recursive.c
> index 9c05eb7f70..cd5367e890 100644
> --- a/merge-recursive.c
> +++ b/merge-recursive.c
> @@ -3256,6 +3256,7 @@ static void merge_recursive_config(struct merge_options *o)
>         git_config_get_int("merge.verbosity", &o->verbosity);
>         git_config_get_int("diff.renamelimit", &o->diff_rename_limit);
>         git_config_get_int("merge.renamelimit", &o->merge_rename_limit);
> +       git_config_get_bool("merge.renames", &o->detect_rename);
>         git_config(git_xmerge_config, NULL);
>  }

I would expect an explicitly passed -Xno-renames or -Xfind-renames to
override the config setting.  Could you check if that's the case?

Also, if someone sets merge.renameLimit (to anything) and sets
merge.renames to false, then they've got a contradictory setup.  Does
it make sense to check and warn about that anywhere?

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Elijah Newren]

On Fri, Apr 20, 2018 at 10:02 AM, Elijah Newren <newren@gmail.com> wrote:
> On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
>> --- a/Documentation/merge-config.txt
>> +++ b/Documentation/merge-config.txt
>> @@ -37,6 +37,11 @@ merge.renameLimit::
>>         during a merge; if not specified, defaults to the value of
>>         diff.renameLimit.
>>
>> +merge.renames::
>> +       Whether and how Git detects renames.  If set to "false",
>> +       rename detection is disabled. If set to "true", basic rename
>> +       detection is enabled. This is the default.
>
> One can already control o->detect_rename via the -Xno-renames and
> -Xfind-renames options.  I think the documentation should mention that
> "false" is the same as passing -Xno-renames, and "true" is the same as
> passing -Xfind-renames.  However, find-renames does take similarity
> threshold as a parameter, so there's a question whether this option
> should provide some way to do the same.  I'm not sure the answer to
> that; it may be that we'd want a separate config option for that, and
> we can wait to add it until someone actually wants it.

I just realized another issue, though it also affects -Xno-renames.
Even if rename detection is turned off for the merge, it is
unconditionally turned on for the diffstat.  In builtin/merge.c,
function finish(), there is the code:

    if (new_head && show_diffstat) {
        ...
        opts.detect_rename = DIFF_DETECT_RENAME;

It seems that this option should affect that line as well.  (Do you
have diffstat turned off by chance?  If not, you may be able to
improve your performance even more...)

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Ben Peart]

On 4/20/2018 1:26 PM, Elijah Newren wrote:
> On Fri, Apr 20, 2018 at 10:02 AM, Elijah Newren <newren@gmail.com> wrote:
>> On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
>>> --- a/Documentation/merge-config.txt
>>> +++ b/Documentation/merge-config.txt
>>> @@ -37,6 +37,11 @@ merge.renameLimit::
>>>          during a merge; if not specified, defaults to the value of
>>>          diff.renameLimit.
>>>
>>> +merge.renames::
>>> +       Whether and how Git detects renames.  If set to "false",
>>> +       rename detection is disabled. If set to "true", basic rename
>>> +       detection is enabled. This is the default.
>>
>> One can already control o->detect_rename via the -Xno-renames and
>> -Xfind-renames options.  I think the documentation should mention that
>> "false" is the same as passing -Xno-renames, and "true" is the same as
>> passing -Xfind-renames.  However, find-renames does take similarity
>> threshold as a parameter, so there's a question whether this option
>> should provide some way to do the same.  I'm not sure the answer to
>> that; it may be that we'd want a separate config option for that, and
>> we can wait to add it until someone actually wants it.
> 
> I just realized another issue, though it also affects -Xno-renames.
> Even if rename detection is turned off for the merge, it is
> unconditionally turned on for the diffstat.  In builtin/merge.c,
> function finish(), there is the code:
> 
>      if (new_head && show_diffstat) {
>          ...
>          opts.detect_rename = DIFF_DETECT_RENAME;
> 
> It seems that this option should affect that line as well.  (Do you
> have diffstat turned off by chance?  If not, you may be able to
> improve your performance even more...)
> 

Seems reasonable to me.  I'll update the patch to do that.

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Ben Peart]

On 4/20/2018 1:02 PM, Elijah Newren wrote:
> On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
>> --- a/Documentation/merge-config.txt
>> +++ b/Documentation/merge-config.txt
>> @@ -37,6 +37,11 @@ merge.renameLimit::
>>          during a merge; if not specified, defaults to the value of
>>          diff.renameLimit.
>>
>> +merge.renames::
>> +       Whether and how Git detects renames.  If set to "false",
>> +       rename detection is disabled. If set to "true", basic rename
>> +       detection is enabled. This is the default.
> 
> One can already control o->detect_rename via the -Xno-renames and
> -Xfind-renames options.  

Yes, but that requires people to know they need to do that and then 
remember to pass it on the command line every time.  We've found that 
doesn't typically happen, we just get someone complaining about slow 
merges. :)

That is why we added them as config options which change the default. 
That way we can then set them on the repo and the default behavior gives 
them better performance.  They can still always override the config 
setting with the command line options.

I think the documentation should mention that
> "false" is the same as passing -Xno-renames, and "true" is the same as
> passing -Xfind-renames.  However, find-renames does take similarity
> threshold as a parameter, so there's a question whether this option
> should provide some way to do the same.  I'm not sure the answer to
> that; it may be that we'd want a separate config option for that, and
> we can wait to add it until someone actually wants it.

I'm of the opinion that we shouldn't bother adding features that we 
aren't sure someone will want/use.  If it comes up, we can certainly add 
it at a later date.

> 
>>   merge.renormalize::
>>          Tell Git that canonical representation of files in the
>>          repository has changed over time (e.g. earlier commits record
>> diff --git a/merge-recursive.c b/merge-recursive.c
>> index 9c05eb7f70..cd5367e890 100644
>> --- a/merge-recursive.c
>> +++ b/merge-recursive.c
>> @@ -3256,6 +3256,7 @@ static void merge_recursive_config(struct merge_options *o)
>>          git_config_get_int("merge.verbosity", &o->verbosity);
>>          git_config_get_int("diff.renamelimit", &o->diff_rename_limit);
>>          git_config_get_int("merge.renamelimit", &o->merge_rename_limit);
>> +       git_config_get_bool("merge.renames", &o->detect_rename);
>>          git_config(git_xmerge_config, NULL);
>>   }
> 
> I would expect an explicitly passed -Xno-renames or -Xfind-renames to
> override the config setting.  Could you check if that's the case?
> 

Yes, command line options override the config settings.  You can see 
that in the code where the call to init_merge_options() which loads the 
config settings is followed by parse_merge_opt() which loads the command 
line options.  I've also verified the behavior in the debugger (it's on 
by default in the code, the config setting turns it off, then the 
command line option turns it back on).

> Also, if someone sets merge.renameLimit (to anything) and sets
> merge.renames to false, then they've got a contradictory setup.  Does
> it make sense to check and warn about that anywhere?
> 

I don't think we need to.  The merge.renameLimit is only used if 
detect_rename it turned on no matter how that gets turned on (default, 
config setting, command line option) so there isn't really a change in 
behavior here.

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Elijah Newren]

Hi Ben,

On Fri, Apr 20, 2018 at 10:59 AM, Ben Peart <peartben@gmail.com> wrote:
>
> On 4/20/2018 1:02 PM, Elijah Newren wrote:
>>
>> On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com>
>> wrote:
>>>
>>> --- a/Documentation/merge-config.txt
>>> +++ b/Documentation/merge-config.txt
>>> @@ -37,6 +37,11 @@ merge.renameLimit::
>>>          during a merge; if not specified, defaults to the value of
>>>          diff.renameLimit.
>>>
>>> +merge.renames::
>>> +       Whether and how Git detects renames.  If set to "false",
>>> +       rename detection is disabled. If set to "true", basic rename
>>> +       detection is enabled. This is the default.
>>
>>
>> One can already control o->detect_rename via the -Xno-renames and
>> -Xfind-renames options.

This statement wasn't meant to be independent of the sentence that
followed it...

> Yes, but that requires people to know they need to do that and then remember
> to pass it on the command line every time.  We've found that doesn't
> typically happen, we just get someone complaining about slow merges. :)
>
> That is why we added them as config options which change the default. That
> way we can then set them on the repo and the default behavior gives them
> better performance.  They can still always override the config setting with
> the command line options.

Sorry, I think I wasn't being clear.  The documentation for the config
options for e.g. diff.renameLimit, fetch.prune, log.abbrevCommit, and
merge.ff all mention the equivalent command line parameters.  Your
patch doesn't do that for merge.renames, but I think it would be
helpful if it did.

Also, a link in the documentation the other way, from
Documentation/merge-strategies.txt under the entries for -Xno-renames
and -Xfind-renames should probably mention this new merge.renames
config setting (much like the -Xno-renormalize flag mentions the
merge.renomralize config option).

(In general, I think having this as a configuration option makes
sense, though I hope my other performance patches would be enough to
make people consider switching back to the defaults and use rename
detection again.)

<snip>
> I'm of the opinion that we shouldn't bother adding features that we aren't
> sure someone will want/use.  If it comes up, we can certainly add it at a
> later date.

Works for me; I was mostly throwing it out there for thought.

> Yes, command line options override the config settings.

Good.  :-)

>> Also, if someone sets merge.renameLimit (to anything) and sets
>> merge.renames to false, then they've got a contradictory setup.  Does
>> it make sense to check and warn about that anywhere?
>
> I don't think we need to.  The merge.renameLimit is only used if
> detect_rename it turned on no matter how that gets turned on (default,
> config setting, command line option) so there isn't really a change in
> behavior here.

I agree that's the pre-existing behavior, but prior to this patch
turning off rename detection could only be done manually with every
invocation.  I'm slightly concerned that users might be confused if
merge.renames was set to false somewhere -- perhaps even in a global
/etc/gitconfig that they had no knowledge of or control over -- and in
an attempt to get rename detection to work they started passing larger
and larger values for renameLimit all to no avail.

The easy fix here may just be documenting the diff.renameLimit and
merge.renameLimit options that they have no effect if rename detection
is turned off.

Or maybe I'm just worrying too much, but we (folks at $dayjob) were
bit pretty hard by renameLimit silently being capped at a value less
than the user specified and in a way that wasn't documented anywhere.

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Junio C Hamano]

Elijah Newren <newren@gmail.com> writes:

>>>> +merge.renames::
>>>> +       Whether and how Git detects renames.  If set to "false",
>>>> +       rename detection is disabled. If set to "true", basic rename
>>>> +       detection is enabled. This is the default.
>>>
>>>
>>> One can already control o->detect_rename via the -Xno-renames and
>>> -Xfind-renames options.
> ...
> Sorry, I think I wasn't being clear.  The documentation for the config
> options for e.g. diff.renameLimit, fetch.prune, log.abbrevCommit, and
> merge.ff all mention the equivalent command line parameters.  Your
> patch doesn't do that for merge.renames, but I think it would be
> helpful if it did.

Yes, and if we are adding a new configuration, we should do so in
such a way that we do not have to come back and extend it when we
know what the command line option does and the configuration being
proposed is less capable already.  I wonder if we can just add a
single configuration whose value can be "never" to pretend as if
"--Xno-renames" were given, and some similarity score like "50" to
pretend as if "--Xfind-renames=50" were given.  

That is, merge.renames does not have to a simple "yes-no to control
the --Xno-renames option".  And it would of course be better to
document it.

I also had to wonder how "merge -s resolve" faired, if the project
is not interested in renamed paths at all.

Thanks.

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Ben Peart]

On 4/21/2018 12:23 AM, Junio C Hamano wrote:
> Elijah Newren <newren@gmail.com> writes:
> 
>>>>> +merge.renames::
>>>>> +       Whether and how Git detects renames.  If set to "false",
>>>>> +       rename detection is disabled. If set to "true", basic rename
>>>>> +       detection is enabled. This is the default.
>>>>
>>>>
>>>> One can already control o->detect_rename via the -Xno-renames and
>>>> -Xfind-renames options.
>> ...
>> Sorry, I think I wasn't being clear.  The documentation for the config
>> options for e.g. diff.renameLimit, fetch.prune, log.abbrevCommit, and
>> merge.ff all mention the equivalent command line parameters.  Your
>> patch doesn't do that for merge.renames, but I think it would be
>> helpful if it did.
> 
> Yes, and if we are adding a new configuration, we should do so in
> such a way that we do not have to come back and extend it when we
> know what the command line option does and the configuration being
> proposed is less capable already.

Between all the different command line options, config settings, merge 
strategies and the interactions between the diff and merge versions, I
was trying to keep things as simple and consistent as possible.  To that 
end 'merge.renames' was modeled after the existing 'diff.renames.'

I wonder if we can just add a
> single configuration whose value can be "never" to pretend as if
> "--Xno-renames" were given, and some similarity score like "50" to
> pretend as if "--Xfind-renames=50" were given.
> 
> That is, merge.renames does not have to a simple "yes-no to control
> the --Xno-renames option".  And it would of course be better to
> document it.
> 

With the existing differences in how these options are passed on the 
command line, I'm hesitant to add yet another pattern in the config 
settings that combines 'renames' and '--find-renames[=<n>]'.

I _have_ wondered why this all isn't configured via find-renames with 
find-renames=0 meaning renames=false (instead of mapping 0 to 32K).  I 
think that could have eliminated the need for splitting rename across 
two different settings (which is what I think you are proposing above). 
I'd then want the config setting and command line option to be the same 
syntax and behavior.

Moving the existing settings to this model and updating the config and 
command line options to be consistent without breaking backwards 
compatibility is outside the intended scope of this patch.

> I also had to wonder how "merge -s resolve" faired, if the project
> is not interested in renamed paths at all.
> 

To be clear, it isn't that we're not interested in detecting renamed 
files and paths.  We're just opposed to it taking an hour to figure that 
out!

> Thanks.
> 

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Junio C Hamano]

Ben Peart <peartben@gmail.com> writes:

>> I also had to wonder how "merge -s resolve" faired, if the project
>> is not interested in renamed paths at all.
>>
>
> To be clear, it isn't that we're not interested in detecting renamed
> files and paths.  We're just opposed to it taking an hour to figure
> that out!

Yeah, but as opposed to passing "oh, let's see if we can get a
reasonable result without rename detection just this time" from the
command line, configuring merge.renames=false in would mean exactly
that: "we don't need rename detection, just want to skip the cycles
spent for it".  That is why I wondered how well the resolve strategy
would have fit your needs.

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Johannes Schindelin]

Hi Junio,

On Tue, 24 Apr 2018, Junio C Hamano wrote:

> Ben Peart <peartben@gmail.com> writes:
> 
> >> I also had to wonder how "merge -s resolve" faired, if the project
> >> is not interested in renamed paths at all.
> >>
> >
> > To be clear, it isn't that we're not interested in detecting renamed
> > files and paths.  We're just opposed to it taking an hour to figure
> > that out!
> 
> Yeah, but as opposed to passing "oh, let's see if we can get a
> reasonable result without rename detection just this time" from the
> command line, configuring merge.renames=false in would mean exactly
> that: "we don't need rename detection, just want to skip the cycles
> spent for it".  That is why I wondered how well the resolve strategy
> would have fit your needs.

Please do not forget that the context is GVFS, where you would cause a lot
of pain and suffering by letting users forget to specify that command-line
option all the time, resulting in several gigabytes of objects having to
be downloaded just for the sake of rename detection.

So there is a pretty good point in doing this as a config option.

Ciao,
Dscho

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Elijah Newren]

Hi Dscho,

On Tue, Apr 24, 2018 at 4:58 AM, Johannes Schindelin
<Johannes.Schindelin@gmx.de> wrote:
> Hi Junio,
>
> On Tue, 24 Apr 2018, Junio C Hamano wrote:
>
>> Yeah, but as opposed to passing "oh, let's see if we can get a
>> reasonable result without rename detection just this time" from the
>> command line, configuring merge.renames=false in would mean exactly
>> that: "we don't need rename detection, just want to skip the cycles
>> spent for it".  That is why I wondered how well the resolve strategy
>> would have fit your needs.
>
> Please do not forget that the context is GVFS, where you would cause a lot
> of pain and suffering by letting users forget to specify that command-line
> option all the time, resulting in several gigabytes of objects having to
> be downloaded just for the sake of rename detection.
>
> So there is a pretty good point in doing this as a config option.

I agree you need a config option, but I think Junio has a good point
that it's worth at least checking out the possibility of a different
one.  In particular, you could add a merge.defaultStrategy (or maybe
merge.twohead to be similar to pull.twohead??) that is set to
'resolve', and use that to avoid rename detection.

Perhaps performance considerations rule out the resolve strategy and
favor recursive, or maybe you need the 'recursive' part of the
recursive strategy (rather than the rename part), or perhaps there's
some other special reason you need to go this route, but since you are
avoiding renames right now it's at least worth considering the resolve
strategy.

Elijah

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Johannes Schindelin]

Hi Elijah,

On Tue, 24 Apr 2018, Elijah Newren wrote:

> On Tue, Apr 24, 2018 at 4:58 AM, Johannes Schindelin
> <Johannes.Schindelin@gmx.de> wrote:
> >
> > On Tue, 24 Apr 2018, Junio C Hamano wrote:
> >
> >> Yeah, but as opposed to passing "oh, let's see if we can get a
> >> reasonable result without rename detection just this time" from the
> >> command line, configuring merge.renames=false in would mean exactly
> >> that: "we don't need rename detection, just want to skip the cycles
> >> spent for it".  That is why I wondered how well the resolve strategy
> >> would have fit your needs.
> >
> > Please do not forget that the context is GVFS, where you would cause a lot
> > of pain and suffering by letting users forget to specify that command-line
> > option all the time, resulting in several gigabytes of objects having to
> > be downloaded just for the sake of rename detection.
> >
> > So there is a pretty good point in doing this as a config option.
> 
> I agree you need a config option, but I think Junio has a good point
> that it's worth at least checking out the possibility of a different
> one.  In particular, you could add a merge.defaultStrategy (or maybe
> merge.twohead to be similar to pull.twohead??) that is set to
> 'resolve', and use that to avoid rename detection.
> 
> Perhaps performance considerations rule out the resolve strategy and
> favor recursive, or maybe you need the 'recursive' part of the
> recursive strategy (rather than the rename part), or perhaps there's
> some other special reason you need to go this route, but since you are
> avoiding renames right now it's at least worth considering the resolve
> strategy.

I would really hesitate to go to a different merge strategy. The recursive
strategy really has the best track record in general, and we have to use
all kinds of branching models (including heavily criss-crossed ones) with
GVFS Git.

Ciao,
Dscho

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Eckhard Maaß]

On Fri, Apr 20, 2018 at 11:34:25AM -0700, Elijah Newren wrote:
> Sorry, I think I wasn't being clear.  The documentation for the config
> options for e.g. diff.renameLimit, fetch.prune, log.abbrevCommit, and
> merge.ff all mention the equivalent command line parameters.  Your
> patch doesn't do that for merge.renames, but I think it would be
> helpful if it did.

I wonder here what the relation to the diff.* options should be in this
regard anyway. There is already diff.renames. Naively, I would assume
that these options are in sync, that is you control the behavior of both
the normal diff family like git show and git merge. The reasoning, at
least for me, is to keep consistency between the outcome of rename
detection while merging and a later simple "git show MERGE_BASE..HEAD".
I would expect those to give me the same style of rename detection.

Hence, I would like to use diff.renames and maybe enhance this option to
also carry the score in backward compatible way (or introduce a second
configuration option?). Is this idea going in a good direction? If yes,
I will try to submit a patch for this.

Ah, by the way: for people that have not touched diff.renames there will
be no visible change in how Git behaves - the default for diff.renames
is a rename with 50% score with is the same for merge. So it will only
change if one has tweaked diff.renames already. But I wonder if one does
that and expect the merge to use a different rename detection anyway.

Greetings,
Eckhard

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Ben Peart]

On 4/22/2018 8:07 AM, Eckhard Maaß wrote:
> On Fri, Apr 20, 2018 at 11:34:25AM -0700, Elijah Newren wrote:
>> Sorry, I think I wasn't being clear.  The documentation for the config
>> options for e.g. diff.renameLimit, fetch.prune, log.abbrevCommit, and
>> merge.ff all mention the equivalent command line parameters.  Your
>> patch doesn't do that for merge.renames, but I think it would be
>> helpful if it did.
> 
> I wonder here what the relation to the diff.* options should be in this
> regard anyway. There is already diff.renames. Naively, I would assume
> that these options are in sync, that is you control the behavior of both
> the normal diff family like git show and git merge. The reasoning, at
> least for me, is to keep consistency between the outcome of rename
> detection while merging and a later simple "git show MERGE_BASE..HEAD".
> I would expect those to give me the same style of rename detection.
> 
> Hence, I would like to use diff.renames and maybe enhance this option to
> also carry the score in backward compatible way (or introduce a second
> configuration option?). Is this idea going in a good direction? If yes,
> I will try to submit a patch for this.

It's a fair question.  If you look at all the options in 
Documentation/merge-config.txt, you will see many merge specific 
settings.  I think the ability to control these settings separately is 
pretty well established.

In commit 2a2ac926547 when merge.renamelimit was added, it was decided 
to have separate settings for merge and diff to give users the ability 
to control that behavior.  In this particular case, it will default to 
the value of diff.renamelimit when it isn't set.  That isn't consistent 
with the other merge settings.

Changing that behavior across the rest of the merge settings is outside 
the scope of this patch.  I don't have a strong opinion as to whether 
that is a good or bad thing.

> 
> Ah, by the way: for people that have not touched diff.renames there will
> be no visible change in how Git behaves - the default for diff.renames
> is a rename with 50% score with is the same for merge. So it will only
> change if one has tweaked diff.renames already. But I wonder if one does
> that and expect the merge to use a different rename detection anyway.
> 
> Greetings,
> Eckhard
> 

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Eckhard Maaß]

On Mon, Apr 23, 2018 at 09:15:09AM -0400, Ben Peart wrote:
> In commit 2a2ac926547 when merge.renamelimit was added, it was decided to
> have separate settings for merge and diff to give users the ability to
> control that behavior.  In this particular case, it will default to the
> value of diff.renamelimit when it isn't set.  That isn't consistent with the
> other merge settings.

However, it seems like a desirable way to do it.

Maybe let me throw in some code for discussion (test and documentation
is missing, mainly to form an idea what the change in options should
be). I admit the patch below is concerned only with diff.renames, but
whatever we come up with for merge should be reflected there, too,
doesn't it?

Greetings,
Eckhard

-- >8 --

From e8a88111f2aaf338a4c19e83251c7178f7152129 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Eckhard=20S=2E=20Maa=C3=9F?= <eckhard.s.maass@gmail.com>
Date: Sun, 22 Apr 2018 23:29:08 +0200
Subject: [PATCH] diff: enhance diff.renames to be able to set rename score
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Signed-off-by: Eckhard S. Maaß <eckhard.s.maass@gmail.com>
---
 diff.c | 35 ++++++++++++++++++++++++++++-------
 1 file changed, 28 insertions(+), 7 deletions(-)

diff --git a/diff.c b/diff.c
index 1289df4b1f..a3cedad5cf 100644
--- a/diff.c
+++ b/diff.c
@@ -30,6 +30,7 @@
 #endif
 
 static int diff_detect_rename_default;
+static int diff_rename_score_default;
 static int diff_indent_heuristic = 1;
 static int diff_rename_limit_default = 400;
 static int diff_suppress_blank_empty;
@@ -177,13 +178,33 @@ static int parse_submodule_params(struct diff_options *options, const char *valu
 	return 0;
 }
 
+int parse_rename_score(const char **cp_p);
+
+static int git_config_rename_score(const char *value)
+{
+	int parsed_rename_score = parse_rename_score(&value);
+	if (parsed_rename_score == -1)
+		return error("invalid argument to diff.renamescore: %s", value);
+	diff_rename_score_default = parsed_rename_score;
+	return 0;
+}
+
 static int git_config_rename(const char *var, const char *value)
 {
-	if (!value)
-		return DIFF_DETECT_RENAME;
-	if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
-		return  DIFF_DETECT_COPY;
-	return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
+	if (!value) {
+		diff_detect_rename_default = DIFF_DETECT_RENAME;
+		return 0;
+	}
+	if (skip_to_optional_arg(value, "copies", &value) || skip_to_optional_arg(value, "copy", &value)) {
+		diff_detect_rename_default = DIFF_DETECT_COPY;
+		return git_config_rename_score(value);
+	}
+	if (skip_to_optional_arg(value, "renames", &value) || skip_to_optional_arg(value, "rename", &value)) {
+		diff_detect_rename_default = DIFF_DETECT_RENAME;
+		return git_config_rename_score(value);
+	}
+	diff_detect_rename_default = git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
+	return 0;
 }
 
 long parse_algorithm_value(const char *value)
@@ -307,8 +328,7 @@ int git_diff_ui_config(const char *var, const char *value, void *cb)
 		return 0;
 	}
 	if (!strcmp(var, "diff.renames")) {
-		diff_detect_rename_default = git_config_rename(var, value);
-		return 0;
+		return git_config_rename(var, value);
 	}
 	if (!strcmp(var, "diff.autorefreshindex")) {
 		diff_auto_refresh_index = git_config_bool(var, value);
@@ -4116,6 +4136,7 @@ void diff_setup(struct diff_options *options)
 	options->add_remove = diff_addremove;
 	options->use_color = diff_use_color_default;
 	options->detect_rename = diff_detect_rename_default;
+	options->rename_score = diff_rename_score_default;
 	options->xdl_opts |= diff_algorithm;
 	if (diff_indent_heuristic)
 		DIFF_XDL_SET(options, INDENT_HEURISTIC);
-- 
2.17.0.252.gfe0a9eaf31

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Ben Peart]

On 4/23/2018 5:32 PM, Eckhard Maaß wrote:
> On Mon, Apr 23, 2018 at 09:15:09AM -0400, Ben Peart wrote:
>> In commit 2a2ac926547 when merge.renamelimit was added, it was decided to
>> have separate settings for merge and diff to give users the ability to
>> control that behavior.  In this particular case, it will default to the
>> value of diff.renamelimit when it isn't set.  That isn't consistent with the
>> other merge settings.
> 
> However, it seems like a desirable way to do it.

I'm just one opinion among many but I personally believe the cascading 
settings are complicated enough just with the various config files and 
command line options and which overwrite the others.  I'd rather not 
complicate them further by having settings inherited from one feature 
(diff) to another (merge).

There are currently ~15 merge specific config settings and only 
merge.renamelimit currently does this inheritance.  That said, at least 
one other person thought it was a good idea. :)

> 
> Maybe let me throw in some code for discussion (test and documentation
> is missing, mainly to form an idea what the change in options should
> be). I admit the patch below is concerned only with diff.renames, but
> whatever we come up with for merge should be reflected there, too,
> doesn't it >
> Greetings,
> Eckhard
> 
> -- >8 --
> 
>  From e8a88111f2aaf338a4c19e83251c7178f7152129 Mon Sep 17 00:00:00 2001
> From: =?UTF-8?q?Eckhard=20S=2E=20Maa=C3=9F?= <eckhard.s.maass@gmail.com>
> Date: Sun, 22 Apr 2018 23:29:08 +0200
> Subject: [PATCH] diff: enhance diff.renames to be able to set rename score
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
> 
> Signed-off-by: Eckhard S. Maaß <eckhard.s.maass@gmail.com>
> ---
>   diff.c | 35 ++++++++++++++++++++++++++++-------
>   1 file changed, 28 insertions(+), 7 deletions(-)
> 
> diff --git a/diff.c b/diff.c
> index 1289df4b1f..a3cedad5cf 100644
> --- a/diff.c
> +++ b/diff.c
> @@ -30,6 +30,7 @@
>   #endif
>   
>   static int diff_detect_rename_default;
> +static int diff_rename_score_default;
>   static int diff_indent_heuristic = 1;
>   static int diff_rename_limit_default = 400;
>   static int diff_suppress_blank_empty;
> @@ -177,13 +178,33 @@ static int parse_submodule_params(struct diff_options *options, const char *valu
>   	return 0;
>   }
>   
> +int parse_rename_score(const char **cp_p);
> +
> +static int git_config_rename_score(const char *value)
> +{
> +	int parsed_rename_score = parse_rename_score(&value);
> +	if (parsed_rename_score == -1)
> +		return error("invalid argument to diff.renamescore: %s", value);
> +	diff_rename_score_default = parsed_rename_score;
> +	return 0;
> +}
> +
>   static int git_config_rename(const char *var, const char *value)
>   {
> -	if (!value)
> -		return DIFF_DETECT_RENAME;
> -	if (!strcasecmp(value, "copies") || !strcasecmp(value, "copy"))
> -		return  DIFF_DETECT_COPY;
> -	return git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
> +	if (!value) {
> +		diff_detect_rename_default = DIFF_DETECT_RENAME;
> +		return 0;
> +	}
> +	if (skip_to_optional_arg(value, "copies", &value) || skip_to_optional_arg(value, "copy", &value)) {
> +		diff_detect_rename_default = DIFF_DETECT_COPY;
> +		return git_config_rename_score(value);
> +	}
> +	if (skip_to_optional_arg(value, "renames", &value) || skip_to_optional_arg(value, "rename", &value)) {
> +		diff_detect_rename_default = DIFF_DETECT_RENAME;
> +		return git_config_rename_score(value);
> +	}
> +	diff_detect_rename_default = git_config_bool(var,value) ? DIFF_DETECT_RENAME : 0;
> +	return 0;
>   }
>   
>   long parse_algorithm_value(const char *value)
> @@ -307,8 +328,7 @@ int git_diff_ui_config(const char *var, const char *value, void *cb)
>   		return 0;
>   	}
>   	if (!strcmp(var, "diff.renames")) {
> -		diff_detect_rename_default = git_config_rename(var, value);
> -		return 0;
> +		return git_config_rename(var, value);
>   	}
>   	if (!strcmp(var, "diff.autorefreshindex")) {
>   		diff_auto_refresh_index = git_config_bool(var, value);
> @@ -4116,6 +4136,7 @@ void diff_setup(struct diff_options *options)
>   	options->add_remove = diff_addremove;
>   	options->use_color = diff_use_color_default;
>   	options->detect_rename = diff_detect_rename_default;
> +	options->rename_score = diff_rename_score_default;
>   	options->xdl_opts |= diff_algorithm;
>   	if (diff_indent_heuristic)
>   		DIFF_XDL_SET(options, INDENT_HEURISTIC);
> 

Re: [PATCH v1 1/2] merge: Add merge.renames config setting

[Ben Peart]

On 4/20/2018 2:34 PM, Elijah Newren wrote:
> Hi Ben,
> 
> On Fri, Apr 20, 2018 at 10:59 AM, Ben Peart <peartben@gmail.com> wrote:
>>
>> On 4/20/2018 1:02 PM, Elijah Newren wrote:
>>>
>>> On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com>
>>> wrote:
>>>>
>>>> --- a/Documentation/merge-config.txt
>>>> +++ b/Documentation/merge-config.txt
>>>> @@ -37,6 +37,11 @@ merge.renameLimit::
>>>>           during a merge; if not specified, defaults to the value of
>>>>           diff.renameLimit.
>>>>
>>>> +merge.renames::
>>>> +       Whether and how Git detects renames.  If set to "false",
>>>> +       rename detection is disabled. If set to "true", basic rename
>>>> +       detection is enabled. This is the default.
>>>
>>>
>>> One can already control o->detect_rename via the -Xno-renames and
>>> -Xfind-renames options.
> 
> This statement wasn't meant to be independent of the sentence that
> followed it...
> 
>> Yes, but that requires people to know they need to do that and then remember
>> to pass it on the command line every time.  We've found that doesn't
>> typically happen, we just get someone complaining about slow merges. :)
>>
>> That is why we added them as config options which change the default. That
>> way we can then set them on the repo and the default behavior gives them
>> better performance.  They can still always override the config setting with
>> the command line options.
> 
> Sorry, I think I wasn't being clear.  The documentation for the config
> options for e.g. diff.renameLimit, fetch.prune, log.abbrevCommit, and
> merge.ff all mention the equivalent command line parameters.  Your
> patch doesn't do that for merge.renames, but I think it would be
> helpful if it did.
> 
> Also, a link in the documentation the other way, from
> Documentation/merge-strategies.txt under the entries for -Xno-renames
> and -Xfind-renames should probably mention this new merge.renames
> config setting (much like the -Xno-renormalize flag mentions the
> merge.renomralize config option).
> 

I'm all in favor of having more information in the documentation.  I'm 
of the opinion that if someone has made the effort to actually _read_ 
the documentation, we should be as descriptive and complete as possible.

I'll take a cut at adding the things you have pointed out would be helpful.

> I agree that's the pre-existing behavior, but prior to this patch
> turning off rename detection could only be done manually with every
> invocation.  I'm slightly concerned that users might be confused if
> merge.renames was set to false somewhere -- perhaps even in a global
> /etc/gitconfig that they had no knowledge of or control over -- and in
> an attempt to get rename detection to work they started passing larger
> and larger values for renameLimit all to no avail.
> 
> The easy fix here may just be documenting the diff.renameLimit and
> merge.renameLimit options that they have no effect if rename detection
> is turned off.

I can add this additional documentation as well.  While some might think 
it is stating the obvious, I'm sure someone will benefit from it being 
explicitly called out.

> 
> Or maybe I'm just worrying too much, but we (folks at $dayjob) were
> bit pretty hard by renameLimit silently being capped at a value less
> than the user specified and in a way that wasn't documented anywhere.
> 

[PATCH v1 2/2] merge: Add merge.aggressive config setting

[Ben Peart]

Add the ability to control the aggressive flag passed to read-tree via a config setting.

Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Ben Peart <benpeart@microsoft.com>
---
 Documentation/merge-config.txt | 4 ++++
 merge-recursive.c              | 1 +
 2 files changed, 5 insertions(+)

diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
index 656f909eb3..5a9ab969db 100644
--- a/Documentation/merge-config.txt
+++ b/Documentation/merge-config.txt
@@ -1,3 +1,7 @@
+merge.aggressive::
+	Passes "aggressive" to read-tree which makes the command resolve
+	a few more cases internally. See "--aggressive" in linkgit:git-read-tree[1].
+
 merge.conflictStyle::
 	Specify the style in which conflicted hunks are written out to
 	working tree files upon merge.  The default is "merge", which
diff --git a/merge-recursive.c b/merge-recursive.c
index cd5367e890..0ca84e4b82 100644
--- a/merge-recursive.c
+++ b/merge-recursive.c
@@ -355,6 +355,7 @@ static int git_merge_trees(struct merge_options *o,
 	o->unpack_opts.fn = threeway_merge;
 	o->unpack_opts.src_index = &the_index;
 	o->unpack_opts.dst_index = &the_index;
+	git_config_get_bool("merge.aggressive", (int *)&o->unpack_opts.aggressive);
 	setup_unpack_trees_porcelain(&o->unpack_opts, "merge");
 
 	init_tree_desc_from_tree(t+0, common);
-- 
2.17.0.windows.1

Re: [PATCH v1 2/2] merge: Add merge.aggressive config setting

[Elijah Newren]

On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
> Add the ability to control the aggressive flag passed to read-tree via a config setting.

This feels like a workaround to the performance problems with index
updates in merge-recursive.c.  That said, it makes sense to me to do
this when rename detection is turned off.  In fact, I think you'd
automatically want to set aggressive to true whenever rename detection
is turned off (whether by your merge.renames option or the
-Xno-renames flag).

I can't think of any reason this setting would be useful separate from
turning rename detection off, and it'd actively harm rename detection
performance improvements I have in the pipeline.  I'd really prefer to
not add this option, and instead combine the setting of aggressive
with the other flag.  Do you have an independent reason for wanting
this?

Thanks,
Elijah

Re: [PATCH v1 2/2] merge: Add merge.aggressive config setting

[Ben Peart]

On 4/20/2018 1:22 PM, Elijah Newren wrote:
> On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
>> Add the ability to control the aggressive flag passed to read-tree via a config setting.
> 
> This feels like a workaround to the performance problems with index
> updates in merge-recursive.c.  

This change wasn't done to solve performance problems.  We turned it on 
because it reduced the number of unmerged entries (from 40K to 1) in the 
particular merge we were looking at.  The additional 3 scenarios that 
--aggressive resolves made that much difference.

That said, it makes sense to me to do
> this when rename detection is turned off.  In fact, I think you'd
> automatically want to set aggressive to true whenever rename detection
> is turned off (whether by your merge.renames option or the
> -Xno-renames flag).
> > I can't think of any reason this setting would be useful separate from
> turning rename detection off, and it'd actively harm rename detection
> performance improvements I have in the pipeline.  I'd really prefer to
> not add this option, and instead combine the setting of aggressive
> with the other flag.  Do you have an independent reason for wanting
> this?
> 

While combining them would work for our specific use scenario (since we 
turn both on already along with turning off merge.stat), I really 
hesitate to tie these two different flags and code paths together with a 
single config setting.

While I don't want to needlessly complicate your optimizations in this 
area (they are already complex enough!) I believe we need to keep the 
option to turn on --aggressive without turning off rename detection as a 
viable option.  Perhaps if that is the case, your optimizations have 
less impact or don't apply but the user should be able to make that 
choice for their specific situation.

> Thanks,
> Elijah
> 

Re: [PATCH v1 2/2] merge: Add merge.aggressive config setting

[Elijah Newren]

Hi Ben,

On Tue, Apr 24, 2018 at 9:45 AM, Ben Peart <peartben@gmail.com> wrote:
> On 4/20/2018 1:22 PM, Elijah Newren wrote:
>> On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com>
>> wrote:
>>>
>>> Add the ability to control the aggressive flag passed to read-tree via a
>>> config setting.
>>
>> This feels like a workaround to the performance problems with index
>> updates in merge-recursive.c.
>
> This change wasn't done to solve performance problems.  We turned it on
> because it reduced the number of unmerged entries (from 40K to 1) in the
> particular merge we were looking at.  The additional 3 scenarios that
> --aggressive resolves made that much difference.
>
> That said, it makes sense to me to do

Um...color me perplexed here.  aggressive exists just to do some
resolutions that higher-level strategies can and totally ought to be
able to handle easily (the rules are almost trivially
straight-forward), but deferring allows the higher level strategies
(either merge-recursive or resolve's git-merge-one-file) to handle
slightly differently (e.g. by detecting renames).  merge-recursive
should be able to resolve anything that the unpack_trees aggressive
setting handles.  If it can't, it sounds like there's a horrible bug
somewhere.

Perhaps fixing that bug is the real problem?

Is there any chance you can dig out more details about any of these
conflicts or come up with a simple testcase where running 'git merge
-X no-renames' gives a merge conflict but running with this option
would run to completion?

>> this when rename detection is turned off.  In fact, I think you'd
>> automatically want to set aggressive to true whenever rename detection
>> is turned off (whether by your merge.renames option or the
>> -Xno-renames flag).
>> > I can't think of any reason this setting would be useful separate from
>> turning rename detection off, and it'd actively harm rename detection
>> performance improvements I have in the pipeline.  I'd really prefer to
>> not add this option, and instead combine the setting of aggressive
>> with the other flag.  Do you have an independent reason for wanting
>> this?
>>
>
> While combining them would work for our specific use scenario (since we turn
> both on already along with turning off merge.stat), I really hesitate to tie
> these two different flags and code paths together with a single config
> setting.
>
> While I don't want to needlessly complicate your optimizations in this area
> (they are already complex enough!) I believe we need to keep the option to
> turn on --aggressive without turning off rename detection as a viable
> option.  Perhaps if that is the case, your optimizations have less impact or
> don't apply but the user should be able to make that choice for their
> specific situation.

I totally buy that you need at least one option to avoid waiting for
(current) rename detection in some fashion, and that you don't want
lots of spurious conflicts.  But I don't understand why you believe
that we need to keep the option to turn on the aggressive flag
independently.  What's the usecase?  It wasn't possible before in the
code, no one else has asked for it, and even you say you don't need it
as a separate option.  Is it a concern that turning on aggressive
whenever rename-detection is turned off will break something?  The
only reason I can see to keep the aggressive codepath in unpack_trees
behind a branch instead of it always running unconditionally for every
single caller throughout the codebase is because of renames.  So the
fact that you're turning renames off, to me, suggests that aggressive
flag should automatically be turned on.  I'd even call pre-existing
code (e.g. the -X no-renames option in merge-recursive) that doesn't
turn on the aggressive flag buggy (even if the only result is
suboptimal-performance).

I don't see how an option to turn on the aggressive flag independently
is possibly useful to anyone.  Further, we have strong reason to
believe it will soon be actively harmful.  So...why?  It's totally
possible I'm just missing something.  If there's a good reason for it,
providing some kind of benefit that the user could weigh in a
tradeoff, then I can get on board with providing it as an option, but
right now I just don't see it.

Re: [PATCH v1 2/2] merge: Add merge.aggressive config setting

[Junio C Hamano]

Ben Peart <peartben@gmail.com> writes:

> That said, it makes sense to me to do
>> this when rename detection is turned off.  In fact, I think you'd
>> automatically want to set aggressive to true whenever rename detection
>> is turned off (whether by your merge.renames option or the
>> -Xno-renames flag).
>> ...
>
> While combining them would work for our specific use scenario (since
> we turn both on already along with turning off merge.stat), I really
> hesitate to tie these two different flags and code paths together with
> a single config setting.

The cases that non-agressive variant leaves unmerged are not
auto-resolved only because marking them as merged will rob the
chance from the rename detection logic to notice which ones are
"new" paths that could be matched with "deleted" ones to turn into
renames.  If rename deteciton is not done, there is no reason to
leave it non aggressive, as "#1 = missing, #2 = something and #3 =
missing" entry (just one example that is not auto-resolved by
non-agressive, but the principle is the same) left unmerged in the
index will get resolved to keep the current entry by the post
processing logic anyway.

In fact, checking git-merge-resolve would tell us that we already
use "aggresive" variant there unconditionally.

So, I think Elijah is correct---there is no reason not to enable
this setting when the other one to refuse rename detection is in
effect.

Re: [PATCH v1 2/2] merge: Add merge.aggressive config setting

[Ben Peart]

On 4/24/2018 7:57 PM, Junio C Hamano wrote:
> Ben Peart <peartben@gmail.com> writes:
> 
>> That said, it makes sense to me to do
>>> this when rename detection is turned off.  In fact, I think you'd
>>> automatically want to set aggressive to true whenever rename detection
>>> is turned off (whether by your merge.renames option or the
>>> -Xno-renames flag).
>>> ...
>>
>> While combining them would work for our specific use scenario (since
>> we turn both on already along with turning off merge.stat), I really
>> hesitate to tie these two different flags and code paths together with
>> a single config setting.
> 
> The cases that non-agressive variant leaves unmerged are not
> auto-resolved only because marking them as merged will rob the
> chance from the rename detection logic to notice which ones are
> "new" paths that could be matched with "deleted" ones to turn into
> renames.  If rename deteciton is not done, there is no reason to
> leave it non aggressive, as "#1 = missing, #2 = something and #3 =
> missing" entry (just one example that is not auto-resolved by
> non-agressive, but the principle is the same) left unmerged in the
> index will get resolved to keep the current entry by the post
> processing logic anyway.
> 
> In fact, checking git-merge-resolve would tell us that we already
> use "aggresive" variant there unconditionally.
> 
> So, I think Elijah is correct---there is no reason not to enable
> this setting when the other one to refuse rename detection is in
> effect.
> 

Thank you. I understand this description and it make sense to me.  I'm 
unfamiliar with the merge code so have to rely on you who are the 
experts for a sanity check.

I will remove the separate merge.aggressive config setting and instead, 
set the aggressive flag when the renames is turned off (whether by the 
new merge.renames setting or by the -Xno-renames flag) as Elijah suggests.

Re: [PATCH v1 0/2] add additional config settings for merge

[Elijah Newren]

On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
> This enables the user to set a couple of additional options for merge.
>
> 1. merge.aggressive - this is to try to resolve a few more trivial
>    merge cases.  It is documented in read-tree and is not something you
>    can pass into merge itself.
>
> 2. merge.renames - this is to save git from having to go through the entire
>    3 trees to see if there were any renames that happened.
>
> For the work item repro that I have been using this drops the merge time
> from ~1 hour to ~5 minutes and the unmerged entries goes down from
> ~40,000 to 1.

Ooh, this is *very* interesting.  Is there any chance I could also get
you to test performing the same merge with the version of git at
https://github.com/newren/git/tree/big-repo-small-cherry-pick and
report on your timings?

The 'big-repo-small-cherry-pick' name could be improved, but that
branch has a number of performance fixes for really poor rename
detection performance during merges.  From your description, I'm
pretty sure it'll apply to your case.  For my specific testcase,  I
got a speedup factor of 30.  Someone else on the list saw a factor of
24[1].  Results are highly dependent on the specific repo, but it's
certainly possible that it gets much of your factor of 12 speedup that
you saw with these new config settings you added.

However, what makes this case even more interesting to me is that my
branch may not be quite as effective as your workarounds.  There are
other other performance issues in merge that I am aware of, but for
which I haven't had the time to write the patches yet (I've been
waiting for the directory rename detection stuff to land and settle
down before working more on the performance aspects).  I do not know
how big a factor those other performance issues are, but your
workarounds (namely the aggressive setting) may get around some of
those other issues as well, so I'm very interested to see how my
current branch compares to the speedups you got with these settings.

Thanks,
Elijah


[1] https://public-inbox.org/git/alpine.DEB.2.00.1711211303290.20686@ds9.cixit.se/

Re: [PATCH v1 0/2] add additional config settings for merge

[Ben Peart]

On 4/20/2018 1:34 PM, Elijah Newren wrote:
> On Fri, Apr 20, 2018 at 6:36 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
>> This enables the user to set a couple of additional options for merge.
>>
>> 1. merge.aggressive - this is to try to resolve a few more trivial
>>     merge cases.  It is documented in read-tree and is not something you
>>     can pass into merge itself.
>>
>> 2. merge.renames - this is to save git from having to go through the entire
>>     3 trees to see if there were any renames that happened.
>>
>> For the work item repro that I have been using this drops the merge time
>> from ~1 hour to ~5 minutes and the unmerged entries goes down from
>> ~40,000 to 1.
> 
> Ooh, this is *very* interesting.  Is there any chance I could also get
> you to test performing the same merge with the version of git at
> https://github.com/newren/git/tree/big-repo-small-cherry-pick and
> report on your timings?
> 

Unfortunately, it isn't quite that simple.  My repo is _really_ big 
(3.2M files and ~100K commits per week) and requires me to use a custom 
fork of git that works with our GVFS solution for it to work at all.

I've been watching your work in this area and am hoping it pays off for 
us if/when we have users that want to do rename detection and override 
our defaults.

> The 'big-repo-small-cherry-pick' name could be improved, but that
> branch has a number of performance fixes for really poor rename
> detection performance during merges.  From your description, I'm
> pretty sure it'll apply to your case.  For my specific testcase,  I
> got a speedup factor of 30.  Someone else on the list saw a factor of
> 24[1].  Results are highly dependent on the specific repo, but it's
> certainly possible that it gets much of your factor of 12 speedup that
> you saw with these new config settings you added.
> 
> However, what makes this case even more interesting to me is that my
> branch may not be quite as effective as your workarounds.  There are
> other other performance issues in merge that I am aware of, but for
> which I haven't had the time to write the patches yet (I've been
> waiting for the directory rename detection stuff to land and settle
> down before working more on the performance aspects).  I do not know
> how big a factor those other performance issues are, but your
> workarounds (namely the aggressive setting) may get around some of
> those other issues as well, so I'm very interested to see how my
> current branch compares to the speedups you got with these settings.
> 
> Thanks,
> Elijah
> 
> 
> [1] https://public-inbox.org/git/alpine.DEB.2.00.1711211303290.20686@ds9.cixit.se/
> 

[PATCH v2 0/2] add additional config settings for merge

[Ben Peart]

Updated in response to feedback.  Mostly documentation changes but the diffstat
at the end of the merge (if on) now honors the new merge.rename setting as well.

Base Ref: master
Web-Diff: https://github.com/benpeart/git/commit/653bfe6e01
Checkout: git fetch https://github.com/benpeart/git merge-options-v2 && git checkout 653bfe6e01


### Interdiff (v1..v2):

diff --git a/Documentation/diff-config.txt b/Documentation/diff-config.txt
index 5ca942ab5e..77caa66c2f 100644
--- a/Documentation/diff-config.txt
+++ b/Documentation/diff-config.txt
@@ -112,7 +112,8 @@ diff.orderFile::
 
 diff.renameLimit::
 	The number of files to consider when performing the copy/rename
-	detection; equivalent to the 'git diff' option `-l`.
+	detection; equivalent to the 'git diff' option `-l`. This setting
+	has no effect if rename detection is turned off.
 
 diff.renames::
 	Whether and how Git detects renames.  If set to "false",
diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
index 5a9ab969db..38492bcb98 100644
--- a/Documentation/merge-config.txt
+++ b/Documentation/merge-config.txt
@@ -39,7 +39,8 @@ include::fmt-merge-msg-config.txt[]
 merge.renameLimit::
 	The number of files to consider when performing rename detection
 	during a merge; if not specified, defaults to the value of
-	diff.renameLimit.
+	diff.renameLimit. This setting has no effect if rename detection
+	is turned off.
 
 merge.renames::
 	Whether and how Git detects renames.  If set to "false",
diff --git a/Documentation/merge-strategies.txt b/Documentation/merge-strategies.txt
index 4a58aad4b8..1e0728aa12 100644
--- a/Documentation/merge-strategies.txt
+++ b/Documentation/merge-strategies.txt
@@ -84,12 +84,14 @@ no-renormalize;;
 	`merge.renormalize` configuration variable.
 
 no-renames;;
-	Turn off rename detection.
+	Turn off rename detection. This overrides the `merge.renames`
+	configuration variable.
 	See also linkgit:git-diff[1] `--no-renames`.
 
 find-renames[=<n>];;
 	Turn on rename detection, optionally setting the similarity
-	threshold.  This is the default.
+	threshold.  This is the default. This overrides the
+	'merge.renames' configuration variable.
 	See also linkgit:git-diff[1] `--find-renames`.
 
 rename-threshold=<n>;;
diff --git a/builtin/merge.c b/builtin/merge.c
index 8746c5e3e8..3be52cd316 100644
--- a/builtin/merge.c
+++ b/builtin/merge.c
@@ -424,6 +424,7 @@ static void finish(struct commit *head_commit,
 		opts.output_format |=
 			DIFF_FORMAT_SUMMARY | DIFF_FORMAT_DIFFSTAT;
 		opts.detect_rename = DIFF_DETECT_RENAME;
+		git_config_get_bool("merge.renames", &opts.detect_rename);
 		diff_setup_done(&opts);
 		diff_tree_oid(head, new_head, "", &opts);
 		diffcore_std(&opts);


### Patches

Ben Peart (2):
  merge: Add merge.renames config setting
  merge: Add merge.aggressive config setting

 Documentation/diff-config.txt      |  3 ++-
 Documentation/merge-config.txt     | 12 +++++++++++-
 Documentation/merge-strategies.txt |  6 ++++--
 builtin/merge.c                    |  1 +
 merge-recursive.c                  |  2 ++
 5 files changed, 20 insertions(+), 4 deletions(-)


base-commit: 0b0cc9f86731f894cff8dd25299a9b38c254569e
-- 
2.17.0.windows.1

[PATCH v2 1/2] merge: Add merge.renames config setting

[Ben Peart]

Add the ability to control rename detection for merge via a config setting.

Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Ben Peart <benpeart@microsoft.com>
---
 Documentation/diff-config.txt      | 3 ++-
 Documentation/merge-config.txt     | 8 +++++++-
 Documentation/merge-strategies.txt | 6 ++++--
 builtin/merge.c                    | 1 +
 merge-recursive.c                  | 1 +
 5 files changed, 15 insertions(+), 4 deletions(-)

diff --git a/Documentation/diff-config.txt b/Documentation/diff-config.txt
index 5ca942ab5e..77caa66c2f 100644
--- a/Documentation/diff-config.txt
+++ b/Documentation/diff-config.txt
@@ -112,7 +112,8 @@ diff.orderFile::
 
 diff.renameLimit::
 	The number of files to consider when performing the copy/rename
-	detection; equivalent to the 'git diff' option `-l`.
+	detection; equivalent to the 'git diff' option `-l`. This setting
+	has no effect if rename detection is turned off.
 
 diff.renames::
 	Whether and how Git detects renames.  If set to "false",
diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
index 12b6bbf591..0540c44e23 100644
--- a/Documentation/merge-config.txt
+++ b/Documentation/merge-config.txt
@@ -35,7 +35,13 @@ include::fmt-merge-msg-config.txt[]
 merge.renameLimit::
 	The number of files to consider when performing rename detection
 	during a merge; if not specified, defaults to the value of
-	diff.renameLimit.
+	diff.renameLimit. This setting has no effect if rename detection
+	is turned off.
+
+merge.renames::
+	Whether and how Git detects renames.  If set to "false",
+	rename detection is disabled. If set to "true", basic rename
+	detection is enabled. This is the default.
 
 merge.renormalize::
 	Tell Git that canonical representation of files in the
diff --git a/Documentation/merge-strategies.txt b/Documentation/merge-strategies.txt
index 4a58aad4b8..1e0728aa12 100644
--- a/Documentation/merge-strategies.txt
+++ b/Documentation/merge-strategies.txt
@@ -84,12 +84,14 @@ no-renormalize;;
 	`merge.renormalize` configuration variable.
 
 no-renames;;
-	Turn off rename detection.
+	Turn off rename detection. This overrides the `merge.renames`
+	configuration variable.
 	See also linkgit:git-diff[1] `--no-renames`.
 
 find-renames[=<n>];;
 	Turn on rename detection, optionally setting the similarity
-	threshold.  This is the default.
+	threshold.  This is the default. This overrides the
+	'merge.renames' configuration variable.
 	See also linkgit:git-diff[1] `--find-renames`.
 
 rename-threshold=<n>;;
diff --git a/builtin/merge.c b/builtin/merge.c
index 8746c5e3e8..3be52cd316 100644
--- a/builtin/merge.c
+++ b/builtin/merge.c
@@ -424,6 +424,7 @@ static void finish(struct commit *head_commit,
 		opts.output_format |=
 			DIFF_FORMAT_SUMMARY | DIFF_FORMAT_DIFFSTAT;
 		opts.detect_rename = DIFF_DETECT_RENAME;
+		git_config_get_bool("merge.renames", &opts.detect_rename);
 		diff_setup_done(&opts);
 		diff_tree_oid(head, new_head, "", &opts);
 		diffcore_std(&opts);
diff --git a/merge-recursive.c b/merge-recursive.c
index 9c05eb7f70..cd5367e890 100644
--- a/merge-recursive.c
+++ b/merge-recursive.c
@@ -3256,6 +3256,7 @@ static void merge_recursive_config(struct merge_options *o)
 	git_config_get_int("merge.verbosity", &o->verbosity);
 	git_config_get_int("diff.renamelimit", &o->diff_rename_limit);
 	git_config_get_int("merge.renamelimit", &o->merge_rename_limit);
+	git_config_get_bool("merge.renames", &o->detect_rename);
 	git_config(git_xmerge_config, NULL);
 }
 
-- 
2.17.0.windows.1

Re: [PATCH v2 1/2] merge: Add merge.renames config setting

[Elijah Newren]

On Tue, Apr 24, 2018 at 10:11 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
> Add the ability to control rename detection for merge via a config setting.

Sweet, thanks for including the documentation updates.

I lean towards the side of the argument that says that since
merge.renameLimit inherits from diff.renameLimit, merge.renames should
inherit default value from diff.renames (allow people to not have to
repeat themselves as much if they want to use the same rename settings
for all cases).  Sounds like you and Johannes disagree.  I don't feel
super strongly about this item, but it'd probably be good to get some
other git folks' opinions on this particular point.

Other than that unresolved question, and the separate one about
whether to go with a different option instead (e.g.
merge.defaultStrategy), as being discussed elsewhere in this thread,
the patch looks good to me.

Re: [PATCH v2 1/2] merge: Add merge.renames config setting

[Elijah Newren]

Sorry, I noticed something else I missed on my last reading...

On Tue, Apr 24, 2018 at 10:11 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
> diff --git a/builtin/merge.c b/builtin/merge.c
> index 8746c5e3e8..3be52cd316 100644
> --- a/builtin/merge.c
> +++ b/builtin/merge.c
> @@ -424,6 +424,7 @@ static void finish(struct commit *head_commit,
>                 opts.output_format |=
>                         DIFF_FORMAT_SUMMARY | DIFF_FORMAT_DIFFSTAT;
>                 opts.detect_rename = DIFF_DETECT_RENAME;
> +               git_config_get_bool("merge.renames", &opts.detect_rename);
>                 diff_setup_done(&opts);
>                 diff_tree_oid(head, new_head, "", &opts);
>                 diffcore_std(&opts);

Shouldn't this also be turned off if either (a) merge.renames is unset
and diff.renames is false, or (b) the user specifies -Xno-renames?

Re: [PATCH v2 1/2] merge: Add merge.renames config setting

[Ben Peart]

On 4/24/2018 2:59 PM, Elijah Newren wrote:
> Sorry, I noticed something else I missed on my last reading...
> 
> On Tue, Apr 24, 2018 at 10:11 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
>> diff --git a/builtin/merge.c b/builtin/merge.c
>> index 8746c5e3e8..3be52cd316 100644
>> --- a/builtin/merge.c
>> +++ b/builtin/merge.c
>> @@ -424,6 +424,7 @@ static void finish(struct commit *head_commit,
>>                  opts.output_format |=
>>                          DIFF_FORMAT_SUMMARY | DIFF_FORMAT_DIFFSTAT;
>>                  opts.detect_rename = DIFF_DETECT_RENAME;
>> +               git_config_get_bool("merge.renames", &opts.detect_rename);
>>                  diff_setup_done(&opts);
>>                  diff_tree_oid(head, new_head, "", &opts);
>>                  diffcore_std(&opts);
> 
> Shouldn't this also be turned off if either (a) merge.renames is unset
> and diff.renames is false, or (b) the user specifies -Xno-renames?
> 

This makes me think that I should probably remove the line that 
overrides the detect_rename setting with the merge config setting.  As I 
look at the code, none of the other merge options are reflected in the 
diffstat; instead, all the settings are pretty much hard coded.  Perhaps 
I shouldn't rock that boat.

Re: [PATCH v2 1/2] merge: Add merge.renames config setting

[Elijah Newren]

On Tue, Apr 24, 2018 at 1:31 PM, Ben Peart <peartben@gmail.com> wrote:
> On 4/24/2018 2:59 PM, Elijah Newren wrote:
>> On Tue, Apr 24, 2018 at 10:11 AM, Ben Peart <Ben.Peart@microsoft.com>
>> wrote:
>>>
>>> diff --git a/builtin/merge.c b/builtin/merge.c
>>> index 8746c5e3e8..3be52cd316 100644
>>> --- a/builtin/merge.c
>>> +++ b/builtin/merge.c
>>> @@ -424,6 +424,7 @@ static void finish(struct commit *head_commit,
>>>                  opts.output_format |=
>>>                          DIFF_FORMAT_SUMMARY | DIFF_FORMAT_DIFFSTAT;
>>>                  opts.detect_rename = DIFF_DETECT_RENAME;
>>> +               git_config_get_bool("merge.renames",
>>> &opts.detect_rename);
>>>                  diff_setup_done(&opts);
>>>                  diff_tree_oid(head, new_head, "", &opts);
>>>                  diffcore_std(&opts);
>>
>>
>> Shouldn't this also be turned off if either (a) merge.renames is unset
>> and diff.renames is false, or (b) the user specifies -Xno-renames?
>>
>
> This makes me think that I should probably remove the line that overrides
> the detect_rename setting with the merge config setting.  As I look at the
> code, none of the other merge options are reflected in the diffstat;
> instead, all the settings are pretty much hard coded.  Perhaps I shouldn't
> rock that boat.

Actually, stat_graph_width respects the diff.statGraphWidth config
option, even though it's slightly hidden due to the magic value of -1,
and being handled from diff.c.

However, trying to get this suggestion of mine hooked up, particularly
with -Xno-renames and -Xfind-renames (the latter because it might need
to override a merge.renames or diff.renames config setting), might be
slightly tricky because the -X options are only passed down to a
single merge strategy but this code is outside of the merge
strategies.  So making it a separate patch, or even a separate patch
series may make sense.  I'm still interested in this change if you
aren't, but I'm fine with it not being part of your series if you
don't want to tackle it.

[PATCH v2 2/2] merge: Add merge.aggressive config setting

[Ben Peart]

Add the ability to control the aggressive flag passed to read-tree via a config setting.

Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Ben Peart <benpeart@microsoft.com>
---
 Documentation/merge-config.txt | 4 ++++
 merge-recursive.c              | 1 +
 2 files changed, 5 insertions(+)

diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
index 0540c44e23..38492bcb98 100644
--- a/Documentation/merge-config.txt
+++ b/Documentation/merge-config.txt
@@ -1,3 +1,7 @@
+merge.aggressive::
+	Passes "aggressive" to read-tree which makes the command resolve
+	a few more cases internally. See "--aggressive" in linkgit:git-read-tree[1].
+
 merge.conflictStyle::
 	Specify the style in which conflicted hunks are written out to
 	working tree files upon merge.  The default is "merge", which
diff --git a/merge-recursive.c b/merge-recursive.c
index cd5367e890..0ca84e4b82 100644
--- a/merge-recursive.c
+++ b/merge-recursive.c
@@ -355,6 +355,7 @@ static int git_merge_trees(struct merge_options *o,
 	o->unpack_opts.fn = threeway_merge;
 	o->unpack_opts.src_index = &the_index;
 	o->unpack_opts.dst_index = &the_index;
+	git_config_get_bool("merge.aggressive", (int *)&o->unpack_opts.aggressive);
 	setup_unpack_trees_porcelain(&o->unpack_opts, "merge");
 
 	init_tree_desc_from_tree(t+0, common);
-- 
2.17.0.windows.1

Re: [PATCH v2 0/2] add additional config settings for merge

[Junio C Hamano]

Ben Peart <Ben.Peart@microsoft.com> writes:

>  diff.renameLimit::
>  	The number of files to consider when performing the copy/rename
> -	detection; equivalent to the 'git diff' option `-l`.
> +	detection; equivalent to the 'git diff' option `-l`. This setting
> +	has no effect if rename detection is turned off.

You mean "turned off via diff.renames"?

This is not meant as a suggestion to rewrite this paragraph
further---but if the answer is "no", then that might be an
indication that the sentence is inviting a misunderstanding.

>  diff.renames::
>  	Whether and how Git detects renames.  If set to "false",
> diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
> index 5a9ab969db..38492bcb98 100644
> --- a/Documentation/merge-config.txt
> +++ b/Documentation/merge-config.txt
> @@ -39,7 +39,8 @@ include::fmt-merge-msg-config.txt[]
>  merge.renameLimit::
>  	The number of files to consider when performing rename detection
>  	during a merge; if not specified, defaults to the value of
> -	diff.renameLimit.
> +	diff.renameLimit. This setting has no effect if rename detection
> +	is turned off.

Ditto.  If your design is to make the merge machinery completely
ignore diff.renames and only pay attention to merge.renames [*1*],
then it probably is a good idea to be more specific here, by saying
"... is turned off via ...", though.

>  merge.renames::
>  	Whether and how Git detects renames.  If set to "false",

[Footnote]

*1* ...which I do not think is such a good idea, by the way.  I'd
personally expect merge.renames to allow overriding and falling back
to diff.renames, just like the {merge,diff}.renameLimit pair does.

Re: [PATCH v2 0/2] add additional config settings for merge

[Ben Peart]

On 4/24/2018 8:13 PM, Junio C Hamano wrote:
> Ben Peart <Ben.Peart@microsoft.com> writes:
> 
>>   diff.renameLimit::
>>   	The number of files to consider when performing the copy/rename
>> -	detection; equivalent to the 'git diff' option `-l`.
>> +	detection; equivalent to the 'git diff' option `-l`. This setting
>> +	has no effect if rename detection is turned off.
> 
> You mean "turned off via diff.renames"?
> 
> This is not meant as a suggestion to rewrite this paragraph
> further---but if the answer is "no", then that might be an
> indication that the sentence is inviting a misunderstanding.
> 

Yes, this is referring to turned off via the config setting 
"diff.renames" but it could also be turned off by passing "--no-renames" 
on the command line.

To be clear, this documentation change isn't trying to document any 
changes to the code or behavior - it is just an attempt to clarify what 
the existing behavior is.  If it isn't helping, I can remove it.

>>   diff.renames::
>>   	Whether and how Git detects renames.  If set to "false",
>> diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
>> index 5a9ab969db..38492bcb98 100644
>> --- a/Documentation/merge-config.txt
>> +++ b/Documentation/merge-config.txt
>> @@ -39,7 +39,8 @@ include::fmt-merge-msg-config.txt[]
>>   merge.renameLimit::
>>   	The number of files to consider when performing rename detection
>>   	during a merge; if not specified, defaults to the value of
>> -	diff.renameLimit.
>> +	diff.renameLimit. This setting has no effect if rename detection
>> +	is turned off.
> 
> Ditto.  If your design is to make the merge machinery completely
> ignore diff.renames and only pay attention to merge.renames [*1*],
> then it probably is a good idea to be more specific here, by saying
> "... is turned off via ...", though.
> 
>>   merge.renames::
>>   	Whether and how Git detects renames.  If set to "false",
> 
> [Footnote]
> 
> *1* ...which I do not think is such a good idea, by the way.  I'd
> personally expect merge.renames to allow overriding and falling back
> to diff.renames, just like the {merge,diff}.renameLimit pair does.
> 

It looks like I'm in the minority on whether the merge settings should 
inherit from the corresponding diff settings so I will submit a new 
patch series that does it the same way as the {merge,diff}.renameLimit 
pair works.

I'll leave it as an exercise for someone else [1] to change any other 
merge settings that should behave that way.

[1] 
https://public-inbox.org/git/20180420133632.17580-1-benpeart@microsoft.com/T/#m52a3dbd0945360bfb873fd3b553472558ef3b796

Re: [PATCH v2 0/2] add additional config settings for merge

[Junio C Hamano]

Ben Peart <peartben@gmail.com> writes:

> To be clear, this documentation change isn't trying to document any
> changes to the code or behavior - it is just an attempt to clarify
> what the existing behavior is.

Yeah, I know, and I do think the new text is a good first step in
the right direction; I merely was trying to help in the clarifying
effort ;-)

[PATCH v3 0/3] add merge.renames config setting

[Ben Peart]

This is a complete rewrite based on the feedback from earlier patches.

Update the documentation to better indicate command line options that override
various config settings related to merge.

Add a new config merge.renames setting to to control the rename detection
behavior of merge.  This setting will default to the value of diff.renames.

Also adds logic so that when rename detection is turned off, the aggressive
flag is passed to read_tree() so that it can auto resolve more cases that would
have been handled by rename detection.

For the repro that I have been using this drops the merge time from ~1 hour to
~5 minutes and the unmerged entries goes down from ~40,000 to 1.

Helped-by: Kevin Willford <kewillf@microsoft.com>
Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Ben Peart <Ben.Peart@microsoft.com>

Base Ref: master
Web-Diff: https://github.com/benpeart/git/commit/6a8372d517
Checkout: git fetch https://github.com/benpeart/git merge-options-v3 && git checkout 6a8372d517

### Patches

Ben Peart (3):
  merge: update documentation for {merge,diff}.renameLimit
  merge: Add merge.renames config setting
  merge: pass aggressive when rename detection is turned off

 Documentation/diff-config.txt             |  3 ++-
 Documentation/merge-config.txt            |  9 +++++++-
 Documentation/merge-strategies.txt        |  6 +++--
 diff.c                                    |  2 +-
 diff.h                                    |  2 ++
 merge-recursive.c                         | 27 +++++++++++++++++------
 merge-recursive.h                         |  8 ++++++-
 t/t3034-merge-recursive-rename-options.sh | 18 +++++++++++++++
 8 files changed, 62 insertions(+), 13 deletions(-)


base-commit: 1f1cddd558b54bb0ce19c8ace353fd07b758510d
-- 
2.17.0.windows.1

[PATCH v3 1/3] merge: update documentation for {merge,diff}.renameLimit

[Ben Peart]

Update the documentation to better indicate that the renameLimit setting is
ignored if rename detection is turned off via command line options or config
settings.

Signed-off-by: Ben Peart <benpeart@microsoft.com>
---
 Documentation/diff-config.txt  | 3 ++-
 Documentation/merge-config.txt | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/Documentation/diff-config.txt b/Documentation/diff-config.txt
index 5ca942ab5e..77caa66c2f 100644
--- a/Documentation/diff-config.txt
+++ b/Documentation/diff-config.txt
@@ -112,7 +112,8 @@ diff.orderFile::
 
 diff.renameLimit::
 	The number of files to consider when performing the copy/rename
-	detection; equivalent to the 'git diff' option `-l`.
+	detection; equivalent to the 'git diff' option `-l`. This setting
+	has no effect if rename detection is turned off.
 
 diff.renames::
 	Whether and how Git detects renames.  If set to "false",
diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
index 12b6bbf591..48ee3bce77 100644
--- a/Documentation/merge-config.txt
+++ b/Documentation/merge-config.txt
@@ -35,7 +35,8 @@ include::fmt-merge-msg-config.txt[]
 merge.renameLimit::
 	The number of files to consider when performing rename detection
 	during a merge; if not specified, defaults to the value of
-	diff.renameLimit.
+	diff.renameLimit. This setting has no effect if rename detection
+	is turned off.
 
 merge.renormalize::
 	Tell Git that canonical representation of files in the
-- 
2.17.0.windows.1

Re: [PATCH v3 1/3] merge: update documentation for {merge,diff}.renameLimit

[Elijah Newren]

On Thu, Apr 26, 2018 at 1:52 PM, Ben Peart <Ben.Peart@microsoft.com> wrote:
> Update the documentation to better indicate that the renameLimit setting is
> ignored if rename detection is turned off via command line options or config
> settings.
>
> Signed-off-by: Ben Peart <benpeart@microsoft.com>
> ---
>  Documentation/diff-config.txt  | 3 ++-
>  Documentation/merge-config.txt | 3 ++-
>  2 files changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/Documentation/diff-config.txt b/Documentation/diff-config.txt
> index 5ca942ab5e..77caa66c2f 100644
> --- a/Documentation/diff-config.txt
> +++ b/Documentation/diff-config.txt
> @@ -112,7 +112,8 @@ diff.orderFile::
>
>  diff.renameLimit::
>         The number of files to consider when performing the copy/rename
> -       detection; equivalent to the 'git diff' option `-l`.
> +       detection; equivalent to the 'git diff' option `-l`. This setting
> +       has no effect if rename detection is turned off.
>
>  diff.renames::
>         Whether and how Git detects renames.  If set to "false",
> diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
> index 12b6bbf591..48ee3bce77 100644
> --- a/Documentation/merge-config.txt
> +++ b/Documentation/merge-config.txt
> @@ -35,7 +35,8 @@ include::fmt-merge-msg-config.txt[]
>  merge.renameLimit::
>         The number of files to consider when performing rename detection
>         during a merge; if not specified, defaults to the value of
> -       diff.renameLimit.
> +       diff.renameLimit. This setting has no effect if rename detection
> +       is turned off.
>
>  merge.renormalize::
>         Tell Git that canonical representation of files in the
> --
> 2.17.0.windows.1

Patch looks fine, but it's hard for me not to notice a separate issue
in this area independent of your series: I'm curious if we should
document that the value of 0 is special here (as per Jonathan Tan's
commit 89973554b52c ("diffcore-rename: make diff-tree -l0 mean
-l<large>", 2017-11-29)), and doesn't actually drop the limit to 0.
cc'ing Jonathan Tan for his thoughts.

Re: [PATCH v3 1/3] merge: update documentation for {merge,diff}.renameLimit

[Jonathan Tan]

On Thu, 26 Apr 2018 16:11:50 -0700
Elijah Newren <newren@gmail.com> wrote:

> Patch looks fine, but it's hard for me not to notice a separate issue
> in this area independent of your series: I'm curious if we should
> document that the value of 0 is special here (as per Jonathan Tan's
> commit 89973554b52c ("diffcore-rename: make diff-tree -l0 mean
> -l<large>", 2017-11-29)), and doesn't actually drop the limit to 0.
> cc'ing Jonathan Tan for his thoughts.

Documenting that the value of 0 is special does make sense to me. I
think this patch can go in as-is, though - it is already an improvement.

[PATCH v3 2/3] merge: Add merge.renames config setting

[Ben Peart]

Add the ability to control rename detection for merge via a config setting.
This setting behaves the same and defaults to the value of diff.renames but only
applies to merge.

Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Ben Peart <benpeart@microsoft.com>
---
 Documentation/merge-config.txt            |  6 ++++++
 Documentation/merge-strategies.txt        |  6 ++++--
 diff.c                                    |  2 +-
 diff.h                                    |  2 ++
 merge-recursive.c                         | 23 +++++++++++++++++------
 merge-recursive.h                         |  8 +++++++-
 t/t3034-merge-recursive-rename-options.sh | 18 ++++++++++++++++++
 7 files changed, 55 insertions(+), 10 deletions(-)

diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
index 48ee3bce77..59848e5634 100644
--- a/Documentation/merge-config.txt
+++ b/Documentation/merge-config.txt
@@ -38,6 +38,12 @@ merge.renameLimit::
 	diff.renameLimit. This setting has no effect if rename detection
 	is turned off.
 
+merge.renames::
+	Whether and how Git detects renames.  If set to "false",
+	rename detection is disabled. If set to "true", basic rename
+	detection is enabled.  If set to "copies" or "copy", Git will
+	detect copies, as well.  Defaults to the value of diff.renames.
+
 merge.renormalize::
 	Tell Git that canonical representation of files in the
 	repository has changed over time (e.g. earlier commits record
diff --git a/Documentation/merge-strategies.txt b/Documentation/merge-strategies.txt
index 4a58aad4b8..1e0728aa12 100644
--- a/Documentation/merge-strategies.txt
+++ b/Documentation/merge-strategies.txt
@@ -84,12 +84,14 @@ no-renormalize;;
 	`merge.renormalize` configuration variable.
 
 no-renames;;
-	Turn off rename detection.
+	Turn off rename detection. This overrides the `merge.renames`
+	configuration variable.
 	See also linkgit:git-diff[1] `--no-renames`.
 
 find-renames[=<n>];;
 	Turn on rename detection, optionally setting the similarity
-	threshold.  This is the default.
+	threshold.  This is the default. This overrides the
+	'merge.renames' configuration variable.
 	See also linkgit:git-diff[1] `--find-renames`.
 
 rename-threshold=<n>;;
diff --git a/diff.c b/diff.c
index 1289df4b1f..5dfc24aa6d 100644
--- a/diff.c
+++ b/diff.c
@@ -177,7 +177,7 @@ static int parse_submodule_params(struct diff_options *options, const char *valu
 	return 0;
 }
 
-static int git_config_rename(const char *var, const char *value)
+int git_config_rename(const char *var, const char *value)
 {
 	if (!value)
 		return DIFF_DETECT_RENAME;
diff --git a/diff.h b/diff.h
index d29560f822..806faee2b3 100644
--- a/diff.h
+++ b/diff.h
@@ -324,6 +324,8 @@ extern int git_diff_ui_config(const char *var, const char *value, void *cb);
 extern void diff_setup(struct diff_options *);
 extern int diff_opt_parse(struct diff_options *, const char **, int, const char *);
 extern void diff_setup_done(struct diff_options *);
+extern int git_config_rename(const char *var, const char *value);
+
 
 #define DIFF_DETECT_RENAME	1
 #define DIFF_DETECT_COPY	2
diff --git a/merge-recursive.c b/merge-recursive.c
index 0c0d48624d..2637d34d87 100644
--- a/merge-recursive.c
+++ b/merge-recursive.c
@@ -555,13 +555,13 @@ static struct string_list *get_renames(struct merge_options *o,
 	struct diff_options opts;
 
 	renames = xcalloc(1, sizeof(struct string_list));
-	if (!o->detect_rename)
+	if (!merge_detect_rename(o))
 		return renames;
 
 	diff_setup(&opts);
 	opts.flags.recursive = 1;
 	opts.flags.rename_empty = 0;
-	opts.detect_rename = DIFF_DETECT_RENAME;
+	opts.detect_rename = merge_detect_rename(o);
 	opts.rename_limit = o->merge_rename_limit >= 0 ? o->merge_rename_limit :
 			    o->diff_rename_limit >= 0 ? o->diff_rename_limit :
 			    1000;
@@ -2232,9 +2232,18 @@ int merge_recursive_generic(struct merge_options *o,
 
 static void merge_recursive_config(struct merge_options *o)
 {
+	char *value = NULL;
 	git_config_get_int("merge.verbosity", &o->verbosity);
 	git_config_get_int("diff.renamelimit", &o->diff_rename_limit);
 	git_config_get_int("merge.renamelimit", &o->merge_rename_limit);
+	if (!git_config_get_string("diff.renames", &value)) {
+		o->diff_detect_rename = git_config_rename("diff.renames", value);
+		free(value);
+	}
+	if (!git_config_get_string("merge.renames", &value)) {
+		o->merge_detect_rename = git_config_rename("merge.renames", value);
+		free(value);
+	}
 	git_config(git_xmerge_config, NULL);
 }
 
@@ -2244,10 +2253,11 @@ void init_merge_options(struct merge_options *o)
 	memset(o, 0, sizeof(struct merge_options));
 	o->verbosity = 2;
 	o->buffer_output = 1;
+	o->diff_detect_rename = -1;
+	o->merge_detect_rename = -1;
 	o->diff_rename_limit = -1;
 	o->merge_rename_limit = -1;
 	o->renormalize = 0;
-	o->detect_rename = 1;
 	merge_recursive_config(o);
 	merge_verbosity = getenv("GIT_MERGE_VERBOSITY");
 	if (merge_verbosity)
@@ -2298,18 +2308,19 @@ int parse_merge_opt(struct merge_options *o, const char *s)
 	else if (!strcmp(s, "no-renormalize"))
 		o->renormalize = 0;
 	else if (!strcmp(s, "no-renames"))
-		o->detect_rename = 0;
+		o->merge_detect_rename = 0;
 	else if (!strcmp(s, "find-renames")) {
-		o->detect_rename = 1;
+		o->merge_detect_rename = 1;
 		o->rename_score = 0;
 	}
 	else if (skip_prefix(s, "find-renames=", &arg) ||
 		 skip_prefix(s, "rename-threshold=", &arg)) {
 		if ((o->rename_score = parse_rename_score(&arg)) == -1 || *arg != 0)
 			return -1;
-		o->detect_rename = 1;
+		o->merge_detect_rename = 1;
 	}
 	else
 		return -1;
+
 	return 0;
 }
diff --git a/merge-recursive.h b/merge-recursive.h
index 80d69d1401..0c5f7eff98 100644
--- a/merge-recursive.h
+++ b/merge-recursive.h
@@ -17,7 +17,8 @@ struct merge_options {
 	unsigned renormalize : 1;
 	long xdl_opts;
 	int verbosity;
-	int detect_rename;
+	int diff_detect_rename;
+	int merge_detect_rename;
 	int diff_rename_limit;
 	int merge_rename_limit;
 	int rename_score;
@@ -28,6 +29,11 @@ struct merge_options {
 	struct hashmap current_file_dir_set;
 	struct string_list df_conflict_file_set;
 };
+inline int merge_detect_rename(struct merge_options *o)
+{
+	return o->merge_detect_rename >= 0 ? o->merge_detect_rename :
+		o->diff_detect_rename >= 0 ? o->diff_detect_rename : 1;
+}
 
 /* merge_trees() but with recursive ancestor consolidation */
 int merge_recursive(struct merge_options *o,
diff --git a/t/t3034-merge-recursive-rename-options.sh b/t/t3034-merge-recursive-rename-options.sh
index b9c4028496..3d9fae68c4 100755
--- a/t/t3034-merge-recursive-rename-options.sh
+++ b/t/t3034-merge-recursive-rename-options.sh
@@ -309,4 +309,22 @@ test_expect_success 'last wins in --find-renames=<m> --rename-threshold=<n>' '
 	check_threshold_0
 '
 
+test_expect_success 'merge.renames disables rename detection' '
+	git read-tree --reset -u HEAD &&
+	git -c merge.renames=false merge-recursive $tail &&
+	check_no_renames
+'
+
+test_expect_success 'merge.renames defaults to diff.renames' '
+	git read-tree --reset -u HEAD &&
+	git -c diff.renames=false merge-recursive $tail &&
+	check_no_renames
+'
+
+test_expect_success 'merge.renames overrides diff.renames' '
+	git read-tree --reset -u HEAD &&
+	test_must_fail git -c diff.renames=false -c merge.renames=true merge-recursive $tail &&
+	$check_50
+'
+
 test_done
-- 
2.17.0.windows.1

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Elijah Newren]

On Thu, Apr 26, 2018 at 1:52 PM, Ben Peart <Ben.Peart@microsoft.com> wrote:

> +merge.renames::
> +       Whether and how Git detects renames.  If set to "false",
> +       rename detection is disabled. If set to "true", basic rename
> +       detection is enabled.  If set to "copies" or "copy", Git will
> +       detect copies, as well.  Defaults to the value of diff.renames.
> +

We shouldn't allow users to force copy detection on for merges  The
diff side of the code will detect them correctly but the code in
merge-recursive will mishandle the copy pairs.  I think fixing it is
somewhere between big can of worms and
it's-a-configuration-that-doesn't-even-make-sense, but it's been a
while since I thought about it.

> diff --git a/merge-recursive.h b/merge-recursive.h
> index 80d69d1401..0c5f7eff98 100644
> --- a/merge-recursive.h
> +++ b/merge-recursive.h
> @@ -17,7 +17,8 @@ struct merge_options {
>         unsigned renormalize : 1;
>         long xdl_opts;
>         int verbosity;
> -       int detect_rename;
> +       int diff_detect_rename;
> +       int merge_detect_rename;
>         int diff_rename_limit;
>         int merge_rename_limit;
>         int rename_score;
> @@ -28,6 +29,11 @@ struct merge_options {
>         struct hashmap current_file_dir_set;
>         struct string_list df_conflict_file_set;
>  };
> +inline int merge_detect_rename(struct merge_options *o)
> +{
> +       return o->merge_detect_rename >= 0 ? o->merge_detect_rename :
> +               o->diff_detect_rename >= 0 ? o->diff_detect_rename : 1;
> +}

Why did you split o->detect_rename into two fields?  You then
recombine them in merge_detect_rename(), and after initial setup only
ever access them through that function.  Having two fields worries me
that people will accidentally introduce bugs by using one of them
instead of the merge_detect_rename() function.  Is there a reason you
decided against having the initial setup just set a single value and
then use it directly?

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Ben Peart]

On 4/26/2018 6:52 PM, Elijah Newren wrote:
> On Thu, Apr 26, 2018 at 1:52 PM, Ben Peart <Ben.Peart@microsoft.com> wrote:
> 
>> +merge.renames::
>> +       Whether and how Git detects renames.  If set to "false",
>> +       rename detection is disabled. If set to "true", basic rename
>> +       detection is enabled.  If set to "copies" or "copy", Git will
>> +       detect copies, as well.  Defaults to the value of diff.renames.
>> +
> 
> We shouldn't allow users to force copy detection on for merges  The
> diff side of the code will detect them correctly but the code in
> merge-recursive will mishandle the copy pairs.  I think fixing it is
> somewhere between big can of worms and
> it's-a-configuration-that-doesn't-even-make-sense, but it's been a
> while since I thought about it.

Color me puzzled. :)  The consensus was that the default value for 
merge.renames come from diff.renames.  diff.renames supports copy 
detection which means that merge.renames will inherit that value.  My 
assumption was that is what was intended so when I reimplemented it, I 
fully implemented it that way.

Are you now requesting to only use diff.renames as the default if the 
value is true or false but not if it is copy?  What should happen if 
diff.renames is actually set to copy?  Should merge silently change that 
to true, display a warning, error out, or something else?  Do you have 
some other behavior for how to handle copy being inherited from 
diff.renames you'd like to see?

Can you write the documentation that clearly explains the exact behavior 
you want?  That would kill two birds with one stone... :)

> 
>> diff --git a/merge-recursive.h b/merge-recursive.h
>> index 80d69d1401..0c5f7eff98 100644
>> --- a/merge-recursive.h
>> +++ b/merge-recursive.h
>> @@ -17,7 +17,8 @@ struct merge_options {
>>          unsigned renormalize : 1;
>>          long xdl_opts;
>>          int verbosity;
>> -       int detect_rename;
>> +       int diff_detect_rename;
>> +       int merge_detect_rename;
>>          int diff_rename_limit;
>>          int merge_rename_limit;
>>          int rename_score;
>> @@ -28,6 +29,11 @@ struct merge_options {
>>          struct hashmap current_file_dir_set;
>>          struct string_list df_conflict_file_set;
>>   };
>> +inline int merge_detect_rename(struct merge_options *o)
>> +{
>> +       return o->merge_detect_rename >= 0 ? o->merge_detect_rename :
>> +               o->diff_detect_rename >= 0 ? o->diff_detect_rename : 1;
>> +}
> 
> Why did you split o->detect_rename into two fields?  You then
> recombine them in merge_detect_rename(), and after initial setup only
> ever access them through that function.  Having two fields worries me
> that people will accidentally introduce bugs by using one of them
> instead of the merge_detect_rename() function.  Is there a reason you
> decided against having the initial setup just set a single value and
> then use it directly?
> 

The setup of this value is split into 3 places that may or may not all 
get called.  The initial values, the values that come from the config 
settings and then any values passed on the command line.

Because the merge value can now inherit from the diff value, you only 
know the final value after you have received all possible inputs.  That 
makes it necessary to be a calculated value.

If you look at diff_rename_limit/merge_rename_limit, detect_rename 
follow the same pattern for the same reasons.  It turns out 
detect_rename was a little more complex because it is used in 3 
different locations (vs just one) which is why I wrapped the inheritance 
logic into the helper function merge_detect_rename().

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Junio C Hamano]

Ben Peart <peartben@gmail.com> writes:

> Color me puzzled. :)  The consensus was that the default value for
> merge.renames come from diff.renames.  diff.renames supports copy
> detection which means that merge.renames will inherit that value.  My
> assumption was that is what was intended so when I reimplemented it, I
> fully implemented it that way.
>
> Are you now requesting to only use diff.renames as the default if the
> value is true or false but not if it is copy?  What should happen if
> diff.renames is actually set to copy?  Should merge silently change
> that to true, display a warning, error out, or something else?  Do you
> have some other behavior for how to handle copy being inherited from
> diff.renames you'd like to see?
>
> Can you write the documentation that clearly explains the exact
> behavior you want?  That would kill two birds with one stone... :)

I think demoting from copy to rename-only is a good idea, at least
for now, because I do not believe we have figured out what we want
to happen when we detect copied files are involved in a merge.

But I am not sure if we even want to fail merge.renames=copy as an
invalid configuration.  So my gut feeling of the best solution to
the above is to do something like:

 - whether the configuration comes from diff.renames or
   merge.renames, turn *.renames=copy to true inside the merge
   recursive machinery.

 - document the fact in "git merge-recursive" documentation (or "git
   merge" documentation) to say "_currently_ asking for rename
   detection to find copies and renames will do the same
   thing---copies are ignored", impliying "this might change in the
   future", in the BUGS section.

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Elijah Newren]

On Thu, Apr 26, 2018 at 7:23 PM, Junio C Hamano <gitster@pobox.com> wrote:
> Ben Peart <peartben@gmail.com> writes:
>
>> Color me puzzled. :)  The consensus was that the default value for
>> merge.renames come from diff.renames.  diff.renames supports copy
>> detection which means that merge.renames will inherit that value.  My
>> assumption was that is what was intended so when I reimplemented it, I
>> fully implemented it that way.
>>
>> Are you now requesting to only use diff.renames as the default if the
>> value is true or false but not if it is copy?  What should happen if
>> diff.renames is actually set to copy?  Should merge silently change
>> that to true, display a warning, error out, or something else?  Do you
>> have some other behavior for how to handle copy being inherited from
>> diff.renames you'd like to see?
>>
>> Can you write the documentation that clearly explains the exact
>> behavior you want?  That would kill two birds with one stone... :)
>
> I think demoting from copy to rename-only is a good idea, at least
> for now, because I do not believe we have figured out what we want
> to happen when we detect copied files are involved in a merge.
>
> But I am not sure if we even want to fail merge.renames=copy as an
> invalid configuration.  So my gut feeling of the best solution to
> the above is to do something like:
>
>  - whether the configuration comes from diff.renames or
>    merge.renames, turn *.renames=copy to true inside the merge
>    recursive machinery.
>
>  - document the fact in "git merge-recursive" documentation (or "git
>    merge" documentation) to say "_currently_ asking for rename
>    detection to find copies and renames will do the same
>    thing---copies are ignored", impliying "this might change in the
>    future", in the BUGS section.

Yes, I agree.  One more thing:

  - It may be best to avoid advertising "copies" as a vaild option for
merge.renames since it doesn't have any current practical use
anywhere.  (Remove the sentence 'If set to "copies" or "copy", Git
will detect copies, as well.' from the documentation)

My rationale for translating "copy" to "true" is a little different
than Junio's, though:

1) The reason we have configuration options around renames and copies
is primarily because they are expensive to compute.  So we let some
users specify that they don't want them, other users are willing to
pay for rename detection, and others are willing to pay for both
rename and copy detection.
2) If rename/copy detection were cheap, every part of git would just
compute whatever level of detection was relevant and use it.
3) The resolve and octopus merge strategies ignores diff.renames and
merge.renames, because they don't have logic to use any rename
information.  diff and log can use both renames and copies.  And the
recursive merge machinery is code which can use renames but not
copies.
4) Therefore, translating from "copy" to "true" inside the merge
recursive machinery is fine and not an error because we are using as
much detection information as is relevant to the algorithm and which
the user is willing to pay for.

To throw one more wrinkle in here, merge.renames could actually be set
to "copy" and make sense, because we compute diffs multiple times.
Twice within the recursive merge machinery (for which we'd want to
translate "copy" to "true"), and once for the diffstat at the end
(which comes from builtin/merge.c, and for which it could make sense
to detect copies).

(Kind of curious whether Junio agrees with my rationale or thinks I'm
out in left field with it...)

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Johannes Schindelin]

Hi,

On Thu, 26 Apr 2018, Elijah Newren wrote:

> On Thu, Apr 26, 2018 at 7:23 PM, Junio C Hamano <gitster@pobox.com> wrote:
> > Ben Peart <peartben@gmail.com> writes:
> >
> >> Color me puzzled. :)  The consensus was that the default value for
> >> merge.renames come from diff.renames.  diff.renames supports copy
> >> detection which means that merge.renames will inherit that value.  My
> >> assumption was that is what was intended so when I reimplemented it, I
> >> fully implemented it that way.
> >>
> >> Are you now requesting to only use diff.renames as the default if the
> >> value is true or false but not if it is copy?  What should happen if
> >> diff.renames is actually set to copy?  Should merge silently change
> >> that to true, display a warning, error out, or something else?  Do you
> >> have some other behavior for how to handle copy being inherited from
> >> diff.renames you'd like to see?
> >>
> >> Can you write the documentation that clearly explains the exact
> >> behavior you want?  That would kill two birds with one stone... :)
> >
> > I think demoting from copy to rename-only is a good idea, at least
> > for now, because I do not believe we have figured out what we want
> > to happen when we detect copied files are involved in a merge.
> >
> > But I am not sure if we even want to fail merge.renames=copy as an
> > invalid configuration.  So my gut feeling of the best solution to
> > the above is to do something like:
> >
> >  - whether the configuration comes from diff.renames or
> >    merge.renames, turn *.renames=copy to true inside the merge
> >    recursive machinery.
> >
> >  - document the fact in "git merge-recursive" documentation (or "git
> >    merge" documentation) to say "_currently_ asking for rename
> >    detection to find copies and renames will do the same
> >    thing---copies are ignored", impliying "this might change in the
> >    future", in the BUGS section.
> 
> Yes, I agree.

Guys, you argued long and hard that one config setting (diff.renames)
should magically imply another one (merge.renames), on the basis that they
essentially do the same.

And now you are suggesting that they *cannot* be essentially the same? Are
you agreeing on such a violation of the Law of Least Surprise?

Please, make up your mind. Inheriting merge.renames from diff.renames is
"magic" enough to be puzzling. Introducing that auto-demoting is just
simply creating a bad user experience. Please don't.

It is fine, of course, to admit at this stage that the whole magic idea of
letting merge.renames inherit from diff.renames was only half thought
through (we're in reviewing stage right now for the purpose of weighing
alternative approaches and trying to come up with the best solution after
all) and that we should go with Ben's original approach, which has the
rather huge and dramatic advantage of being very, very clear and easy to
understand to the user. We could even document that (and why) the 'copy'
value in merge.renames is not allowed for the time being.

Ciao,
Dscho

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Elijah Newren]

Hi Dsco,

On Fri, Apr 27, 2018 at 12:23 AM, Johannes Schindelin
<Johannes.Schindelin@gmx.de> wrote:
> Hi,
<snip>
>
> Guys, you argued long and hard that one config setting (diff.renames)
> should magically imply another one (merge.renames), on the basis that they
> essentially do the same.

I apologize for any frustration; I've probably caused a good deal of
it by repeatedly catching or noticing things after one review and then
bringing them up later.  I just didn't catch it all at first.

Your restatement of the basis of the argument doesn't match my
rationale, though.  My reasoning was that (1) the configuration
options exist to allow the user to specify how much of a performance
cost they're willing to pay, (2) the two options are separate because
some users might want to configure one more loosely or tightly than
the other, but (3) many users will want to just specify once how much
they are willing to pay without being forced to repeat themselves for
different parts of git (diff, merge, possible future commands).

I'll add to my former basis by stating that I think the diffstat at
the end of the merge should be controlled by these variables, even if
that's not implemented as part of Ben's series.  And both of those
configuration options clearly feed into whether that diffstat should
be doing rename or copy or no detection.  While they could be kept
separate options, forcing us to somehow decide which one overrides
which, I think it's much more natural if we already have merge.renames
inheriting from and overriding diff.renames.

> And now you are suggesting that they *cannot* be essentially the same? Are
> you agreeing on such a violation of the Law of Least Surprise?
>
> Please, make up your mind. Inheriting merge.renames from diff.renames is
> "magic" enough to be puzzling. Introducing that auto-demoting is just
> simply creating a bad user experience. Please don't.

From my view, resolve and octopus merges auto-demote diff.renames and
merge.renames to false.  In fact, they optimize the work involved in
that demotion by not even checking the value.  I think demoting "copy"
to "true" in the recursive merge machinery is no more surprising than
that is.  You can find more details to this argument in the portion of
my email that you responded to but snipped out.

But to add to that argument, if auto-demotion is a violation of the
Law of Least Surprise, as you claim, then naming this option
merge.renames is also a violation of the Law of Least Surprise.  You
would instead need to rename it to something like
merge.recursive.renames.  (And then when a new merge strategy comes
along that also does renames, we'll need to add a merge.ort.renames.)

> It is fine, of course, to admit at this stage that the whole magic idea of
> letting merge.renames inherit from diff.renames was only half thought
> through (we're in reviewing stage right now for the purpose of weighing
> alternative approaches and trying to come up with the best solution after
> all) and that we should go with Ben's original approach, which has the
> rather huge and dramatic advantage of being very, very clear and easy to
> understand to the user. We could even document that (and why) the 'copy'
> value in merge.renames is not allowed for the time being.

It was only half thought through, yes, at least by me.  But the more I
think through it, the more I like the inheritance personally.  I see
no problem with the demotion and think the inheritance has the
advantage of being easier to understand, because I see your proposal
as causing questions like:
  - "Why does merge.renameLimit inherit from diff.renameLimit but
merge.renames doesn't from diff.renames?"
  - "Why can't I just specify in one place that I {am, am not} willing
to pay for {full, both copy and} rename detection wherever it makes
sense?"
  - "How do I control the detection for the diffstat at the end of the merge?"


Hope that helps,
Elijah

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Eckhard Maaß]

On Fri, Apr 27, 2018 at 11:23:56AM +0900, Junio C Hamano wrote:
> I think demoting from copy to rename-only is a good idea, at least
> for now, because I do not believe we have figured out what we want
> to happen when we detect copied files are involved in a merge.

Does anyone know some threads concerning that topic? I tried to search
for it, but somehow "merge copy detection" did not find me useful
threads so far. I would be interested in that topic anyway, so I would
like to know what the ideas are that floated so far.

Greetings,
Eckhard

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Elijah Newren]

On Fri, Apr 27, 2018 at 11:37 AM, Eckhard Maaß
<eckhard.s.maass@googlemail.com> wrote:
> On Fri, Apr 27, 2018 at 11:23:56AM +0900, Junio C Hamano wrote:
>> I think demoting from copy to rename-only is a good idea, at least
>> for now, because I do not believe we have figured out what we want
>> to happen when we detect copied files are involved in a merge.
>
> Does anyone know some threads concerning that topic? I tried to search
> for it, but somehow "merge copy detection" did not find me useful
> threads so far. I would be interested in that topic anyway, so I would
> like to know what the ideas are that floated so far.
>

I doubt it has ever been discussed before this thread.  But, if you're
curious, I'll try to dump a few thoughts.  Let's say we have branches
A and B, and:
   A: modifies file z
   B: copies z to y

Should the modifications to z done in A propagate to both z and y?  If
not, what good is copy detection?  If so, then there are several
ramifications...

- If B not only copied z but also first modified it, then do we have
  potential conflicts with both z and y -- possibly the exact same
  conflicts, making the user resolve them repeatedly?

- What if A copied z to x?  Do changes to z propagate to all three of
  z and x and y?  Do changes to either x or y affect z?  Do they
  affect each other?

- If A deleted z, does that give us a copy/delete conflict for y?  Do
  we also have to worry about copy/add conflicts?  copy/add/delete?
  rename/copy (multiple variants)?  copy/copy?

- Extra degrees of freedom may mean new conflict types:

  - The extra degrees of freedom from renames introduced multiple new
    conflict types (e.g. rename/add, rename/rename(1to2),
    rename/rename(2to1)).

  - Directory rename detection added more (rename/rename/rename(1to3),
    rename/rename(Nto1), add/add/add, directory/file/file, n-fold
    transitive rename, etc.), -- and forced us to specify various
    rules to limit the possibility space so that conflicts could be
    representable in the index and understandable by the user.

  - I suspect adding copies would add new types of conflicts we
    haven't thought of yet.

The more I think about it, the more I think that attempting to detect
copies in a merge algorithm just doesn't make sense.  Anything I can
think of that someone might attempt to use detected copies for would
just surprise users in a bad way...and it'd open up a big can of edge
and corner case problems as well.

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Eckhard Maaß]

On Fri, Apr 27, 2018 at 01:23:20PM -0700, Elijah Newren wrote:
> I doubt it has ever been discussed before this thread.  But, if you're
> curious, I'll try to dump a few thoughts.

Thank you, I try to dump some of mine, too. Maybe let me first stress
that for me copy detection without --find-copies-harder is much more a
"find content extracted" (like methods being factored out). In a way
this is nearer to a rename than to a real copy.


> [...] Let's say we have branches
> A and B, and:
>    A: modifies file z
>    B: copies z to y
> 
> Should the modifications to z done in A propagate to both z and y?  If
> not, what good is copy detection?  If so, then there are several
> ramifications...

If one just assumes the most likely outcome is that something from z wad
factored out to y, it might just be sufficient to see whether the
modifications of the two branches apply cleanly - if A touched the parts
of B that have been factored out there would be a normal merge conflict
(where one could be nice and give a hint that some content was copied to
y on the B branch), if A did not touched the parts touched (or moved) by
B, then there is no problem. If A exactly deleted the content moved by
B, there will be no conflict - but this is seems to be strange anyway.

I admit that a "real" copy would get unnoticed that way. But the
semantics of such a copy isn't too clear for me either - did I copy the
other part to make it independent of the other or did I just employ a
copy and paste tactic? The former does not want the changes, the later
does. But I am happy catering to the former here.

To sum up:
- fail as before for conflicting merges, but give a hint that one has
  copied to quicken up resolution.


> - If B not only copied z but also first modified it, then do we have
>   potential conflicts with both z and y -- possibly the exact same
>   conflicts, making the user resolve them repeatedly?

With the above suggestion, if there are conflicts, you fail and give a
hint.

> - What if A copied z to x?  Do changes to z propagate to all three of
>   z and x and y?  Do changes to either x or y affect z?  Do they
>   affect each other?

A copy on branch to x and one another to y seems strange even if z
merges cleanly. Did both sides try to factor the same thing out to
different files? Or did they try to make something independent, but
managed to make it to different files? For this I would be inclined to
just suggest fail with a copy/copy(somewhere else). But this is a real
corner case after all. Has anyone seen just thing in practice?

> - If A deleted z, does that give us a copy/delete conflict for y?  Do
>   we also have to worry about copy/add conflicts?  copy/add/delete?
>   rename/copy (multiple variants)?  copy/copy?

We do have the modified/deleted conflict where we could hint that
content also has been copied and then not try to do more.

> - Extra degrees of freedom may mean new conflict types:
> 
>   - The extra degrees of freedom from renames introduced multiple new
>     conflict types (e.g. rename/add, rename/rename(1to2),
>     rename/rename(2to1)).

For renaming one side and coping the other, I would think doing the same
as above is sensible enough: if there are conflicts one can give an
additional hint of the one part having been copied, but not change the
kind of conflicts much.

> The more I think about it, the more I think that attempting to detect
> copies in a merge algorithm just doesn't make sense.  Anything I can
> think of that someone might attempt to use detected copies for would
> just surprise users in a bad way...

Hm, it didn't sound like that. Would you think that users would be
surprised by my suggestions? Or are they all too corner casey to be
worth implementing anyway?

Greetings,
Eckhard

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Elijah Newren]

Hi Eckhard,

On Mon, Apr 30, 2018 at 1:03 AM, Eckhard Maaß
<eckhard.s.maass@googlemail.com> wrote:
> On Fri, Apr 27, 2018 at 01:23:20PM -0700, Elijah Newren wrote:
>> I doubt it has ever been discussed before this thread.  But, if you're
>> curious, I'll try to dump a few thoughts.
>
> Thank you, I try to dump some of mine, too. Maybe let me first stress
> that for me copy detection without --find-copies-harder is much more a
> "find content extracted" (like methods being factored out). In a way
> this is nearer to a rename than to a real copy.

Ooh, if you wanted to detect movement of code between files (as blame
does, I think searches for PICKAXE_BLAME_MOVE would point you in the
right direction) and then try to use that during merge to allow
applied changes to move with the code, that would be awesome.
Expensive, and might be a lot of work to wire it all up, but it'd be
very interesting.  I was only discussing DIFF_DETECT_COPY in my
previous email, which was all about duplicating entire files; that's
something I don't see utility in right now for resolving merges.

<snip>
> I admit that a "real" copy would get unnoticed that way. But the
> semantics of such a copy isn't too clear for me either - did I copy the
> other part to make it independent of the other or did I just employ a
> copy and paste tactic? The former does not want the changes, the later
> does. But I am happy catering to the former here.

Right, if you have to assume that the copy was made to make the code
independent, then there's no value for merge resolution to having
detected the copy in the first place.  That has the advantage of
side-stepping the possible new edge and corner cases I mentioned in
the rest of my email, but it means we shouldn't even spend time
detecting copies -- whether whole file (via DIFF_DETECT_COPY) or
individual lines (via PICKAXE_BLAME_COPY and variants).


Elijah

Re: [PATCH v3 2/3] merge: Add merge.renames config setting

[Elijah Newren]

On Thu, Apr 26, 2018 at 5:54 PM, Ben Peart <peartben@gmail.com> wrote:
> On 4/26/2018 6:52 PM, Elijah Newren wrote:
>> On Thu, Apr 26, 2018 at 1:52 PM, Ben Peart <Ben.Peart@microsoft.com>
>> wrote:
>>
>>> diff --git a/merge-recursive.h b/merge-recursive.h
>>> index 80d69d1401..0c5f7eff98 100644
>>> --- a/merge-recursive.h
>>> +++ b/merge-recursive.h
>>> @@ -17,7 +17,8 @@ struct merge_options {
>>>          unsigned renormalize : 1;
>>>          long xdl_opts;
>>>          int verbosity;
>>> -       int detect_rename;
>>> +       int diff_detect_rename;
>>> +       int merge_detect_rename;
>>>          int diff_rename_limit;
>>>          int merge_rename_limit;
>>>          int rename_score;
>>> @@ -28,6 +29,11 @@ struct merge_options {
>>>          struct hashmap current_file_dir_set;
>>>          struct string_list df_conflict_file_set;
>>>   };
>>> +inline int merge_detect_rename(struct merge_options *o)
>>> +{
>>> +       return o->merge_detect_rename >= 0 ? o->merge_detect_rename :
>>> +               o->diff_detect_rename >= 0 ? o->diff_detect_rename : 1;
>>> +}
>>
>>
>> Why did you split o->detect_rename into two fields?  You then
>> recombine them in merge_detect_rename(), and after initial setup only
>> ever access them through that function.  Having two fields worries me
>> that people will accidentally introduce bugs by using one of them
>> instead of the merge_detect_rename() function.  Is there a reason you
>> decided against having the initial setup just set a single value and
>> then use it directly?
>>
> The setup of this value is split into 3 places that may or may not all get
> called.  The initial values, the values that come from the config settings
> and then any values passed on the command line.
>
> Because the merge value can now inherit from the diff value, you only know
> the final value after you have received all possible inputs.  That makes it
> necessary to be a calculated value.
>
> If you look at diff_rename_limit/merge_rename_limit, detect_rename follow
> the same pattern for the same reasons.  It turns out detect_rename was a
> little more complex because it is used in 3 different locations (vs just
> one) which is why I wrapped the inheritance logic into the helper function
> merge_detect_rename().

Ah, you're following the precedent set by
diff_rename_limit/merge_rename_limit; that makes sense.  Thanks for
the explanation.  I believe another possibility here is that for both
the {merge,diff}_rename_limit pair of variables and the
{diff,merge}_renames pair of variables, since the code parses all
inputs before ever using the result, we could calculate the result
once and store it rather than storing the constituent pieces of the
calculation.  That would also prevent people from trying to use one of
the pieces of the calculation instead of treating it as a coherent
whole.  However, while I would have preferred that the rename_limit
pair of variables also went away in favor of just one field which is
updated as it parses each input option, what you have is fine for this
series.

(no subject)

[Elijah Newren]

From: Elijah Newren <newren@gmail.com>

On Thu, Apr 26, 2018 at 5:54 PM, Ben Peart <peartben@gmail.com> wrote:

> Can you write the documentation that clearly explains the exact behavior you
> want?  That would kill two birds with one stone... :)

Sure, something like the following is what I envision, and I've tried to
include the suggestion from Junio to document the copy behavior in the
merge-recursive documentation.

-- 8< --
Subject: [PATCH] fixup! merge: Add merge.renames config setting

---
 Documentation/merge-config.txt     | 3 +--
 Documentation/merge-strategies.txt | 5 +++--
 merge-recursive.c                  | 8 ++++++++
 3 files changed, 12 insertions(+), 4 deletions(-)

diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
index 59848e5634..662c2713ca 100644
--- a/Documentation/merge-config.txt
+++ b/Documentation/merge-config.txt
@@ -41,8 +41,7 @@ merge.renameLimit::
 merge.renames::
 	Whether and how Git detects renames.  If set to "false",
 	rename detection is disabled. If set to "true", basic rename
-	detection is enabled.  If set to "copies" or "copy", Git will
-	detect copies, as well.  Defaults to the value of diff.renames.
+	detection is enabled.  Defaults to the value of diff.renames.
 
 merge.renormalize::
 	Tell Git that canonical representation of files in the
diff --git a/Documentation/merge-strategies.txt b/Documentation/merge-strategies.txt
index 1e0728aa12..aa66cbe41e 100644
--- a/Documentation/merge-strategies.txt
+++ b/Documentation/merge-strategies.txt
@@ -23,8 +23,9 @@ recursive::
 	causing mismerges by tests done on actual merge commits
 	taken from Linux 2.6 kernel development history.
 	Additionally this can detect and handle merges involving
-	renames.  This is the default merge strategy when
-	pulling or merging one branch.
+	renames, but currently cannot make use of detected
+	copies.  This is the default merge strategy when pulling
+	or merging one branch.
 +
 The 'recursive' strategy can take the following options:
 
diff --git a/merge-recursive.c b/merge-recursive.c
index 6cc4404144..b618f134d2 100644
--- a/merge-recursive.c
+++ b/merge-recursive.c
@@ -564,6 +564,14 @@ static struct string_list *get_renames(struct merge_options *o,
 	opts.flags.recursive = 1;
 	opts.flags.rename_empty = 0;
 	opts.detect_rename = merge_detect_rename(o);
+	/*
+	 * We do not have logic to handle the detection of copies.  In
+	 * fact, it may not even make sense to add such logic: would we
+	 * really want a change to a base file to be propagated through
+	 * multiple other files by a merge?
+	 */
+	if (opts.detect_rename > DIFF_DETECT_RENAME)
+		opts.detect_rename = DIFF_DETECT_RENAME;
 	opts.rename_limit = o->merge_rename_limit >= 0 ? o->merge_rename_limit :
 			    o->diff_rename_limit >= 0 ? o->diff_rename_limit :
 			    1000;
-- 
2.17.0.294.g8e3b83c0e3

Re:

[Ben Peart]

On 4/27/2018 2:19 PM, Elijah Newren wrote:
> From: Elijah Newren <newren@gmail.com>
> 
> On Thu, Apr 26, 2018 at 5:54 PM, Ben Peart <peartben@gmail.com> wrote:
> 
>> Can you write the documentation that clearly explains the exact behavior you
>> want?  That would kill two birds with one stone... :)
> 
> Sure, something like the following is what I envision, and I've tried to
> include the suggestion from Junio to document the copy behavior in the
> merge-recursive documentation.
> 
> -- 8< --
> Subject: [PATCH] fixup! merge: Add merge.renames config setting
> 
> ---
>   Documentation/merge-config.txt     | 3 +--
>   Documentation/merge-strategies.txt | 5 +++--
>   merge-recursive.c                  | 8 ++++++++
>   3 files changed, 12 insertions(+), 4 deletions(-)
> 
> diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
> index 59848e5634..662c2713ca 100644
> --- a/Documentation/merge-config.txt
> +++ b/Documentation/merge-config.txt
> @@ -41,8 +41,7 @@ merge.renameLimit::
>   merge.renames::
>   	Whether and how Git detects renames.  If set to "false",
>   	rename detection is disabled. If set to "true", basic rename
> -	detection is enabled.  If set to "copies" or "copy", Git will
> -	detect copies, as well.  Defaults to the value of diff.renames.
> +	detection is enabled.  Defaults to the value of diff.renames.
>   
>   merge.renormalize::
>   	Tell Git that canonical representation of files in the
> diff --git a/Documentation/merge-strategies.txt b/Documentation/merge-strategies.txt
> index 1e0728aa12..aa66cbe41e 100644
> --- a/Documentation/merge-strategies.txt
> +++ b/Documentation/merge-strategies.txt
> @@ -23,8 +23,9 @@ recursive::
>   	causing mismerges by tests done on actual merge commits
>   	taken from Linux 2.6 kernel development history.
>   	Additionally this can detect and handle merges involving
> -	renames.  This is the default merge strategy when
> -	pulling or merging one branch.
> +	renames, but currently cannot make use of detected
> +	copies.  This is the default merge strategy when pulling
> +	or merging one branch.
>   +
>   The 'recursive' strategy can take the following options:
>   
> diff --git a/merge-recursive.c b/merge-recursive.c
> index 6cc4404144..b618f134d2 100644
> --- a/merge-recursive.c
> +++ b/merge-recursive.c
> @@ -564,6 +564,14 @@ static struct string_list *get_renames(struct merge_options *o,
>   	opts.flags.recursive = 1;
>   	opts.flags.rename_empty = 0;
>   	opts.detect_rename = merge_detect_rename(o);
> +	/*
> +	 * We do not have logic to handle the detection of copies.  In
> +	 * fact, it may not even make sense to add such logic: would we
> +	 * really want a change to a base file to be propagated through
> +	 * multiple other files by a merge?
> +	 */
> +	if (opts.detect_rename > DIFF_DETECT_RENAME)
> +		opts.detect_rename = DIFF_DETECT_RENAME;
>   	opts.rename_limit = o->merge_rename_limit >= 0 ? o->merge_rename_limit :
>   			    o->diff_rename_limit >= 0 ? o->diff_rename_limit :
>   			    1000;
> 

Thanks Elijah. I've applied this patch and reviewed and tested it.  It 
works and addresses the concerns around the settings inheritance from 
diff.renames.  I still _prefer_ the simpler model that doesn't do the 
partial inheritance but I can use this model as well.

I'm unsure on the protocol here.  Should I incorporate this patch and 
submit a reroll or can it just be applied as is?

Re:

[Elijah Newren]

On Mon, Apr 30, 2018 at 6:11 AM, Ben Peart <peartben@gmail.com> wrote:
> On 4/27/2018 2:19 PM, Elijah Newren wrote:
>>
>> From: Elijah Newren <newren@gmail.com>
>>
>> On Thu, Apr 26, 2018 at 5:54 PM, Ben Peart <peartben@gmail.com> wrote:
>>
>>> Can you write the documentation that clearly explains the exact behavior
>>> you
>>> want?  That would kill two birds with one stone... :)
>>
>>
>> Sure, something like the following is what I envision, and I've tried to
>> include the suggestion from Junio to document the copy behavior in the
>> merge-recursive documentation.
>>
<snip>
>
> Thanks Elijah. I've applied this patch and reviewed and tested it.  It works
> and addresses the concerns around the settings inheritance from
> diff.renames.  I still _prefer_ the simpler model that doesn't do the
> partial inheritance but I can use this model as well.
>
> I'm unsure on the protocol here.  Should I incorporate this patch and submit
> a reroll or can it just be applied as is?

I suspect you'll want to re-roll anyway, to base your series on
en/rename-directory-detection-reboot instead of on master.  (Junio
plans to merge it down to next, and your series has four different
merge conflicts with it.)

There are two other loose ends with this series that Junio will need
to weigh in on:

- I'm obviously a strong proponent of the inherited setting, but Junio
may change his mind after reading Dscho's arguments against it (or
after reading my arguments for it).

- I like the setting as-is, and think we could allow a "copy" setting
for merge.renames to specify that the post-merge diffstat should
detect copies (not part of your series, but a useful addition I'd like
to tackle afterwards).  However, Junio had comments in
xmqqwox19ohw.fsf@gitster-ct.c.googlers.com about merge.renames
handling the scoring as well, like -Xfind-renames.  Those sound
incompatible to me for a single setting, and I'm unsure if Junio would
resolve them the way I do or still feels strongly about the scoring.

Re:

[Ben Peart]

On 4/30/2018 12:12 PM, Elijah Newren wrote:
> On Mon, Apr 30, 2018 at 6:11 AM, Ben Peart <peartben@gmail.com> wrote:
>> On 4/27/2018 2:19 PM, Elijah Newren wrote:
>>>
>>> From: Elijah Newren <newren@gmail.com>
>>>
>>> On Thu, Apr 26, 2018 at 5:54 PM, Ben Peart <peartben@gmail.com> wrote:
>>>
>>>> Can you write the documentation that clearly explains the exact behavior
>>>> you
>>>> want?  That would kill two birds with one stone... :)
>>>
>>>
>>> Sure, something like the following is what I envision, and I've tried to
>>> include the suggestion from Junio to document the copy behavior in the
>>> merge-recursive documentation.
>>>
> <snip>
>>
>> Thanks Elijah. I've applied this patch and reviewed and tested it.  It works
>> and addresses the concerns around the settings inheritance from
>> diff.renames.  I still _prefer_ the simpler model that doesn't do the
>> partial inheritance but I can use this model as well.
>>
>> I'm unsure on the protocol here.  Should I incorporate this patch and submit
>> a reroll or can it just be applied as is?
> 
> I suspect you'll want to re-roll anyway, to base your series on
> en/rename-directory-detection-reboot instead of on master.  (Junio
> plans to merge it down to next, and your series has four different
> merge conflicts with it.)
> 
> There are two other loose ends with this series that Junio will need
> to weigh in on:
> 
> - I'm obviously a strong proponent of the inherited setting, but Junio
> may change his mind after reading Dscho's arguments against it (or
> after reading my arguments for it).
> 
> - I like the setting as-is, and think we could allow a "copy" setting
> for merge.renames to specify that the post-merge diffstat should
> detect copies (not part of your series, but a useful addition I'd like
> to tackle afterwards).  However, Junio had comments in
> xmqqwox19ohw.fsf@gitster-ct.c.googlers.com about merge.renames
> handling the scoring as well, like -Xfind-renames.  Those sound
> incompatible to me for a single setting, and I'm unsure if Junio would
> resolve them the way I do or still feels strongly about the scoring.
> 

I think this patch series (including Elijah's fixup!) improves the 
situation from where we were and it provides the necessary functionality 
to solve the problem I started out to solve.  While there are other 
changes that could be made, I think they should be done in separate 
follow up patches.

I'm happy to reroll this incorporating the fixup! so that we can make 
progress.  Junio, would you prefer I reroll this based on 
en/rename-directory-detection-reboot or master?

[PATCH v3 3/3] merge: pass aggressive when rename detection is turned off

[Ben Peart]

Set aggressive flag in git_merge_trees() when rename detection is turned off.
This allows read_tree() to auto resolve more cases that would have otherwise
been handled by the rename detection.

Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Ben Peart <benpeart@microsoft.com>
---
 merge-recursive.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/merge-recursive.c b/merge-recursive.c
index 2637d34d87..6cc4404144 100644
--- a/merge-recursive.c
+++ b/merge-recursive.c
@@ -276,6 +276,7 @@ static void init_tree_desc_from_tree(struct tree_desc *desc, struct tree *tree)
 }
 
 static int git_merge_trees(int index_only,
+			   int aggressive,
 			   struct tree *common,
 			   struct tree *head,
 			   struct tree *merge)
@@ -294,6 +295,7 @@ static int git_merge_trees(int index_only,
 	opts.fn = threeway_merge;
 	opts.src_index = &the_index;
 	opts.dst_index = &the_index;
+	opts.aggressive = aggressive;
 	setup_unpack_trees_porcelain(&opts, "merge");
 
 	init_tree_desc_from_tree(t+0, common);
@@ -1993,7 +1995,7 @@ int merge_trees(struct merge_options *o,
 		return 1;
 	}
 
-	code = git_merge_trees(o->call_depth, common, head, merge);
+	code = git_merge_trees(o->call_depth, !merge_detect_rename(o), common, head, merge);
 
 	if (code != 0) {
 		if (show(o, 4) || o->call_depth)
-- 
2.17.0.windows.1

Re: [PATCH v3 3/3] merge: pass aggressive when rename detection is turned off

[Elijah Newren]

On Thu, Apr 26, 2018 at 1:52 PM, Ben Peart <Ben.Peart@microsoft.com> wrote:
> Set aggressive flag in git_merge_trees() when rename detection is turned off.
> This allows read_tree() to auto resolve more cases that would have otherwise
> been handled by the rename detection.
>
> Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> Signed-off-by: Ben Peart <benpeart@microsoft.com>
> ---
>  merge-recursive.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/merge-recursive.c b/merge-recursive.c
> index 2637d34d87..6cc4404144 100644
> --- a/merge-recursive.c
> +++ b/merge-recursive.c
> @@ -276,6 +276,7 @@ static void init_tree_desc_from_tree(struct tree_desc *desc, struct tree *tree)
>  }
>
>  static int git_merge_trees(int index_only,
> +                          int aggressive,
>                            struct tree *common,
>                            struct tree *head,
>                            struct tree *merge)
> @@ -294,6 +295,7 @@ static int git_merge_trees(int index_only,
>         opts.fn = threeway_merge;
>         opts.src_index = &the_index;
>         opts.dst_index = &the_index;
> +       opts.aggressive = aggressive;
>         setup_unpack_trees_porcelain(&opts, "merge");
>
>         init_tree_desc_from_tree(t+0, common);
> @@ -1993,7 +1995,7 @@ int merge_trees(struct merge_options *o,
>                 return 1;
>         }
>
> -       code = git_merge_trees(o->call_depth, common, head, merge);
> +       code = git_merge_trees(o->call_depth, !merge_detect_rename(o), common, head, merge);
>
>         if (code != 0) {
>                 if (show(o, 4) || o->call_depth)
> --
> 2.17.0.windows.1

Patch looks fine but as a heads up -- since merge_options is a
parameter in git_merge_trees after the
en/rename-directory-detection-reboot lands, we'll be able to switch
this patch to set opts.aggressive directly instead of needing to pass
it in as a parameter.

Re: [PATCH v3 0/3] add merge.renames config setting

[Elijah Newren]

Hi Ben,

On Thu, Apr 26, 2018 at 1:52 PM, Ben Peart <Ben.Peart@microsoft.com> wrote:
> This is a complete rewrite based on the feedback from earlier patches.

Thanks for pushing forward on this.

> Update the documentation to better indicate command line options that override
> various config settings related to merge.
>
> Add a new config merge.renames setting to to control the rename detection
> behavior of merge.  This setting will default to the value of diff.renames.
>
> Also adds logic so that when rename detection is turned off, the aggressive
> flag is passed to read_tree() so that it can auto resolve more cases that would
> have been handled by rename detection.
>
> For the repro that I have been using this drops the merge time from ~1 hour to
> ~5 minutes and the unmerged entries goes down from ~40,000 to 1.
>
> Helped-by: Kevin Willford <kewillf@microsoft.com>
> Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> Signed-off-by: Ben Peart <Ben.Peart@microsoft.com>
>
> Base Ref: master

We may need to figure out how to coordinate amongst a few topics.
Looking over your patches, there are going to be a few conflicts with
en/rename-directory-detection-reboot, so this won't apply to pu.
Martin's series to introduce clear_unpack_trees_porcelain()[1], which
he was waiting to submit until mine went through, will also conflict
with this, if he uses the changes I suggested for the handling in
merge-recursive[2].  These aren't major conflicts, but I'm just
flagging it.

[1] https://public-inbox.org/git/cover.1524545557.git.martin.agren@gmail.com/
[2] https://public-inbox.org/git/20180424162939.20956-1-newren@gmail.com/

[PATCH v4 0/3] add additional config settings for merge

[Ben Peart]

This version incorporates Elijah's fixup patch and is now based on
en/rename-directory-detection in next.

Base Ref: en/rename-directory-detection
Web-Diff: https://github.com/benpeart/git/commit/16b175c25f
Checkout: git fetch https://github.com/benpeart/git merge-renames-v4 && git checkout 16b175c25f

### Patches

Ben Peart (3):
  merge: update documentation for {merge,diff}.renameLimit
  merge: Add merge.renames config setting
  merge: pass aggressive when rename detection is turned off

 Documentation/diff-config.txt             |  3 ++-
 Documentation/merge-config.txt            |  8 +++++-
 Documentation/merge-strategies.txt        | 11 +++++---
 diff.c                                    |  2 +-
 diff.h                                    |  1 +
 merge-recursive.c                         | 31 ++++++++++++++++++-----
 merge-recursive.h                         |  8 +++++-
 t/t3034-merge-recursive-rename-options.sh | 18 +++++++++++++
 8 files changed, 68 insertions(+), 14 deletions(-)


base-commit: c5b761fb2711542073cf1906c0e86a34616b79ae
-- 
2.17.0.windows.1

[PATCH v4 1/3] merge: update documentation for {merge,diff}.renameLimit

[Ben Peart]

Update the documentation to better indicate that the renameLimit setting is
ignored if rename detection is turned off via command line options or config
settings.

Signed-off-by: Ben Peart <benpeart@microsoft.com>
---
 Documentation/diff-config.txt  | 3 ++-
 Documentation/merge-config.txt | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/Documentation/diff-config.txt b/Documentation/diff-config.txt
index 5ca942ab5e..77caa66c2f 100644
--- a/Documentation/diff-config.txt
+++ b/Documentation/diff-config.txt
@@ -112,7 +112,8 @@ diff.orderFile::
 
 diff.renameLimit::
 	The number of files to consider when performing the copy/rename
-	detection; equivalent to the 'git diff' option `-l`.
+	detection; equivalent to the 'git diff' option `-l`. This setting
+	has no effect if rename detection is turned off.
 
 diff.renames::
 	Whether and how Git detects renames.  If set to "false",
diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
index 12b6bbf591..48ee3bce77 100644
--- a/Documentation/merge-config.txt
+++ b/Documentation/merge-config.txt
@@ -35,7 +35,8 @@ include::fmt-merge-msg-config.txt[]
 merge.renameLimit::
 	The number of files to consider when performing rename detection
 	during a merge; if not specified, defaults to the value of
-	diff.renameLimit.
+	diff.renameLimit. This setting has no effect if rename detection
+	is turned off.
 
 merge.renormalize::
 	Tell Git that canonical representation of files in the
-- 
2.17.0.windows.1

[PATCH v4 2/3] merge: Add merge.renames config setting

[Ben Peart]

Add the ability to control rename detection for merge via a config setting.
This setting behaves the same and defaults to the value of diff.renames but only
applies to merge.

Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Helped-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Ben Peart <benpeart@microsoft.com>
---
 Documentation/merge-config.txt            |  5 ++++
 Documentation/merge-strategies.txt        | 11 ++++++---
 diff.c                                    |  2 +-
 diff.h                                    |  1 +
 merge-recursive.c                         | 30 ++++++++++++++++++-----
 merge-recursive.h                         |  8 +++++-
 t/t3034-merge-recursive-rename-options.sh | 18 ++++++++++++++
 7 files changed, 63 insertions(+), 12 deletions(-)

diff --git a/Documentation/merge-config.txt b/Documentation/merge-config.txt
index 48ee3bce77..662c2713ca 100644
--- a/Documentation/merge-config.txt
+++ b/Documentation/merge-config.txt
@@ -38,6 +38,11 @@ merge.renameLimit::
 	diff.renameLimit. This setting has no effect if rename detection
 	is turned off.
 
+merge.renames::
+	Whether and how Git detects renames.  If set to "false",
+	rename detection is disabled. If set to "true", basic rename
+	detection is enabled.  Defaults to the value of diff.renames.
+
 merge.renormalize::
 	Tell Git that canonical representation of files in the
 	repository has changed over time (e.g. earlier commits record
diff --git a/Documentation/merge-strategies.txt b/Documentation/merge-strategies.txt
index fd5d748d1b..30acc99232 100644
--- a/Documentation/merge-strategies.txt
+++ b/Documentation/merge-strategies.txt
@@ -23,8 +23,9 @@ recursive::
 	causing mismerges by tests done on actual merge commits
 	taken from Linux 2.6 kernel development history.
 	Additionally this can detect and handle merges involving
-	renames.  This is the default merge strategy when
-	pulling or merging one branch.
+	renames, but currently cannot make use of detected
+	copies.  This is the default merge strategy when pulling
+	or merging one branch.
 +
 The 'recursive' strategy can take the following options:
 
@@ -84,12 +85,14 @@ no-renormalize;;
 	`merge.renormalize` configuration variable.
 
 no-renames;;
-	Turn off rename detection.
+	Turn off rename detection. This overrides the `merge.renames`
+	configuration variable.
 	See also linkgit:git-diff[1] `--no-renames`.
 
 find-renames[=<n>];;
 	Turn on rename detection, optionally setting the similarity
-	threshold.  This is the default.
+	threshold.  This is the default. This overrides the
+	'merge.renames' configuration variable.
 	See also linkgit:git-diff[1] `--find-renames`.
 
 rename-threshold=<n>;;
diff --git a/diff.c b/diff.c
index fb22b19f09..e744b35cdc 100644
--- a/diff.c
+++ b/diff.c
@@ -177,7 +177,7 @@ static int parse_submodule_params(struct diff_options *options, const char *valu
 	return 0;
 }
 
-static int git_config_rename(const char *var, const char *value)
+int git_config_rename(const char *var, const char *value)
 {
 	if (!value)
 		return DIFF_DETECT_RENAME;
diff --git a/diff.h b/diff.h
index 7cf276f077..966fc8fce6 100644
--- a/diff.h
+++ b/diff.h
@@ -321,6 +321,7 @@ extern int git_diff_ui_config(const char *var, const char *value, void *cb);
 extern void diff_setup(struct diff_options *);
 extern int diff_opt_parse(struct diff_options *, const char **, int, const char *);
 extern void diff_setup_done(struct diff_options *);
+extern int git_config_rename(const char *var, const char *value);
 
 #define DIFF_DETECT_RENAME	1
 #define DIFF_DETECT_COPY	2
diff --git a/merge-recursive.c b/merge-recursive.c
index 5f42c677d5..372ffbbacc 100644
--- a/merge-recursive.c
+++ b/merge-recursive.c
@@ -1524,7 +1524,15 @@ static struct diff_queue_struct *get_diffpairs(struct merge_options *o,
 	diff_setup(&opts);
 	opts.flags.recursive = 1;
 	opts.flags.rename_empty = 0;
-	opts.detect_rename = DIFF_DETECT_RENAME;
+	opts.detect_rename = merge_detect_rename(o);
+	/*
+	 * We do not have logic to handle the detection of copies.  In
+	 * fact, it may not even make sense to add such logic: would we
+	 * really want a change to a base file to be propagated through
+	 * multiple other files by a merge?
+	 */
+	if (opts.detect_rename > DIFF_DETECT_RENAME)
+		opts.detect_rename = DIFF_DETECT_RENAME;
 	opts.rename_limit = o->merge_rename_limit >= 0 ? o->merge_rename_limit :
 			    o->diff_rename_limit >= 0 ? o->diff_rename_limit :
 			    1000;
@@ -2564,7 +2572,7 @@ static int handle_renames(struct merge_options *o,
 	ri->head_renames = NULL;
 	ri->merge_renames = NULL;
 
-	if (!o->detect_rename)
+	if (!merge_detect_rename(o))
 		return 1;
 
 	head_pairs = get_diffpairs(o, common, head);
@@ -3241,9 +3249,18 @@ int merge_recursive_generic(struct merge_options *o,
 
 static void merge_recursive_config(struct merge_options *o)
 {
+	char *value = NULL;
 	git_config_get_int("merge.verbosity", &o->verbosity);
 	git_config_get_int("diff.renamelimit", &o->diff_rename_limit);
 	git_config_get_int("merge.renamelimit", &o->merge_rename_limit);
+	if (!git_config_get_string("diff.renames", &value)) {
+		o->diff_detect_rename = git_config_rename("diff.renames", value);
+		free(value);
+	}
+	if (!git_config_get_string("merge.renames", &value)) {
+		o->merge_detect_rename = git_config_rename("merge.renames", value);
+		free(value);
+	}
 	git_config(git_xmerge_config, NULL);
 }
 
@@ -3256,7 +3273,8 @@ void init_merge_options(struct merge_options *o)
 	o->diff_rename_limit = -1;
 	o->merge_rename_limit = -1;
 	o->renormalize = 0;
-	o->detect_rename = 1;
+	o->diff_detect_rename = -1;
+	o->merge_detect_rename = -1;
 	merge_recursive_config(o);
 	merge_verbosity = getenv("GIT_MERGE_VERBOSITY");
 	if (merge_verbosity)
@@ -3307,16 +3325,16 @@ int parse_merge_opt(struct merge_options *o, const char *s)
 	else if (!strcmp(s, "no-renormalize"))
 		o->renormalize = 0;
 	else if (!strcmp(s, "no-renames"))
-		o->detect_rename = 0;
+		o->merge_detect_rename = 0;
 	else if (!strcmp(s, "find-renames")) {
-		o->detect_rename = 1;
+		o->merge_detect_rename = 1;
 		o->rename_score = 0;
 	}
 	else if (skip_prefix(s, "find-renames=", &arg) ||
 		 skip_prefix(s, "rename-threshold=", &arg)) {
 		if ((o->rename_score = parse_rename_score(&arg)) == -1 || *arg != 0)
 			return -1;
-		o->detect_rename = 1;
+		o->merge_detect_rename = 1;
 	}
 	else
 		return -1;
diff --git a/merge-recursive.h b/merge-recursive.h
index d863cf8867..c1d9b5b3d9 100644
--- a/merge-recursive.h
+++ b/merge-recursive.h
@@ -18,7 +18,8 @@ struct merge_options {
 	unsigned renormalize : 1;
 	long xdl_opts;
 	int verbosity;
-	int detect_rename;
+	int diff_detect_rename;
+	int merge_detect_rename;
 	int diff_rename_limit;
 	int merge_rename_limit;
 	int rename_score;
@@ -55,6 +56,11 @@ struct collision_entry {
 	struct string_list source_files;
 	unsigned reported_already:1;
 };
+inline int merge_detect_rename(struct merge_options *o)
+{
+	return o->merge_detect_rename >= 0 ? o->merge_detect_rename :
+		o->diff_detect_rename >= 0 ? o->diff_detect_rename : 1;
+}
 
 /* merge_trees() but with recursive ancestor consolidation */
 int merge_recursive(struct merge_options *o,
diff --git a/t/t3034-merge-recursive-rename-options.sh b/t/t3034-merge-recursive-rename-options.sh
index b9c4028496..3d9fae68c4 100755
--- a/t/t3034-merge-recursive-rename-options.sh
+++ b/t/t3034-merge-recursive-rename-options.sh
@@ -309,4 +309,22 @@ test_expect_success 'last wins in --find-renames=<m> --rename-threshold=<n>' '
 	check_threshold_0
 '
 
+test_expect_success 'merge.renames disables rename detection' '
+	git read-tree --reset -u HEAD &&
+	git -c merge.renames=false merge-recursive $tail &&
+	check_no_renames
+'
+
+test_expect_success 'merge.renames defaults to diff.renames' '
+	git read-tree --reset -u HEAD &&
+	git -c diff.renames=false merge-recursive $tail &&
+	check_no_renames
+'
+
+test_expect_success 'merge.renames overrides diff.renames' '
+	git read-tree --reset -u HEAD &&
+	test_must_fail git -c diff.renames=false -c merge.renames=true merge-recursive $tail &&
+	$check_50
+'
+
 test_done
-- 
2.17.0.windows.1

Re: [PATCH v4 2/3] merge: Add merge.renames config setting

[Junio C Hamano]

Ben Peart <Ben.Peart@microsoft.com> writes:

I'd downcase the verb on the subject.

> Add the ability to control rename detection for merge via a config setting.
> This setting behaves the same and defaults to the value of diff.renames but only
> applies to merge.
>
> Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
> Helped-by: Elijah Newren <newren@gmail.com>
> Signed-off-by: Ben Peart <benpeart@microsoft.com>
> ...
> diff --git a/merge-recursive.h b/merge-recursive.h
> index d863cf8867..c1d9b5b3d9 100644
> --- a/merge-recursive.h
> +++ b/merge-recursive.h
> @@ -55,6 +56,11 @@ struct collision_entry {
>  	struct string_list source_files;
>  	unsigned reported_already:1;
>  };
> +inline int merge_detect_rename(struct merge_options *o)
> +{
> +	return o->merge_detect_rename >= 0 ? o->merge_detect_rename :
> +		o->diff_detect_rename >= 0 ? o->diff_detect_rename : 1;
> +}

I'll tweak the above to leave a blank before the function, and make
it "static inline", to ensure that the output from

    $ git grep -e '\<inline\>' --and --not -e 'static inline' -- \*.h

is empty.

[PATCH v4 3/3] merge: pass aggressive when rename detection is turned off

[Ben Peart]

Set aggressive flag in git_merge_trees() when rename detection is turned off.
This allows read_tree() to auto resolve more cases that would have otherwise
been handled by the rename detection.

Reviewed-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Ben Peart <benpeart@microsoft.com>
---
 merge-recursive.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/merge-recursive.c b/merge-recursive.c
index 372ffbbacc..cea054cfd4 100644
--- a/merge-recursive.c
+++ b/merge-recursive.c
@@ -355,6 +355,7 @@ static int git_merge_trees(struct merge_options *o,
 	o->unpack_opts.fn = threeway_merge;
 	o->unpack_opts.src_index = &the_index;
 	o->unpack_opts.dst_index = &the_index;
+	o->unpack_opts.aggressive = !merge_detect_rename(o);
 	setup_unpack_trees_porcelain(&o->unpack_opts, "merge");
 
 	init_tree_desc_from_tree(t+0, common);
-- 
2.17.0.windows.1

Re: [PATCH v4 0/3] add additional config settings for merge

[Elijah Newren]

Hi Ben,

Thanks for your persistence.

On Wed, May 2, 2018 at 9:01 AM, Ben Peart <Ben.Peart@microsoft.com> wrote:
> This version incorporates Elijah's fixup patch and is now based on
> en/rename-directory-detection in next.

en/rename-directory-detection was in master but reverted.
en/rename-directory-detection-reboot is the new series, and isn't
quite in next yet; Junio just said in the last "What's cooking" that
he intends to merge it down.  However, the two series are obviously
similar, so your rebasing even onto the old one does cut down on the
conflicts considerably.  Rebasing this latest series onto
en/rename-directory-detection-reboot leaves just one minor conflict.

Anyway, I've looked over this latest re-roll and it looks good to me:
Reviewed-by: Elijah Newren <newren@gmail.com>