Re: [PATCH v2 0/2] Restrict when prefetcing occurs

From: Junio C Hamano <gitster@pobox.com>
To: Jonathan Tan <jonathantanmy@google.com>
Cc: git@vger.kernel.org, stolee@gmail.com,
	Garima Singh <garima.singh@microsoft.com>
Subject: Re: [PATCH v2 0/2] Restrict when prefetcing occurs
Date: Thu, 02 Apr 2020 13:28:53 -0700	[thread overview]
Message-ID: <xmqq369l3a4a.fsf@gitster.c.googlers.com> (raw)
In-Reply-To: <cover.1585854639.git.jonathantanmy@google.com> (Jonathan Tan's message of "Thu, 2 Apr 2020 12:19:15 -0700")

Jonathan Tan <jonathantanmy@google.com> writes:

> Thanks, everyone, for your review.
>
> New in v2:
>  - added restriction on fetching rename_src's blob, following Stolee's
>    comment
>  - folded oid_nr==0 check into promisor_remote_get_direct(), following
>    Stolee's comment
>  - used "mv server/b server/c", following Stolee's comment
>  - made diff_add_if_missing() public function, following Junio's comment
>
> I didn't change the "continue" part that Stolee suggested [1].
>
> [1] https://lore.kernel.org/git/xmqqlfng75cl.fsf@gitster.c.googlers.com/
>
> Jonathan Tan (2):
>   promisor-remote: accept 0 as oid_nr in function
>   diff: restrict when prefetching occurs
>
>  builtin/index-pack.c          |  5 ++--
>  diff.c                        | 49 +++++++++++++++++++++++------------
>  diffcore-rename.c             | 37 +++++++++++++++++++++++++-
>  diffcore.h                    | 10 ++++++-
>  promisor-remote.c             |  3 +++
>  promisor-remote.h             |  8 ++++++
>  t/t4067-diff-partial-clone.sh | 48 ++++++++++++++++++++++++++++++++++
>  unpack-trees.c                |  5 ++--
>  8 files changed, 141 insertions(+), 24 deletions(-)

I notice that a439b4ef (diff: skip batch object download when
possible, 2020-03-30) by Garima seems to aim for something similar.

I'll for now keep both topics with conflict resolution, but it may
make sense for you two to compare notes.  

I especially like the way this series enumerates the formats that
matter to prefetching and the way the change is localized in
diffcore_std(); the other topic splits a similar logic (with
different criteria) between diff_setup_done() and diffcore_std(),
which I found suboptimal.

Thanks.