git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH v2 0/2] Update sha1dc from upstream & optionally make it a submodule

From
Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Date
May 20, 2017, 11:54 UTC
Message-ID
<20170520115429.12289-1-avarab@gmail.com>
In-Reply-To
<xmqqpof3srw4.fsf@gitster.mtv.corp.google.com>
On Sat, May 20, 2017 at 1:13 PM, Junio C Hamano <gitster@pobox.com> wrote:
Show 49 quoted lines
> Ævar Arnfjörð Bjarmason  <avarab@gmail.com> writes:
>
>> Replace the forked sha1dc directory with a copy of the upstream code
>> imported as a submodule. This is the exact same code as now exists in
>> the sha1dc/ directory.
>>
>> The initial reason for copy/pasting the code into sha1dc and locally
>> modifying it was that it needed to be altered to work with the git
>> project.
>>
>> The upstream project has accepted my code changes to allow us to use
>> their code as-is, see the preceding commit for details. So import the
>> code as a submodule instead, this will make it easier to keep
>> up-to-date with any upstream fixes or improvements.
>>
>> Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
>> ---
>>  .gitmodules            | 4 ++++
>>  Makefile               | 4 ++--
>>  hash.h                 | 2 +-
>>  sha1collisiondetection | 1 +
>>  4 files changed, 8 insertions(+), 3 deletions(-)
>>  create mode 100644 .gitmodules
>>  create mode 160000 sha1collisiondetection
>
> I am not sure how prepared our .travis.yml is to deal with a
> submodule, I'd prefer to have this step broken down to two step
> process.
>
> That is, [PATCH 2.1/3] first adds an otherwise unused submodule, so
> that people can optionally do "git submodule init && git submodule
> update" so that they can compare the contents of sha1dc/ that has
> been updated by [PATCH 1/3] with the up-to-date upstream.  Then
> [PATCH 2.2/3] would update the Makefile and hash.h to use the code
> in the submodule.
>
> I actually would want to see us proceed even more cautiously---if
> the latter-half, i.e. [PATCH 2.2/3], is arranged so that it uses the
> new sha1collisiondetection/ only when the submodule is initialized
> and populated, and otherwise it uses sha1dc/ as before, I would feel
> a lot safer.  I wouldn't be this paranoid if this "let's start using
> submodule ourselves" were done to some optional corner (like compat/
> or contrib/ somewhere), but this is the default hash function.  I do
> want to have something like this to force us (and submodule folks)
> to get any kinks out, but I do not want to see many people not even
> be able to build while this new arrangement is eased in.  Once
> people are comfortable with the new arrangement to use code from
> submodule, we can then take [PATCH 3/3] to remove the old sha1dc/
> directory and the migration will be complete.

Makes sense to take it slow. Hopefully this addresses your comments. I dropped the 3rd patch to remove sha1dc/ and the 2nd patch adds sha1collisiondetection/ as submodule, but it's not used unless you specify DC_SHA1_SUBMODULE in addition to DC_SHA1.

Both patches should be safe to include & not cause any disruption, but now those interested in making the submodule experience in git.git better can init/update & set DC_SHA1_SUBMODULE=Y to play with it.

Note that both patches update to a newer version of the upstream code. I sent them another pull request with some cleanups, one of which is to ignore .depends in their .gitignore file.

> I also am not very happy with .gitmodules pointing at a single point
> of failure.  It would be nice if you can arrange a couple of mirrors
> and have a comment in .gitmodules file to tell folks that they can
> use these alternates by insteadOf or some other mechanism.

I liked the suggestion to make the URL a relative path, but this would require you to maintain a mirror in the same places you push git.git to, is that something you'd be willing to do?

For now having no-mirror isn't a big issue with my new 2/2 since it'se something you have to opt-in to with a build flag, which I suspect only I/Brandon/Stefan & a few others will use.

Ævar Arnfjörð Bjarmason (2):
  sha1dc: update from upstream
  sha1dc: optionally use sha1collisiondetection as a submodule
 .gitmodules            |  4 +++
 Makefile               | 21 +++++++++++-
 hash.h                 |  4 +++
 sha1collisiondetection |  1 +
 sha1dc/sha1.c          | 91 +++++++++++++++++++++++++++++++++-----------------
 sha1dc/sha1.h          | 90 ++++++++++++++++++++++---------------------------
 sha1dc/ubc_check.c     | 13 ++++++--
 sha1dc/ubc_check.h     | 14 ++++++--
 sha1dc_git.c           | 24 +++++++++++++
 sha1dc_git.h           | 19 +++++++++++
 10 files changed, 193 insertions(+), 88 deletions(-)
 create mode 100644 .gitmodules
 create mode 160000 sha1collisiondetection
 create mode 100644 sha1dc_git.c
 create mode 100644 sha1dc_git.h
-- 
2.13.0.303.g4ebf302169
Previous: Junio C HamanoNext: Ævar Arnfjörð Bjarmason
Message 5 of 20 in “Update sha1dc from upstream & optionally make it a submodule”
  1. 0/3 Update sha1dc from upstream & optionally make it a submoduleÆvar Arnfjörð Bjarmason, May 18, 2017
  2. 1/3 sha1dc: update from upstreamÆvar Arnfjörð Bjarmason, May 18, 2017
  3. 2/3 sha1dc: use sha1collisiondetection as a submoduleÆvar Arnfjörð Bjarmason, May 18, 2017
  4. Junio C HamanoMay 20, 2017
  5. 0/2 Update sha1dc from upstream & optionally make it a submoduleÆvar Arnfjörð Bjarmason, May 20, 2017
  6. 1/2 sha1dc: update from upstreamÆvar Arnfjörð Bjarmason, May 20, 2017
  7. 2/2 sha1dc: optionally use sha1collisiondetection as a submoduleÆvar Arnfjörð Bjarmason, May 20, 2017
  8. Junio C HamanoMay 22, 2017
  9. Junio C HamanoMay 22, 2017
  10. Ævar Arnfjörð BjarmasonMay 22, 2017
  11. Junio C HamanoMay 22, 2017
  12. Junio C HamanoMay 22, 2017
  13. Stefan BellerMay 22, 2017
  14. Junio C HamanoMay 23, 2017
  15. Ævar Arnfjörð BjarmasonMay 23, 2017
  16. Junio C HamanoMay 23, 2017
  17. Ævar Arnfjörð BjarmasonMay 25, 2017
  18. Junio C HamanoMay 25, 2017
  19. Ævar Arnfjörð BjarmasonMay 25, 2017
  20. 3/3 sha1dc: remove the unused sha1dc/ directoryÆvar Arnfjörð Bjarmason, May 18, 2017

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.