Re: [PATCH 0/3] update sha1dc from PR #36

From: Stefan Beller <sbeller@google.com>
Date: Jun 27, 2017, 18:37 UTC
Message-ID: <CAGZ79kYbPfzH2vDhDuo5mePSSMcWZJgKNT19U4Nb20tpG6xJXg@mail.gmail.com>
In-Reply-To: <20170627121718.12078-1-avarab@gmail.com>

On Tue, Jun 27, 2017 at 5:17 AM, Ævar Arnfjörð Bjarmason <avarab@gmail.com> wrote:

Show 16 quoted lines

> This hopefully fixes the Solaris SPARC issue & doesn't cause
> regressions elsewhere, e.g. on Cygwin. Adam, it would be great if you
> could test that platform.
>
> I've already confirmed with Michael Kebe + another SPARC user
> (CosmicDJ on Freenode #Solaris) that it works on Solaris SPARC. The
> question is whether it breaks anything else.
>
> Per the upstream pull request:
> https://github.com/cr-marcstevens/sha1collisiondetection/pull/36
>
> Marc would (understandably) like some wider testing of this before
> merging it into the upstream project.
>
> WRT the submodule URL & branch changing: I have no idea how
> git-submodule handles this in the general case,

It is trying a couple of things, see cmd_update in git-submodule.sh look for the call sites of fetch_in_submodule which tries to fetch the default branch, by-sha1 if that doesn't help.

> but it Just Works with
> GitHub because it allows fetching arbitrary SHA1s that any ref
> (including pull req refs) point to.

Which is only half the story (but it works most of the time). They may have uploadpack.allowTipSHA1InWant but what is really interesting is uploadpack.allowReachableSHA1InWant.