On Mon, Jul 18, 2022 at 10:01:37AM -0400, Jonathan S. Katz wrote:
> On 7/18/22 4:37 AM, Julien Rouhaud wrote:
> > On Mon, Jul 18, 2022 at 10:27:47AM +0200, Magnus Hagander wrote:
> > >
> > > AFAICT this is not functionality that can be turned off, it's all
> > > hardcoded both in what it searches and when it does it. :/
> > >
> > > I doubt it's worth forking and maintaining a fork just to handle this
> > > situation given how seldom it shows up.
> >
> > +1, especially since it was introduced 17 years ago and this is probably the
> > first time someone notice it in our instance.
>
> Maybe it's worth reporting upstream? It seems like it's doing something very
> opinionated that may not make sense to projects using git.
Trying to guess what that code was originally supposed to do (as I can't find
anything apart from the "v203" original commit message), I'd say that it was
originally thought to shorten addresses in commit messages like
Merge master.kernel.org:/...
Merge rsync://rsync.kernel.org/...
(or other domains) but as written it will easily lead to nonsensical shortened
title.
Naively, something as simple as adding a leading whitespace in the regex would
remove 90% of wrong matches in our repo, and making sure it's not followed by
another whitespace would remove all of them AFAICS.
I will try to see if upstream would welcome such a change.