Re: [PATCH 11/17] t5509: specify bare repository path explicitly

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

Hi Junio,

On Fri, 3 Apr 2026, Junio C Hamano wrote:

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> 
> > It would be indeed more conservative, and it even results in less changes
> > when done in a more elegant fashion, by appending `--git-dir=.` after the
> > `-C pushee` [...]
> 
> Makes sense.
> 
> Assuming that we want to tighten the rule and prepare for the tightening
> before it happens, that is.  I personally do not think it is a bad move,
> but I do not recall we had much discussion to gain a community consensus
> to go in that direction.

Well, I wanted to contribute a patch to that extent, as a discussion
starter. And then the CI build broke. And then I found myself almost
overwhelmed with the changes required to let the CI build pass. And then I
thought I'd go through the effort of making those changes just to gauge
how much of a problem changing the default would be for affected parties.
And then I split up the preparatory patches into multiple sub-branches so
that reviewers wouldn't get overwhelmed. And you're looking at the first
sub-branch broken out from that PR.

Note that this here patch series, while it would be a necessary
prerequisite for changing the `safe.bareRepository` default in Git 3.0,
does have merit on its own, and I do not intend these patches to make such
a change of default behavior more or less likely in Git 3.0.

By introducing that setting, we declared that while not necessary at the
moment, it is better, really, to specify the location of bare repositories
explcitly than to rely on the implicit discovery. And this patch series
addresses a couple of places where Git's own test suite does not follow
our own advice.

For that reason I consider this patch series strictly a spring cleaning.

Ciao,
Johannes