Re: [PATCH 0/2] Add Windows support to the new RUNTIME_PREFIX design

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

Hi Dan,

On Tue, 27 Mar 2018, Daniel Jacques wrote:

> On Mon, Mar 26, 2018 at 5:31 PM Johannes Schindelin <
> johannes.schindelin@gmx.de> wrote:
> 
> > Even if the RUNTIME_PREFIX feature originates from Git for Windows, the
> > current patch series is different enough in its design that it leaves the
> > Windows-specific RUNTIME_PREFIX handling in place: On Windows, we still
> > have to override argv[0] with the absolute path of the current `git`
> > executable.
> 
> > Let's just port the Windows-specific code over to the new design and get
> > rid of that argv[0] overwriting.
> 
> > This also partially addresses a very obscure problem reported on the Git
> > for Windows bug tracker, where misspelling a builtin command using a
> > creative mIxEd-CaSe version could lead to an infinite ping-pong between
> > git.exe and Git for Windows' "Git wrapper" (that we use in place of
> > copies when on a file system without hard-links, most notably FAT).
> 
> > Dan, I would be delighted if you could adopt these patches into your patch
> > series.
> 
> Great, I'm glad this patch set could be useful to you! I'm happy to apply
> this to the patch series. They applied cleanly, so I'll push a new version
> after Travis validates the candidate.
> 
> I don't have a Windows testing facility available, so I'm hoping that you
> verified that this works locally. I suppose that's what the unstable branch
> series is for.

Yes, I performed manual testing.

I guess we should add a test where we copy the `git` executable into a
subdirectory with the name "git" and call `git/git --exec-path` and verify
that its output matches our expectation?

Ciao,
Dscho