Re: [PATCH v3 4/4] exec_cmd: RUNTIME_PREFIX on some POSIX systems

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Nov 28, 2017, 14:08 UTC
Message-ID: <alpine.DEB.2.21.1.1711281330150.6482@virtualbox>
In-Reply-To: <20171128032538.98071-1-dnj@google.com>

Hi Dan, and (based on the timezone recorded in your mail:) good morning!

On Mon, 27 Nov 2017, Dan Jacques wrote:

Show 14 quoted lines

> > In Git for Windows, we have an almost identical patch:
> >
> > https://github.com/git-for-windows/git/commit/bdd739bb2b0b
> >
> > We just guard the call to system_path() behind a test whether podir is
> > already absolute, but these days, system_path() does that itself.
> >
> > I am too little of a Perl expert to be helpful with the other patches, but
> > I would gladly runa build & test on Windows if you direct me to an
> > easily-pullable branch.
> 
> Oh interesting - I've only peripherally looked at Git-for-Windows code,
> since Chromium uses its packages verbatim (thanks, BTW!). I think you're
> correct though - this patch set seems to be doing the same thing.

Excellent, thanks for confirming.

Show 10 quoted lines

> I've been force-pushing my changes to the "runtime-prefix" branch of my
> Git fork for travis.ci testing. The latest commit on that branch adds a
> "config.mak" for testing, so one commit from the branch head will
> contain the sum set of this patch series applied at (or near) Git's
> master branch:
> 
> https://github.com/danjacques/git/tree/runtime-prefix~1
> 
> Let me know if this is what you are looking for, and if I can offer any
> help with Windows testing. Thanks!

Thank you, that was exactly what I was looking for.

BTW I think that your config.mak settings are partially unnecessary: the gitexecdir and template_dir should be identical, and sysconfdir, too (at least unless you override prefix).

I triggered a build (with the config.mak commit, because it does not really matter), and it failed immediately due to a quoting issue: the path separator is a semicolon on Windows and therefore must be quoted (so that it is not misinterpreted as ending the command). There were a couple of other issues, too, and I opened a PR here:

	https://github.com/danjacques/git/pull/1

Ciao, Dscho