Re: [PATCH] use only the PATH for exec'ing git commands

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Oct 22, 2007, 23:34 UTC
Message-ID: <Pine.LNX.4.64.0710230026380.25221@racer.site>
In-Reply-To: <20071022224015.GG23714@steel.home>

Hi,

On Tue, 23 Oct 2007, Alex Riesen wrote:

Show 12 quoted lines

> Scott R Parish, Tue, Oct 23, 2007 00:12:02 +0200:
> > > Alex Riesen, Mon, Oct 22, 2007 12:01
> > > Scott R Parish, Mon, Oct 22, 2007 19:01:48 +0200:
> > > > +                strbuf_addch(out, ':');
> > >
> > > Shouldn't it break MingW32 native port?
> > 
> > What can i do here to better accommodate MingW32? You're right, just 
> > because the original code did it this way isn't a good excuse for me 
> > not to do it better.
> 
> someone here mentioned "higher abstractions".

That someone may well have mentioned that term, but what he meant was special case those code paths which have to be treated differently for Windows, and Windows alone.

(Yeah, I know, a response to that was that Nokia phones have drive letters, too, but that was an invalid objection to that particular statement, since Nokia phones expect the full path at any time, so I did not even bother answering to that mail. Well, in a way, I guess I just did.)

All this boils down to adding code dedicated to Windows. And I am a huge fan of the principle that you should cross bridges when you get to them, and not before (i.e. avoid over-engineering).

So my vote is: get it in as-is, us msysGit lot will cope with it in one way or another. (We already have at least one other place where ":" is used, which we have to cope with, in shell.c:do_cvs_command().)

Ciao, Dscho