From: Jeff King <peff@peff.net>
Date: Tue, 12 Apr 2016 17:22:47 GMT
Subject: Re: Hardcoded #!/bin/sh in t5532 causes problems on Solaris
Message-ID: <20160412172247.GA2856@sigill.intra.peff.net>
In-Reply-To: <xmqqvb3mrcgj.fsf@gitster.mtv.corp.google.com>

On Tue, Apr 12, 2016 at 09:58:20AM -0700, Junio C Hamano wrote:

> > Looks good and is the minimal change. I kind of wonder if the example
> > would be more clear, though, as just:
> >
> >   write_script .git/hooks/pre-commit <<-\EOF &&
> >   exit 1
> >   EOF
> >   echo whatever >file1 &&
> >   ...
> >
> > I don't think we ever actually need the pre-commit check to pass, as we
> > simply override it with --no-verify. But I dunno. Maybe people find it
> > easier to read with a pseudo-realistic example (it took me a minute to
> > realize the trailing whitespace in the content was important).
> 
> I was mostly worried about closing the door for future enhancement
> where there are multiple commits to be replayed, some of which fail
> and others pass the test.  Unconditional "exit 1" would have to be
> reverted when it happens.

Yeah, that's fair. It is at least trying to re-create a real-world
situation.

-Peff