From: Eric Sunshine <sunshine@sunshineco.com>
Date: Thu, 18 Jul 2013 20:49:47 GMT
Subject: Re: [PATCH 1/2] git_mkstemps: correctly test return value of open()
Message-ID: <CAPig+cRyRq3depm+eBzZp-nRa2iPKbK-JU8cXgZdzQnCduFTMA@mail.gmail.com>
In-Reply-To: <201307182032.r6IKWtWC016218@freeze.ariadne.com>

On Thu, Jul 18, 2013 at 4:32 PM, Dale R. Worley <worley@alum.mit.edu> wrote:
> I've been looking into writing a proper test for this patch.  My first
> attempt tests the symptom that was seen initially, that "git commit"
> fails if fd 0 is closed.
>
> One problem is how to arrange for fd 0 to be closed.  I could use the
> bash redirection "<&-", but I think you want to be more portable than
> that.  This version uses execvp() inside a small C program, and
> execvp() is a Posix function.

"<&-" is POSIX.

http://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18_07_05

> I've tested that this test does what it should:  If you remove the
> fix, "fd >= 0", the test fails.  If you then remove "test-close-fd-0"
> from before "git init" in the test, the test is nullified and succeeds
> again.
>
> Here is the diff.  What do people think of it?