Re: [PATCH] Fix start_command() pipe bug when stdin is closed.

From: Karl Chen <quarl@cs.berkeley.edu>
Date: Aug 25, 2008, 15:56 UTC
Message-ID: <quack.20080825T0856.lth63ppulyu@roar.cs.berkeley.edu>
In-Reply-To: <48B28CF8.2060306@viscovery.net>

>>>>> On 2008-08-25 03:44 PDT, Johannes Sixt writes:

    Johannes> When I try these instructions I don't get an error;
    Johannes> instead the command runs successfully.

Weird. I see the symptom on two machines, both 1.5.6 and tracking master. The 1.5.6 system installation could be interfering even though I used PATH=/path/to/git:$PATH.

    Johannes> While I do see that there is a problem, it is only
    Johannes> half of the story, and your patch addresses only
    Johannes> this half.

    Johannes> What if stdout is closed, too? Then the ends of the
    Johannes> first allocated pipe would go to fds 0 and 1, and
    Johannes> then the pipe end at 1 would be closed by a
    Johannes> subsequent dup2(xxx, 1).

My patch was intended to fix the problem for any renaming where fd_from==fd_to, including target stdout. I didn't say so in the changelog though.