Re: Segfault in the attr stack

From: Stefan Beller <sbeller@google.com>
Date: Jun 2, 2016, 16:56 UTC
Message-ID: <CAGZ79ka7sXKwOdY4_WK15c3HHOp2pRtcUgeZ6rDAt8hSvcx1=Q@mail.gmail.com>
In-Reply-To: <xmqqshwvvaxq.fsf@gitster.mtv.corp.google.com>

On Thu, Jun 2, 2016 at 8:59 AM, Junio C Hamano <gitster@pobox.com> wrote:

Show 11 quoted lines

> Junio C Hamano <gitster@pobox.com> writes:
>
>>>> Gaah, of course.
>>>>
>>>> This is coming from the cache preload codepath, where multiple threads
>>>> try to run ce_path_match().
>>>> It used to be OK because pathspec magic never looked at attributes,
>>>> but now it does, and attribute system is not thread-safe.
>
> I'll look into teaching a threadble interface to the attribute
> subsystem, but for now, this should get you unstuck, I think.

Thanks for looking into this! (I was about to do that if you would not have stepped up, as the bug only surfaces when using the pathspec labels.)

>
> +       /* Do not preload when pathspec uses non-threadable subsystems */
> +       if (pathspec && pathspec_is_non_threadable(pathspec))
> +               return; /* for now ... */

This fixes the problem for now; I'll look into the comma escaping then.

Thanks, Stefan