Re: [PATCH/RFC] grep: turn off threading when recursing submodules

From: Martin Ågren <martin.agren@gmail.com>
Date: Oct 19, 2017, 19:50 UTC
Message-ID: <CAN0heSpxJDxmSXu=P0stq+CsKVqjsvOnjDpOoMT9GdWxeZU2Pw@mail.gmail.com>
In-Reply-To: <20171019193413.prx5pkmwaxk6mpdt@sigill.intra.peff.net>

On 19 October 2017 at 21:34, Jeff King <peff@peff.net> wrote:

Show 17 quoted lines

> On Thu, Oct 19, 2017 at 12:26:18PM -0700, Brandon Williams wrote:
>
>> One alternative to turning off threading would be to employ proper
>> locking (like I failed to do) by wrapping the call the
>> 'add_to_alternates_memory()' in calls to grep_read_lock/unlock:
>>
>>   +       grep_read_lock();
>>           add_to_alternates_memory(submodule.objectdir);
>>   +       grep_read_unlock();
>>
>> That way adding the submodule to the object database won't happen at the
>> same time another thread is trying to read from the object store.
>
> Yeah, I think that is the right approach. Many of Git's APIs aren't
> thread-safe, so we need a big mutex around "I am doing anything except
> actually looking at the already-read blob contents". We just missed this
> spot.

That's a big glaring lock I missed, right in front of me. Looks like the right solution!