From: Matt McCutchen <matt@mattmccutchen.net>
Date: Tue, 15 Nov 2016 19:12:52 GMT
Subject: Re: Protecting old temporary objects being reused from concurrent "git gc"?
Message-ID: <1479237172.2406.91.camel@mattmccutchen.net>
In-Reply-To: <20161115174028.zvohfcw4jse3jrmm@sigill.intra.peff.net>

On Tue, 2016-11-15 at 12:40 -0500, Jeff King wrote:
> On Tue, Nov 15, 2016 at 12:33:04PM -0500, Matt McCutchen wrote:
> 
> > 
> > On Tue, 2016-11-15 at 12:06 -0500, Jeff King wrote:
> > > 
> > >  - when an object write is optimized out because we already have the
> > >    object, git will update the mtime on the file (loose object or
> > >    packfile) to freshen it
> > 
> > FWIW, I am not seeing this happen when I do "git read-tree --prefix"
> > followed by "git write-tree" using the current master (3ab2281).  See
> > the attached test script.
> 
> The optimization I'm thinking about is the one from write_sha1_file(),
> which learned to freshen in 33d4221c7 (write_sha1_file: freshen existing
> objects, 2014-10-15).
> 
> I suspect the issue is that read-tree populates the cache-tree index
> extension, and then write-tree omits the object write before it even
> gets to write_sha1_file(). The solution is that it should probably be
> calling one of the freshen() functions (possibly just replacing
> has_sha1_file() with check_and_freshen(), but I haven't looked).
> 
> I'd definitely welcome patches in this area.

Cool, it's nice to have an idea of what's going on.  I don't think I'm
going to try to fix it myself though.

By the way, thanks for the fast response to my original question!

Matt