From: Duy Nguyen <pclouds@gmail.com>
Date: Wed, 30 Dec 2015 10:31:41 GMT
Subject: Re: Segfault in git reflog
Message-ID: <CACsJy8Db-SjNj_tOgERMuRyydq6XC3Fq4pEKNRr8bqgpaUxKig@mail.gmail.com>
In-Reply-To: <20151230092400.GA9319@spirit>

On Wed, Dec 30, 2015 at 4:24 PM, Dennis Kaarsemaker
<dennis@kaarsemaker.net> wrote:
> I've hit a segfault in git reflog with latest git, reproducable in git.git:
>
> spirit:~/code/git (master)$ ./git describe
> v2.7.0-rc3
>
> I've minimized the reflog to:
>
> spirit:~/code/git (master)$ cat .git/logs/HEAD
> 2635c2b8bfc9aec07b7f023d8e3b3d02df715344 54bc41416c5d3ecb978acb0df80d57aa3e54494c Dennis Kaarsemaker <dennis@kaarsemaker.net> 1446765642 +0100
> 74c855f87d25a5b5c12d0485ec77c785a1c734c5 54bc41416c5d3ecb978acb0df80d57aa3e54494c Dennis Kaarsemaker <dennis@kaarsemaker.net> 1446765951 +0100  checkout: moving from 3c3d3f629a6176b401ebec455c5dd59ed1b5f910 to master
>
> ...which I realize looks a bit broken. I think at the time I was playing with
> some patches that also caused segfaults, causing gaps in the reflog.
> Nevertheless, I think segfaulting is bad. All objects in the reflog are
> reachable.
>
> gdb has the following to say:
>
> spirit:~/code/git (master)$ gdb --args ./git --no-pager reflog

If it helps, if add_ref_decoration is not called at all, there's no
segfault. Definitely something with reflog parsing, probably leaving
an uninitialized pointer. But I have not gotten there yet.
-- 
Duy