From: Junio C Hamano <gitster@pobox.com>
Date: Tue, 24 Apr 2012 01:51:48 GMT
Subject: Re: Potential problem in git-add may corrupt the index file
Message-ID: <xmqq4ns9j3rf.fsf@junio.mtv.corp.google.com>
In-Reply-To: <CAEJt7hYYJh5dT1Tk99Oze6Hg4QaOqX5ERi0faWjyifQ6dr-pWw@mail.gmail.com>

Gang Hu <ganghu@cs.columbia.edu> writes:

> If a crash happens after the rename(), the data may have not been
> completely flushed into the index file, so the user may face an empty
> or corrupted index file.

Yeah, probably we should add fsync_or_die() at the end of write_index()
in read-cache.c, possibly protect it with fsync_object_files just like
we do in close_sha1_file() in sha1_file.c for loose object files.

I think the real fix would be to update the write_index() codepath to
use the csum-file API, though.  Then sha1close() will give us the fsync
behaviour for free.