From: Junio C Hamano <gitster@pobox.com>
Date: Sun, 03 Feb 2008 23:58:57 GMT
Subject: Re: [PATCH 2/3] fsck_commit: remove duplicate tests
Message-ID: <7v3as9r3r2.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <12020737601683-git-send-email-mkoegler@auto.tuwien.ac.at>

Martin Koegler <mkoegler@auto.tuwien.ac.at> writes:

> All tests on the commit buffer in fsck_cmd are ready done by
> parse_commit_buffer.
>
> This patch rips out all redundant tests.

As I believe in belt-and-suspenders when it comes to validation,
I am somewhat uneasy with this change.  If it ever turns out
that we would need to make parse_commit_buffer() more lenient
during the normal operation, some of this would need to be
reverted (for example, you may not need to have a valid "tree"
to salvage the log messages from a corrupt history by running
"git log"), and forgetting to do so will result in fsck that
does not validate enough.

The validation done by parse_commit_buffer() and fsck serve two
different purposes, and they can have different leniency
requirements.  I do not necessarily agree that it is a good idea
to tie them together like this patch.