From: Thomas Rast <trast@student.ethz.ch>
Date: Fri, 10 Aug 2012 14:24:25 GMT
Subject: Re: [PATCH/RFC v3 01/13] Move index v2 specific functions to their own file
Message-ID: <87hasan7ty.fsf@thomas.inf.ethz.ch>
In-Reply-To: <7vtxwbn2qe.fsf@alter.siamese.dyndns.org>

Junio C Hamano <gitster@pobox.com> writes:

> If you found that an entry you read halfway has an inconsistent crc,
> and if you suspect that is because somebody else was writing to the
> same index, it is a _sure_ sign that you are not alone, and all the
> entries you read so far to the core, even if they weren't touched by
> that sombody else when you read them, may be stale, and worse yet,
> what you are going to read may be inconsistent with what you've read
> and have in-core (e.g. you may have read "f" before somebody else
> that is racing with you have turned it into a directory, and your
> next read may find "f/d" in the index without crc error).

The intention for v5 (admittedly this probably requires a lot more
documentation) was to only allow an in-place update in two cases:

* updating the data fields (*not* the name) of a file entry,

* adding or removing conflict entries at the end.

The latter probably requires a bit more thought to make it safe, too.
But I think the idea always was that any write that changes the basic
layout of the file (so that you would read something wrong) will need a
full rewrite.  Otherwise we're too far in DB land.  Most updates will be
of the "update the stat and/or sha1 of a file" kind, anyway.

-- 
Thomas Rast
trast@{inf,student}.ethz.ch