From: Junio C Hamano <gitster@pobox.com>
Date: Tue, 19 Aug 2008 00:11:51 GMT
Subject: Re: [JGIT PATCH 01/14] Detect path names which overflow the name length field in the index
Message-ID: <7v7iadlv7c.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <1219103602-32222-2-git-send-email-spearce@spearce.org>

"Shawn O. Pearce" <spearce@spearce.org> writes:

> C Git allows a path name to be longer than 4095 bytes by storing 4095
> into the path name length field within flags and then searching for a
> null terminator at the end of the path name, instead of relying on the
> length indicatior.

This reminds me.

In the longer term, we should make this "CE_NAMEMASK gives the real length
for sane names but otherwise we need to count" merely a property of the
on-disk index structure.  In-core index should gain a new ce_namelen field
that records the real name (even when it is longer than the mask would
permit).  IOW, the knowledge of CE_NAMEMASK should be confined to
read_index_from() and ce_write_entry().

I expect this to be a relatively easy janitor project; hint, hint...