Subject: Re: bzip2 data corruption
To: Simon Burge <email@example.com>
From: Frank van der Linden <firstname.lastname@example.org>
Date: 11/02/1999 22:21:19
On Wed, Nov 03, 1999 at 02:06:37AM +1100, Simon Burge wrote:
> I just tried -current alpha (Oct 22), -current pmax (Oct 28) and 1.4.1
> i386, all with no problems. I'll try a -current (Aug 8) i386 (PII400)
> at work tomorrow...
With src/lib/libbz2/decompress.c version 1.6, you introduced a heuristic
to check for corruption, based on vectors growing too large. This
method gave false positives with some files (-current i386). I have
reverted the change.
The problem I have with such a heuristic, is that it seems hard to
be absolutely sure that you're not dealing with huge/strange data
in stead of an error. How would you know?