[Bio-bwa-help] [lh3@sanger.ac.uk: Re: BWA 0.5.7 seqfault fixed]

[Heng Li <lh...@sa...>]

----- Forwarded message from Heng Li <lh...@sa...> -----

Date: Fri, 16 Apr 2010 09:17:53 -0400
From: Heng Li <lh...@sa...>
To: Vaughn Iverson <vsi@u.washington.edu>
Cc: Chris Berthiaume <chrisbee@u.washington.edu>
Subject: Re: BWA 0.5.7 seqfault fixed
User-Agent: Mutt/1.5.20 (2009-06-14)

Many thanks for the fix. This is indeed a bug.

Best,

Heng

On Thu, Apr 15, 2010 at 04:42:28PM -0700, Vaughn Iverson wrote:
> Hi Heng, Just wanted to let you know that I found and fixed a
> seqfault in BWA 0.5.7
> 
> The problem is a simple bug in the new code to handle outputting
> multiple matches for a read in the new SAM format.
> 
> In bwase.c line: 59
> 
> Change:
> 
> s->multi = calloc(rest, rest * sizeof(bwt_multi1_t));
> 
> to:
> 
> s->multi = calloc(rest, sizeof(bwt_multi1_t));
> 
> As I'm sure you know, calloc automatically multiples the first
> (count) parameter with the second (size) parameter to get the bytes
> of memory allocated.
> The line with the bug effectively requests the square of the actual
> number of structures needed, with will quickly exhaust the memory of
> any machine when the number of matches gets large.
> 
> This calloc will fail in those cases, btu there is no code to check
> that it succeeds, resulting in a NULL pointer exception when that
> case arises.  This bug could also cause other undesirable effects
> such as hitting the ULIMIT for a job on a cluster node
> unnecessarily, causing a job to be killed by the kernal, even when
> the physical memory is available, etc.
> 
> We will be patching our own private version of BWA, but I wanted you
> to have this for the next release.
> 
> Cheers, -Vaughn
> 
> 

----- End forwarded message -----


-- 
 The Wellcome Trust Sanger Institute is operated by Genome Research 
 Limited, a charity registered in England with number 1021457 and a 
 company registered in England with number 2742969, whose registered 
 office is 215 Euston Road, London, NW1 2BE.