Dear Granger,
I've looked into the problem of the TIGR assembler crashing on the
Silicon
Graphics unix boxes, and found that the problem comes from the SGI cc
compiler, which treats the char type as unsigned by default
(while Sun's compiler defines them signed by default).
That causes problems in several parts of the program,
like for instance when reading sequences at the beginning
(the trans_char codes for end of lines are positives in an unsigned char
environment, so \n chars contaminate the sequence. The unsigned char
value
for -2 is 254, so the quad_val that will be computed will be out of
range of
hash_array, which makes SGI crash).
Once compiled with the special -signed flag on the cc command line,
the program runs fine.
You will be glad to know as well that we could compile and run it
without
problems on a Pentium Pro Windows NT machine, as well as on a Power
PC mac (although the mac version needs some polishing work since there
is no true console on macs).
Thanks again for giving us access to the sources; we will of course give
you
feedback on our usage, eventual problems encountered as well as any
changes or development based on your code.
Regards,
Massimo
----------------------------------------
Massimo de Francesco, PhD
Geneva Biomedical Research Institute
Bioinformatics Group
14, rue des Aulx
CH-1228 Plan les Ouates
Switzerland
----- End Included Message -----
> From arkerlav at tigr.org Tue Feb 10 08:45 EST 1998
> X-Sender: arkerlav at mail.tigr.org> Mime-Version: 1.0
> Date: Tue, 10 Feb 1998 08:49:01 -0500
> To: grange
> From: Peter Bjarke Olsen <pbo at novo.dk> (by way of Anthony R. Kerlavage)
> Subject: Bug in TIGR assembler?
> X-Lines: 19
>> Dear Sir,
>> I have tried using TIGR assembler to assemble various EST libraries (In
> the range from 500 to 2000 sequences) and in most cases the run results
> in segmentation violation. The run always results in segment violation
> when the program is compiled under Irix 6.3 and most of the times when
> compiled under linux 2.0. I would be very happy if you could help me
> with the problem. Is there an updated version of the source code?
>> Best regards,
>> Peter Bjarke Olsen
> Building 1as31
> 2880 Bagsvaerd
> Novo Nordisk A/S
> Denmark
> Phone +4544423326
>>>
Yes there is a new version of TIGR Assembler which we plan to release in
the near future but we have been optimizing it for genomic projects and
still use the old version for our EST assemblies. We might do a new EST
version as well.
Granger Sutton
