>>August 24, 1996

New programs - tfastx3, tfastx3_t, compare a protein sequence to
forward and reverse translations of a DNA sequence database.  An excellent
replacement for tfasta3.

Sun multiprocessing - change in thr_create() to use all CPU's if available.

GCG formats - now can search with simple GCG-format query sequences and
results with GCG format Swissprot and Genpept are more readable.

>>August 26, 1996

Fixed bugs in tfastx3(_t) and fastx3(_t) including an ancient problem
with aatran().  Less redundancy in gcg_ranlib().


>>August 31, 1996

Included support for BLOSUM62 (-s BL62) as per documentation.

Rearranged Makefile's so that they would make everything in one pass.

>>September 6, 1996

Corrected yet another problem with the fastx/tfastx code.

Noticed that searching without optimized scores gave no optimized
scores on the final list of scores - fixed this.

The pvm version now does alignments - not thoroughly tested.

>>September 13, 1996

Fixed display of best scores to stdout.

Fixed problem with alignments when -o flag used.

pvcompfa/pvcompsw have now been tested on DEC Alpha, Solaris X86, and
SGI PVM implementations.  Several bugs were corrected.

>>September 18, 1996

Fixed bug selectbestz() that caused core dumps in pvcomplib.c
(changes to pvcomplib.c, comp_thr.c, complib.c).

>>September 23, 1996

Corrected showalign.c/pvm_showalign.c addressing bug found and fixed
by Erik Wallin. (erikw@biokemi.su.se).

>>October 15, 1996

Corrected bug so alternative scoring matrices are used.

>>October 22, 1996

Remove singularities from regression routine.

-z 0 now means no statistics (same as -z -1).

No longer show alignment for 0 score.

>>October 26, 1996

Fix problem with -b, -d when Z-values disabled.

>>November 1, 1996

Altschul-Gish statistical estimates (-z 3) now work properly.

Fix problem with mean_var==0.0.