[Debian-med-packaging] Bug#848735: pynast: FTBFS: Test failures

[Andreas Tille]

Hi Aaron,

On Tue, Dec 20, 2016 at 12:31:07PM -0500, Aaron M. Ucko wrote:
> Andreas Tille <tille at debian.org> writes:
> 
> >    ncbi-blast+-legacy (>= 2.5.0-2~) | blast2
> 
> Yes, this is an appropriate dependency for anything that historically
> listed blast2 and continues to use legacy command names such as bl2seq.

OK.
 
> > Could you give some hint for a patch to fix this test suite error:
> 
> Good question.  The error from the underlying blastn command (buried in
> the middle of the log) is
> 
>   BLAST engine error: Error: Gap existence and extension values 0 and 0 are not supported for substitution scores 1 and -1
>   3 and 2 are supported existence and extension values
>   2 and 2 are supported existence and extension values
>   1 and 2 are supported existence and extension values
>   0 and 2 are supported existence and extension values
>   4 and 1 are supported existence and extension values
>   3 and 1 are supported existence and extension values
>   2 and 1 are supported existence and extension values
>   4 and 2 are supported existence and extension values
>   Any values more stringent than 4 and 2 are supported
> 
> I've tried fiddling with parameters to address this error, but so far
> haven't been able to get blastn to report *any* hits for these test
> cases, perhaps because the sequences are too short.
> 
> I'm also concerned that pynast.util.blast_align_unaligned_seqs invokes
> bl2seq with the -VT flag, which historically requested the old BLAST
> engine.

When thinking about this just disabling the affected test might be a
reasonable thing to do.  I'm not sure about the consequences and I can
report this upstream but the last code changes were 3 years ago the
latest issue in Github issue tracker was opened about 2 years ago.

In any case it is one of the more freqently used programs in the Debian
Med scope.

Kind regards

       Andreas. 

-- 
http://fam-tille.de