[Debian-med-packaging] Bug#842794: autopkgtests fail (since perl 5.24?): Failed test 'Testing vcf-fix-ploidy .. cat ...

[Andreas Tille]

Hi Gregor,

thanks for your analysis.

On Wed, Dec 14, 2016 at 08:00:00PM +0100, gregor herrmann wrote:
> > > I noticed in Ubuntu, where we run autopkgtests as part of britney
> > > migration, that vcftools fails now. You can see on ci.debian.net[0]
> > > 
> > > > not ok 26 - Testing vcf-fix-ploidy .. cat fix-ploidy.vcf | perl -I../. -MVcf /usr/bin/vcf-fix-ploidy -s fix-ploidy.samples -p fix-ploidy.txt 2>/dev/null | vcf-query -f '%POS[\t%SAMPLE %GTR %PL]\n'
> > > > #   Failed test 'Testing vcf-fix-ploidy .. cat fix-ploidy.vcf | perl -I../. -MVcf /usr/bin/vcf-fix-ploidy -s fix-ploidy.samples -p fix-ploidy.txt 2>/dev/null | vcf-query -f '%POS[\t%SAMPLE %GTR %PL]\n''
> > > > #   at ./test.t line 452.
> > > > #     Structures begin differing at:
> > > > #          $got->[0] = '61098   M1 0/1 0,9,72,5,6,7     M2 0/0 0,15,140,5,6,7   F3 1 147,0,5    F4 0 0,131,5    M5 0/0 0,9,83,5,6,7     M6 0/0 0,6,56,5,6,7
> > > > #     '
> > > > #     $expected->[0] = '61098   M1 0 0,9,72,5,6,7       M2 0 0,15,140,5,6,7     F3 1/1 147,0,5  F4 0/0 0,131,5  M5 0 0,9,83,5,6,7       M6 0 0,6,56,5,6,7
> > > > #     '
> > 
> > Since the bug is menitoning the Perl version bump I wonder whether the
> > perl team would have an explanation / fix for the issue.
> 
> The tests started failing on 2016-09-01.
> perl 5.24 entered unstable on 23 Sep 2016.

Ahhh, that's the changelog date from 1.3.1-3 with the following patch
included:

   https://anonscm.debian.org/cgit/debian-med/htslib.git/tree/debian/patches/define_PATH_MAX.patch

While there is some time coincicence I can not imagine that this
specific patch would be able to cause this result.  To be really sure I
tried with a tabix where I reverted the patch and there is the same test
failure.
 
> What did happen around the time was the removal of '.' from @INC.
> 
> But still, I think the failure is unrelated;
> https://ci.debian.net/data/packages/unstable/amd64/v/vcftools/20160901_192825.log
> (the first failure) doesn't show any changes in dependencies. But
> has perl 5.22.2-4. But re-adding . to @INC in
> /etc/perl/sitecustomize.pl doesn't change anything.
> 
> The log from the failed test before is:
> https://ci.debian.net/data/packages/unstable/amd64/v/vcftools/20160901_164433.log
> 
> -tabix 1.3.1-2
> +tabix 1.3.1-3
> 
> Not sure if and how this is related.

I think its a nice hint anyway even if the actual change in this package
does not have caused the failed test.  I even fetched tabix 1.3.1-2 +
libhts1 1.3.1-2 from snapshot.d.o and the test failure remains.
 
> The test call can be shortened to:
> 
> # vcf-fix-ploidy -s fix-ploidy.samples < fix-ploidy.vcf | grep 61098
> 20  61098   .   C   A,T 999 PASS    AC1=41;AF1=0.2104;DP4=209,284,67,76;DP=658;FQ=999;MQ=45;PV4=0.39,4.4e-10,0.0034,0.2 GT:PL:DP:SP:GQ  0/1:0,9,72,5,6,7:3:212:12   0/0:0,15,140,5,6,7:5:458752:18  1:147,0,5:7:384:24  0:0,131,5:5:208:18  0/0:0,9,83,5,6,7:3:392:12   0/0:0,6,56,5,6,7:2:204:9
> 
> Which already has the 0/1:0,9,72,5,6,7 sequence which is not the
> expected "0 0,9,72,5,6,7". And I don't know what either of them mean
> ...

I confirm that I have no idea what this means but it somehow looks
"quite similar".  I'll ask upstream about this.
 
> And /usr/share/perl5/Vcf.pm has 3523 lines which I'm not going to
> read and try to understand now :)

How comes? ;-)
 
Thanks for your very helpful analysis

     Andreas.

-- 
http://fam-tille.de