[Debian-med-packaging] Bug#890993: (Bug#890993) primer3 FTBFS on big endian: test failure
Andreas Tille
andreas at an3as.eu
Fri May 25 12:03:36 BST 2018
Hi Liubov,
On Fri, May 25, 2018 at 12:36:54PM +0200, Liubov Chuprikova wrote:
> I tried to figure out what exactly could cause the failure. The error
> refers to the test files in the kmer_lists directory that are in binary
> format. It seems they were generated on a little-endian architecture, that
> is why the test fails on big-endian when it tries to verify the file by
> comparing its first bytes with a given magic.
>
> kmer_list/readme.txt says that the test files could be downloaded from
> http://primer3.ut.ee/lists.htm, which says the files can also be created by
> *glistmaker* from GenomeTester4 package.
>
> So one solution could be to generate a set of similar test files for
> big-endian architectures. For this, only genome sequences in Fasta format
> would be necessary. But at this point, I get stuck as I have never faced a
> need to emulate big-endian. Are there other ways to deal with that type of
> problem?
While I think your suggestion is the proper solution I do not think that
it is worth the effort. I would rather do the following:
1. Exclude those test that are failing (it is only a small number)
for the architectures where the test fails). Please do not hesitate
to ask here if you are lacking ideas how to do this. May be the
python-cogent package can serve as an exemple. Its probably
sufficient to do something like this:
override_dh_auto_test:
ifeq ($(BUILDARCH),...)
mkdir tempdir
mv test/... tempdir
endif
dh_auto_test --sourcedirectory=test
dh clean --sourcedirectory=test
ifeq ($(BUILDARCH),...)
move back files to restore original source state
rmdir tempdir
endif
2. Since you have saved time to work out proper tests on rarely
used architectures you have time to package
https://github.com/bioinfo-ut/GenomeTester4
While the later is half joking this could be quite useful. You would
have need the tool for your original plan - may be for other things as
well and you could learn packaging.
Kind regards
Andreas.
--
http://fam-tille.de
More information about the Debian-med-packaging
mailing list