Hello all,
Recently I've been debugging various problems with my Galaxy tool TravisCI tests as part of moving to using "planemo test" rather than the fragile system I first created which mimicked a pre-ToolShed manual install:
https://blastedbio.blogspot.co.uk/2013/09/using-travis-ci-for-testing-galaxy...
Progress:
https://travis-ci.org/peterjc/galaxy_mira - passing now :) https://travis-ci.org/peterjc/pico_galaxy - probably a loc file problem https://travis-ci.org/peterjc/galaxy_blast - different loc file problem
This email is about the problems testing the BLAST wrappers where I am using a BLAST database defined in test-data/*.loc files.
(I wasted some time today trying moving/copying the loc files and databases to different places before I noticed the issue below)
Problem lines quoted are from this run, galaxy master or dev branch: https://travis-ci.org/peterjc/galaxy_blast/builds/150072050 as of this commit: https://github.com/peterjc/galaxy_blast/commit/03f81e7fb1e0562b46b8b2a2079c9...
Some of the command expect a single BLAST database from the test-data/*.loc files, and the test should do that, but we get:
blastdbcmd ... -db path,path ... rpsblast ... -db path,path ... rpstblastn ... -db path,path ...
Here we should get ... -db path (once only), not repeated with a comma, e.g.
galaxy.jobs.command_factory INFO 2016-08-05 14:55:08,543 Built script [/tmp/tmp9NURC9/job_working_directory/000/58/tool_script.sh] for tool command[blastdbcmd -version > /tmp/tmp9NURC9/tmp/GALAXY_VERSION_STRING_58 2>&1; blastdbcmd -dbtype prot -db "/home/travis/build/peterjc/galaxy_blast/test-data/four_human_proteins.fasta,/home/travis/build/peterjc/galaxy_blast/galaxy-dev/test-data/four_human_proteins.fasta" -info -out "/tmp/tmp9NURC9/files/000/dataset_58.dat"]
For other tools you can select multiple databases as input, and the blastn tests try that, e.g. a working test gave this (where the space is deliberate as that's how the NCBI handle multiple DB arguments):
blastn ... -db "path1 path2"
galaxy.jobs.command_factory INFO 2016-08-05 14:58:01,783 Built script [/tmp/tmp9NURC9/job_working_directory/000/76/tool_script.sh] for tool command[blastn -version > /tmp/tmp9NURC9/tmp/GALAXY_VERSION_STRING_76 2>&1; blastn -query "/tmp/tmp9NURC9/files/000/dataset_75.dat" -db "/home/travis/build/peterjc/galaxy_blast/test-data/three_human_mRNA.fasta /home/travis/build/peterjc/galaxy_blast/galaxy-dev/test-data/three_human_mRNA.fasta" -task megablast -evalue 0.001 -out "/tmp/tmp9NURC9/files/000/dataset_76.dat" -outfmt 6 -num_threads "${GALAXY_SLOTS:-8}"]
But when a single database is given we get a repeated path:
blastn ... -db "path path"
galaxy.jobs.command_factory INFO 2016-08-05 14:58:22,901 Built script [/tmp/tmp9NURC9/job_working_directory/000/78/tool_script.sh] for tool command[blastn -version > /tmp/tmp9NURC9/tmp/GALAXY_VERSION_STRING_78 2>&1; blastn -query "/tmp/tmp9NURC9/files/000/dataset_77.dat" -db "/home/travis/build/peterjc/galaxy_blast/test-data/rhodopsin_nucs.fasta /home/travis/build/peterjc/galaxy_blast/galaxy-dev/test-data/rhodopsin_nucs.fasta" -task megablast -evalue 0.001 -out "/tmp/tmp9NURC9/files/000/dataset_78.dat" -outfmt 6 -num_threads "${GALAXY_SLOTS:-8}"]
Any thoughts on where this is breaking? In planemo?
These tests are working locally via "planemo test" (older version of planemo, older version of Galaxy, Python 2.6).
Peter
On Tue, Aug 9, 2016 at 6:33 PM, Peter Cock p.j.a.cock@googlemail.com wrote:
Hello all,
Recently I've been debugging various problems with my Galaxy tool TravisCI tests as part of moving to using "planemo test" rather than the fragile system I first created which mimicked a pre-ToolShed manual install:
https://blastedbio.blogspot.co.uk/2013/09/using-travis-ci-for-testing-galaxy...
Progress:
https://travis-ci.org/peterjc/galaxy_mira - passing now :) https://travis-ci.org/peterjc/pico_galaxy - probably a loc file problem
I now understand why this was breaking and have a workaround, filed as:
https://github.com/galaxyproject/planemo/issues/530
https://travis-ci.org/peterjc/galaxy_blast - different loc file problem
This email is about the problems testing the BLAST wrappers where I am using a BLAST database defined in test-data/*.loc files.
(I wasted some time today trying moving/copying the loc files and databases to different places before I noticed the issue below)
Problem lines quoted are from this run, galaxy master or dev branch: https://travis-ci.org/peterjc/galaxy_blast/builds/150072050 as of this commit: https://github.com/peterjc/galaxy_blast/commit/03f81e7fb1e0562b46b8b2a2079c9...
Some of the command expect a single BLAST database from the test-data/*.loc files, and the test should do that, but we get:
blastdbcmd ... -db path,path ... rpsblast ... -db path,path ... rpstblastn ... -db path,path ...
Here we should get ... -db path (once only), not repeated with a comma, e.g.
I think I now understand how this happens, see:
https://github.com/galaxyproject/galaxy/pull/2644
Peter
galaxy-dev@lists.galaxyproject.org