Dear all,
While updating my MIRA wrapper for MIRA v3.4.0, with a test case that takes about half an hour to run, I was sometimes getting the following error message (on stderr):
uname: write error: Broken pipe
My hunch is this is a side effect of adding a version_command tag, <version_command>mira --version | head -n 1</version_command> https://bitbucket.org/peterjc/galaxy-central/changeset/d2c80e51d1fc
This isn't down to the dataset or very long run times - I've tried a small phage which assembles in about 5 minutes with the same error on stdout. When I have repeated the exact same command at the terminal, it has run without issue, e.g.
$ mira --version | head -n 1 &> /home/pjcock/repositories/galaxy-central/database/tmp/GALAXY_VERSION_STRING_192; python /home/pjcock/repositories/galaxy-central/tools/sr_assembly/mira.py mira /home/pjcock/repositories/galaxy-central/database/files/000/dataset_403.dat /home/pjcock/repositories/galaxy-central/database/files/000/dataset_404.dat /home/pjcock/repositories/galaxy-central/database/files/000/dataset_406.dat /home/pjcock/repositories/galaxy-central/database/files/000/dataset_405.dat /home/pjcock/repositories/galaxy-central/database/files/000/dataset_407.dat /home/pjcock/repositories/galaxy-central/database/files/000/dataset_408.dat mira --job=denovo,genome,accurate 454_SETTINGS -LR:lsd=yes -LR:mxti=no -LR:ft=fastq -FN:fqi=/home/pjcock/repositories/galaxy-central/database/files/000/dataset_402.dat COMMON_SETTINGS -OUT:rrot=yes -OUT:rtd=yes
MIRA took 4.98 minutes Done
One point which may be relevant is that the MIRA binary produces quite a lot of output on stdout (often megabytes), which my Python wrapper scripts is redirecting to a file (to collect it as a Galaxy output). Perhaps this issue is related to lots of data on stdout/stderr?
For the moment I am removing the version_command tag, which appears to solve this problem, and I will report back if the broken pipe error still happens.
Has anyone else noticed anything similar?
Peter