Open clintval opened 3 weeks ago
Local and remote tests hang forever when you install bwa-aln-interactive
from bioconda and run tests. The executable from bioconda is either broken or this library isn't compatible with it yet.
My full stack trace is:
EDIT: I solved the issue. See later comments.
I've narrowed the issue down to this "hack" which does not appear to work for me:
Increasing the records for warmup doesn't resolve the issue.
This is difficult to debug because stderr is suppressed for everything (bwa, htslib, etc.).
Forcing flush/unbufferd IO with PYTHONUNBUFFERED=1
doesn't work either.
Although it initially appears pysam AlignmentFile is to blame (hence the hack), I have discovered the issue is not actually with pysam at all. If I remove the dependency on AlignmentFile
and process lines into AlignmentHeader
and AlignedSegment
objects directly (using the from_text()
and from_string()
class methods) then I see that bwa-aln-interactive
does not flush output immediately. It appears to buffer and hang at times. The number of threads you assign to the executable do matter!
Removing the "hack" is easy and I suggest we do it anyway.
But it exposes an issue with bwa-aln-interactive
.
I figured it out. The GitHub release of bwa-aln-interactive
is broken and we are shipping the non-interactive bwa. Next, I will fix up the public repository and the bioconda package by incrementing the build
number for bwa-aln-interactive
.
As a side quest, I was able to remove the AlignmentFile
"hack"!
I wish prymer
had better logging redirects for the underlying code it calls. Debugging this was painful without knowing where all the standard errors go (hint... it's /dev/null
and the user can't change that).
I wish prymer had better logging redirects for the underlying code it calls. Debugging this was painful without knowing where all the standard errors go (hint... it's /dev/null and the user can't change that).
Can you open a follow-up issue for this? I agree
I wish prymer had better logging redirects for the underlying code it calls. Debugging this was painful without knowing where all the standard errors go (hint... it's /dev/null and the user can't change that).
Can you open a follow-up issue for this? I agree
You have one open for Primer3:
I wrote one up for bwa:
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 97.25%. Comparing base (
a132a52
) to head (a6205c9
). Report is 1 commits behind head on main.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
@msto's tested the developer docs and tests passed.
I think this is ready for review!
Closes https://github.com/fulcrumgenomics/prymer/issues/5 Closes https://github.com/fulcrumgenomics/prymer/issues/30 Closes https://github.com/fulcrumgenomics/prymer/issues/27
This PR also: