https://nf-co.re logo
Join Slack
Powered by
Merged `master` into this PR branch to try and get...
# modules
a
Merged 
master
into this PR branch to try and get it ready to merge again, and now seemingly randomly the 
.tbi
files for only the gvcf are reporting different 
md5sum
! The VCF/versions etc of all the other snapshots are fine, but for some reason the indexes are changed. Anyone have any thoughts on how that can happen? github.com/nf-core/modules/pull/11859 Git diff also shows no difference between the files under 
subworkflows
or 
modules
from the commit that previously ran these tests passing 🤯
s
tbi files aren't always stable for some unknown reason
party this 1
a
shared a GIF
It’s odd because I see the same issue when I regenerate the snapshots locally (and they match the remote regenerated ones) so whatever is causing it is some semi-deterministic seed or something
s
A timestamp? 😄
can you read what the header contains?
a
In the index but not the actual VCF?! That would be hell
The actual VCF files are fine - same exact 
md5sum
as months ago. Do 
tbi
have a human-readable form at all?
s
Claude thinks it might be related to parallelised bgzip writing different block layouts
a
Going to assume it’s that. It’s so odd it doesn’t affect the VCFs or the VCF/TBIs for the non-gVCFs in the same exact process.
s
maybe the vcf tbis aren't being parallelised?
a
Possibly. I think that the runners have been given more CPUs since the last run so genuinely might be why
s
could be
m
tbis are werid
2 Views
Open in Slack
PreviousNext
Powered by