it might have been the multi-build container that was broken
which I fixed by bumping the samtools version
all good :)
I’m rewriting the involucro part of mulled-build in python, couldn’t figure out how to capture container output and use it in a subsequent step… if I fdo this we could also produce proper docker files using the multi-stage strategy. would that be something worth publishing somewhere ?
but why use this overly complex thing that isn’t designed to interact and capture container output
if we just template out a dockerfile we can do the same for singularity, giving us on-the-fly singularity builds without needing docker, that would be another advantage
and we don’t need host mounts either that aren’t properly clean up
there’s lot’s of advantages to be had if we simplify this
this wouldn’t change anything whatsoever in how you use the mulled-build ecosystem
same commandline flags etc
and we can also template out the ENV vars that conda activate puts into the environemnt
I'm not worried about that. Many people complained about the lua stuff, so there is probably some appeal to use multi-stanged-build - I guess lua was a wrong choice. I just think its a bad timing. This really needs to be tested carefully before we can roll this out. This has a large impact to not only our ecosystem.
the lua file is quite simple in what it does, I will produce tests for all cases
Sorry, I'm in an EOSC meeting, I can not concentrate enough on this topic as it deserves