1. A tool that gives different md5sums on docker v...
# rants
j
1. A tool that gives different md5sums on docker vs conda? Because it embeds the command in the header of the output file! Not bad for reproducibility right? BUT WAIT, BECAUSE OF SOME STUPID PYTHON BEHAVIOUR, IT RESOLVES THE FULL PATHS OF THE EXECUTION SCRIPTS! SO YES OF COURSE ITS GOING TO BE DIFFERENT BETWEEN AND DOCKER AND SINGULARITY EVEN THOUGH THE COMMAND IS EXACTLY THE SAME AND SAME OUTPUT?! 2. And for another tool... SOMEHOW FOR REASONS THAT ARE THAT IS BEYOND ME, RUNNING LOCALLY PRODUCES AN OUTPUT FILE WHERE IN THE LAST COLUMN THERE IS AN INTEGER!? RUN THE SAME COMMAND ON GHA? YES LETS CONVERT THAT COLUMN TO A FLOAT?! YET BOTH TESTS ARE RUNNING IN DOCKER
elmofire 2
n
1. Is the file readable? Can you parse out the command line? 2. well that's just bad tool design 😱
j
1. Just skipping that line now 2. But it's both on docker! It should be the same
n
I've never seen 2 happen before, what tool is that??! (so I can avoid it all cost 😁)
j
hamronization
n
run away
s
There are many many tools that print the command in the header, probably the #1 cause of docker/conda differences
(some tools will print the full file path to the file itself, so they aren't stable in anything due to the work directory)
j
I hate them all
l
@Sofia Stamouli 👆