These are chat archives for nextflow-io/nextflow
-resumeerrors, but I frequently see processes re-run that I don't expect. There are simple reasons like input files getting touched even if the content hasn't changed. If any of your input channels are reduced to a list, then the order of the list may change, even if nothing upstream has actually changed. And don't forget transforming channel operators like
groupTuplethat can have similar ordering issues. Using the
cache 'deep'directive can help with file problems and things like
.toSortedList()can help with enforcing an ordering. Finally, to debug @pditommaso recently added the
-dump-hashescommand line flag that will dump the hashes used to determine whether to resume to stdout. It'll take some digging to work through what's what, but the answer is probably there.