WDL output files can be mis-organized in the output directory if they come via the call cache

[adamnovak]

I ran a WDL workflow. It failed after filling some of the cache. I fixed the workflow and ran it again, using the same call cache. I managed to get a workflow output file like this:

"AcceptanceTest.candidate_aardvark_summary": "/private/groups/patenlab/anovak/trash/acceptance-test/out/AcceptanceTest.candidateCalling.DeepVariant.AardvarkEvaluation/miniwdl-cache/compareCallsAardvark/woqqzwgmk5m6bmlfevhdwugtr7fu2mcu/stxl7uans3lxx4zsqcieyrc3tgpzitqi/9d6aea74-6234-4bed-a29b-0d74ccc34357/AcceptanceTest.candidateCalling.DeepVariant.AardvarkEvaluation.compareCallsAardvark/HG002/summary.tsv",

I think that that file was pulled in from the cache entry from the earlier workflow run, and then ended up using its path within the cache (including the miniwdl-cache directory name for where I put my cache) in the part of the output filename that's meant to reflect where the file came from within the task that generated it.

I want this file to have ended up at:

/private/groups/patenlab/anovak/trash/acceptance-test/out/AcceptanceTest.candidateCalling.DeepVariant.AardvarkEvaluation.compareCallsAardvark/HG002/summary.tsv

This would be under a different task's name in the output (the same one it was under in the cache). I think that's where it would have ended up if it hadn't been stored in and retrieved from the call cache.

┆Issue is synchronized with this Jira Story
┆Issue Number: TOIL-1860

[adamnovak]

This is with Toil c6c2b1e.