-
Notifications
You must be signed in to change notification settings - Fork 100
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Simplify reconstruction of fetch actions #376
Conversation
✅ Deploy Preview for conda-lock ready!
To edit notification comments on pull requests, go to your Netlify site settings. |
pre-commit.ci autofix |
Seems like this does improve the md5 test issue? |
cf13dc2
to
5e0fd20
Compare
☝️ squash @mariusvniekerk, yes, this resolves it. In the process I noticed a whole bunch of illogical stuff regarding the formation of LockedDependencies which I wanted to clean up, but it goes deeper than I originally thought. I think it makes sense to merge this so that CI works, and follow up with the remaining stuff in future PRs. 👇 rebase |
5e0fd20
to
b39bea4
Compare
As a brief explanation, my prevailing theory (based on substantial evidence) is that when Micromamba computes the dry-run This PR abandons the approach of filling in missing Micromamba |
@mariusvniekerk, all tests are green, so if you approve, then we can merge this, fixing most of the CI issues. As a follow-up to this, we should make sure that the approaches in |
cc @wolfv there is probably some stuff in this discussion that you should be aware of |
@maresb can you share a bit more of the symptoms? What I understand is that the repodata_record.json contains the correct complete values for MD5 and SHA256, but micromamba is somehow not reading them? |
Ya, I will try to open a proper issue soon in mamba-org/mamba with more details. It relates to some apparent race condition that occurs on a regular basis on our test suite but is not reliably reproducible. We have multiple parallelized tests spawning Conda/mamba/micromamba, but I am not sure whether or not this is a multiprocess problem. The problem seems to occur when we do a dry-run solve to get LINK and FETCH action JSON data, and the md5 and sha256 fields of a link-only package are missing from this output. When this triggers, I immediately look (programmatically) at the installation location of the package in question, and it looks like extraction is incomplete. It completes after sleeping for 100 ms, and if I then repeat the dry run it is successful. |
I'm very interested to see if this has any affect on #338.