regro / cf-scripts

Flagship repo for cf-regro-autotick-bot
Other
53 stars 74 forks source link

Py313 migration stuck on networkx - wrongly still "Awaiting parents"? #3191

Closed h-vetinari closed 1 hour ago

h-vetinari commented 14 hours ago

The 3.13 migration has networkx as the top missing dependency, yet, paradoxically, as still "Awaiting parents". I've checked the "Done" feedstocks as well for whether they're parents of networkx (and whether they were built without spurious CI failures), but I cannot tell where this is hanging.

The problem is that even a manual PR to networkx will fix this, as the bot still thinks it needs to wait for something.

CC @beckermr

beckermr commented 14 hours ago

Hrmmmm ok. We need to check the node attrs and see what they look like. Could be a v1 parsing error.

beckermr commented 14 hours ago

I think the issue is here: https://github.com/regro/cf-scripts/blob/main/conda_forge_tick/migrators/migration_yaml.py#L615

It appears that v1 recipes parsed by rattler build do not come out with the outputs build section filled in. On networkx, the output has a build section of null in the json and so the output is wrongly assumed to not be noarch.

What is the semantics for v1 recipes? Does the parent build section get overrides from the outputs and then get used for the output? We can add w/e logic is correct to the function.

h-vetinari commented 14 hours ago

What is the semantics for v1 recipes? Does the parent build section get overrides from the outputs and then get used for the output? We can add w/e logic is correct to the function.

Pinging @wolfv @baszalmstra @nichmor :)