regro / conda-forge-feedstock-check-solvable

A mamba- or rattler-based package to check if a conda-forge feedstock is solvable.
BSD 3-Clause "New" or "Revised" License
1 stars 7 forks source link

fmt11 migration erred on mamba #51

Closed jjerphan closed 2 weeks ago

jjerphan commented 1 month ago

The bot err'ed on mamba for the fmt 11 migration with the following error:

bot error (
bot CI job
): main:
Error running check-solvable in container - error ValueError raised:
"Very odd pinning: pin_compatible('fmt',max_pin='x.x')! Package fmt not found in host {}!"

fmt is indeed only present in the host section of libmamba (and libmambapy's) and not in mamba, because mamba does not depend on it.

I will open a PR to perform the migration manually.

This was previously discussed with @h-vetinari which invited me to report this issue.

h-vetinari commented 1 month ago

fmt is indeed only present in the host section of libmamba

This is not correct, see line 55. However the bot is also not correct, in that it diagnoses a use of pin_compatible without a host dependency, yet the host dependency is there.

There are a bunch of superfluous run-requirements (already covered by run-exports), but those are benign, and certainly shouldn't fail the bot, which is why I suggested raising an issue.

jjerphan commented 1 month ago

I should have been clearer: I meant to highlight that the relevant host sections were libmamba's and libmambapy's, not that other pins were not present in other sections like the pin_compatible ones.

beckermr commented 1 month ago

Even weirder is that if you dig directly into the bot ci job, you'll see an actual solver error and some other random missing host package. I'm not actually sure what the error string the bot was supposed to report was.

beckermr commented 1 month ago

Looks like the pr failed which means I think maybe the error string on the status page is wrong?

jjerphan commented 1 month ago

We need the changes upstream to be part of a new release for this PR to pass.