-
-
Notifications
You must be signed in to change notification settings - Fork 503
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
cuda string incompatibility #2270
Comments
cc @conda-forge/core any ideas on this one? |
I think it's related to this: https://github.com/conda-forge/faiss-split-feedstock/blob/master/recipe/meta.yaml#L248 |
What I don't understand is how faiss (much less tsnecuda, which is brandnew) would affect the root build? |
Nevermind, just saw conda-forge/conda-forge-repodata-patches-feedstock#194. |
This is a failure in the final packaging step, so I don't know where the dependency tree is coming from. But I put in some debug statements and it's definitely the faiss/tsnecuda constraint conflict. But I believe I can close this as it's being addressed in conda-forge/conda-forge-repodata-patches-feedstock#194 and conda-forge/tsnecuda-feedstock#5 |
Well, beyond the fact that it happened now, we should maybe use this issue to discuss how to avoid it in the future? It was a simple oversight to use Also, given the fall-out (of affecting global resolution in completely unrelated feedstocks), I'm now worried to touch this at all (e.g. remove the underscore to follow a uniform approach, if that's what we think would be best). |
yeah that this broke the entire thing is kind of nuts, right? Does that mean there is a bug in conda/mamba somewhere? |
This showed up in conda build, completely independent of mamba, fwiw. But the fact that it broke everything is nuts, and also the error message is ... less detailed than it could be. |
Meaning |
Meaning conda build crashed in this strange way, and I didn't use mambabuild because I don't know if it works with multi-outputs yet? |
Ok. Was wondering because if they have different behaviors that might help in isolating the bug. |
Ugh, I have seen that error in recipes that merge build and host, since if a package name is present in both envs, the actual dist needs to be exactly the same. But that recipe doesn't do that so why is conda-build trying to merge specs like that? 🤷 |
Yeah exactly! That seems like a bug. |
I don't think any of us are following. Where in conda build is this merging done? |
It's not a conda-build issue. It's conda. See conda-forge/tsnecuda-feedstock#4 (comment) |
Right, it's in the "packaging" phase of the build process:
|
Hm, it's called by Mamba does not use that afaik, so if mambabuild also patches that phase, it should be ok there? |
Just to be clear, this is happening in a global sense? Because the package here doesn't depend on |
@erykoff, could you reopen this? I think we still haven't followed the underlying issue to the end, and the context of this thread is perhaps better than a new issue? |
This issue will go away if conda/conda#11612 is merged. However, the general solution for "compatible globs" is not trivial and requires regex look-aheads. Btw, that PR taught me that conda indeed supports full regex queries o.o |
Issue:
Building today in CI I am getting a cuda string error:
Example CI build here: https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=421595&view=logs&j=517fe804-fa30-5dc2-1413-330699242c05&t=c10fa5f2-fdf6-5338-3bdb-c4bea7c23412
from this PR: conda-forge/root-feedstock#151
This was rerendered with the following commit: "MNT: Re-rendered with conda-build 3.21.4, conda-smithy 3.15.0, and conda-forge-pinning 2021.12.02.17.12.32"
cc @beckermr
Environment (
conda list
):Details about
conda
and system (conda info
):The text was updated successfully, but these errors were encountered: