Test with mambabuild

[mariusvniekerk]

…0, and conda-forge-pinning 2021.10.13.20.28.53

Checklist

• Used a personal fork of the feedstock to propose changes
• Bumped the build number (if the version is unchanged)
• Reset the build number to 0 (if the version changed)
• Re-rendered with the latest conda-smithy (Use the phrase @conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• Ensured the license file is being packaged.

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[mariusvniekerk]

This PR should not be merged, but exists to validate that mambabuild works as intended for a default build system

[isuruf]

@mariusvniekerk, what's the purpose of this PR?
Please note that this feedstock is really expensive in terms of CPU hours.

[mariusvniekerk]

@beckermr requested that I use this one as a stress test before we roll out the new smithy with mambabuild

[mariusvniekerk]

cc @wolfv seems that conda mambabuild doesn't work here and fails with some circular reference parts.

Could that also just have been caused by the removal of the defaults channel here?

[beckermr]

Yeah this is a red flag for me that we have some latent bugs that need fixing before we roll out mambabuild.

[wolfv]

this circular dependency notification seems to come from conda-build. Not sure if it's really related to mamba in this case (I don't see mamba invoked before the circular dep error happens -- mamba usually prints a transaction).

Is there a circular dependency?

conda_build.exceptions.RecipeError: Circular dependencies in recipe: 
    libgcc-devel_linux-aarch64 <-> gcc_impl_linux-aarch64
    libgfortran5 <-> gcc_impl_linux-aarch64
    libgfortran5 <-> gfortran_impl_linux-aarch64
    libgomp <-> gcc_impl_linux-aarch64
    libstdcxx-devel_linux-aarch64 <-> gxx_impl_linux-aarch64
    libgcc-ng <-> gcc_impl_linux-aarch64
    libgcc-ng <-> gfortran_impl_linux-aarch64
    libsanitizer <-> gcc_impl_linux-aarch64
    libstdcxx-ng <-> gcc_impl_linux-aarch64
    libstdcxx-ng <-> gfortran_impl_linux-aarch64
    gcc_impl_linux-aarch64 <-> gfortran_impl_linux-aarch64
    gcc_impl_linux-aarch64 <-> gxx_impl_linux-aarch64
    gfortran_impl_linux-aarch64 <-> gxx_impl_linux-aarch64

I will have a more in-depth look tomorrow.

[beckermr]

This package builds when mamba is not used, so something is happening there.

[beckermr]

Looking back on it, this feedstock passed two days ago without mamba: #76

AFAIK there has not been a release of any relevant packages for the build since then.

So I do think somehow mamba is tripping a bug or has one.

[wolfv]

I am happy to look into this more tomorrow (quite later over here in Europe). Still unsure what's going on -- maybe there is some pre-processing that's missed in the mambabuild -- or maybe what Isuru just pushed will fix it, let's see! :)

[beckermr]

Sorry @wolfv! Not trying to push you at all. Just writing down my thinking for why this is odd. Get some sleep and thank you for all of the hard work!

[wolfv]

No worries, I am really curious about this one. It doesn't really look like a "mamba"-specific bug right now, but I need to investigate a bit more.

[isuruf]

It's not. It's a conda-build that I fixed a long time ago, but waiting for a conda-build release.

[wolfv]

This is where the error seems to come from: https://github.com/conda/conda-build/blob/13aa64b3701499215c80847e097945441c410135/conda_build/metadata.py#L286-L305

It may be an issue with cross-compilation ... I'll have a look tomorrow! We can release boa fairly quickly, too :)

[beckermr]

It's not. It's a conda-build that I fixed a long time ago, but waiting for a conda-build release.

Something doesn't make sense though. Something about mamba is causing the build to hit this error again (as it did not a few days ago). That means it is shipping sufficiently different environments in the build to trigger the code path that conda did not. That in and of itself is a cause for concern.

[isuruf]

This line in master was removed in rerendering. https://github.com/conda-forge/ctng-compilers-feedstock/blob/master/.scripts/build_steps.sh#L15

[wolfv]

There is also the defaults channel that's removed though

[beckermr]

Ahhh thanks @isuruf. Assuming this build passes then we should go ahead and switch over!

[mariusvniekerk]

Nicely found @isuruf