cdt?

[hmaarrfk]

I could always adapt them from https://github.com/AnacondaRecipes/aggregate
but I don't feel like writing custom scripts for them.

[hmaarrfk]

@jjhelmus did you do something special to your CDT packages?

I can find your tk package, but not your libxcb-cos7-aarch64 package for example.

[root@c9be9729e2b9 /]# conda install tk
Solving environment: done

## Package Plan ##

  environment location: /root/archiconda3

  added / updated specs: 
    - tk


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    tk-8.6.8                   |       hbc83047_0         3.2 MB  c4aarch64

The following NEW packages will be INSTALLED:

    tk: 8.6.8-hbc83047_0 c4aarch64

Proceed ([y]/n)? n


CondaSystemExit: Exiting.

[root@c9be9729e2b9 /]# conda install libxcb-cos7-aarch64
Solving environment: failed

PackagesNotFoundError: The following packages are not available from current channels:

  - libxcb-cos7-aarch64

[hmaarrfk]

hmmm, even mine are weird: https://anaconda.org/archiarm/libxcb-cos7-aarch64/files

but i can find them locally on my qemu machine. I guess i'll just build tk like that.

[jjhelmus]

I was not able to upload CDT packages either, I'm looking into the issue.

As for creating the recipes I needed to modify conda-build with a new URL for the source packages. Building needed to be done with an older version of conda-build due to conda/conda-build#3347 which you reported. I'm planning on submitting PRs to conda-build to fix both issue.

[jjhelmus]

After applying conda/conda-build#3350 I was able to update the CDT recipes using conda build 3.17.6. The resulting recipes had no issue building with the same version of conda-build and could be uploaded without issue. CDT packages are now available in the c4aarch64 channel

[hmaarrfk]

Cool. conda 3.17 was having trouble testing split packages. I don't know if that is still an issue.
That isn't usually an issue, but as we have a few hiccups with aarch64, it is probably best not to skip tests

I might have to make a run and just build all the conda-build packages. Is there a reason why they aren't set to noarch: python?

I guess AD wants to include them in miniconda/Anaconda, so having them be noarch would be annoying for AD.