No build for Mac found via mamba/conda - Mac installation fails.

[bskubi]

It looks like there are only versions of hictk available for the linux-64 platform, so the installation instructions you give for Mac are not working. To check, I ran mamba search hictk --info and conda search hictk --info which only list bioconda/linux-64 in the URL.

Would you be able to publish builds for Mac? This is useful software, it would be very nice to have an easy install for my non-Linux-running collaborators! Thank you!.

[robomics]

Hi!
Uhm, that's weird. We do build a package for macos as part of the recipe on bioconda, however for the time being the package targets x86 (as far as I know, bioconda does not yet allow building native arm packages for macos).

I don't have access to a mac at the moment, so I've setup a simple CI pipeline that installs conda and then tries to install hictk on various platforms (the yaml for the CI is hosted here: https://github.com/robomics/hictk-conda-macos-dbg/blob/main/.github/workflows/ci.yml).

Here is a successful run of the above pipeline: https://github.com/robomics/hictk-conda-macos-dbg/actions/runs/8207731977/job/22449672110

In order to make conda find hictk when running on apple silicon I had to export CONDA_SUBDIR=osx-64 before running conda install (see this issue for more details: bioconda/bioconda-recipes#41702 (comment)).

I assume that your colleagues are running recent macs. Could you please double check if they are exporting CONDA_SUBDIR before runnning conda install?

However note that installing hictk in this way will lead to performance that is suboptimal, as hictk will be executed using Rosetta2.

If performance is important, unfortunately the only solution is either compile hictk from source, or using containers.

When bioconda will start supporting building packages for apple silicon I will make sure to update the recipe to also target this platform.

[bskubi]

Thanks for the quick response. Prepending CONDA_SUBDIR=osx-64 worked for my colleague. May be worth adding that to the installation page :) We'll try docker if it turns out to be too slow.

[robomics]

Thanks for confirming the fix worked!
Will keep this issue open as a reminder to update the docs :)