Skip to content

How to install on osx-arm64? #112

Closed
Closed

Description

@paddyroddy

Apologies, I'm aware that this is the last stable version but I just got a new laptop and now cannot install this great library.

Clearly this is related to libigl/libigl#1686. Is there a way and I can make the setup.py use header only mode if I compile from scratch i.e. libigl/libigl#1686 (comment)?

Activity

  1. changed the title [-]How install on macos-arm64?[/-] [+]How to install on macos-arm64?[/+] on Nov 7, 2021
  2. paddyroddy commented on Nov 7, 2021

    @paddyroddy
    ContributorAuthor

    The error I get after running python setup.py develop is like this
    image

  3. changed the title [-]How to install on macos-arm64?[/-] [+]How to install on osx-arm64?[/+] on Nov 29, 2021
  4. EJShim commented on Dec 10, 2021

    @EJShim

    Is that latest merged PR indicates that I can install and use igl-python in M1 mac?

    conda install -c conda-forge igl does not work for me now.

  5. paddyroddy commented on Dec 10, 2021

    @paddyroddy
    ContributorAuthor

    That's what I thought, doesn't seem to work yet. Someone told me about the conda-forge/conda-forge-pinning-feedstock repo

  6. paddyroddy commented on Mar 23, 2022

    @paddyroddy
    ContributorAuthor

    For me, now at least, a local installation works i.e. python setup.py develop

  7. merlinyx commented on May 11, 2022

    @merlinyx

    Thanks for the post! It's very helpful. I just want to add that I think currently in setup.py it's possible to fail to get the correct environment info.
    https://github.com/libigl/libigl-python-bindings/blob/master/setup.py#L89
    I would get errors like:

    (mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64e')

    I changed this line to
    tmp = os.getenv("target_platform", "") + platform.machine()
    and it worked out fine (added the correct cmake flag for osx arch).

  8. rubenwgs commented on May 30, 2022

    @rubenwgs

    Are there any updates on this on how to install libigl via conda on M1 architecture? igl is not found on the noarch or osx-arm64 conda repo, forcing the install via the osx-64 repo works but then obviously fails when trying to use libigl in python...

  9. paddyroddy commented on May 30, 2022

    @paddyroddy
    ContributorAuthor

    That's what I thought, doesn't seem to work yet. Someone told me about the conda-forge/conda-forge-pinning-feedstock repo

    So I've now realised that a PR is made here conda-forge/igl-feedstock#27, but looks like it's gone stale. Presumably because there are challenges in actually getting it working

  10. alecjacobson commented on Feb 5, 2023

    @alecjacobson
    Contributor

    fixed with #164 (use pip install libigl)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions