Conda-build: noarch: python path lengths appear to cause breaks on CircleCI

[pmlandwehr]

I've been rebuilding some conda-forge recipes using the new noarch: python flag and have been seeing some build breaks due to path lengths getting too long. e.g. fortqdm I get:
`FileNotFoundError: [Errno 2] No such file or directory: '/feedstock_root/build_artefacts/tqdm_1505374729453/_b_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placeho/bin/tqdm=tqdm._main:mai'

I've seen the same error when rebuilding faker, pyjwt, and datadotworld-py. Has anyone else noticed an uptick in these path issues?

[mingwandroid]

I would stay away from noarch: python personally. Too many corner cases + bugs.

[msarahan]

I would like to support noarch better, but your timing as it stands
coincides with our next Anaconda release. PRs to fix anything are always
welcome, but I won't be able to look at anything more than a PR for the
next couple of weeks.

On Fri, Sep 15, 2017 at 3:23 AM, Ray Donnelly [e-mail]
wrote:

I would stay away from noarch: python personally. Too many corner cases +
bugs.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
https://github.com/conda/conda-build/issues/2378#issuecomment-329715813,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AACV-bw4sKPsOQ-PHSkptXeT7XwX0H22ks5sijP3gaJpZM4PYj2N
.

[jjhelmus]

@pmlandwehr In conda-build 2.x noarch: python packages are a bit more picky about the format of the entry_points in the meta.yaml file. Specifically, there need to be a space surrounding the equal sign. So change the tqdm recipe to - tqdm = tqdm._main:main.

This was also reported in #1893 and fixed in #1894 but has not yet been backported to conda-build 2.x. I'll submit a PR shortly for a backport.

[jjhelmus]

No PR is needed, the above change was already backported to the 2.1.x branch in #2348. I will be available in the 2.1.18 release.

[jakirkham]

My bad. Should have read the issue more closely before redirecting here. Already buried in notifications. 😧

[jakirkham]

Circling back to this one.

With faker and datadotworld-py, it looks like the previously described issue. So those should be easy to fix.

However with pyjwt the entry point should have an acceptable format, but it still fails. Here is the build log. Not sure why it can't find the entry point, but it misses it.

[pmlandwehr]

Thanks everyone for the quick feedback on this!

[jakirkham]

After looking at pyjwt it looks like the name of the entry point was wrong, which caused it issues. I think this issue can be closed.

[msarahan]

Thanks.