search

OpenAstronomy / conda-channel

An open source community conda channel
8 stars 8 forks source link

Try to build 0.7rc1 of SunPy #15

Closed Cadair closed 7 years ago

Cadair commented 8 years ago

@mwcraig any way to upload to a different label for pre-release versions?!

mwcraig commented 8 years ago

Not at the moment, but (relatively) easy to implement, I think. Does logic something like this work?

if letters_are_in_version_number:
    upload_package_to_alternate_label
else:
    upload_package_to_main_label

Also, do you see a need for separate labels for dev releases (e.g. 0.7.dev1234) and release candidates/alpha/beta (e.g. 0.7.rc1 or 0.7.beta1)?

mwcraig commented 8 years ago

Builds are failing because it can't download the course tarball...

Cadair commented 8 years ago

@mwcraig yes maybe that's because pypi put dev versions somewhere else?!

mwcraig commented 8 years ago

@Cadair -- more likely the fact pypi.io is still in development. Can you try changing the url to something like this:

https://pypi.python.org/packages/source/s/sunpy/sunpy-{{version}}.tar.gz

Cadair commented 8 years ago

This URL does not work either: https://pypi.python.org/packages/source/s/sunpy/sunpy-0.7rc1.tar.gz

mwcraig commented 8 years ago

I really hope this isn't a permanent change in how one resolves links to pypi packages...

mwcraig commented 8 years ago

I did, incidentally, get Windows 64-bit building last night, still need to merge the change, though. A couple more changes to make that I'll do that now, and then, if you want, you can remove the skip on Windows.

Cadair commented 8 years ago

the py35 windows skip is because of glymur.

mwcraig commented 8 years ago

Windows build is going to fail on import sunpy again, so canceling appveyor. I'll ping you to rebase in a bit, and then Windows should succeed.

mwcraig commented 8 years ago

@Cadair -- can you please rebase?

Cadair commented 8 years ago

@mwcraig done

mwcraig commented 8 years ago

@Cadair -- looking into the source of the latest fails. Could you do me a favor and also push your branch to this repo (openastronomy/conda-channel)?

That would help determine whether the problem really is with not having a BINSTAR_TOKEN set, as I suspect.

In the meantime I'm restarting the appveyor build in hope it was just an intermittent thing a anaconda.org.

Cadair commented 8 years ago

pushed to a branch.

Cadair commented 8 years ago

say waaaa https://ci.appveyor.com/project/openastronomy/conda-channel/build/1.0.56/job/06lq94t0o70p0cyu#L2037

mwcraig commented 8 years ago

say waaaa

yeah....that. Oddly, that is what was happening when I used an outdated version of conda-build in python 2.7 locally. Updating conda-build seemed to fix it (on 2.7).

Will try to reproduce locally, but could you also try altering https://github.com/OpenAstronomy/conda-channel/blob/sunpy-07rc1/appveyor.yml#L81 to use conda-build 1.20.3 just in case that fixes?

It looks like there is a separate issue with conda-build-all in cases when a valid binstar token is not defined, but we can work around that for now by using branches in the openastronomy repo.

mwcraig commented 8 years ago

I have no idea what the underlying issue is. When I build locally on python 3.4/win64 I don't see the error. I can, at least for now, upload my local build if you'd like.

mwcraig commented 8 years ago

For reference, this discussion of handling pre-release uploads: https://github.com/astropy/conda-build-tools/issues/13

bsipocz commented 7 years ago

This is probably need to be closed as there are newer releases of sunpy in this channel.