search

conda-forge / iris-feedstock

A conda-smithy repository for iris.
BSD 3-Clause "New" or "Revised" License
0 stars 17 forks source link

Restrict to matplotlib<2 #19

Closed bjlittle closed 7 years ago

bjlittle commented 7 years ago

The cartopy-feedstock recipe is limited to use matplotlib <2.

cartopy is not yet ready or tested against matplotlib 2.0.0 and iris is only tested up to matplotlib 1.5.3.

At the moment iris allows matplotlib 2.0.0 to be pulled in. We should throttle this back (for now) on this until iris and cartopy are tested and good to go with the latest matplotlib.

conda-forge-linter commented 7 years ago

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

bjlittle commented 7 years ago

@ocefpaf Hi! Any idea why this is failing ... thanks

ocefpaf commented 7 years ago

Reason: Incompatible library version: _datetime.cpython-36m-darwin.so requires version 12.0.0 or later, but libnetcdf.11.dylib provides version 11.0.0

My guess is that something needs to be re-built. I will investigate this further. Since this affects only Python 3.6 I am OK if you want to merge this for now.

bjlittle commented 7 years ago

@ocefpaf I'd be interested to know whether it's an easy fix that somehow I can blend into this PR on your recommendation ... otherwise do you think it would be best to merge a fix in a separate PR and then merge this one? That way we're not knowingly merging something that results in different builds available for different platforms ...

ocefpaf commented 7 years ago

@ocefpaf I'd be interested to know whether it's an easy fix that somehow I can blend into this PR on your recommendation ...

The thing is that I am not sure right now if that is an easy fix. I need to rebuild the dependencies locally (without a Mac BTW) to check. But that error has nothing to do with iris, so there is nothing to do in this PR.

otherwise do you think it would be best to merge a fix in a separate PR and then merge this one? That way we're not knowingly merging something that results in different builds available for different platforms ...

If I find the culprit all we need to do is to restart the CI that failed here. In this PR or in master if you choose to merge this.

@bjlittle I cannot look at this week but I will try to fix this over the weekend. I recommend merging as few people moved to Python 3.6 and this won't hurt many users.

ocefpaf commented 7 years ago

@bjlittle I tweaked the missing dependencies and this is passing now, but I am not merging this to avoid having to bump the build number twice when I merge #20

PS: my best guess the issue was a missing netcdf4 1.2.4 for numpy 1.12. It would be nice if that limitation was fixed in iris..