Python 3.5 for 0.22.X #49
Conversation
|
I guess we can just fetch the py35 artifacts and upload them to pypi. However we need to revert back right after since this ci is also used for the nightly builds and the dev version dropped support for py35 |
|
yeah, not clean. We really need a better release process. @thomasjpfan help us lolol (azure pipelines? :D :P) |
azure-pipelines.yml
Outdated
| @@ -75,6 +87,9 @@ jobs: | |||
| name: macOS | |||
| vmImage: xcode9-macos10.13 | |||
| matrix: | |||
| py_3.5_64: | |||
| MB_PYTHON_VERSION: "3.5" | |||
| NP_BUILD_DEP: "numpy==1.13.3" | |||
There was a problem hiding this comment.
Why not 1.12.1 as for Windows?
We should try to use the even older numpy 1.11.0 for building as noted below.
There was a problem hiding this comment.
They were different for py36, I just used the same as py36 for each platform.
There was a problem hiding this comment.
This is also consistent with https://pypi.org/project/oldest-supported-numpy/ BTW
There was a problem hiding this comment.
yes for macos, but not for windows. And we don't force the version for linux. I don't know why
| MB_PYTHON_VERSION: "3.5" | ||
| PLAT: "i686" | ||
| py_3.5_64: | ||
| MB_PYTHON_VERSION: "3.5" |
There was a problem hiding this comment.
I think we should force the build against the oldest supported of numpy for Python 3.5.
There was a problem hiding this comment.
Yes but the existing ones are not following oldest supported numpy. I thought that was intentional
There was a problem hiding this comment.
It's probably because I used the version numbers for 0.23 and not 0.22.X
There was a problem hiding this comment.
Anyway that should be fine as it is.
This is already Azure Pipelines. Actually it's my fault when we recently migrated to Azure Pipelines I removed older versions of Python to speed things up but I forgot that we still needed Python 3.5 for the 0.22.X branch. |
|
In the old travis config we used to build the Python 3.5 wheels against numpy 1.11.0: |
|
Note that for 0.23 and following, we should try to find a way to automatically fetch the oldest supported version of numpy and scipy from the |
Python 3.5 was removed from the matrix during the move to azure pipelines but 0.22.X still support py35.