Closed jarrodmillman closed 9 months ago
The PIP_FLAGS
variable wasn't being used. Setting CIBW_ENVIRONMENT
didn't work. I also don't think the comment about 'numpy>=1.26.0b1; python_version>="3.12.0.rc1"',
is true. Most of the other projects including scipy and pandas are using "numpy>=1.26.0b1; python_version>='3.12'",
which seems to work fine in all the projects. If you see an instance where that doesn't work, let me know.
See https://github.com/matplotlib/matplotlib/issues/26736. Also you may want to look at my comments about the version numbering. I suspect other projects are likely to be using the wrong numbering scheme, since I've seen other projects do similar things.
They are planning matplotlib 3.8 final release this week. That should simplify this PR, so we may want to wait for that release. But I am also happy merging this now and fixing things later.
Most of the other projects including scipy and pandas are using
"numpy>=1.26.0b1; python_version>='3.12'",
which seems to work fine in all the projects.
Hmm, I thought it wouldn't - but maybe tools have special-cased Python version numbers now in a way that deviates from PEP 440 for convenience. If so, great - fewer things to tweak.
That should simplify this PR, so we may want to wait for that release. But I am also happy merging this now and fixing things later.
If you need a final release of pywavelets now to make life easier for scikit-image & co, then we can merge this now. Otherwise, maybe easier to wait? No strong preference from me.
Let's give it a few days and see how things progress. It would be good to have a pywt release early next week. I have a few more cleanup PRs I would like to make, so maybe we can get those in as well.
I'm on holiday next week, so if you need it soon it may be useful to do the release by Friday. @grlee77 is that fine with you? Anything you'd like to get in?
The numpy 1.26.0 release will be out today if things go as planned, so let's hold off on modifying the wheel build jobs to deal with pre-releases.
To pick up the right
numpy
pre-release version, see https://github.com/numpy/numpy/issues/23808#issuecomment-1678677094 andPIP_PRE
a few comments up from that.