Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

REV: 1.13.x: revert changes to f2py and tempita handling in meson.build files #20567

Merged

Conversation

rgommers
Copy link
Member

Reverts the last two commits of gh-20121. Those were not robust against Meson picking up a different Python interpreter for which numpy and/or cython were not installed. This is rare, but can happen when a distro builds for a non-default version of Python and disables build isolation.

Addresses the issue reported in gh-20535 in the 1.13.x branch. Leaving the issue open because I'd like a different change in main. In a release branch a straight revert seems best though, since we had no issues opened related to this for 1.12.x.

Copy link
Contributor

@tylerjereddy tylerjereddy left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I spent some time trying to reproduce and actually failed even with various attempts at aliasing python3 to something else and disabling build isolation. I think things are fairly robust already with i.e., venv on Ubuntu even if I make sure that python3 points to something without TPL installs.

Anyway, probably not worth me trying to repro in docker or something. I'll go ahead and merge.

There are some CI failures that are not related, some are connectivity we can ignore, but others may need investigating on the release branch so I'll open an issue for those.

@tylerjereddy tylerjereddy merged commit dd1e525 into scipy:maintenance/1.13.x Apr 24, 2024
21 of 29 checks passed
@tylerjereddy
Copy link
Contributor

thanks

@rgommers
Copy link
Member Author

I think I could write a CI job that covers this, and I may do so at some point - but it'd be time-consuming to fiddle with for sure.

@bnavigator
Copy link
Contributor

I spent some time trying to reproduce and actually failed even with various attempts at aliasing python3 to something else and disabling build isolation. I think things are fairly robust already with i.e., venv on Ubuntu even if I make sure that python3 points to something without TPL installs.

I gave you a reproducer in gh-20535, didn't I? Make sure the shebang hardcoded interpreter /usr/bin/python3 does not have numpy and f2py installed (/usr/lib.../site-packages) and a non-isolated build will fail.

@tylerjereddy
Copy link
Contributor

I gave you a reproducer in #20535, didn't I?

Not really, I spent quite some time trying to reproduce in a venv on Ubuntu Linux, and meson was smart enough to do the right thing time and time again. Same thing on MacOS too, to be honest. I understood your basic complaint, but I don't buy that the reproducer is fully trivial and portable to venvs with a naive python3 more broadly just yet.

@tylerjereddy tylerjereddy mentioned this pull request May 2, 2024
10 tasks
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Build issues Issues with building from source, including different choices of architecture, compilers and OS Meson Items related to the introduction of Meson as the new build system for SciPy
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

3 participants