This email is clearly for post branch and maybe even much longer, but as we just ran into (I think) new py-twisted doesn't work with current py-OpenSSL and new py-OpenSSL drops 2.7, I wonder where we are on the long slow removal of 2.7. I am in no rush to do this prematurely in terms of breaking packages, even if by definition those packages do not have functioning upstreams. But in the end we'll need to make the call to just drop 27 as surely some projects are never going to deal. I feel like it's "not for 2023Q1" at least, and maybe not for 2023 at all. By 2025, it's hard to think we'd still have 2.7; part of this is being slightly slower to drop 27 than Debian etc. It seems like there are still a fair bit of things that need python2.7. The only one that comes to mind is ham/chirp, and yes, upstream knows that python3 support is desired and not present. More and more libraries are dropping 2.7 leading to more versioned dependencies and that I think is causing update friction but maybe the py-pkgsrc-experts are adding versioned deps so fast it's not. Do people think we have a few stragglers, or a lot? Is somebody easily able to run a script over a bulk build and list packages that are py27 only (excluding the py27 half of versioned_dependencies)?
Attachment:
signature.asc
Description: PGP signature