[Including core@ because there is a specific arrangement to invoke them
on situations like this.]
Am 09.12.25 um 19:00 schrieb Antoine Brodin:
On Tue, Dec 9, 2025 at 6:06 PM Charlie Li <[email protected]> wrote:
George Mitchell wrote:
Thanks! Still interested in hearing consensus (if any) on whether it's
time to upgrade the default version ... -- George
3.13 will most likely be the next DEFAULT_VERSIONS bump at this point,
but there are a number of major consumers broken with it. These bumps
are a lot more difficult than what meets the eye, and have been getting
more difficult with the types of changes that every new major version
brings.
Hello,
With my exp-run hat, this is not a good idea to jump from 3.11 to
3.13, last time it took more than 1 year to complete the exp-run ,
while going from 3.11 to 3.12 would require a few weeks / months
Thanks for sharing that bad experience.
We've been unable to manage Python 3.13 until I pushed 3.14 into the
tree before 3.13 landed, and we've been unable to make 3.12 the default
Python before it went from "bugfix support" life^W "security support
only" (where limitation to source releases isn't disturbing for FreeBSD).
If things turn out bad, we need to leave some consumers behind and move
forward; in the interim, the backported fixes should be applied or we
upgrade to a point release if AND ONLY IF it's really imminent.
It does not make sense to waste brain or CPU cycles from the EOL 3.11 to
the EOL 3.12, and it also does not make sense to waste brain or CPU
cycles on getting a 100% collection if it's unfeasible.
So my take is try to fix as much as possible for three weeks, putting
the word out as a call for help (coordinate through the Wiki if needed)
then move before 2026Q1 branches, marking everything BROKEN and
DEPRECATED that doesn't work by then, and end of January everything
that's still broken with 3.13 or 3.14 gets EXPIRATION_DATE=2026-03-31 so
it doesn't appear in 2026Q2.
I am aware some ports will NOT work with 3.14 made the default because
we'd enter a bootstrapping circular dependency for now, but I hear
python@ seem to be wanting to make 3.13 (N-1) the default.
We can't let the museum of py-* ports hog the entire Python stuff in the
the ports tree down. We've eaten enough shit for our Python situation
already, it's tainting our reputation and we need to get our act together.
If someone proposes otherwise, get the backport patches from the
upstream repo in within the week for 3.12 and 3.11 and possibly either
forbid 3.10 or fix it too.
We might want to change policies though and push for Python 3.15 more
aggressively and make that the default in December 2026 with the
bimonthly regular 3.15.X release (likely 3.15.1 unless they need to
expedite security fixes as this year with 3.14.2).
--
Matthias Andree
FreeBSD ports committer
