When the python module is in scope, usually the WANTLIB lists are quite
big and it's very easy to miss a hardcoded python314 or similar, as shown
by recent mistakes I made. Having this expanded and at the start of the
WANTLIB list gives a hint that one needs to pay attention and might help
reduce mishaps.
tested & ok sthen
failure if junking occurs. This was being avoided by adding a BDEP on
pythran to ports depending on cython. Change this to using nojunk
DPB_PROPERTIES instead as the pythran dep is fairly heavy to build
(using gfortran) and this on the path to building rust (via clang
-> py-sphinx -> py-stemmer -> cython) which is indirectly used by
a quarter of the ports tree.
manually; typically this would only be used for building a python extension
that's part of some other distribution in a subdir so this is what you
usually want)
with an old version of py3-bsddb3 installed (otherwise python is blocked).
has to be done this way rather than as quirks due to quirks leaving the
old packages still installed.
lang/python/3 via @pkgpath + @conflict (no quirks entry required
for that, because it is part of the update-set anyway).
glade in 7.8 depended on python 3.12 and in order that upgrades to
-current (and later 7.9) can work, we need to provide an upgrade path
that makes it disappear if installed. listing a port as obsolete does
not do this, it just leaves it installed, along with any packages
which it depends on (i.e. python 3.12 here), resulting in problems
if that causes a conflict.
3.13.10 was a rather massive update and .11 adds a few fixes on top,
some of them for not too terrible security issues:
https://www.python.org/downloads/release/python-31310/https://www.python.org/downloads/release/python-31311/
The annoying bit about this update is that our homegrown expat build
system does not install expat_config.h, which, strangely, is autoconf
results exposed in a public header (next to a few XML_ things). Until
that's sorted in base we get to patch the Python code since the expat
maintainer made the mitigation knobs only available behind some config
knobs, which the new Python code assumes to be enabled.
longer handle py2+3 in the same port
(I'll remove them from existing ports sometime, but probably alongside
a switch to a newer python branch, as a sweep touches a lot of PLISTs)
autodetected there anyway), otherwise python tries to use it and fails
(currently with SIGILL).
explicitly set --with-lto=no when we want it disabled, to make it clear.
clear up old .if block that just had comments in, talking about
--enable-optimizations. that wants revisiting now that we have newer
toolchain, but needs per-arch tests.
found by phessler, diff following discussions with phessler tb jca
daniel