* 006-3.7-ftm.patch: code no longer exists and builds fine without
* 010-3.8-nis-cygwin.patch: code no longer exists and builds fine without
* 012-3.8-pep3149-cygwin.patch: refreshed
* 013-3.8-tkinter-cygwin.patch: code no longer exists and builds fine without
* 28881-backport.patch, 930-fix-missing-tcp-include.patch: included in the new release
* 970-ossaudiodev.patch: work around "error: initializer element is not constant"
remove useless permissions change
* remove sed for cgi, cgi is deprecated and about to be dropped
* remove CPPFLAGS/LDFLAGS. it leaks into sysconfig and builds fine without it
* don't set CXX (defaults to g++ otherwise). nure sure why an absolute path is
needed here and not for CC. It builds fine without it.
* remove ac_cv_func_bind_textdomain_codeset override, gets detected correctly
* drop LC_CTYPE, we default to utf-8
* remove EXTRA_CFLAGS, cpython supports CFLAGS since forever
* remove some optdepends commented code
In some cases this will introduce cycles which could likely be cleaned
up by only depending on some unix tools instead of the whole autotools.
Rebuilding doesn't make much sense at this point since it just adds a subset
of base-devel which is still installed by default.
The main change here is that sys.platform == 'cygwin', which allows
us to drop various changes for extending cygwin checks. Fewer patches
and less likely that we miss to patch a new cygwin check on updates.
If one really needs to check for msys Python then this still works:
sysconfig.get_platform().startswith("msys")
We currently have an issue where python was replaced in the pacman repo, leading
to install errors with existing caches containing the old version (CI caches etc).
Bump pkgrel to fix things.
To prevent this in the future see https://github.com/msys2/msys2-autobuild/issues/16
This doesn't contain anything and just depends on python itself and
libcrypt-devel which it uses in Python.h.
This allows packages building against Python to makedepend on this without
having to specify libcrypt-devel etc.
It mostly worked, except for some ln failures in make install due to not
being able to create a broken symlink to a file to be created later.
Fixed by making the make rule that ln's depend on the rule that creates
the files.
Tees stdout from make to a file (in a way that doesn't suppress an error
exit from make from failing the build), and runs a sed script over that
file which looks for a list of failed modules, outputs it to ensure it
isn't lost in the build output, and exits with an error code to cause
the build to fail.
