Re: Fixing lang/python311 by splitting off NIS module (Re: using pkg-config modules from host (Re: wip/py-numpy + blas))

[Niclas Rosenvik <nros%netbsd.org@localhost>]

On Wed, 27 Dec 2023 18:49:49 -0500
Greg Troxel <gdt%lexort.com@localhost> wrote:

> Joerg Sonnenberger <joerg%bec.de@localhost> writes:
> 
> > On Wednesday, December 27, 2023 4:02:53 PM CET Greg Troxel wrote:
> >> This seems like a very big change this late in the freeze.
> >
> > On the contrary, it actually simplifies the situation a lot. I'm not
> > convinced that anyone actually uses py-nis and until that point,
> > I would prefer to just disable it and wait for someone to
> > complain...
> 
> Withdrawals are supposed to have a proposal and chance for discussion.
> And, 12h notice to detect, report etc. before branch is just not a
> reasonable way to treat people.
> 
> And, I'm not so concerned about the intended change in behavior.  It's
> far more than code changes that intend to do one thing often do
> something else.

Greg, Thomas and other would this be okay:

.if ${OPSYS} == "Linux"
CONFIGURE_ARGS+= ac_cv_search_yp_match=no
.else
here the large configure parts that looks for nis header files and sets
PLIST.nis .
.endif

The thing is that the pkgconfig check in configure already breaks
thanks to PKG_CONFIG_LIBDIR being set by pkgsrc, so libnsl is guaranteed
to not be found. If it is not found the configure script runs
AC_SEARCH_LIBS for yp_match in libnsl, this works well on linux machines
where libnsl is part of glibc. By setting ac_cv_search_yp_match to "no"
AC_SEARCH_LIBS(yp_match ... has already failed and nis.so won't be
built.
So no PLIST churn on Linux.
I have tested it on a GLIBC system with nis in GLIBC (Slackware 14) and
on one with nsl as a separate package (Fedora).

Advantages of this patch:
No need to patch python sources, so it probably does not cause problems
on other platforms.
Rather small compared to Thomas patch.
Disadvantages:
Linux users won't have the nis module, no matter what.

Don't feel pressured to accept this, just want to provide a possible
quick fix for the python on Linux mess.

Regards, Niclas