Avoid hardcoding of the version in lang/gcc in Uses/fortran.mk (PING)
Tijl Coosemans
tijl at FreeBSD.org
Mon Jan 26 10:19:24 UTC 2015
On Mon, 26 Jan 2015 01:04:40 +0100 (CET) Gerald Pfeifer wrote:
> On Sunday 2015-01-25 22:14, Antoine Brodin wrote:
>> I didn't test but this looks ok.
>> I think the LANG_GCC_IS assignement can be moved before the .endif
>
> Yes, this is a good point! I made this change before committing
> the patch.
I'm not going to ask for this patch to be backed out, but I do want it
to be noted (again) that I disagree with the way lang/gcc* is organised.
The only reason for this change is because the gcc ports are fighting
the default-versions mechanism instead of conforming to it and no good
explanation has been given for why they should be allowed to do so.
This patch is a hack. None of the other ports that use default-versions
need it.
In my opinion one of the lang/gccXY ports has to become the default
instead of lang/gcc. That's how default-versions works. lang/gcc can
stay as a meta-port with a RUN_DEPENDS on the default lang/gccXY,
similar to lang/python. This way "pkg add gcc" still works. And just
like lang/python, lang/gcc could install symlinks like gcc -> gccXY,
g++ -> g++XY,... which would be really convenient now that /usr/bin/gcc
is gone.
This patch would not be necessary then because the .if in fortran.mk
would disappear (what's currently in the .else branch would be all
that's needed). A similar .if in bsd.gcc.mk would disappear as well,
along with the LANG_GCC_IS variable.
More information about the freebsd-fortran
mailing list