tech-pkg archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: pkgsrc gcc discussion #3874



On 12/30/17 19:24, Greg Troxel wrote:
Jason Bacon <bacon4000%gmail.com@localhost> writes:

On 12/30/17 12:30, Greg Troxel wrote:
coypu%sdf.org@localhost writes:

Options:
- globally use a newer compiler
- use the newest compiler that exists (I recall having a patch for it
    that wasn't well-received because it was too much magic)
Can you comment against the analysis and plan in
    https://wiki.netbsd.org/pkgsrc/gcc/

this has been pretty confusing and I've tried to distll all the previous
discussion.  We have more or less settled on picking a version for
C, and for C++, and always using those versions.

I think the big question is whether the C version and C++ version have
to match.  Or if it's ok to use e.g. base 4.8 for C and pkgsrc 5 for C++
on NetBSD 7.

But, if it's mostly ok, and we later decide it's not really ok (maybe
when we deal with fortran), that probably won't hurt much, compared to
not starting.
Is there any advantage to using base 4.8 for C besides avoiding the
gcc dependency until a C++ package is built?
I believe that is the only advantage.  I don't know how much it matters.

C seems to support version mixing.  The big trouble is programs in C
that link with libraries that provide a C interface but are implemented
in C++ (e.g. libspatialite in C which links again geos).  So the
question is how icky patching around that (to use the higher version)
is, compared to the saved pain of others building a compiler they didn't
really need.

But how many people use pkgsrc and never build a C++ package?

I can imagine that the "others" you're referring to are rapidly going extinct, except in a few special use cases

On any platform with binary packages, the cost of the "extra" compiler suite is especially trivial.

--
Earth is a beta site.



Home | Main Index | Thread Index | Old Index