pkgsrc-wip/gcc7snapshot/TODO

48 lines
2.0 KiB
Plaintext

TODO
====
I won't work on gcc any longer after gcc6 has been released in about
April 2016. So please step up as maintainer of gcc6/gcc7 for NetBSD.
0) This package aims to keep the latest gcc usable on NetBSD, although
any work on gcc should happen upstream as far as possible. But it is
hopeless without enough NetBSD developers working on it upstream. I have sent
some patches for NetBSD to GCC a few month ago but there is no progress yet:
PR 64271, the minimal patch set to bootstrap, extracted from pkgsrc.
Jonathan Wakely is working on the libstdc++ patches involved.
Without acceptance of patches for PR 64271, no other patches
will be accepted upstream.
PR 39570, this bug causes all gfortran run tests to fail. I have developed
the necessary patches and sent them upstream. For gcc50snapshot it means
replacing a short patch by a bunch of complex patches, but they are gcc
conforming and can be reused for other patches as well.
PR 48244, for iso_c_binding support in gfortran. The patches are
there but it is delayed by PR 64271.
1) On NetBSD a working cabsl function must be present in libm to build gfortran.
This has been added to NetBSD7 on 2014/10/13. Before this date it was defined
in complex.h but missing from libm.
2) Unless NetBSD gets proper support for C99, the number of test failures
will stay high. The dozens of missing or inaccurate long double functions
causes a misconception of NetBSD being an obscure platform, causing other
projects like gcc or R to abandon support for NetBSD. And the race goes on,
now we need support for C11 functions like aligned_alloc and the plethora
of thread functions in our libc/libm.
3) This package and the lang/gccXY need a cleanup or a complete new start. I
think that SUBST_VARS could handle the platform dependencies in a much cleaner
way.
4) Patches from lang/gcc5 should be integrated as far as possible, especially
patches for other platforms. Right now I just look at it from time to time.
Have fun,
Kai-Uwe Eckhardt