On 11/26/17 10:50 AM, Andreas K. Huettel wrote:
> Am Samstag, 25. November 2017, 15:01:20 CET schrieb Anthony G. Basile:
>> Hi everyone,
>>
>> With the stabilization of gcc-6.4.0, the uclibc build broke because the
>> eclass requires UCLIBC_VER to be define on uclibc systems else it will
>> die(). Since uclibc specific patches are no longer needed in gcc-6 and
>> above, we don't want to error out in the eclass when the patchset is not
>> found.
>>
>
> I'd guard this so it only applies to gcc-6 and later... for the simple reason
> that
> otherwise people will try to emerge some historical gcc versions and fail..
I don't think that's necessary because the ebuild is supposed to provide
a value of UCLIBC_VER if and only if a patchset is needed, and writing
the ebuild is up to us toolchain folks. I can see the possibility of
upstream porting the fix to versions of gcc previous to 6 and then we'd
have to go back and hack away at the toolchain.eclass.
I'm planning to use the same logic for musl specific patches: if
MUSL_VER is provided in the gcc ebuild, then there is a musl patchset to
be applied, otherwise there isn't.
This seems to be the cleanest approach without littering the eclass with
tc_version_is_at_least.
Comment?
>
> Otherwise WFM
>
> diff --git a/eclass/toolchain.eclass b/eclass/toolchain.eclass
> index 503f7dbe94f..58d859dfaf3 100644
> --- a/eclass/toolchain.eclass
> +++ b/eclass/toolchain.eclass
> @@ -378,9 +378,6 @@ toolchain_pkg_pretend() {
> "in your make.conf if you want to use this
> version."
> fi
>
> - [[ -z ${UCLIBC_VER} ]] && [[ ${CTARGET} == *-uclibc* ]] && \
> - die "Sorry, this version does not support uClibc"
> -
> if ! use_if_iuse cxx ; then
> use_if_iuse go && ewarn 'Go requires a C++ compiler,
> disabled due to USE="-cxx"'
> use_if_iuse objc++ && ewarn 'Obj-C++ requires a C++
> compiler, disabled due to USE="-cxx"'
>
--
Anthony G. Basile, Ph.D.
Gentoo Linux Developer [Hardened]
E-Mail : bluen...@gentoo.org
GnuPG FP : 1FED FAD9 D82C 52A5 3BAB DC79 9384 FA6E F52D 4BBA
GnuPG ID : F52D4BBA
