Created attachment 686529 [details] build.log Happens twice on the runs marked with [*], the other failures are probably blocked by use flags and irrelevant for this. USE='-doc -fortran mpi -openmp -quad -threads' failed for =sci-libs/fftw-3.3.9 * USE='doc fortran -mpi openmp quad -threads' failed for =sci-libs/fftw-3.3.9 USE='-doc fortran -mpi -openmp -quad threads' succeeded for =sci-libs/fftw-3.3.9 USE='doc fortran -mpi -openmp -quad threads' succeeded for =sci-libs/fftw-3.3.9 USE='doc -fortran mpi -openmp -quad threads' failed for =sci-libs/fftw-3.3.9 USE='doc fortran mpi -openmp -quad threads' failed for =sci-libs/fftw-3.3.9 USE='-doc fortran -mpi openmp -quad threads' succeeded for =sci-libs/fftw-3.3.9 USE='doc -fortran mpi openmp -quad threads' failed for =sci-libs/fftw-3.3.9 USE='-doc fortran mpi openmp -quad threads' failed for =sci-libs/fftw-3.3.9 USE='-doc fortran mpi -openmp quad threads' failed for =sci-libs/fftw-3.3.9 * USE='-doc fortran -mpi openmp quad threads' failed for =sci-libs/fftw-3.3.9 USE='-doc -fortran mpi openmp quad threads' failed for =sci-libs/fftw-3.3.9 Looks like USE="quad -mpi" to me.
Isn't gcc's libquadmath missing everywhere except amd64/x86/ia64? I don't see it in hppa's stage3s anyway. So, shouldn't USE=quad get some masks?
But I'm having a hard time to figure out why "quad mpi" would work, but I don't have the hardware to check.
It doesn't, virtual/mpi is not stable on hppa so all of them failed for that reason.
(In reply to Rolf Eike Beer from comment #3) > It doesn't, virtual/mpi is not stable on hppa so all of them failed for that > reason. Oh I misread, everything makes sense then. quad might as well receive a base mask + unmask only on amd64+ia64 (already masked on x86 for other reasons).
(In reply to Ionen Wolkens from comment #1) > Isn't gcc's libquadmath missing everywhere except amd64/x86/ia64? I don't > see it in hppa's stage3s anyway. > > So, shouldn't USE=quad get some masks? Yeah. Looking at https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libquadmath/configure.ac;h=f9d745e60ca4102e1dba260adb0218e69d59f4c9;hb=HEAD#l251 libquadmath requires _Float128 support from from target: AM_CONDITIONAL(BUILD_LIBQUADMATH, [test "x$libquad_cv_have_float128" = xyes]) https://gcc.gnu.org/onlinedocs/gcc/Floating-Types.html says: As an extension, GNU C and GNU C++ support additional floating types, which are not supported by all targets. __float128 is available on i386, x86_64, IA-64, and hppa HP-UX, as well as on PowerPC GNU/Linux targets that enable the vector scalar (VSX) instruction set. __float128 supports the 128-bit floating type. On i386, x86_64, PowerPC, and IA-64 other than HP-UX, __float128 is an alias for _Float128. On hppa and IA-64 HP-UX, __float128 is an alias for long double. For Gentoo that practically means these tuples: $ find /usr/lib/gcc/ -name libquadmath.so.0 /usr/lib/gcc/powerpc64le-unknown-linux-gnu/10.2.0/libquadmath.so.0 /usr/lib/gcc/ia64-unknown-linux-gnu/10.2.0/libquadmath.so.0 /usr/lib/gcc/x86_64-pc-linux-gnu/10.2.0/libquadmath.so.0 Fun fact: profiles/arch/x86/package.use.mask already masks the flag even on x86: # Christoph Junghans <junghans@gentoo.org> (2013-09-08) # restrict USE=quad for x86 as there are too less registers sci-libs/fftw quad Possible options are: o1: make flag automagic on _Float128 presence for target (similar to libquadmath's own test) o2: drop flag completely until users appear o3: mask flag everywhere except amd64 o4: mask flag everywhere except amd64, ia64 and some of ppc64-le profiles To me o2/o3 sounds simplest (worth doing now). o1 sounds most right (worth doing later).
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=6a6c139e8f92d1fe4e140e7096853f5d32a257be commit 6a6c139e8f92d1fe4e140e7096853f5d32a257be Author: David Seifert <soap@gentoo.org> AuthorDate: 2021-02-13 18:30:44 +0000 Commit: David Seifert <soap@gentoo.org> CommitDate: 2021-02-13 18:30:44 +0000 sci-libs/fftw: Remove IUSE=quad * The feature isn't useful enough right now to warrant complicated masking. Closes: https://bugs.gentoo.org/770346 Package-Manager: Portage-3.0.14, Repoman-3.0.2 Signed-off-by: David Seifert <soap@gentoo.org> sci-libs/fftw/fftw-3.3.9.ebuild | 31 ++++--------------------------- 1 file changed, 4 insertions(+), 27 deletions(-)
