log1pl: Work around musl libc bug

2021-01-31 Thread Bruno Haible
log1pl lacks precision on musl libc 1.2.2/arm64, musl libc 1.2.2/s390x. This patch adds a configure test, that enables a gnulib workaround. 2021-01-31 Bruno Haible log1pl: Work around musl libc bug. * doc/posix-functions/log1pl.texi: Document musl libc bug. *

fmaf: Document QEMU bug

2021-01-31 Thread Bruno Haible
On my Alpine Linux 3.13/powerpcle system, the fmaf configure test is reported as "checking whether fmaf works... no". As far as I can see, it's a QEMU bug. Let me document it here. 2021-01-31 Bruno Haible fmaf: Document QEMU bug. * doc/posix-functions/fmaf.texi: Add note

logl: Document musl libc bug

2021-01-31 Thread Bruno Haible
logl lacks precision on musl libc 1.2.2/arm64, musl libc 1.2.2/s390x. Fortunately the configure test meant for NetBSD already catches it. 2021-01-31 Bruno Haible logl: Document musl libc bug. * doc/posix-functions/logl.texi: Document musl libc bug. * m4/logl.m4

log10l: Document musl libc bug

2021-01-31 Thread Bruno Haible
log10l lacks precision on musl libc 1.2.2/arm64, musl libc 1.2.2/s390x. Fortunately the configure test meant for NetBSD already catches it. 2021-01-31 Bruno Haible log10l: Document musl libc bug. * doc/posix-functions/log10l.texi: Document musl libc bug. *

remainderl: Work around musl libc bug

2021-01-31 Thread Bruno Haible
remainderl lacks precision on musl libc 1.2.2/arm64, musl libc 1.2.2/s390x. This leads to a test failure: ../../gltests/test-remainder.h:54: assertion '(z > - L_(2.0) * L_(16.0) / TWO_MANT_DIG && z < L_(2.0) * L_(16.0) / TWO_MANT_DIG) || (z > y - L_(2.0) * L_(16.0) / TWO_MANT_DIG && z < y +

expl: Document musl libc bug

2021-01-31 Thread Bruno Haible
expl lacks precision musl libc 1.2.2/arm64, musl libc 1.2.2/s390x. Fortunately the configure test meant for NetBSD already catches it. 2021-01-31 Bruno Haible expl: Document musl libc bug. * doc/posix-functions/expl.texi: Document musl libc bug. * m4/expl.m4

expm1l: Document musl libc bug

2021-01-31 Thread Bruno Haible
expm1l lacks precision on musl libc 1.2.2/arm64, musl libc 1.2.2/s390x. Fortunately the configure test meant for Mac OS X and NetBSD already catches it. 2021-01-31 Bruno Haible expm1l: Document musl libc bug. * doc/posix-functions/expm1l.texi: Document musl libc bug.

log2l: Work around musl libc bug

2021-01-31 Thread Bruno Haible
log2l lacks precision on musl libc 1.2.2/arm64, musl libc 1.2.2/s390x. This patch adds a configure test, that enables a gnulib workaround. 2021-01-31 Bruno Haible log2l: Work around musl libc bugs. * doc/posix-functions/log2l.texi: Document musl libc bugs. *

Re: [PATCH 4/7] stdlib: putenv() needs a cast on OS/2 kLIBC

2021-01-31 Thread KO Myung-Hun
Hi/2. Bruno Haible wrote: > Hi KO, > >> On OS/2 kLIBC, the first parameter of putenv () is `const char *string' >> not `char *string'. > > Can't you fix that in OS/2 kLIBC? Then a workaround in Gnulib would not be > needed. > This was rejected by the maintainer of kLIBC. See details here:

canonicalize-lgpl: Simplify

2021-01-31 Thread Bruno Haible
Since canonicalize-lgpl.c no longer invokes malloc() nor realloc() directly (it does its memory allocation through the 'scratch_buffer' module now), defining _GL_USE_STDLIB_ALLOC is a redundant no-op now. 2021-01-31 Bruno Haible canonicalize-lgpl: Simplify. *

relocatable-prog-wrapper: Update after recent changes

2021-01-31 Thread Bruno Haible
The module 'relocatable-prog-wrapper' includes a flattened view of the 'canonicalize-lgpl' module. Since the 'canonicalize-lgpl' module has changed a month ago, the 'relocatable-prog-wrapper' may produce compilation errors on platforms where it is needed (AIX, Cygwin, as well as older versions of