Mercurial > gnulib
view m4/rint.m4 @ 40140:81f075eaa990
ptsname_r: Work around bug on Android 4.3.
* m4/ptsname_r.m4 (gl_FUNC_PTSNAME_R): Define
HAVE_ESSENTIALLY_WORKING_PTSNAME_R. Test whether the return value is
correct.
* lib/ptsname_r.c (__ptsname_r): If HAVE_ESSENTIALLY_WORKING_PTSNAME_R
is defined, just fix the return value.
* doc/glibc-functions/ptsname_r.texi: Mention the Android bug. Reword:
The behaviour of musl libc is nothing to be "fixed", since it is
compliant with the next POSIX standard.
author | Bruno Haible <bruno@clisp.org> |
---|---|
date | Sat, 26 Jan 2019 15:23:19 +0100 |
parents | b06060465f09 |
children |
line wrap: on
line source
# rint.m4 serial 4 dnl Copyright (C) 2011-2019 Free Software Foundation, Inc. dnl This file is free software; the Free Software Foundation dnl gives unlimited permission to copy and/or distribute it, dnl with or without modifications, as long as this notice is preserved. AC_DEFUN([gl_FUNC_RINT], [ AC_REQUIRE([gl_MATH_H_DEFAULTS]) dnl Determine RINT_LIBM. dnl The reason for the extra code is: dnl In AIX 7.1 with gcc 4.2, when optimization is turned on, calls to dnl rint() with simple arguments are turned into rintf() calls by the dnl compiler. But while rint() is resides in libc, rintf() is in libm. gl_MATHFUNC([rint], [double], [(double)], [], [y += rint (2.0);]) if test $gl_cv_func_rint_no_libm = no \ && test $gl_cv_func_rint_in_libm = no; then HAVE_RINT=0 RINT_LIBM= fi AC_SUBST([RINT_LIBM]) ])