kumquat-buildroot/package/glibc/0006-Let-fpclassify-use-the-builtin-when-optimizing-for-s.patch
Romain Naour d6a4807f94 package/glibc: bump to 2.26
Drop upstream patch:

0001-sh-Fix-building-with-gcc5-6.patch
https://sourceware.org/git/?p=glibc.git;a=commit;h=d40dbe722f004f999b589de776f7e57e564dda01

0002-CVE-2017-1000366-Ignore-LD_LIBRARY_PATH-for-AT_SECUR.patch
https://sourceware.org/git/?p=glibc.git;a=commit;h=f6110a8fee2ca36f8e2d2abecf3cba9fa7b8ea7d

0003-ld.so-Reject-overly-long-LD_PRELOAD-path-elements.patch
https://sourceware.org/git/?p=glibc.git;a=commit;h=6d0ba622891bed9d8394eef1935add53003b12e8

0004-ld.so-Reject-overly-long-LD_AUDIT-path-elements.patch

https://sourceware.org/git/?p=glibc.git;a=commit;h=81b82fb966ffbd94353f793ad17116c6088dedd9

0005-fix-binutils-2-29-build.patch
https://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commitdiff_plain;h=388b4f1a02f3a801965028bbfcd48d905638b797

0006-i686-Add-missing-IS_IN-libc-guards-to-vectorized-strcspn.patch
https://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commitdiff_plain;h=69052a3a95da37169a08f9e59b2cc1808312753c

0006-sh4-trap.patch
https://sourceware.org/git/gitweb.cgi?p=glibc.git;a=commitdiff_plain;h=db3d848e154b00071f4a5e729d5884efad410109

But 2.26 version introduced quite a few regressions on x86, x86_64 and
powerpc64le when float128 support was added. All these issues are now
fixed in the glibc 2.26 stable branch.

Backport patches from glibc 2.26 stable branch related to bug 21930.
Take all patches listed in this bug report in order to have all fix
for this issue [1] [2].

Fixes gnuradio build:
[...]/sysroot/usr/include/boost/functional/hash/detail/hash_float.hpp: In function ‘std::size_t boost::hash_detail::float_hash_value(T)’:
[...]/sysroot/usr/include/boost/functional/hash/detail/hash_float.hpp:210:21: error: expected primary-expression before ‘float’
             switch (fpclassify(v))
                     ^
[...]/sysroot/usr/include/boost/functional/hash/detail/hash_float.hpp:210:21: error: expected primary-expression before ‘long’
             switch (fpclassify(v))
                     ^
[...]/sysroot/usr/include/boost/functional/hash/detail/hash_float.hpp:210:21: error: found ‘:’ in nested-name-specifier, expected ‘::’
             switch (fpclassify(v))
                     ^
[...]/sysroot/usr/include/boost/functional/hash/detail/hash_float.hpp:210:21: error: ‘_Float128’ is not a class or namespace
             switch (fpclassify(v))
                     ^

[...]/sysroot/usr/include/boost/functional/hash/detail/hash_float.hpp:210:21: error: ‘_Generic’ was not declared in this scope
             switch (fpclassify(v))
                     ^

Also backport fix for glibc bug 22146.
Without this patch some C++ applications (jsoncpp, mesa3d and kodi) fail
to build due to an issue while building the toolchain.

A test in the libstdc++ configure script fail when -Os and float128 are
used.

See the bug report for details [3].

[1] https://sourceware.org/ml/libc-alpha/2017-08/msg00586.html
[2] https://sourceware.org/bugzilla/show_bug.cgi?id=21930
[3] https://sourceware.org/bugzilla/show_bug.cgi?id=22146

Signed-off-by: Romain Naour <romain.naour@gmail.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Tested-by: Bernd Kuhls <bernd.kuhls@t-online.de>
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
2017-09-27 22:58:52 +02:00

64 lines
2.5 KiB
Diff

From 33f54cf4f81a51e5c8cbfb2408edd439bdee9435 Mon Sep 17 00:00:00 2001
From: "Gabriel F. T. Gomes" <gabriel@inconstante.eti.br>
Date: Wed, 20 Sep 2017 15:10:26 -0300
Subject: [PATCH] Let fpclassify use the builtin when optimizing for size in
C++ mode (bug 22146)
When optimization for size is on (-Os), fpclassify does not use the
type-generic __builtin_fpclassify builtin, instead it uses __MATH_TG.
However, when library support for float128 is available, __MATH_TG uses
__builtin_types_compatible_p, which is not available in C++ mode.
On the other hand, libstdc++ undefines (in cmath) many macros from
math.h, including fpclassify, so that it can provide its own functions.
However, during its configure tests, libstdc++ just tests for the
availability of the macros (it does not undefine them, nor does it
provide its own functions).
Finally, when libstdc++ is configured with optimization for size
enabled, its configure tests include math.h and get the definition of
fpclassify that uses __MATH_TG (and __builtin_types_compatible_p).
Since libstdc++ does not undefine the macros during its configure tests,
they fail.
This patch lets fpclassify use the builtin in C++ mode, even when
optimization for size is on. This allows the configure test in
libstdc++ to work.
Tested for powerpc64le and x86_64.
[BZ #22146]
math/math.h: Let fpclassify use the builtin in C++ mode, even
when optimazing for size.
(cherry picked from commit c5c4a626098ec884b8527356abdf2a4bb7b6bf27)
[Romain rebase on glibc 2.26)
Signed-off-by: Romain Naour <romain.naour@gmail.com>
---
math/math.h | 8 +++++++-
1 file changed, 7 insertions(+), 1 deletion(-)
diff --git a/math/math.h b/math/math.h
index 7c0fc6d..f9348ec 100644
--- a/math/math.h
+++ b/math/math.h
@@ -402,7 +402,13 @@ enum
/* Return number of classification appropriate for X. */
# if __GNUC_PREREQ (4,4) && !defined __SUPPORT_SNAN__ \
- && !defined __OPTIMIZE_SIZE__
+ && (!defined __OPTIMIZE_SIZE__ || defined __cplusplus)
+ /* The check for __cplusplus allows the use of the builtin, even
+ when optimization for size is on. This is provided for
+ libstdc++, only to let its configure test work when it is built
+ with -Os. No further use of this definition of fpclassify is
+ expected in C++ mode, since libstdc++ provides its own version
+ of fpclassify in cmath (which undefines fpclassify). */
# define fpclassify(x) __builtin_fpclassify (FP_NAN, FP_INFINITE, \
FP_NORMAL, FP_SUBNORMAL, FP_ZERO, x)
# else
--
2.9.5