platform/upstream/glibc.git
11 years agoFix MIPS n32 cancellation in static libc (bug 15506).
Joseph Myers [Tue, 21 May 2013 20:27:45 +0000 (20:27 +0000)]
Fix MIPS n32 cancellation in static libc (bug 15506).

11 years agoTest getaddrinfo return value
Dmitry V. Levin [Thu, 16 May 2013 21:55:09 +0000 (21:55 +0000)]
Test getaddrinfo return value

This helps testing for regression of BZ#15339.  Creation of network
isolated environments is a privileged operation and therefore is not
included to the test.

11 years agoSet EAI_SYSTEM only when h_errno is NETDB_INTERNAL
Siddhesh Poyarekar [Tue, 21 May 2013 16:24:41 +0000 (21:54 +0530)]
Set EAI_SYSTEM only when h_errno is NETDB_INTERNAL

Fixes BZ #15339.

NSS_STATUS_UNAVAIL may mean that a necessary input resource is not
available.  This could occur in a number of cases including when the
network is down, system runs out of file descriptors, etc.  The
correct differentiator in such a case is the h_errno, which gives the
nature of failure.  In case of failures other than a simple 'not
found', we set h_errno as NETDB_INTERNAL and let errno be the
identifier for the exact error.

11 years agoFix parsing of numeric hosts in gethostbyname_r
Andreas Schwab [Mon, 21 Jan 2013 16:41:28 +0000 (17:41 +0100)]
Fix parsing of numeric hosts in gethostbyname_r

11 years agoAdd a README for benchtests
Siddhesh Poyarekar [Tue, 21 May 2013 09:29:50 +0000 (14:59 +0530)]
Add a README for benchtests

Move instructions from the Makefile here and expand on them.

11 years agoAArch64: Don't clobber argument for tail call to __sigjmp_save in sigsetjmp
Andreas Schwab [Mon, 20 May 2013 08:19:31 +0000 (10:19 +0200)]
AArch64: Don't clobber argument for tail call to __sigjmp_save in sigsetjmp

11 years ago[AArch64] Update libm-test-ulps.
Marcus Shawcroft [Tue, 21 May 2013 08:01:19 +0000 (09:01 +0100)]
[AArch64] Update libm-test-ulps.

11 years agoPowerPC: update missing CL/NEWS bug reference
Adhemerval Zanella [Mon, 20 May 2013 14:35:01 +0000 (09:35 -0500)]
PowerPC: update missing CL/NEWS bug reference

Update ChangLog bugzilla number and NEWS for commit
13d3b41a36c4f28d171a144f8a9baad3a8835981 (PowerPC: fix hypot/hypotf
check for -INF).

11 years agoFaster memset on x64
Ondrej Bilka [Mon, 20 May 2013 06:26:00 +0000 (08:26 +0200)]
Faster memset on x64

This implementation speed up memset in several ways. First is avoiding
expensive computed jump. Second is using fact that arguments of memset
are most of time aligned to 8 bytes.

Benchmark results on:
kam.mff.cuni.cz/~ondra/benchmark_string/memset_profile_result27_04_13.tar.bz2

11 years agoFaster memcpy on x64.
Ondrej Bilka [Mon, 20 May 2013 06:20:00 +0000 (08:20 +0200)]
Faster memcpy on x64.

We add new memcpy version that uses unaligned loads which are fast
on modern processors. This allows second improvement which is avoiding
computed jump which is relatively expensive operation.

Tests available here:
http://kam.mff.cuni.cz/~ondra/memcpy_profile_result27_04_13.tar.bz2

11 years agoFix nearbyint scheduling of arithmetic past fesetenv (bug 15490).
Joseph Myers [Sun, 19 May 2013 18:40:25 +0000 (18:40 +0000)]
Fix nearbyint scheduling of arithmetic past fesetenv (bug 15490).

11 years agoHandle sincos with generic libm-test logic.
Joseph Myers [Sun, 19 May 2013 14:45:41 +0000 (14:45 +0000)]
Handle sincos with generic libm-test logic.

11 years agoSimplify gen-libm-test.pl handling of tests with extra outputs.
Joseph Myers [Sun, 19 May 2013 14:40:40 +0000 (14:40 +0000)]
Simplify gen-libm-test.pl handling of tests with extra outputs.

11 years agoDon't disable CMPLXL macro for __NO_LONG_DOUBLE_MATH (bug 15488).
Joseph Myers [Sat, 18 May 2013 12:12:38 +0000 (12:12 +0000)]
Don't disable CMPLXL macro for __NO_LONG_DOUBLE_MATH (bug 15488).

11 years agoMake libm-test look up ulps by name at runtime.
Joseph Myers [Sat, 18 May 2013 12:10:59 +0000 (12:10 +0000)]
Make libm-test look up ulps by name at runtime.

11 years agoFix remainder exceptions and directed-rounding results (bugs 15480, 15485).
Joseph Myers [Fri, 17 May 2013 19:04:08 +0000 (19:04 +0000)]
Fix remainder exceptions and directed-rounding results (bugs 15480, 15485).

11 years agoSimplify libm-test extra-output initialization.
Joseph Myers [Fri, 17 May 2013 19:02:19 +0000 (19:02 +0000)]
Simplify libm-test extra-output initialization.

11 years agoDon't handle ulps for integer tests in libm-test.inc.
Joseph Myers [Fri, 17 May 2013 19:01:27 +0000 (19:01 +0000)]
Don't handle ulps for integer tests in libm-test.inc.

11 years agoTest more cases of "inexact" exceptions in libm-test.inc.
Joseph Myers [Fri, 17 May 2013 14:45:50 +0000 (14:45 +0000)]
Test more cases of "inexact" exceptions in libm-test.inc.

11 years agoPrevent optimizing out of benchmark function call
Siddhesh Poyarekar [Fri, 17 May 2013 13:40:33 +0000 (19:10 +0530)]
Prevent optimizing out of benchmark function call

Resolves: #15424

The compiler would optimize the benchmark function call out of the
loop and call it only once, resulting in blazingly fast times for some
benchmarks (notably atan, sin and cos).  Mark the inputs as volatile
so that the code is forced to read again from the input for each
iteration.

11 years agoPowerPC: fix hypot/hypotf check for -INF
Adhemerval Zanella [Fri, 17 May 2013 13:12:16 +0000 (08:12 -0500)]
PowerPC: fix hypot/hypotf check for -INF

11 years agoMIPS: soft-fp NaN representation corrections
Maciej W. Rozycki [Thu, 16 May 2013 22:33:55 +0000 (23:33 +0100)]
MIPS: soft-fp NaN representation corrections

[BZ #15442] This adds support for the inverse interpretation of the
quiet bit of IEEE 754 floating-point NaN data that some processors
use.  This includes in particular MIPS architecture processors; the
payload used for the canonical qNaN encoding is updated accordingly
so as not to interfere with the quiet bit.

11 years agoTest for errno setting in more pole error cases.
Joseph Myers [Thu, 16 May 2013 21:57:25 +0000 (21:57 +0000)]
Test for errno setting in more pole error cases.

11 years agoRemove ENOSYS tests in libm-test.inc.
Joseph Myers [Thu, 16 May 2013 19:09:54 +0000 (19:09 +0000)]
Remove ENOSYS tests in libm-test.inc.

11 years agoRemove libm-test support for TEST_* inside functions.
Joseph Myers [Thu, 16 May 2013 19:09:03 +0000 (19:09 +0000)]
Remove libm-test support for TEST_* inside functions.

11 years agoAdd #include <stdint.h> for uint[32|64]_t usage (except installed headers).
Ryan S. Arnold [Wed, 1 May 2013 15:46:34 +0000 (10:46 -0500)]
Add #include <stdint.h> for uint[32|64]_t usage (except installed headers).

11 years agoConvert TEST_extra tests from code to data.
Joseph Myers [Thu, 16 May 2013 15:53:40 +0000 (15:53 +0000)]
Convert TEST_extra tests from code to data.

11 years agoDe-stringify constants in math/atest-exp2.c.
Richard Henderson [Thu, 16 May 2013 13:34:01 +0000 (06:34 -0700)]
De-stringify constants in math/atest-exp2.c.

11 years agoSupport testing "inexact" exceptions in libm-test.inc.
Joseph Myers [Thu, 16 May 2013 13:19:38 +0000 (13:19 +0000)]
Support testing "inexact" exceptions in libm-test.inc.

11 years agoTest for errno setting in more overflow error cases.
Joseph Myers [Thu, 16 May 2013 13:18:48 +0000 (13:18 +0000)]
Test for errno setting in more overflow error cases.

11 years agoConvert TEST_cc_c tests from code to data.
Joseph Myers [Thu, 16 May 2013 13:17:55 +0000 (13:17 +0000)]
Convert TEST_cc_c tests from code to data.

11 years agoConvert TEST_f_L tests from code to data.
Joseph Myers [Thu, 16 May 2013 13:16:56 +0000 (13:16 +0000)]
Convert TEST_f_L tests from code to data.

11 years agoRemove const attribute on get_log2
Peter Collingbourne [Mon, 13 May 2013 15:36:12 +0000 (08:36 -0700)]
Remove const attribute on get_log2

This function is not const, as it can modify log2_m and log2_m_inited.

11 years agohppa: Cleanup libm-test-ulps.
Carlos O'Donell [Wed, 15 May 2013 21:19:20 +0000 (17:19 -0400)]
hppa: Cleanup libm-test-ulps.

Joseph Myers noted that there were several old and really very
incorrect values in the hppa libm-test-ulps. This patch removes
all of the ulps values for ceil, floor, rint, round, trun,
llrint, and llround, all of which were previously incorreclty
added (including some negative values which are really wrong).

---
ports/

2013-05-15  Carlos O'Donell  <carlos@redhat.com>

* sysdeps/hppa/fpu/libm-test-ulps: Remove old values for ceil, floor,
rint, round, trunc, llrint, and llround.

11 years agoConvert TEST_f_l tests from code to data.
Joseph Myers [Wed, 15 May 2013 19:48:17 +0000 (19:48 +0000)]
Convert TEST_f_l tests from code to data.

11 years agoUse movq for 64-bit operations
Peter Collingbourne [Wed, 15 May 2013 18:33:45 +0000 (20:33 +0200)]
Use movq for 64-bit operations

The EXTRACT_WORDS64 and INSERT_WORDS64 macros use movd for a 64-bit
operation.  Somehow gcc manages to turn this into movq, but LLVM won't.

2013-05-15  Peter Collingbourne  <pcc@google.com>

* sysdeps/x86_64/fpu/math_private.h (MOVQ): New macro.
(EXTRACT_WORDS64) Use where appropriate.
(INSERT_WORDS64) Likewise.

11 years agoUse x constraints for operands to vfmaddss and vfmaddsd
Peter Collingbourne [Wed, 15 May 2013 18:31:53 +0000 (20:31 +0200)]
Use x constraints for operands to vfmaddss and vfmaddsd

While these instructions accept memory operands, only one operand
may be a memory operand.  Giving two operands xm constraints gives
the compiler the option of using memory for both operands, which
would result in invalid assembly code.  Using x for all operands is
more appropriate, as most x86_64 calling conventions will pass the
arguments in registers anyway.

2013-05-15  Peter Collingbourne  <pcc@google.com>

* sysdeps/x86_64/fpu/multiarch/s_fma.c (__fma_fma4): Replace xm
constraints with x constraints.
* sysdeps/x86_64/fpu/multiarch/s_fmaf.c (__fmaf_fma4): Likewise.

11 years agoMove _obstack_compat out of common
Peter Collingbourne [Wed, 15 May 2013 18:28:08 +0000 (20:28 +0200)]
Move _obstack_compat out of common

it is impossible to create an alias of a common symbol (as
compat_symbol does), because common symbols do not have a section or
an offset until linked.  GNU as tolerates aliases of common symbols by
simply creating another common symbol, but other assemblers (notably
LLVM's integrated assembler) are less tolerant.

2013-05-15  Peter Collingbourne  <pcc@google.com>

* malloc/obstack.c (_obstack_compat): Add initializer.
-

11 years agoReserve new TLS field for x86 and x86_64
Andreas Jaeger [Wed, 15 May 2013 18:20:54 +0000 (20:20 +0200)]
Reserve new TLS field for x86 and x86_64

[BZ #10686]
* sysdeps/x86_64/tls.h (struct tcbhead_t): Add __private_ss
field.
* sysdeps/i386/tls.h (struct tcbhead_t): Likewise.

11 years agoUpdate s390/bits/siginfo.h
Edjunior Machado [Wed, 15 May 2013 18:15:19 +0000 (20:15 +0200)]
Update s390/bits/siginfo.h

2013-05-15  Edjunior Machado  <emachado@linux.vnet.ibm.com>

* sysdeps/unix/sysv/linux/s390/bits/siginfo.h (siginfo_t):
Remove si_trapno and add si_addr_lsb to _sifields.sigfault.
(si_trapno): Remove macro.
(si_addr_lsb): Define new macro.
(BUS_MCEERR_AR, BUS_MCEERR_AO): Define new values.

11 years agoConsistently use TEST_f_L in tests of llrint and llround.
Joseph Myers [Wed, 15 May 2013 16:58:41 +0000 (16:58 +0000)]
Consistently use TEST_f_L in tests of llrint and llround.

11 years agohppa: Update libm-test-ulps
Carlos O'Donell [Wed, 15 May 2013 16:42:59 +0000 (12:42 -0400)]
hppa: Update libm-test-ulps

Update libm-test-ulps for hppa. There are a few entries
with 4 or 5 ulps, but these appear to be expected. A more
thorough review will be required if hppa switches long-double
to a different type.

---
ports/

2013-05-15  Carlos O'Donell  <carlos@redhat.com>

* sysdeps/hppa/fpu/libm-test-ulps: Regenerate.

11 years agohppa: Fix _FPU_GETCW and _FPU_SETCW.
Carlos O'Donell [Wed, 15 May 2013 15:47:47 +0000 (11:47 -0400)]
hppa: Fix _FPU_GETCW and _FPU_SETCW.

The following patch fixes both _FPU_GETCW and
_FPU_SETCW for hppa. The initial implementation was
flawed and not well tested. We failed to set cw,
and passed in the value of a register to fldd.
This patch fixes both of those errors and allows
the libm tests to pass without failure.

Signed-off-by: Guy Martin <gmsoft@tuxicoman.be>
Signed-off-by: Carlos O'Donell <carlos@redhat.com>
---

2013-05-15  Guy Martin  <gmsoft@tuxicoman.be>
    Carlos O'Donell  <carlos@redhat.com>

[BZ# 15000]
* ports/sysdeps/hppa/fpu/fpu_control.h (_FPU_GETCW): Set cw.
(_FPU_SETCW): Pass address to fldd.

11 years agoConvert TEST_f_b tests from code to data.
Joseph Myers [Wed, 15 May 2013 12:04:07 +0000 (12:04 +0000)]
Convert TEST_f_b tests from code to data.

11 years agoTest for errno setting in more domain error cases.
Joseph Myers [Wed, 15 May 2013 10:43:19 +0000 (10:43 +0000)]
Test for errno setting in more domain error cases.

11 years agoDisable libm-test errno testing for TEST_INLINE.
Joseph Myers [Wed, 15 May 2013 10:42:18 +0000 (10:42 +0000)]
Disable libm-test errno testing for TEST_INLINE.

11 years agoConvert TEST_ff_i tests from code to data.
Joseph Myers [Wed, 15 May 2013 10:41:39 +0000 (10:41 +0000)]
Convert TEST_ff_i tests from code to data.

11 years agoUpdate Sparc ULPs.
David S. Miller [Tue, 14 May 2013 22:44:29 +0000 (15:44 -0700)]
Update Sparc ULPs.

* sysdeps/sparc/fpu/libm-test-ulps: Update.

11 years agoFix whitespace in RUN_TEST_LOOP_2_f.
Joseph Myers [Tue, 14 May 2013 21:17:19 +0000 (21:17 +0000)]
Fix whitespace in RUN_TEST_LOOP_2_f.

11 years agoCorrect types of fields in libm-test.inc structures.
Joseph Myers [Tue, 14 May 2013 21:00:56 +0000 (21:00 +0000)]
Correct types of fields in libm-test.inc structures.

11 years agoConvert TEST_ffI_f1 tests from code to data.
Joseph Myers [Tue, 14 May 2013 20:36:03 +0000 (20:36 +0000)]
Convert TEST_ffI_f1 tests from code to data.

11 years agoStop ARM setjmp/longjmp saving/restoring fpscr (bug 14908).
Joseph Myers [Tue, 14 May 2013 19:50:14 +0000 (19:50 +0000)]
Stop ARM setjmp/longjmp saving/restoring fpscr (bug 14908).

11 years agoAdd test for setjmp / longjmp and floating-point state.
Joseph Myers [Tue, 14 May 2013 19:49:09 +0000 (19:49 +0000)]
Add test for setjmp / longjmp and floating-point state.

11 years agoConvert TEST_f_i tests from code to data.
Joseph Myers [Tue, 14 May 2013 19:13:43 +0000 (19:13 +0000)]
Convert TEST_f_i tests from code to data.

11 years agoFix function name in scalbln_test call to END.
Joseph Myers [Tue, 14 May 2013 19:12:54 +0000 (19:12 +0000)]
Fix function name in scalbln_test call to END.

11 years agoConvert TEST_fI_f1 tests from code to data.
Joseph Myers [Tue, 14 May 2013 19:12:10 +0000 (19:12 +0000)]
Convert TEST_fI_f1 tests from code to data.

11 years agoConvert TEST_fF_f1 tests from code to data.
Joseph Myers [Tue, 14 May 2013 14:49:26 +0000 (14:49 +0000)]
Convert TEST_fF_f1 tests from code to data.

11 years agoConvert TEST_f_f1 tests from code to data.
Joseph Myers [Tue, 14 May 2013 12:01:19 +0000 (12:01 +0000)]
Convert TEST_f_f1 tests from code to data.

11 years agoAdd comments to vDSO hwcap loading process.
Carlos O'Donell [Tue, 14 May 2013 04:06:35 +0000 (00:06 -0400)]
Add comments to vDSO hwcap loading process.

Loading of the vDSO pseudo-hwcap from the type 2 GNU note is
a rather arcane and poorly documented process. Given that I had
a chance to review this code today I thought I would add all
of the things I had to lookup to verify the validity of the
process.

With a single .note.GNU the vDSO can register up to 64 flags,
though in practice you are limited to 64 - _DL_FIRST_EXTRA
bits which on x86 is 12 bits.

The only use of this that I know of is in the Xen support
in Linux where they use the 1st bit to indicate "nosegneg".
I see "We use bit 1 to avoid bugs in some versions of glibc
when bit 0 is used; the choice is otherwise arbitrary.", but
no reference to a glibc bug anywhere. The code as-is should
support bit zero, so we still have that free for future use.

The kernel, glibc, and ld.so.cache must coordinate to ensure
that bit values don't go too high and are used consistently.

---

2013-05-13  Carlos O'Donell  <carlos@redhat.com>

* elf/dl-hwcaps.c (_dl_important_hwcaps): Comment vDSO hwcap loading.
* elf/ldconfig.c (is_hwcap_platform): Comment each hwcap check.
(main): Comment "tls" pseudo-hwcap.

11 years agoARM: Make multiarch memcpy always use NEON when compiler does
Roland McGrath [Mon, 13 May 2013 22:41:46 +0000 (15:41 -0700)]
ARM: Make multiarch memcpy always use NEON when compiler does

11 years agoConvert TEST_fl_f tests from code to data.
Joseph Myers [Mon, 13 May 2013 22:38:15 +0000 (22:38 +0000)]
Convert TEST_fl_f tests from code to data.

11 years agoConvert TEST_fi_f tests from code to data.
Joseph Myers [Mon, 13 May 2013 19:45:36 +0000 (19:45 +0000)]
Convert TEST_fi_f tests from code to data.

11 years agoConvert TEST_c_f tests from code to data.
Joseph Myers [Mon, 13 May 2013 18:58:17 +0000 (18:58 +0000)]
Convert TEST_c_f tests from code to data.

11 years agoConvert TEST_if_f tests from code to data.
Joseph Myers [Mon, 13 May 2013 18:57:35 +0000 (18:57 +0000)]
Convert TEST_if_f tests from code to data.

11 years agoConsistently use TEST_fl_f in tests of scalbln.
Joseph Myers [Mon, 13 May 2013 18:56:50 +0000 (18:56 +0000)]
Consistently use TEST_fl_f in tests of scalbln.

11 years agoDefine decimal constant M_1_DIV_El in libm-test.inc
Adhemerval Zanella [Mon, 13 May 2013 17:08:42 +0000 (12:08 -0500)]
Define decimal constant M_1_DIV_El in libm-test.inc

This patch replaces the 1.0 / M_El by the decimal constant M_1_DIV_El.

11 years agoUse HP_TIMING for benchmarks if available
Siddhesh Poyarekar [Mon, 13 May 2013 08:14:32 +0000 (13:44 +0530)]
Use HP_TIMING for benchmarks if available

HP_TIMING uses native timestamping instructions if available, thus
greatly reducing the overhead of recording start and end times for
function calls.  For architectures that don't have HP_TIMING
available, we fall back to the clock_gettime bits.  One may also
override this by invoking the benchmark as follows:

  make USE_CLOCK_GETTIME=1 bench

and get the benchmark results using clock_gettime.  One has to do
`make bench-clean` to ensure that the benchmark programs are rebuilt.

11 years agoMake _LIB_VERSION a weak symbol
Siddhesh Poyarekar [Mon, 13 May 2013 06:16:36 +0000 (11:46 +0530)]
Make _LIB_VERSION a weak symbol

That way it can live alongside _LIB_VERSION in libieee.a for
statically compiled programs.

Resolves #14582.

11 years agoConvert TEST_fff_f tests from code to data.
Joseph Myers [Sun, 12 May 2013 21:07:58 +0000 (21:07 +0000)]
Convert TEST_fff_f tests from code to data.

11 years agoConvert TEST_c_c tests from code to data.
Joseph Myers [Sun, 12 May 2013 17:38:26 +0000 (17:38 +0000)]
Convert TEST_c_c tests from code to data.

11 years ago[AArch64] Fix out of range branch from ioctl() and clone()
Marcus Shawcroft [Sun, 12 May 2013 17:09:33 +0000 (18:09 +0100)]
[AArch64] Fix out of range branch from ioctl() and clone()

2013-05-12  Marcus Shawcroft  <marcus.shawcroft@linaro.org>

* sysdeps/unix/sysv/linux/aarch64/clone.S (__clone):
Do not call sycall_error directly with a confitional branch.

* sysdeps/unix/sysv/linux/aarch64/ioctl.S (__ioctl):
Do not call sycall_error directly with a confitional branch.

11 years agoConvert TEST_ff_f tests from code to data.
Joseph Myers [Sun, 12 May 2013 13:17:09 +0000 (13:17 +0000)]
Convert TEST_ff_f tests from code to data.

11 years agoMake fma and nextafter tests use max_value instead of fltmax.
Joseph Myers [Sat, 11 May 2013 11:10:31 +0000 (11:10 +0000)]
Make fma and nextafter tests use max_value instead of fltmax.

11 years agoConvert TEST_f_f tests from code to data.
Joseph Myers [Sat, 11 May 2013 11:09:30 +0000 (11:09 +0000)]
Convert TEST_f_f tests from code to data.

11 years agoAdd BZ #10191 to ChangeLog/NEWS
Andreas Jaeger [Sat, 11 May 2013 08:28:21 +0000 (10:28 +0200)]
Add BZ #10191 to ChangeLog/NEWS

BZ #10191 was fixed some time ago without noticing at that time.

11 years agoAdd getgid.c for SH
Christian Grönke [Fri, 10 May 2013 18:40:48 +0000 (20:40 +0200)]
Add getgid.c for SH

[BZ #12387]
* sysdeps/unix/sysv/linux/sh/getgid.c: New file.

11 years agoHurd: add ST_NOATIME
Pino Toscano [Fri, 10 May 2013 18:39:32 +0000 (20:39 +0200)]
Hurd: add ST_NOATIME

11 years agoFix integer overflow in sysdeps/unix/sysv/linux/bits/sched.h
Andreas Jaeger [Fri, 10 May 2013 18:28:40 +0000 (20:28 +0200)]
Fix integer overflow in sysdeps/unix/sysv/linux/bits/sched.h

[BZ #15448]
* sysdeps/unix/sysv/linux/bits/sched.h (__CPU_SET_S)
(__CPU_CLR_S, __CPU_ISSET_S): Avoid integer overflow.

11 years agoStart converting libm tests from code to data.
Joseph Myers [Fri, 10 May 2013 16:45:04 +0000 (16:45 +0000)]
Start converting libm tests from code to data.

11 years agoFix coding style
Siddhesh Poyarekar [Fri, 10 May 2013 12:14:27 +0000 (17:44 +0530)]
Fix coding style

11 years agoFix tgamma errno setting on domain error (bug 6809).
Joseph Myers [Fri, 10 May 2013 11:35:11 +0000 (11:35 +0000)]
Fix tgamma errno setting on domain error (bug 6809).

11 years agoUse *stat64 instead of *stat in installed programs
Florian Weimer [Fri, 10 May 2013 09:41:53 +0000 (11:41 +0200)]
Use *stat64 instead of *stat in installed programs

This ensures reliable operation on file systems with inode numbers
which do not fit into 32 bits.

11 years agoLocalize rpcgen
Andreas Jaeger [Fri, 10 May 2013 06:06:21 +0000 (08:06 +0200)]
Localize rpcgen

[BZ #15395]
* sunrpc/rpc_main.c (main): Invoke setlocale and textdomain for
localization.
Include <locale.h>.

11 years agoAdd more comments to dlclose() algorithm.
Carlos O'Donell [Thu, 9 May 2013 21:37:15 +0000 (17:37 -0400)]
Add more comments to dlclose() algorithm.

The algorithm for scanning dependencies upon dlclose is
less than immediately obvious. This patch adds two bits
of comments that explain why you start the dependency
search at l_initfini[1], and why you need to restart
the search.

---

2013-05-09  Carlos O'Donell  <carlos@redhat.com>

* elf/dl-close.c (_dl_close_worker): Add comments.

11 years agoFix ldbl-128ibm cos range reduction near pi/2 (bug 15359).
Joseph Myers [Thu, 9 May 2013 21:30:08 +0000 (21:30 +0000)]
Fix ldbl-128ibm cos range reduction near pi/2 (bug 15359).

11 years agoFix ldbl-128 cos range reduction near pi/2 (bug 15429).
Joseph Myers [Thu, 9 May 2013 21:28:54 +0000 (21:28 +0000)]
Fix ldbl-128 cos range reduction near pi/2 (bug 15429).

11 years agoUse M_SQRT1_2l instead of local M_SQRT_2_2 in libm-test.inc.
Joseph Myers [Thu, 9 May 2013 14:33:30 +0000 (14:33 +0000)]
Use M_SQRT1_2l instead of local M_SQRT_2_2 in libm-test.inc.

11 years agoFix tst-mutexpi8
Andi Kleen [Thu, 9 May 2013 14:12:26 +0000 (16:12 +0200)]
Fix tst-mutexpi8

2013-05-09  Andi Kleen  <ak@linux.intel.com>

* tst-mutex8.c (do_test): Check for ENABLE_PI.

11 years agoUse M_PI_34l consistently in libm-test.inc.
Joseph Myers [Thu, 9 May 2013 14:13:10 +0000 (14:13 +0000)]
Use M_PI_34l consistently in libm-test.inc.

11 years agoUse decimal constants in defining M_* in libm-test.inc.
Joseph Myers [Thu, 9 May 2013 12:24:44 +0000 (12:24 +0000)]
Use decimal constants in defining M_* in libm-test.inc.

11 years ago Update powerpc libm-test ULPs
Adhemerval Zanella [Wed, 8 May 2013 20:06:56 +0000 (15:06 -0500)]
 Update powerpc libm-test ULPs

11 years agoFactor out initializers for libm-test.inc constants.
Joseph Myers [Wed, 8 May 2013 16:10:44 +0000 (16:10 +0000)]
Factor out initializers for libm-test.inc constants.

11 years agoUse correct TEST_* macros for jn, ldexp and yn tests.
Joseph Myers [Wed, 8 May 2013 16:09:54 +0000 (16:09 +0000)]
Use correct TEST_* macros for jn, ldexp and yn tests.

11 years agoARM: Add Cortex-A15 optimized NEON and VFP memcpy routines, with IFUNC.
Will Newton [Wed, 8 May 2013 12:06:34 +0000 (12:06 +0000)]
ARM: Add Cortex-A15 optimized NEON and VFP memcpy routines, with IFUNC.

11 years agoMove some libm-test logic for running tests from gen-libm-test.pl to libm-test.inc.
Joseph Myers [Wed, 8 May 2013 12:02:22 +0000 (12:02 +0000)]
Move some libm-test logic for running tests from gen-libm-test.pl to libm-test.inc.

11 years agoImprove tgamma accuracy (bugs 2546, 2560, 5159, 15426).
Joseph Myers [Wed, 8 May 2013 11:58:18 +0000 (11:58 +0000)]
Improve tgamma accuracy (bugs 2546, 2560, 5159, 15426).

11 years agoPreheat CPU in benchtests.
Ondrej Bilka [Wed, 8 May 2013 06:21:05 +0000 (08:21 +0200)]
Preheat CPU in benchtests.

A benchmark could be skewed by CPU initialy working on minimal
frequency and speeding up later. We first run code in loop
to partialy fix this issue.

11 years agoOnly defined DEV_BSIZE if not already defined.
Aurelien Jarno [Tue, 7 May 2013 22:00:46 +0000 (00:00 +0200)]
Only defined DEV_BSIZE if not already defined.

11 years agoDeclare _dl_skip_args in ldsodefs.h header.
Roland McGrath [Tue, 7 May 2013 21:49:26 +0000 (14:49 -0700)]
Declare _dl_skip_args in ldsodefs.h header.

11 years agoARM: Rewrite elf_machine_dynamic in pure C.
Roland McGrath [Tue, 7 May 2013 17:29:13 +0000 (10:29 -0700)]
ARM: Rewrite elf_machine_dynamic in pure C.