platform/upstream/glibc.git
6 years agohurd: Build stubs for new gnumach.defs
Samuel Thibault [Sun, 28 Jan 2018 18:38:09 +0000 (19:38 +0100)]
hurd: Build stubs for new gnumach.defs

* mach/Makefile (user-interfaces): Add mach/gnumach.
* sysdeps/mach/configure.ac (mach_interface_list): Add gnumach.
* sysdeps/mach/configure (mach_interface_list): Regenerate.

6 years agohurd: Include <sigsetops.h> from hurd/hurd/signal.h
Samuel Thibault [Sun, 28 Jan 2018 18:36:29 +0000 (19:36 +0100)]
hurd: Include <sigsetops.h> from hurd/hurd/signal.h

* hurd/hurd/signal.h [__USE_EXTERN_INLINES][_LIBC][IS_IN(libc) ||
IS_IN(libpthread)]: Include <sigsetops.h>.

6 years agohurd: Fix allocalim build
Samuel Thibault [Sun, 28 Jan 2018 18:33:37 +0000 (19:33 +0100)]
hurd: Fix allocalim build

* sysdeps/pthread/allocalim.h [!defined PTHREAD_STACK_MIN]: Do not
check size against PTHREAD_STACK_MIN.

6 years agohurd: fix typo
Samuel Thibault [Sun, 28 Jan 2018 18:27:19 +0000 (19:27 +0100)]
hurd: fix typo

6 years agohurd: Fix posix options
Samuel Thibault [Sun, 28 Jan 2018 18:13:55 +0000 (19:13 +0100)]
hurd: Fix posix options

_POSIX_CHOWN_RESTRICTED and _POSIX_NO_TRUNC should be always defined.

* sysdeps/mach/hurd/bits/posix_opt.h (_POSIX_CHOWN_RESTRICTED,
_POSIX_NO_TRUNC): Define to 0.

6 years agoRevert "hurd: Fix building io/tst-fchownat.c"
Samuel Thibault [Sun, 28 Jan 2018 18:13:20 +0000 (19:13 +0100)]
Revert "hurd: Fix building io/tst-fchownat.c"

This reverts commit fc221145a734d128d8f56f40d40fd4bb70a1d151.

6 years agohurd: Fix getifaddrs / freeifaddrs exposition
Samuel Thibault [Sun, 28 Jan 2018 18:06:15 +0000 (19:06 +0100)]
hurd: Fix getifaddrs / freeifaddrs exposition

400669754de4 ('hurd: Fix nscd build') had the side effect of making
libc's freeaddrinfo expose freeifaddrs through __check_pf.  We can just
move the renames to gai.c itself, along others.

* sysdeps/mach/hurd/check_pf.c (__getifaddrs, __freeifaddrs): Do not
define macros.
* nscd/gai.c (__getifaddrs): Define macro to getifaddrs.
(__freeifaddrs): Define macro to freeifaddrs.

6 years agohurd: take __USE_EXTERN_INLINES into account and restrict inlines
Samuel Thibault [Sun, 28 Jan 2018 17:43:10 +0000 (18:43 +0100)]
hurd: take __USE_EXTERN_INLINES into account and restrict inlines

* hurd/hurd.h (__hurd_fail): Always declare function, and provide inline
version only if __USE_EXTERN_INLINES is defined.
* hurd/hurd/fd.h (_hurd_fd_error_signal, _hurd_fd_error, __hurd_dfail,
__hurd_sockfail): Likewise.
(_hurd_fd_get): Always declare functions, and provide inline versions
only if __USE_EXTERN_INLINES and _LIBC are defined and IS_IN(libc).
* hurd/hurd/port.h (_hurd_port_init, _hurd_port_locked_get,
_hurd_port_get, _hurd_port_free, _hurd_port_locked_set,
_hurd_port_set): Always declare functions, and provide inline versions
only if __USE_EXTERN_INLINES and _LIBC are defined and
IS_IN(libc).
* hurd/hurd/signal.h (_hurd_self_sigstate, _hurd_critical_section_lock,
_hurd_critical_section_unlock): Likewise.
* hurd/hurd/threadvar.h (__hurd_threadvar_location_from_sp,
* __hurd_threadvar_location): Likewise.
* hurd/hurd/userlink.h (_hurd_userlink_link, _hurd_userlink_unlink,
_hurd_userlink_clear): Likewise.
* mach/lock-intern.h (__spin_lock_init, __spin_lock, __mutex_lock,
__mutex_unlock, __mutex_trylock): Always declare functions, and provide
inline versions only if __USE_EXTERN_INLINES and _LIBC are defined.
* mach/mach/mig_support.h (__mig_strncpy): Likewise.
* sysdeps/generic/machine-lock.h (__spin_unlock, __spin_try_lock,
__spin_lock_locked): Likewise.
* sysdeps/mach/i386/machine-lock.h (__spin_unlock, __spin_try_lock,
__spin_lock_locked): Likewise.
* mach/spin-lock.c (__USE_EXTERN_INLINES): Define to 1.
* hurd/Versions (libc: GLIBC_2.27): Add _hurd_fd_error_signal,
_hurd_fd_error, __hurd_dfail, __hurd_sockfail, _hurd_port_locked_set,
__hurd_threadvar_location_from_sp, __hurd_threadvar_location,
_hurd_userlink_link, _hurd_userlink_unlink, _hurd_userlink_clear.

6 years agohurd: Fix building io/tst-copy_file_range.c
Samuel Thibault [Sun, 28 Jan 2018 16:56:49 +0000 (17:56 +0100)]
hurd: Fix building io/tst-copy_file_range.c

* io/tst-copy_file_range.c [!defined CLONE_NEWNS]: Do not include
<sys/mount.h>.

6 years agohurd: Fix building io/tst-fchownat.c
Samuel Thibault [Sun, 28 Jan 2018 16:52:34 +0000 (17:52 +0100)]
hurd: Fix building io/tst-fchownat.c

* io/tst-fchownat.c [!defined _POSIX_CHOWN_RESTRICTED]: Act like when
_POSIX_CHOWN_RESTRICTED is defined to 0.

6 years agohurd: Fix building libio/tst-memstream3.c
Samuel Thibault [Sun, 28 Jan 2018 16:45:41 +0000 (17:45 +0100)]
hurd: Fix building libio/tst-memstream3.c

FWRITE is already an fcntl.h macro.

* libio/tst-memstream3.c (FWRITE): Rename to _FWRITE.
(do_test_bz20181): Rename accordingly.
* libio/tst-wmemstream3.c (FWRITE): Rename accordingly.

6 years agohurd: add base abilist files
Samuel Thibault [Sun, 28 Jan 2018 16:26:35 +0000 (17:26 +0100)]
hurd: add base abilist files

* sysdeps/mach/hurd/libhurduser.abilist: New file.
* sysdeps/mach/libmachuser.abilist: New file.

6 years agohurd: Add missing fcntl-internal.h
Thomas Schwinge [Sun, 28 Jan 2018 15:53:32 +0000 (16:53 +0100)]
hurd: Add missing fcntl-internal.h

* hurd/fcntl-internal.h: New file.

6 years agohurd: Fix nscd build
Samuel Thibault [Sun, 28 Jan 2018 15:28:48 +0000 (16:28 +0100)]
hurd: Fix nscd build

nscd won't find check_* from inet/, it needs a sysdeps file.

* sysdeps/mach/hurd/check_native.c: New file.
* sysdeps/mach/hurd/check_pf.c: New file.

6 years agohurd: Fix warnings
Samuel Thibault [Sun, 28 Jan 2018 15:23:59 +0000 (16:23 +0100)]
hurd: Fix warnings

* hurd/hurd/lookup.h (__hurd_file_name_lookup, hurd_file_name_lookup,
__hurd_file_name_split, hurd_file_name_split,
__hurd_directory_name_split, hurd_directory_name_split,
__hurd_file_name_lookup_retry, hurd_file_name_lookup_retry,
hurd_file_name_path_lookup): Make lookup function parameter take a
const char *name instead of char *name.
* hurd/hurdlookup.c (__hurd_file_name_lookup, __hurd_file_name_split,
__hurd_directory_name_split): Likewise.
* hurd/lookup-retry.c (__hurd_file_name_lookup_retry): Likewise.
* hurd/path-lookup.c (hurd_file_name_path_lookup): Likewise.

6 years agohurd: Fix warning
Samuel Thibault [Sun, 28 Jan 2018 15:21:28 +0000 (16:21 +0100)]
hurd: Fix warning

* sysdeps/mach/hurd/spawni.c (__spawni): Make relpath and abspath
const char * instead of char *.

6 years agohurd: Make build-many-glibcs.py use mainline gnumach
Samuel Thibault [Sun, 28 Jan 2018 14:48:04 +0000 (15:48 +0100)]
hurd: Make build-many-glibcs.py use mainline gnumach

Some warnings need a couple of fixes in the gnumach headers.

        * scripts/build-many-glibcs.py (checkout_vcs): Add gnumach
        repository URLs, run autoreconf, and make it the default for now.

6 years agohurd: fix warning
Samuel Thibault [Sat, 27 Jan 2018 22:49:18 +0000 (23:49 +0100)]
hurd: fix warning

* sysdeps/mach/hurd/getresgid.c (__getresgid): Set result from
critical section to make code simpler and avoid warning.
* sysdeps/mach/hurd/getresuid.c (__getresuid): Set result from
critical section to make code simpler and avoid warning.

6 years agohurd: Fix enabling vm_copy
Samuel Thibault [Sat, 27 Jan 2018 21:09:30 +0000 (22:09 +0100)]
hurd: Fix enabling vm_copy

Benchmarked on
http://lists.gnu.org/archive/html/bug-hurd/2014-12/msg00081.html

* sysdeps/mach/pagecopy.h (PAGE_THRESHOLD): Rename to
PAGE_COPY_THRESHOLD and set to benchmarked 16384.

6 years agohurd: fix warning
Samuel Thibault [Sat, 27 Jan 2018 21:01:34 +0000 (22:01 +0100)]
hurd: fix warning

Making `special_profil_failure' both avoids warning "variable
'special_profil_failure' set but not used", and makes it easier to
access with gdb.

* sysdeps/mach/hurd/profil.c (special_profil_failure): Move variable
to global scope.

6 years agohurd: fix warning
Samuel Thibault [Sat, 27 Jan 2018 21:00:23 +0000 (22:00 +0100)]
hurd: fix warning

* sysdeps/mach/hurd/dl-sysdep.c (_exit): Call LOSE and abort() if
__task_terminate would ever return successfully.

6 years agohurd: make build-many-glibcs.py use mainline mig
Samuel Thibault [Sat, 27 Jan 2018 20:44:57 +0000 (21:44 +0100)]
hurd: make build-many-glibcs.py use mainline mig

Some warnings come from code generated by mig, so we need a very recent
version for now.

        * scripts/build-many-glibcs.py (checkout_vcs): Add mig repository
        URL, and run autoreconf, make it the default for now.

6 years agohurd: drop elder MACH_IPC_COMPAT handling
Samuel Thibault [Sat, 27 Jan 2018 20:40:08 +0000 (21:40 +0100)]
hurd: drop elder MACH_IPC_COMPAT handling

This was dropped from GNU Mach in 2006.

* mach/Machrules (MIGFLAGS): Do not set -DMACH_IPC_COMPAT=0.
* mach/mach/mach_traps.h: Drop comment about MACH_IPC_COMPAT.
* sysdeps/mach/hurd/fork.c (__fork): Drop special casing
MACH_IPC_COMPAT.

6 years agohurd: fix warning
Samuel Thibault [Sat, 27 Jan 2018 20:37:29 +0000 (21:37 +0100)]
hurd: fix warning

* sysdeps/mach/hurd/Makefile [$(subdir)==nis]: Add
-DUSE_BINDINGDIR=1 to CFLAGS-ypclnt.c.

6 years agohurd: fix gcc build in build-many-glibcs.py
Samuel Thibault [Sat, 27 Jan 2018 18:53:44 +0000 (19:53 +0100)]
hurd: fix gcc build in build-many-glibcs.py

gcc's libcilkrts has never actually supported GNU/Hurd, and doesn't
automatically disable it, and the support was actually removed in gcc trunk,
so that will never actually be fixed there.

* scripts/build-many-glibcs.py [os == gnu] (build_gcc): Pass
--disable-libcilkrts to gcc configure.

6 years agohurd: fix warning
Samuel Thibault [Sat, 27 Jan 2018 16:17:49 +0000 (17:17 +0100)]
hurd: fix warning

timer_ptr2id and timer_id2ptr are used to convert between
application-visible timer_t and struct timer_node *. timer_ptr2id was made
to use void * instead of timer_t in 49b650430eb5 ('Update.') for no reason.
It happens that on Linux timer_t is void *, so both that change and this
commit are no-ops there, but not on systems where timer_t is not void *.

Using timer_ptr2id for filling sival_ptr also does not make sense since that
actually is a void *.

* sysdeps/pthread/posix-timer.h (timer_ptr2id): Cast to timer_t
instead of void *.
* sysdeps/pthread/timer_create.c (timer_create): Do not use
timer_ptr2id to cast struct timer_node * to void *.

6 years agohurd: fix warning
Samuel Thibault [Sat, 27 Jan 2018 15:49:05 +0000 (16:49 +0100)]
hurd: fix warning

* sysdeps/mach/hurd/dl-sysdep.c (_dl_sysdep_start): Cast
vm_address_t * to ElfW(Addr) * for dl_main parameter.

6 years agohurd: fix warnings
Samuel Thibault [Sat, 27 Jan 2018 15:47:36 +0000 (16:47 +0100)]
hurd: fix warnings

* sysdeps/generic/sigsetops.h (__sigemptyset, __sigfillset,
__sigandset, __sigorset, __sigaddset, __sigdelset): Make them really
return 0.

6 years agohurd: fix warning
Samuel Thibault [Sat, 27 Jan 2018 15:45:55 +0000 (16:45 +0100)]
hurd: fix warning

* sysdeps/generic/sigset-cvt-mask.h: Include <sigsetops.h>.

6 years agohurd: fix warnings
Samuel Thibault [Sat, 27 Jan 2018 15:45:10 +0000 (16:45 +0100)]
hurd: fix warnings

* sysdeps/generic/not-cancel.h: Include <fcntl.h>, <unistd.h>,
<sys/wait.h>, <time.h>, <sys/uio.h>.
(NOT_CANCEL_H): Add inclusion guard.

6 years agohurd: fix warning
Samuel Thibault [Sat, 27 Jan 2018 15:42:58 +0000 (16:42 +0100)]
hurd: fix warning

* resolv/res-close.c: Include <stdlib.h>.

6 years agohurd: fix warning
Samuel Thibault [Sat, 27 Jan 2018 15:33:13 +0000 (16:33 +0100)]
hurd: fix warning

* sysdeps/mach/hurd/profil.c: Reuse `a' variable instead of introducing
a `c' variable.

6 years agohurd: fix warning
Samuel Thibault [Sat, 27 Jan 2018 15:32:16 +0000 (16:32 +0100)]
hurd: fix warning

* sysdeps/mach/hurd/xmknodat.c: Include <sys/sysmacros.h>.

6 years agoUpdate hppa libm-test-ulps
James Clarke [Sat, 27 Jan 2018 01:52:54 +0000 (01:52 +0000)]
Update hppa libm-test-ulps

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

Signed-off-by: James Clarke <jrtc27@jrtc27.com>
6 years agoUpdate Alpha libm-test-ulps
James Clarke [Sat, 27 Jan 2018 01:17:45 +0000 (01:17 +0000)]
Update Alpha libm-test-ulps

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

Signed-off-by: James Clarke <jrtc27@jrtc27.com>
6 years agoAdd ChangeLog entry for "pl_PL: Add alternative month names".
Rafal Luzynski [Sat, 27 Jan 2018 01:03:34 +0000 (02:03 +0100)]
Add ChangeLog entry for "pl_PL: Add alternative month names".

6 years agoBuild only shared libnsl objects if NIS is disabled (bug 22701)
Andreas Schwab [Fri, 26 Jan 2018 18:01:03 +0000 (19:01 +0100)]
Build only shared libnsl objects if NIS is disabled (bug 22701)

6 years agoREADME: hppa no longer requires patches.
Carlos O'Donell [Fri, 26 Jan 2018 17:49:54 +0000 (09:49 -0800)]
README: hppa no longer requires patches.

6 years agoUpdate sparc ULPs.
Patrick McGehearty [Fri, 26 Jan 2018 16:23:01 +0000 (08:23 -0800)]
Update sparc ULPs.

        * sysdeps/sparc/fpu/libm-test-ulps: Update
        cpow, ctan, ctanh, j0, j1, y0, yn ulps.

6 years agoRevert Intel CET changes to __jmp_buf_tag (Bug 22743)
Carlos O'Donell [Thu, 25 Jan 2018 04:35:22 +0000 (20:35 -0800)]
Revert Intel CET changes to __jmp_buf_tag (Bug 22743)

In commit cba595c350e52194e10c0006732e1991e3d0803b and commit
f81ddabffd76ac9dd600b02adbf3e1dac4bb10ec, ABI compatibility with
applications was broken by increasing the size of the on-stack
allocated __pthread_unwind_buf_t beyond the oringal size.
Applications only have the origianl space available for
__pthread_unwind_register, and __pthread_unwind_next to use,
any increase in the size of __pthread_unwind_buf_t causes these
functions to write beyond the original structure into other
on-stack variables leading to segmentation faults in common
applications like vlc. The only workaround is to version those
functions which operate on the old sized objects, but this must
happen in glibc 2.28.

Thank you to Andrew Senkevich, H.J. Lu, and Aurelien Jarno, for
submitting reports and tracking the issue down.

The commit reverts the above mentioned commits and testing on
x86_64 shows that the ABI compatibility is restored. A tst-cleanup1
regression test linked with an older glibc now passes when run
with the newly built glibc. Previously a tst-cleanup1 linked with
an older glibc would segfault when run with an affected glibc build.

Tested on x86_64 with no regressions.

Signed-off-by: Carlos O'Donell <carlos@redhat.com>
6 years agohurd build-many-glibcs.py: use hurd mainline for now
Samuel Thibault [Fri, 26 Jan 2018 01:35:10 +0000 (02:35 +0100)]
hurd build-many-glibcs.py: use hurd mainline for now

* scripts/build-many-glibcs.py (checkout_vcs): Make hurd
mainline the default for now.

6 years agoAdd ChangeLog entry for "uk_UA: Add alternative month names".
Rafal Luzynski [Thu, 25 Jan 2018 23:21:16 +0000 (00:21 +0100)]
Add ChangeLog entry for "uk_UA: Add alternative month names".

6 years agoAdd support for the RISC-V-specific ELF flags
Palmer Dabbelt [Sat, 13 Jan 2018 09:59:12 +0000 (01:59 -0800)]
Add support for the RISC-V-specific ELF flags

The RISC-V port defines ELF flags that enforce compatibility between
various objects.  This adds the shared support necessary for these
flags.

2018-01-25  Palmer Dabbelt  <palmer@sifive.com>

        * elf/cache.c (print_entry): Add FLAG_RISCV_FLOAT_ABI_SOFT and
        FLAG_RISCV_FLOAT_ABI_DOUBLE.
        * elf/elf.h (EF_RISCV_RVC): New define.
        (EF_RISCV_FLOAT_ABI): Likewise.
        (EF_RISCV_FLOAT_ABI_SOFT): Likewise.
        (EF_RISCV_FLOAT_ABI_SINGLE): Likewise.
        (EF_RISCV_FLOAT_ABI_DOUBLE): Likewise.
        (EF_RISCV_FLOAT_ABI_QUAD): Likewise.
        * sysdeps/generic/ldconfig.h (FLAG_RISCV_FLOAT_ABI_SOFT): New
        define.
        (FLAG_RISCV_FLOAT_ABI_DOUBLE): Likewise.

6 years agoLIBC_SLIBDIR_RTLDDIR: substitute arguments in single quotes
Andreas Schwab [Thu, 25 Jan 2018 09:41:27 +0000 (10:41 +0100)]
LIBC_SLIBDIR_RTLDDIR: substitute arguments in single quotes

The arguments of the LIBC_SLIBDIR_RTLDDIR macro are used both in unquoted
and single quoted context, so that neither shell nor makefile variable
references work.  Consistently put them in single quotes so that they can
refer to makefile variables.

6 years agouk_UA: Add alternative month names (bug 10871).
Rafal Luzynski [Tue, 5 Jan 2016 02:06:25 +0000 (03:06 +0100)]
uk_UA: Add alternative month names (bug 10871).

Primary month names are in a genitive case now, alternative month names
are in a nominative case.

The alternative digits hack is no longer needed and has been removed.

[BZ #10871]
* localedata/locales/uk_UA (mon): Renamed to...
(alt_mon): This.
(alt_digits): "0" removed and then renamed to...
(mon): This.
(date_fmt): Definition changed not to use the alternative
digits hack.

6 years agobuild-many-glibcs.py: Fix hurd-vcs checkout
Samuel Thibault [Thu, 25 Jan 2018 02:14:22 +0000 (03:14 +0100)]
build-many-glibcs.py: Fix hurd-vcs checkout

* scripts/build-many-glibcs.py (checkout_vcs): Run autoconf.

6 years agobuild-many-glibcs.py: Add hurd vcs support
Samuel Thibault [Thu, 25 Jan 2018 02:01:59 +0000 (03:01 +0100)]
build-many-glibcs.py: Add hurd vcs support

* scripts/build-many-glibcs.py (checkout_vcs): Add hurd repository URL.

6 years agoAdd soft-float ColdFire to build-many-glibcs.py.
Joseph Myers [Wed, 24 Jan 2018 23:48:52 +0000 (23:48 +0000)]
Add soft-float ColdFire to build-many-glibcs.py.

Since it turns out soft-float ColdFire has a different glibc ABI to
hard-float ColdFire, as well as various differences in which glibc
code gets built, this patch adds such a configuration to
build-many-glibcs.py to (hopefully) complete the set of ABIs being
tested.  (Note that the build for soft-float ColdFire is currently
broken even with GCC mainline - I have a glibc patch to fix this, but
it needs before-and-after build-many-glibcs.py comparison of stripped
binaries for all configurations before being committed.)

* scripts/build-many-glibcs.py (Context.add_all_configs): Add
soft-float ColdFire configuration.

6 years agoAdd localplt.data for ColdFire.
Joseph Myers [Wed, 24 Jan 2018 23:43:33 +0000 (23:43 +0000)]
Add localplt.data for ColdFire.

The sole failure for ColdFire in the compilation part of the glibc
testsuite is the localplt test.  This patch adds a localplt baseline
for ColdFire to eliminate that failure.  The difference from the
existing m68k baseline is that no PLT entry for _Unwind_Find_FDE is
expected, because ColdFire does not set
libc_cv_gcc_unwind_find_fde=yes.

Tested with build-many-glibcs.py.

* sysdeps/unix/sysv/linux/m68k/localplt.data: Move to ....
* sysdeps/unix/sysv/linux/m68k/m680x0/localplt.data: ... here.
* sysdeps/unix/sysv/linux/m68k/coldfire/localplt.data: New file.

6 years agoAdd no-FPU ColdFire math_private.h.
Joseph Myers [Wed, 24 Jan 2018 23:33:03 +0000 (23:33 +0000)]
Add no-FPU ColdFire math_private.h.

As with some other soft-float configurations, no-FPU ColdFire needs
various fenv.h functions and glibc-internal macros overridden in
math_private.h to avoid references to undefined FE_* macros when
building glibc.  This patch adds a suitable math_private.h, based on
the MicroBlaze one (Nios II and Tile also have similar files).

There's a case for having such a file in sysdeps/ieee754/soft-fp so
this logic is applied more generally to configurations without
exceptions and rounding modes, even when the relevant macros are
defined in fenv.h - the only case where that might be inappropriate is
ARM soft-float (where the fenv.h functions might or might not work at
runtime, depending on whether the processor used at runtime supports
VFP).  There's also a case that soft-float configurations (on
processors with both hard-float and soft-float) should more
consistently avoid defining FE_* macros in bits/fenv.h when not
actually supported.  But both of those are separate potential
cleanups.

This allows the no-FPU ColdFire build to get further (another fix is
needed to allow the build to complete).

* sysdeps/m68k/coldfire/nofpu/math_private.h: New file.  Based on
MicroBlaze file.

6 years agoAdd jmp_buf-macros.h for no-FPU ColdFire.
Joseph Myers [Wed, 24 Jan 2018 23:22:28 +0000 (23:22 +0000)]
Add jmp_buf-macros.h for no-FPU ColdFire.

Continuing the fixes for ColdFire glibc build with
build-many-glibcs.py, given a GCC patch for the libgcc build failure,
this patch adds jmp_buf-macros.h for no-FPU ColdFire.  This allows the
no-FPU build to progress further than without the patch (although
other fixes are still needed for the build to complete).

* sysdeps/unix/sysv/linux/m68k/coldfire/jmp_buf-macros.h: Move to
....
* sysdeps/unix/sysv/linux/m68k/coldfire/fpu/jmp_buf-macros.h:
... here.
* sysdeps/unix/sysv/linux/m68k/coldfire/nofpu/jmp_buf-macros.h:
New file.

6 years agoAdd jmp_buf-macros.h for ColdFire.
Joseph Myers [Wed, 24 Jan 2018 22:26:24 +0000 (22:26 +0000)]
Add jmp_buf-macros.h for ColdFire.

This patch adds a jmp_buf-macros.h for ColdFire.  In conjunction with
a GCC patch to fix the libgcc build failure for ColdFire
<https://gcc.gnu.org/ml/gcc-patches/2018-01/msg02064.html> this
suffices to restore the build (tested with build-many-glibcs.py).  A
further patch will be needed for soft-float ColdFire (while the
function-calling ABI is the same for hard-float and soft-float
ColdFire, it turns out the glibc ABI is not - so another ColdFire
variant will be needed in build-many-glibcs.py), but I'll deal with
that separately.

Tested with build-many-glibcs.py (m68k-linux-gnu and
m68k-linux-gnu-coldfire).  (There's a localplt test failure for
coldfire; that's the only failure in the compilation part of the
testsuite.)

* sysdeps/unix/sysv/linux/m68k/jmp_buf-macros.h: Move to ....
* sysdeps/unix/sysv/linux/m68k/m680x0/jmp_buf-macros.h: ... here.
* sysdeps/unix/sysv/linux/m68k/coldfire/jmp_buf-macros.h: New
file.

6 years agoaarch64: Revert the change of the __reserved member of mcontext_t
Szabolcs Nagy [Mon, 22 Jan 2018 11:13:31 +0000 (11:13 +0000)]
aarch64: Revert the change of the __reserved member of mcontext_t

The uc_mcontext.__reserved member of ucontext_t is a user visible API,
that should not be changed, because this is the only way to access cpu
states of various extensions of linux asm/sigcontext.h, it does not
violate namespace rules either, so revert this part of the commit

commit 4fa9b3bfe6759c82beb4b043a54a3598ca467289
Commit:     Joseph Myers <joseph@codesourcery.com>

    Fix mcontext_t sigcontext namespace (bug 21457).

(In principle the user can type cast &uc_mcontext to struct sigcontext*
to use the linux sigcontext fields, but that's not the existing practice
since mcontext_t used to be a typedef of struct sigcontext.)

[BZ #22742]
* sysdeps/unix/sysv/linux/aarch64/sys/ucontext.h (__glibc_reserved1):
Rename to __reserved and add comment.
* sysdeps/unix/sysv/linux/aarch64/ucontext_i.sym (__glibc_reserved1):
Rename to __reserved.

6 years agoAdd build-many-glibcs.py support for Hurd.
Joseph Myers [Wed, 24 Jan 2018 01:18:54 +0000 (01:18 +0000)]
Add build-many-glibcs.py support for Hurd.

This patch adds build-many-glibcs.py support for GNU Hurd.  Builds of
the i686-gnu configuration will fail until sufficient support is
merged to master, so completing build-many-glibcs.py coverage of all
glibc ABIs and making results accurately reflect the broken state of
builds for Hurd.

* scripts/build-many-glibcs.py (Context.add_all_configs): Add
i686-gnu configurations.
(Context.run_builds): Include mig, gnumach and hurd in components
considered.
(Context.checkout): Add mig, gnumach and hurd to components.
(Context.checkout_tar): Add URL mappings for mig, gnumach and
hurd.
(Context.bot_cycle): Check for changes to mig, gnumach and hurd.
(Config.build): Install gnumach headers, build mig and install
hurd headers for 'gnu' OS.
(Config.install_gnumach_headers): New function.
(Config.install_hurd_headers): Likewise.
(Glibc.build_glibc): Do not use /usr for 'gnu' OS.  Specifiy MIG
when building for 'gnu' OS.

6 years agoNEWS: Fix spelling mistake: "%ob" -> "%Ob".
Rafal Luzynski [Tue, 23 Jan 2018 21:52:39 +0000 (22:52 +0100)]
NEWS: Fix spelling mistake: "%ob" -> "%Ob".

6 years agoNEWS: fix spelling mistake
Dmitry V. Levin [Tue, 23 Jan 2018 20:31:13 +0000 (20:31 +0000)]
NEWS: fix spelling mistake

6 years agoNEWS: Fix spelling mistake.
Carlos O'Donell [Tue, 23 Jan 2018 20:22:51 +0000 (12:22 -0800)]
NEWS: Fix spelling mistake.

6 years agoNEWS: Document static application consequenes of %OB/%Ob.
Carlos O'Donell [Tue, 23 Jan 2018 20:20:01 +0000 (12:20 -0800)]
NEWS: Document static application consequenes of %OB/%Ob.

6 years agomanual: Fix spelling of "Auxiliary."
Tobias Klauser [Tue, 23 Jan 2018 19:40:44 +0000 (11:40 -0800)]
manual: Fix spelling of "Auxiliary."

6 years agomanual: Touch up documentation for strftime and ALTMON_* constants.
Rical Jasan [Tue, 23 Jan 2018 03:24:24 +0000 (19:24 -0800)]
manual: Touch up documentation for strftime and ALTMON_* constants.

"%OB" is considered a conversion specifier ("B" is the format
specifier), and the list of format specifiers for months in the
description of the optional "O" modifier was incomplete.  A
cross-reference from the ALTMON_* constants to the strftime section
is also provided.  Lastly, some grammatical fixes (commas) are made
and paragraphs refactored (rewrapped).

* manual/locale.texi (ALTMON_1, ALTMON_2, ALTMON_3, ALTMON_4,
ALTMON_5, ALTMON_6, ALTMON_7, ALTMON_8, ALTMON_9, ALTMON_10,
ALTMON_11, ALTMON_12): Improve documentation.
* manual/time.texi (strftime): Likewise.

6 years agoFix the wording in NEWS to use "or later"
Szabolcs Nagy [Mon, 22 Jan 2018 15:54:02 +0000 (15:54 +0000)]
Fix the wording in NEWS to use "or later"

6 years agoUpdate NEWS with binutils version requirement for static PIE support
Szabolcs Nagy [Mon, 22 Jan 2018 10:22:24 +0000 (10:22 +0000)]
Update NEWS with binutils version requirement for static PIE support

6 years agopl_PL: Add alternative month names (bug 10871).
Rafal Luzynski [Tue, 5 Jan 2016 02:03:10 +0000 (03:03 +0100)]
pl_PL: Add alternative month names (bug 10871).

[BZ #10871]
* localedata/locales/pl_PL: Alternative month names added,
primary month names are genitive now.
* time/tst-strptime.c (day_tests): Actually use a genitive case
of a month name in Polish language.

6 years agoDocumentation to the above changes (bug 10871).
Rafal Luzynski [Mon, 18 Sep 2017 22:00:22 +0000 (00:00 +0200)]
Documentation to the above changes (bug 10871).

[BZ #10871]
* manual/locale.texi: Document ALTMON_1..12 constants for
nl_langinfo.  Specify when to use ALTMON instead of MON.
* manual/time.texi (strftime, strptime): Document GNU extension
permitting O modifier with %B and %b.  Specify when to use
%OB instead of %B.

Reviewed-by: Carlos O'Donell <carlos@redhat.com>
6 years agoAbbreviated alternative month names (%Ob) also added (bug 10871).
Rafal Luzynski [Tue, 14 Nov 2017 01:22:38 +0000 (02:22 +0100)]
Abbreviated alternative month names (%Ob) also added (bug 10871).

All the previous changes also repeated to support abbreviated
alternative month names.  In most languages which have declension and
need nominative/genitive month names the abbreviated forms for both
cases are the same.  An example where they do differ is May in Russian:
this name is too short to be abbreviated so even the abbreviated form
features the declension suffixes.

[BZ #10871]
* locale/C-time.c (_nl_C_LC_TIME): Add abbreviated alternative month
names, define them as the same as abbreviated month names explicitly.
* locale/categories.def (LC_TIME): Add ab_alt_mon and wide-ab_alt_mon.
* locale/langinfo.h: (_NL_ABALTMON_1, _NL_ABALTMON_2, _NL_ABALTMON_3,
_NL_ABALTMON_4, _NL_ABALTMON_5, _NL_ABALTMON_6, _NL_ABALTMON_7,
_NL_ABALTMON_8, _NL_ABALTMON_9, _NL_ABALTMON_10, _NL_ABALTMON_11,
_NL_ABALTMON_12, _NL_WABALTMON_1, _NL_WABALTMON_2, _NL_WABALTMON_3,
_NL_WABALTMON_4, _NL_WABALTMON_5, _NL_WABALTMON_6, _NL_WABALTMON_7,
_NL_WABALTMON_8, _NL_WABALTMON_9, _NL_WABALTMON_10, _NL_WABALTMON_11,
_NL_WABALTMON_12): New enum constants.
* locale/programs/ld-time.c (struct locale_time_t): Add ab_alt_mon,
wab_alt_mon, and ab_alt_mon_defined members.
(time_output): Output ab_alt_mon and wab_alt_mon members.
(time_read): Read them, initialize them as copies of abmon and wabmon
respectively if they are missing, initialize ab_alt_mon_defined.
* locale/programs/locfile-kw.gperf (ab_alt_mon): Define.
* locale/programs/locfile-kw.h: Regenerate.
* locale/programs/locfile-token.h (tok_ab_alt_mon): New enum constant.
* time/Makefile [$(run-built-tests) = yes] (LOCALES): Add es_ES.UTF-8
and ru_RU.UTF-8.
* time/strftime_l.c (a_altmonth, aam_len): New macros.
[!COMPILE_WIDE] (ABALTMON_1): New macro.
(__strftime_internal): Handle %Ob and %Oh formats.
* time/strptime_l.c [_LIBC] (ab_alt_month_name): New macro.
(__strptime_internal): Handle %Ob and %Oh formats.
* time/tst-strptime.c (day_tests): Add more tests to parse different
forms of month names including the new %Ob format specifier.

Reviewed-by: Carlos O'Donell <carlos@redhat.com>
6 years agoImplement alternative month names (bug 10871).
Rafal Luzynski [Tue, 14 Nov 2017 01:19:35 +0000 (02:19 +0100)]
Implement alternative month names (bug 10871).

Some languages (Slavic, Baltic, etc.) require a genitive case of the
month name when formatting a full date (with the day number) while
they require a nominative case when referring to the month standalone.
This requirement cannot be fulfilled without providing two forms for
each month name.  From now it is specified that nl_langinfo(MON_1)
series (up to MON_12) and strftime("%B") generate the month names in
the grammatical form used when the month is a part of a complete date.
If the grammatical form used when the month is named by itself is needed,
the new values nl_langinfo(ALTMON_1) (up to ALTMON_12) and
strftime("%OB") are supported.  This new feature is optional so the
languages which do not need it or do not yet provide the updated
locales simply do not use it and their behaviour is unchanged.

[BZ #10871]
* locale/C-time.c (_nl_C_LC_TIME): Add alternative month names,
define them as the same as primary full month names explicitly.
* locale/categories.def (LC_TIME): Add alt_mon and wide-alt_mon.
* locale/langinfo.h (__ALTMON_1, __ALTMON_2, __ALTMON_3, __ALTMON_4,
__ALTMON_5, __ALTMON_6, __ALTMON_7, __ALTMON_8, __ALTMON_9, __ALTMON_10,
__ALTMON_11, __ALTMON_12, _NL_WALTMON_1, _NL_WALTMON_2, _NL_WALTMON_3,
_NL_WALTMON_4, _NL_WALTMON_5, _NL_WALTMON_6, _NL_WALTMON_7,
_NL_WALTMON_8, _NL_WALTMON_9, _NL_WALTMON_10, _NL_WALTMON_11,
_NL_WALTMON_12): New enum constants.
[__USE_GNU] (ALTMON_1, ALTMON_2, ALTMON_3, ALTMON_4, ALTMON_5, ALTMON_6,
ALTMON_7, ALTMON_8, ALTMON_9, ALTMON_10, ALTMON_11, ALTMON_12): New
macros.
* locale/programs/ld-time.c (struct locale_time_t): Add alt_mon,
walt_mon, and alt_mon_defined members.
(time_output): Output alt_mon and walt_mon members.
(time_read): Read them, initialize them as copies of mon and wmon
respectively if they are missing, initialize alt_mon_defined.
* locale/programs/locfile-kw.gperf (alt_mon): Define.
* locale/programs/locfile-kw.h: Regenerate.
* locale/programs/locfile-token.h (tok_alt_mon): New enum constant.
* localedata/tst-langinfo.c (map): Add tests for the new constants
ALTMON_1 .. ALTMON_12.
* time/Makefile [$(run-built-tests) = yes] (LOCALES): Add fr_FR.UTF-8
and pl_PL.UTF-8.
* time/strftime_l.c (f_altmonth): New macro.
(__strftime_internal): Handle %OB format.
* time/strptime_l.c [_LIBC] (alt_month_name): New macro.
(__strptime_internal): Handle %OB format.
* time/tst-strptime.c (day_tests): Add tests to parse different forms
of month names including the new %OB format specifier.

Reviewed-by: Carlos O'Donell <carlos@redhat.com>
6 years agopowerpc: Fix syscalls during early process initialization [BZ #22685]
Tulio Magno Quites Machado Filho [Fri, 12 Jan 2018 20:50:33 +0000 (18:50 -0200)]
powerpc: Fix syscalls during early process initialization [BZ #22685]

The tunables framework needs to execute syscall early in process
initialization, before the TCB is available for consumption.  This
behavior conflicts with powerpc{|64|64le}'s lock elision code, that
checks the TCB before trying to abort transactions immediately before
executing a syscall.

This patch adds a powerpc-specific implementation of __access_noerrno
that does not abort transactions before the executing syscall.

Tested on powerpc{|64|64le}.

[BZ #22685]
* sysdeps/powerpc/powerpc32/sysdep.h (ABORT_TRANSACTION_IMPL): Renamed
from ABORT_TRANSACTION.
(ABORT_TRANSACTION): Redirect to ABORT_TRANSACTION_IMPL.
* sysdeps/powerpc/powerpc64/sysdep.h (ABORT_TRANSACTION,
ABORT_TRANSACTION_IMPL): Likewise.
* sysdeps/unix/sysv/linux/powerpc/not-errno.h: New file.  Reuse
Linux code, but remove the code that aborts transactions.

Signed-off-by: Tulio Magno Quites Machado Filho <tuliom@linux.vnet.ibm.com>
Tested-by: Aurelien Jarno <aurelien@aurel32.net>
6 years agogeneric if_ether.h: Fix build
Samuel Thibault [Fri, 19 Jan 2018 00:47:45 +0000 (00:47 +0000)]
generic if_ether.h: Fix build

* sysdeps/generic/netinet/if_ether.h: Include <stdint.h>.

6 years agolocales gu_IN, lo_LA: Fix obvious typos in dates.
Rafal Luzynski [Fri, 19 Jan 2018 00:09:12 +0000 (01:09 +0100)]
locales gu_IN, lo_LA: Fix obvious typos in dates.

Reported-by: Robert Pluim <rpluim@gmail.com>
* localedata/locales/gu_IN (LC_IDENTIFICATION): Fix an obvious typo
in date: "2004-14-09" should be "2004-09-14".
* localedata/locales/lo_LA: Fix an obvious typo in date in the header:
"2003-15-09" should be "2003-09-15".

6 years agoFix integer overflows in internal memalign and malloc functions [BZ #22343]
Arjun Shankar [Thu, 18 Jan 2018 16:47:06 +0000 (16:47 +0000)]
Fix integer overflows in internal memalign and malloc functions [BZ #22343]

When posix_memalign is called with an alignment less than MALLOC_ALIGNMENT
and a requested size close to SIZE_MAX, it falls back to malloc code
(because the alignment of a block returned by malloc is sufficient to
satisfy the call).  In this case, an integer overflow in _int_malloc leads
to posix_memalign incorrectly returning successfully.

Upon fixing this and writing a somewhat thorough regression test, it was
discovered that when posix_memalign is called with an alignment larger than
MALLOC_ALIGNMENT (so it uses _int_memalign instead) and a requested size
close to SIZE_MAX, a different integer overflow in _int_memalign leads to
posix_memalign incorrectly returning successfully.

Both integer overflows affect other memory allocation functions that use
_int_malloc (one affected malloc in x86) or _int_memalign as well.

This commit fixes both integer overflows.  In addition to this, it adds a
regression test to guard against false successful allocations by the
following memory allocation functions when called with too-large allocation
sizes and, where relevant, various valid alignments:
malloc, realloc, calloc, reallocarray, memalign, posix_memalign,
aligned_alloc, valloc, and pvalloc.

6 years agoChangeLog: Fix an entry for [BZ #22657]
Rafal Luzynski [Thu, 18 Jan 2018 00:40:15 +0000 (01:40 +0100)]
ChangeLog: Fix an entry for [BZ #22657]

Remove myself, give the full credit to Egmont Koblinger.

When committing someone else's changes, one should put the contributor's
name in the change log entry.

6 years agolocales bho_NP, mai_IN, mai_NP: Fix an obvious typo in date.
Rafal Luzynski [Thu, 18 Jan 2018 00:27:10 +0000 (01:27 +0100)]
locales bho_NP, mai_IN, mai_NP: Fix an obvious typo in date.

* localedata/locales/bho_NP (LC_IDENTIFICATION): Fix an obvious typo
in date: "2017-24-07" should be "2017-07-24".
* localedata/locales/mai_IN: Likewise.
* localedata/locales/mai_NP: Likewise.

6 years agoUpdate translations from the Translation Project
Dmitry V. Levin [Wed, 17 Jan 2018 22:00:50 +0000 (22:00 +0000)]
Update translations from the Translation Project

* po/ru.po: Update translations.

6 years agoFix backtrace for hppa (bug 22719).
Joseph Myers [Wed, 17 Jan 2018 13:31:47 +0000 (13:31 +0000)]
Fix backtrace for hppa (bug 22719).

The only architecture in glibc that uses the generic debug/backtrace.c
is hppa.  The debug/tst-backtrace* tests fail for hppa, so in fact the
generic debug/backtrace.c is not functional anywhere.  Instead, the
x86_64 version is a reasonably generic version that uses
_Unwind_Backtrace from libgcc to backtrace using unwind info, and is
used by several architectures.  This patch adds hppa to the
architectures using it (leaving open the possibility of a subsequent
cleanup for 2.28 of moving the x86_64 version to debug/backtrace.c,
and removing all the frame.h files that are now unused).

Reported by Adhemerval in
<https://sourceware.org/ml/libc-alpha/2018-01/msg00564.html> that this
does fix the backtrace test failures for hppa.

[BZ #22719]
* sysdeps/hppa/backtrace.c: New file.

6 years agox86-64: Properly align La_x86_64_retval to VEC_SIZE [BZ #22715]
H.J. Lu [Wed, 17 Jan 2018 12:31:51 +0000 (04:31 -0800)]
x86-64: Properly align La_x86_64_retval to VEC_SIZE [BZ #22715]

_dl_runtime_profile calls _dl_call_pltexit, passing a pointer to
La_x86_64_retval which is allocated on stack.  The lrv_vector0
field in La_x86_64_retval must be aligned to size of vector register.
When allocating stack space for La_x86_64_retval, we need to make sure
that the address of La_x86_64_retval + RV_VECTOR0_OFFSET is aligned to
VEC_SIZE.  This patch checks the alignment of the lrv_vector0 field
and pads the stack space if needed.

Tested with x32 and x86-64 on SSE4, AVX and AVX512 machines.  It fixed

FAIL: elf/tst-audit10
FAIL: elf/tst-audit4
FAIL: elf/tst-audit5
FAIL: elf/tst-audit6
FAIL: elf/tst-audit7

on x32 AVX512 machine.

[BZ #22715]
* sysdeps/x86_64/dl-trampoline.h (_dl_runtime_profile): Properly
align La_x86_64_retval to VEC_SIZE.

6 years agoUse LIBGCC_S_SO in x86_64 backtrace.
Joseph Myers [Tue, 16 Jan 2018 20:53:03 +0000 (20:53 +0000)]
Use LIBGCC_S_SO in x86_64 backtrace.

The x86_64 backtrace implementation is used as a generic
implementation (unwinding via unwind info and _Unwind_Backtrace) by
various other architectures.  This patch makes it more generic by
making it use LIBGCC_S_SO from gnu/lib-names.h instead of hardcoding
the libgcc_s.so.1 name, so that it can also be used on hppa which uses
libgcc_s.so.4.

Tested for x86_64.

* sysdeps/x86_64/backtrace.c: Include <gnu/lib-names.h>.
(init): Use LIBGCC_S_SO not hardcoded "libgcc_s.so.1".

6 years agonptl/tst-thread-exit-clobber: Run with any C++ compiler
Florian Weimer [Tue, 16 Jan 2018 15:46:00 +0000 (16:46 +0100)]
nptl/tst-thread-exit-clobber: Run with any C++ compiler

We do not need thread_local support in the C++11 comiler, and the
minimum GCC version for glibc has C++11 support (if it has C++ support).

6 years agoaarch64: Update bits/hwcap.h for Linux 4.15.
Szabolcs Nagy [Thu, 11 Jan 2018 17:08:51 +0000 (17:08 +0000)]
aarch64: Update bits/hwcap.h for Linux 4.15.

Define new HWCAP bits and add their name to dl-procinfo.c following
the linux definitions. Synchronizing with v4.15-rc8 version of linux,
these are not expected to change before the 4.15 release.

* sysdeps/unix/sysv/linux/aarch64/bits/hwcap.h (HWCAP_SHA3): Define.
(HWCAP_SM3, HWCAP_SM4, HWCAP_ASIMDDP, HWCAP_SHA512, HWCAP_SVE): Define.
* sysdeps/unix/sysv/linux/aarch64/dl-procinfo.c
(_dl_aarch64_cap_flags): Update.
(_DL_HWCAP_COUNT): Update.

6 years agoaarch64: make HWCAP updates less error prone
Szabolcs Nagy [Thu, 11 Jan 2018 17:21:46 +0000 (17:21 +0000)]
aarch64: make HWCAP updates less error prone

Remove unused _DL_HWCAP_LAST definition and move _DL_HWCAP_COUNT
where it is needed (dl-procinfo.h always includes dl-procinfo.c).

* sysdeps/unix/sysv/linux/aarch64/dl-procinfo.h
(_DL_HWCAP_LAST): Remove.
(_DL_HWCAP_COUNT): Move to ...
* sysdeps/unix/sysv/linux/aarch64/dl-procinfo.c
(_DL_HWCAP_COUNT): ... here.

6 years agonptl/tst-minstack-throw: Compile in C++11 mode with GNU extensions
Florian Weimer [Tue, 16 Jan 2018 06:19:28 +0000 (07:19 +0100)]
nptl/tst-minstack-throw: Compile in C++11 mode with GNU extensions

6 years agoAdd NT_ARM_SVE to elf.h
Alan Hayward [Fri, 12 Jan 2018 16:56:19 +0000 (16:56 +0000)]
Add NT_ARM_SVE to elf.h

This definition is for AArch64 SVE registers in elf core dumps.

* elf/elf.h (NT_ARM_SVE): Define.

6 years agonptl: Add PTHREAD_MIN_STACK C++ throw test [BZ #22636]
Florian Weimer [Mon, 15 Jan 2018 14:30:00 +0000 (15:30 +0100)]
nptl: Add PTHREAD_MIN_STACK C++ throw test [BZ #22636]

6 years agoUpdate build-many-glibcs.py binutils, mpfr, mpc versions.
Joseph Myers [Mon, 15 Jan 2018 14:02:54 +0000 (14:02 +0000)]
Update build-many-glibcs.py binutils, mpfr, mpc versions.

* scripts/build-many-glibcs.py (Context.checkout): Default
binutils version to 2.30 branch, MPFR version to 4.0.0 and MPC
version to 1.1.0.

6 years agoSynchronize DF_1_* flags with binutils (Bug 22707)
Carlos O'Donell [Sun, 14 Jan 2018 04:21:57 +0000 (20:21 -0800)]
Synchronize DF_1_* flags with binutils (Bug 22707)

This patch synchronizes DF_1_* flags with binutils
and ensures that all DF_1_* flags defined in binutil's
include/elf/common.h are also defined glibc's elf/elf.h.
This is a user visible change since elf/elf.h is installed
by default as /usr/include/elf.h.

Signed-off-by: Carlos O'Donell <carlos@redhat.com>
6 years agopowerpc: Fix llround spurious inexact on 32-bit POWER4 [BZ #22697]
Tulio Magno Quites Machado Filho [Fri, 12 Jan 2018 22:39:42 +0000 (20:39 -0200)]
powerpc: Fix llround spurious inexact on 32-bit POWER4 [BZ #22697]

This issue is similar to BZ #19235, where spurious exceptions are
created from adding 0.5 then converting to an integer.
The solution is based on Joseph's fix for BZ #19235.

[BZ #22697]
* sysdeps/powerpc/powerpc32/power4/fpu/s_llround.S (__llround):
Do not add 0.5 to integer or out-of-range arguments.

6 years agoUpdate translations from the Translation Project
Dmitry V. Levin [Thu, 11 Jan 2018 15:00:00 +0000 (15:00 +0000)]
Update translations from the Translation Project

* po/bg.po: Update translations.
* po/cs.po: Likewise.
* po/de.po: Likewise.
* po/ko.po: Likewise.
* po/pl.po: Likewise.
* po/sv.po: Likewise.
* po/uk.po: Likewise.
* po/vi.po: Likewise.

6 years agoaarch64: fix static pie enabled libc when main is in a shared library
Szabolcs Nagy [Tue, 19 Dec 2017 18:43:04 +0000 (18:43 +0000)]
aarch64: fix static pie enabled libc when main is in a shared library

In the static pie enabled libc, crt1.o uses the same position independent
code as rcrt1.o and crt1.o is used instead of Scrt1.o when -no-pie
executables are linked.  When main is not defined in the executable, but
in a shared library crt1.o is currently broken, it assumes main is local.
(glibc has a test for this but i missed it in my previous testing.)

To make both rcrt1.o and crt1.o happy with the same code, a wrapper is
introduced around main: with this crt1.o works with extern main symbol
while rcrt1.o does not depend on GOT relocations. (The change only
affects static pie enabled libc. Further simplification of start.S is
possible in the future by using the same approach for Scrt1.o too.)

* aarch64/start.S (_start): Use __wrap_main.
(__wrap_main): New local symbol.

6 years agolinux: make getcwd(3) fail if it cannot obtain an absolute path [BZ #22679]
Dmitry V. Levin [Sun, 7 Jan 2018 02:03:41 +0000 (02:03 +0000)]
linux: make getcwd(3) fail if it cannot obtain an absolute path [BZ #22679]

Currently getcwd(3) can succeed without returning an absolute path
because the underlying getcwd syscall, starting with linux commit
v2.6.36-rc1~96^2~2, may succeed without returning an absolute path.

This is a conformance issue because "The getcwd() function shall
place an absolute pathname of the current working directory
in the array pointed to by buf, and return buf".

This is also a security issue because a non-absolute path returned
by getcwd(3) causes a buffer underflow in realpath(3).

Fix this by checking the path returned by getcwd syscall and falling
back to generic_getcwd if the path is not absolute, effectively making
getcwd(3) fail with ENOENT.  The error code is chosen for consistency
with the case when the current directory is unlinked.

[BZ #22679]
CVE-2018-1000001
* sysdeps/unix/sysv/linux/getcwd.c (__getcwd): Fall back to
generic_getcwd if the path returned by getcwd syscall is not absolute.
* io/tst-getcwd-abspath.c: New test.
* io/Makefile (tests): Add tst-getcwd-abspath.

6 years agomalloc: Ensure that the consolidated fast chunk has a sane size.
Istvan Kurucsai [Fri, 12 Jan 2018 14:26:20 +0000 (15:26 +0100)]
malloc: Ensure that the consolidated fast chunk has a sane size.

6 years agosupport: Preserve errno in write_message, TEST_VERIFY and other checks
Florian Weimer [Fri, 12 Jan 2018 07:18:30 +0000 (08:18 +0100)]
support: Preserve errno in write_message, TEST_VERIFY and other checks

These facilities could clobber errno, which makes it difficult to write
certain checks because a specific order has to be used.

6 years agoAdd missing reference to bug 20532
Florian Weimer [Fri, 12 Jan 2018 11:15:38 +0000 (12:15 +0100)]
Add missing reference to bug 20532

6 years agolibnsl: Do not install libnsl.so, libnsl.a if NIS is disabled [BZ #22701]
Florian Weimer [Fri, 12 Jan 2018 07:42:56 +0000 (08:42 +0100)]
libnsl: Do not install libnsl.so, libnsl.a if NIS is disabled [BZ #22701]

This also skips building the .o files for libnsl.a.

Reviewed-by: Carlos O'Donell <carlos@redhat.com>
6 years agohu_HU locale: Avoid double space (bug 22657).
Egmont Koblinger [Mon, 1 Jan 2018 20:40:11 +0000 (20:40 +0000)]
hu_HU locale: Avoid double space (bug 22657).

The current date format prefixes one-digit days with a space, resulting
in ugly two spaces:

$ LC_ALL=hu_HU.UTF-8 date
2018. jan.  1., hétfő, 21:25:35 CET
          ^^

The official orthography rules doesn't contain an explicit rule about
this (which already gives no sane reason for double space), and an
implicit example of "1848. március 9." under bullet point 296 at
http://helyesiras.mta.hu/helyesiras/default/akh12 contains a single
space only. It's sure not convincing on an HTML page, but I confirm
that the official book edition (e.g.
https://www.libri.hu/en/konyv/a-magyar-helyesiras-szabalyai-32.html)
also contains a single space there.

[BZ #22657]
* localedata/locales/hu_HU (d_t_fmt): Avoid a leading space
before the day number which may produce a double space.
(date_fmt): Likewise.

6 years agoFix s390 linknamespace fallout of bug 22702 patch.
Joseph Myers [Fri, 12 Jan 2018 00:12:57 +0000 (00:12 +0000)]
Fix s390 linknamespace fallout of bug 22702 patch.

My fix for bug 22702 introduced linknamespace test failures on
s390x-linux-gnu and s390-linux-gnu because it made remainder call
__feholdexcept, and the s390 __feholdexcept calls fegetenv, and
remainder is in Unix98 and XPG4.2 but fegetenv isn't.  This patch
makes __feholdexcept call __fegetenv instead to avoid that namespace
issue.

Tested (compilation) with build-many-glibcs.py for s390x-linux-gnu,
where it resolves the test failures.

* sysdeps/s390/fpu/feholdexcpt.c (__feholdexcept): Call __fegetenv
instead of fegetenv.

6 years agoMake default libc_feholdsetround_noex_ctx use __feholdexcept (bug 22702).
Joseph Myers [Thu, 11 Jan 2018 18:18:46 +0000 (18:18 +0000)]
Make default libc_feholdsetround_noex_ctx use __feholdexcept (bug 22702).

For soft-float powerpc, the math/test-nearbyint-except-2 test fails
because nearbyintl traps when traps on "inexact" are enabled on entry
(and an "inexact" exception is generated internally, though cleared
for the final return).

The problem is the default implementation of
libc_feholdsetround_noex_ctx, which does not disable exception traps.
There is some ambiguity about whether the *noex* interfaces are
required to do so or only permitted to do so.  But given that we
support fe* interfaces to enable and disable traps (on architectures
with that functionality), functions that must not raise an exception
(must not leave the flag set on exit if not set on entry) should also
not trap on it when traps on that exception are enabled.  So it is
appropriate to define these interfaces to have the feholdexcept effect
of disabling exception traps; this patch updates the default
implementation and comments accordingly.

At least some architecture versions already disable traps; there are
few uses of the *noex* interfaces at all, and while it's possible
there are bugs on any architecture versions failing to disable traps
that appear in the exp2 and remainder implementations, there are
currently no tests, other than this one for nearbyintl (where only the
ldbl-128ibm implementation uses SET_RESTORE_ROUND_NOEX), that would
fail as a result of such a bug.  (Hard-float powerpc does disable
traps here, hence the nearbyintl failure not appearing there.)

Tested for powerpc (soft-float).  This brings that configuration to
clean math/ test results, provided you build with GCC 8 to get the fix
for GCC bug 64811.

[BZ #22702]
* sysdeps/generic/math_private.h (libc_feresetround_noex): Update
comment to say exceptions are discarded.
(libc_feholdsetround_noex_ctx): Use __feholdexcept instead of
__fegetenv.
(SET_RESTORE_ROUND_NOEX): Update comment to say non-stop mode must
be enabled.

6 years agocsu: Update __libgcc_s_init comment
Florian Weimer [Thu, 11 Jan 2018 12:13:28 +0000 (13:13 +0100)]
csu: Update __libgcc_s_init comment

Reviewed-by: Carlos O'Donell <carlos@redhat.com>
6 years agonptl: Add tst-minstack-cancel, tst-minstack-exit [BZ #22636]
Florian Weimer [Thu, 11 Jan 2018 12:13:14 +0000 (13:13 +0100)]
nptl: Add tst-minstack-cancel, tst-minstack-exit [BZ #22636]

I verified that without the guard accounting change in commit
630f4cc3aa019ede55976ea561f1a7af2f068639 (Fix stack guard size
accounting) and RTLD_NOW for libgcc_s introduced by commit
f993b8754080ac7572b692870e926d8b493db16c (nptl: Open libgcc.so with
RTLD_NOW during pthread_cancel), the tst-minstack-cancel test fails on
an AVX-512F machine.  tst-minstack-exit still passes, and either of
the mentioned commit by itself frees sufficient stack space to make
tst-minstack-cancel pass, too.

Reviewed-by: Carlos O'Donell <carlos@redhat.com>
6 years agoFix ldbl-128ibm log1pl (-qNaN) spurious "invalid" exception (bug 22693).
Joseph Myers [Wed, 10 Jan 2018 17:59:01 +0000 (17:59 +0000)]
Fix ldbl-128ibm log1pl (-qNaN) spurious "invalid" exception (bug 22693).

The ldbl-128ibm implementation of log1pl does ordered comparisons on a
negative qNaN argument, so resulting in spurious "invalid" exceptions
(for soft-float powerpc; hard-float only avoids this because of GCC
bug 58684 meaning ordered comparison instructions never get
generated).  This patch fixes this by arranging for the test for NaN
or infinity arguments to handle negative arguments as well.

Tested for powerpc (soft float).

[BZ #22693]
* sysdeps/ieee754/ldbl-128ibm/s_log1pl.c (__log1pl): Handle
negative arguments in test for NaN or infinity argument.

6 years agoRegenerate libc.pot
Dmitry V. Levin [Wed, 10 Jan 2018 15:00:00 +0000 (15:00 +0000)]
Regenerate libc.pot

6 years agonptl: Open libgcc.so with RTLD_NOW during pthread_cancel [BZ #22636]
Florian Weimer [Wed, 10 Jan 2018 12:18:04 +0000 (13:18 +0100)]
nptl: Open libgcc.so with RTLD_NOW during pthread_cancel [BZ #22636]

Disabling lazy binding reduces stack usage during unwinding.

Note that RTLD_NOW only makes a difference if libgcc.so has not
already been loaded, so this is only a partial fix.

Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>