Fix tls support for glibc 2.25

* csu/libc-start.c (LIBC_START_MAIN) [__GNU__]: Do not call
__libc_setup_tls.
* sysdeps/mach/hurd/i386/init-first.c (init): Call __libc_setup_tls.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg2.25-tls.diff

cvs-libc_init_secure

commit 8ee1abdb4b2c92fb5f189aabec34d1b7d54ac332
Author: Samuel Thibault <samuel.thibault@ens-lyon.org>
Date:   Wed Aug 2 23:29:57 2017 +0200

    [hurd]: Add __libc_init_secure stub

    csu/libc-start.c now insists on calling __libc_init_secure, while the Hurd
    port already implements it "very early" in dl-sysdep.c and init-first.c

            * sysdeps/mach/hurd/enbl-secure.c (__libc_init_secure): Define
            function.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name cvs-libc_init_secure.diff

submitted-shm_open_pthread

Add missing include for pthread_setcancelstate

* sysdeps/posix/shm_open.c: Include <pthread.h>.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name submitted-shm_open_pthread.diff

local-no_unsupported_ioctls

These ioctls are not actually supported (and will probably not be in the
close future), and are not available on Linux either, so don't expose
them to application at the risk of them complaining that they don't work
(e.g. xterm using TIOCLSET).

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name local-no_unsupported_ioctls.diff

Add mlockall support

* sysdeps/mach/hurd/mlockall.c: New file
* sysdeps/mach/hurd/munlockall.c: New file

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-mlockall.diff

hurd: Handle `pid' magical lookup retry

        * hurd/lookup-retry.c: Handle `pid' magical lookup
        retry.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-magic-pid.diff

XXX: make libc_rwlock recursive

Without making the rwlocks recursive, running fakeroot-tcp gets this:

#0  0x0106e91c in mach_msg_trap () at /usr/src/glibc-2.24/build-tree/hurd-i386-libc/mach/mach_msg_trap.S:2
#1  0x0106f090 in __mach_msg (msg=0x20034a0, option=3, send_size=64, rcv_size=32, rcv_name=421, timeout=0, notify=0) at msg.c:110
#2  0x0125a241 in __gsync_wait (task=1, addr=19101080, val1=2, val2=0, msec=0, flags=0)
    at /usr/src/glibc-2.24/build-tree/hurd-i386-libc/mach/RPC_gsync_wait.c:175
#3  0x010b0743 in __dcigettext (domainname=0x8050740 <_libc_intl_domainname@@GLIBC_2.2.6> "libc",
    msgid1=0x8051d88 "undefined symbol: acl_get_fd", msgid2=0x0, plural=0, n=0, category=5) at dcigettext.c:527
#4  0x010af776 in __dcgettext (domainname=0x8050740 <_libc_intl_domainname@@GLIBC_2.2.6> "libc",
    msgid=0x8051d88 "undefined symbol: acl_get_fd", category=5) at dcgettext.c:47
#5  0x0124e427 in __dlerror () at dlerror.c:94
#6  0x01035ae3 in load_library_symbols () from /usr/lib/i386-gnu/libfakeroot/libfakeroot-tcp.so
#7  0x01035cc3 in tmp___fxstat64 () from /usr/lib/i386-gnu/libfakeroot/libfakeroot-tcp.so
#8  0x01036cd6 in __fxstat64 () from /usr/lib/i386-gnu/libfakeroot/libfakeroot-tcp.so
#9  0x010ad831 in _nl_load_locale_from_archive (category=category@entry=0, namep=namep@entry=0x200399c) at loadarchive.c:211
#10 0x010ac45b in _nl_find_locale (locale_path=0x0, locale_path_len=0, category=category@entry=0, name=0x200399c) at findlocale.c:154
#11 0x010abde7 in setlocale (category=0, locale=0x804c2e4 "") at setlocale.c:417
#12 0x0804947f in main (argc=2, argv=0x2003ad4) at programs/locale.c:191

That's very unfortunate: libfakeroot gets initialized from a section
where __libc_setlocale_lock is already locked, and thus the dlerror()
call hangs inside __dcigettext. It happens that Linux doesn't have
the problem probably because pthread_rwlock_wrlock returns a EDEADLK
error instead of hanging, and then the first unlock unlocks, and the
second unlock probably returns an EINVAL. This is all very unsafe, but
that's fakeroot-tcp's matter (see http://bugs.debian.org/845930 for the
follow-up)...

We only use it when constructing the debian installer for -s -r options
anyway.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-libc_rwlock_recursive.diff

hurd: Fix O_DIRECTORY | O_NOFOLLOW

Appending / to the path to be looked up makes us always follow a final
symlink, even with O_NOTRANS (since the final resolution is after the
'/').  In the O_DIRECTORY | O_NOFOLLOW case, we thus have to really open
the node and stat it, which we already do anyway, and check for
directory type.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-NOFOLLOW-DIRECTORY.diff

hurd: Fix O_NOFOLLOW

The error code documented by POSIX for opening a symlink with O_NOFOLLOW
is ELOOP.

Also, if the translator does not expose symlink as a symlink translator but
as a S_IFLNK file, O_NOFOLLOW needs to return ELOOP too.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-NOFOLLOW.diff

tg-libpthread-gsync-mutex

commit f6839074d25aba7e93b8672a9ed9893c0baa5e3d
Author: Agustina Arzille <avarzille@riseup.net>
Date:   Mon Oct 17 00:56:58 2016 +0200

    Make pthread_mutex use gsync

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-libpthread-gsync-mutex.diff

Fix SS_ONSTACK support

* sysdeps/mach/hurd/i386/sigreturn.c (__sigreturn2): New function,
unlocks SS and returns to the saved PC.
(__sigreturn): Do not unlock SS, and "return" into __sigreturn2 on the
thread stack instead of the saved PC.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-ONSTACK.diff

tg-libpthread-gsync-spin

commit cd7e69c545200949cfd1b3c595568556ff9a76e3
Author: Agustina Arzille <avarzille@riseup.net>
Date:   Tue Oct 18 00:20:45 2016 +0200

    Make pthread_spinlock use gsync

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-libpthread-gsync-spin.diff

t/pthread_deps

libpthread needs some internal symbols to avoid exposing the plain ones

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-pthread_deps.diff

tg-gsync-libc

t/gsync-libc
From: Agustina Arzille <avarzille@riseup.net>
Subject: [PATCH] Introduce gsync-based locks to glibc.

* hurd/Makefile: Add hurdlock.
* hurd/Versions: Added new entry to export the above interface.
* hurd/hurdlock.c: New file.
* hurd/hurdlock.h: New file.
* hurd/hurdpid.c: Include <lowlevellock.h>
  (_S_msg_proc_newids): Use lll_wait to synchronize.
* hurd/hurdsig.c: (reauth_proc): Use __mutex_lock and __mutex_unlock.
* hurd/setauth.c: Include <hurdlock.h>, use integer for synchronization.
* hurd/sysvshm.c: Include <hurdlock.h>, use integer for synchronization.
* mach/Makefile: Remove unneeded file
* mach/lock-intern.h: Use lll to implement spinlocks.
* mach/lowlevellock.h: New file
* mach/mutex-init.c: Rewrite mutex initialization.
* sysdeps/mach/Makefile: Add libmachuser as dependencies for some libs.
* sysdeps/mach/libc-lock.h: Reimplemented libc internal locks
  with lll, cleanup routines now use gcc's cleanup attribute
* sysdeps/mach/hurd/bits/errno.h: New errno values.
* sysdeps/mach/hurd/libc-lock.h: Removed file.
* sysdeps/mach/hurd/malloc-machine.h: Reimplemented malloc locks.
* sysdeps/mach/hurd/setpgid.c: (setpgid): Use gsync for synchronization.
* sysdeps/mach/hurd/setsid.c: (setsid): Likewise.

t/libc_cleanup
From: Agustina Arzille <avarzille@riseup.net>
Subject: [PATCH] Use gcc's cleanup attributes

cleanup routines now use gcc's cleanup attribute

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-gsync-libc.diff

tg-eintr

commit 230b85f414291ac955827aba15cfbd103ab6ebdd
Author: Samuel Thibault <samuel.thibault@ens-lyon.org>
Date:   Thu Jun 9 01:15:10 2016 +0200

    Fix pipe() call returning EINTR sometimes

    because it uses a critical section

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-eintr.diff

hurd: take __USE_EXTERN_INLINES into account

* 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_get, _hurd_fd_error_signal, _hurd_fd_error,
__hurd_dfail, __hurd_sockfail): Likewise.
* mach/lock-intern.h (__spin_lock_init, __spin_lock, __mutex_lock,
__mutex_unlock, __mutex_trylock): Likewise.
* mach/mach/mig_support.h (__mig_strncpy): Likewise.
* sysdeps/generic/machine-lock.h (__spin_unlock, __spin_try_lock,
__spin_lock_locked): Likewise.
* sysdeps/generic/machine-sp.h (__thread_stack_pointer): Likewise.
* sysdeps/mach/i386/machine-lock.h (__spin_unlock, __spin_try_lock,
__spin_lock_locked): Likewise.

* 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 version
only if __USE_EXTERN_INLINES is defined and _LIBC is defined and
NOT_IN_libc is not defined.
* 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/spin-lock.c (__USE_EXTERN_INLINES): Define to 1.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-extern_inline.diff

Implement faccessat without AT_EACCESS flag

* hurd/hurd/fd.h: Include <fcntl.h>
(__hurd_at_flags): New function.
* hurd/lookup-at.c (__file_name_lookup_at): Replace flag computation
with call to __hurd_at_flags.
* include/unistd.h (__faccessat): Add declaration.
* sysdeps/mach/hurd/access.c (__access): Move implementation to
__faccessat, and replace it with a call to __faccessat.
* sysdeps/mach/hurd/euidaccess.c (__euidaccess): Replace implementation
with a call to __faccessat.
* sysdeps/mach/hurd/faccessat.c (faccessat): Rename into...
(__faccessat): ... this. Move implementation of __access into it when
AT_FLAGS does not contain AT_EACCESS. Make it call __hurd_at_flags, add
reauthenticate_cwdir_at helper to implement AT mechanism.
(faccessat): Define weak alias.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-faccessat.diff

local-versions

We introduced them in 2.21 in Debian

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name local-versions.diff

submitted-hurd-abilist

2016-02-29  Aurelien Jarno  <aurelien@aurel32.net>

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

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name submitted-hurd-abilist.diff

local-mach_print

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name local-mach_print.diff

tg-bootstrap

This dependency is missing, but would pose problem on Darwin

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-bootstrap.diff

libpthread_pthread_types

===================================================================

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name libpthread_pthread_types.diff

tg-allocalim

===================================================================

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-allocalim.diff

cvs-revert-gnu-gnu-cleanup

commit ebf27d12e602b428a316c105ed10371ed84d2d3d
Author: Samuel Thibault <samuel.thibault@ens-lyon.org>
Date:   Wed Feb 11 23:27:50 2015 +0000

     revert part of ba90e05.

     * configure.ac, configure: Revert ba90e05: modify gnu-* host_os back
     into gnu-gnu, and update comment to refer to abi-tags.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name cvs-revert-gnu-gnu-cleanup.diff

Lookup the startup server through /servers/startup

* sysdeps/mach/hurd/reboot.c: Include <hurd/paths.h>
(reboot): Lookup _SERVERS_STARTUP instead of calling proc_getmsgport to get a
port to the startup server.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-reboot-startup.diff

Make F_RDLCK/F_WRLCK atomic

lockf(LOCK_EX) would for instance drop any existing shared lock before taking
the exclusive lock. F_RDLCK/F_WRLCK need atomic changes, so introduce and use
__LOCK_ATOM

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
* misc/sys/file.h (__LOCK_ATOMIC): New macro.
* sysdeps/mach/hurd/fcntl.c (__libc_fcntl): Use __LOCK_ATOMIC along LOCK_SH and
LOCK_EX.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-WRLCK-upgrade.diff

local-libpthread-stacksize

Set the default stack size to 8MiB like on Linux, to avoid surprises
with packages which assume the Linuxish default.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name local-libpthread-stacksize.diff

submitted-bind_umask2

2014-08-27  Samuel Thibault  <samuel.thibault@ens-lyon.org>

Fix bind when umask is e.g. 0777.

        * sysdeps/mach/hurd/bind.c (__bind): Pass mode 0666 to __dir_mkfile
        instead of final mode, so that call __ifsock_getsockaddr can always
        succeed, before calling __file_chmod to fix the mode according to umask,
        before calling __dir_link to show the file.

Part of the original fix was commited, the other hasn't been yet, see Roland's
"Harumph" reply to
https://sourceware.org/ml/libc-alpha/2014-08/msg00408.html

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name submitted-bind_umask2.diff

Hurd: fix port leak in TLS

Depending on whether the thread is the main thread or not, the threading
library can have trouble determining whether the thread reference in the
TCB is valid. The simple solution is to let the threading library
initialize the TCB, and use a temporary reference when initializing TLS.

* sysdeps/mach/hurd/i386/tls.h (_hurd_tls_init): Use a temporary thread
reference.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-tls_thread_leak.diff

Hurd: make sigstates hold a reference on thread ports

This change is required in order to correctly release per-thread
resources. Directly reusing the threading library reference isn't
possible since the sigstate is also used early in the main thread,
before threading is initialized.

* hurd/hurd/signal.h (_hurd_self_sigstate): Drop thread reference after
calling _hurd_thread_sigstate.
(_hurd_critical_section_lock): Likewise.
* hurd/hurdsig.c (_hurd_thread_sigstate): Add a reference on the thread.
(_hurd_sigstate_delete): Drop thread reference.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-sigstate_thread_reference.diff

Add futimens support

* sysdeps/mach/hurd/futimens.c: New file.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-futimens.diff

tg-libc_getspecific

Drop once updated to a libpthread which doesn't define __libc_getspecific

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-libc_getspecific.diff

pipe2

2008-12-17  Thomas Schwinge  <tschwinge@gnu.org>

pipe2 for GNU Hurd.
* sysdeps/mach/hurd/pipe2.c: New file, copy from pipe.c.  Evolve it to
implement __pipe2.
* sysdeps/mach/hurd/pipe.c (__pipe): Reimplement using __pipe2.
* sysdeps/mach/hurd/kernel-features.h (__ASSUME_PIPE2): Define.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-pipe2.diff

socketpair_flags

2008-12-17  Thomas Schwinge  <tschwinge@gnu.org>

SOCK_CLOEXEC and SOCK_NONBLOCK for socketpair on GNU Hurd.
* sysdeps/mach/hurd/socketpair.c (__socketpair): Handle SOCK_CLOEXEC
and SOCK_NONBLOCK.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-socketpair_flags.diff

socket_flags

2008-12-17  Thomas Schwinge  <tschwinge@gnu.org>

SOCK_CLOEXEC and SOCK_NONBLOCK for socket on GNU Hurd.
* sysdeps/mach/hurd/socket.c (__socket): Handle SOCK_CLOEXEC and
SOCK_NONBLOCK.
* sysdeps/mach/hurd/kernel-features.h (__ASSUME_SOCK_CLOEXEC): Define.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-socket_flags.diff

submitted-path_mounted

To: libc-alpha@sourceware.org
Subject: [PATCH 2/2] Define _PATH_MOUNTED as "/etc/mtab"
Date: Thu, 15 Aug 2013 10:43:35 +0200

Change the definition of _PATH_MOUNTED to "/etc/mtab". This is the
value used on Linux.

The change is motivated by the fact that on Debian /etc/mtab is a
symbolic link to /proc/mounts. This patch adjusts the macro for
non-linux systems such as Hurd. Changing this using
sysdeps/mach/hurd/paths.h causes build problems because
/usr/include/hurd/paths.h is shadowed by this file. This change is
proposed in the hope that aligning the non-linux targets with the
glibc for Linux is perceived as a good thing while fixing this problem
on Debian/Hurd along the way.

* sysdeps/generic/paths.h (_PATH_MOUNTED): Change value to "/etc/mtab".

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name submitted-path_mounted.diff

unsubmitted-clock_t_centiseconds

Some applications assume centisecond precision, or at most millisecond precision
(e.g. guile).  This is a work-around for them.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name unsubmitted-clock_t_centiseconds.diff

Fix poll and select POSIX compliancy details about errors

This fixes the following:

- On error, poll must not return without polling, including EBADF, and instead
report POLLHUP/POLLERR/POLLNVAL
- Select must report EBADF if some set contains an invalid FD.

The idea is to move error management to after all select calls, in the
poll/select final treatment. The error is instead recorded in a new `error'
field, and a new SELECT_ERROR bit set.

Thanks Svante Signell for the initial version of the patch.

* hurd/hurdselect.c (SELECT_ERROR): New macro.
(_hurd_select):
- Add `error' field to `d' structures array.
- If a poll descriptor is bogus, set EBADF, but continue with a zero timeout.
- Go through the whole fd_set, not only until _hurd_dtablesize. Return EBADF
there is any bit set above _hurd_dtablesize.
- Do not request io_select on bogus descriptors (SELECT_ERROR).
- On io_select request error, record the error.
- On io_select bogus reply, use EIO error code.
- On io_select bogus or error reply, record the error.
- Do not destroy reply port for bogus FDs.
- On error, make poll set POLLHUP in the EPIPE case, POLLNVAL in the EBADF
case, or else POLLERR.
- On error, make select simulated readiness.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-poll_errors_fixes.diff

Fix timeout handling in _hurd_select

Rely on servers to implement timeouts, so that very short values (including
0) don't make mach_msg return before valid replies can be received. The
purpose of this scheme is to guarantee a full client-server round-trip,
whatever the timeout value.

This change depends on the new io_select_timeout RPC being implemented by
servers.

* hurd/Makefile (user-interfaces): Add io_reply and io_request.
* hurd/hurdselect.c: Include <sys/time.h>, <hurd/io_request.h> and <limits.h>.
(_hurd_select): Replace the call to __io_select with either __io_select_request
or __io_select_timeout_request, depending on the timeout. Count the number of
ready descriptors (replies for which at least one type bit is set). Implement
the timeout locally when there is no file descriptor.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-io_select_timeout.diff

t/EGREGIOUS-fr

Fix grammar in french translation for EGREGIOUS

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-EGREGIOUS-fr.diff

Correction traduction « Computer bought the farm »

* po/fr.po: Fix EIEIO french translation.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-EIEIO-fr.diff

Fix boot with statically-linked exec server.

* sysdeps/mach/hurd/i386/init-first.c (init): Also find ELF headers by oneself
when the pointer given in D is nul (as set by ext2fs).

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-exec-static.diff

Allow unknown root directory.

To be efficient, the remap translator simply returns ports from the underlying
filesystem, and thus the root directory found through browsing '..' is the
underlying root, not the remap root. This should not be a reason for getcwd to
fail.

* sysdeps/mach/hurd/getcwd.c (_hurd_canonicalize_directory_name_internal): Do
not remove the heading slash if we got an unknown root directory.
(__getcwd): Do not fail with EGRATUITOUS if we got an unknown root directory.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-remap_getcwd.diff

Workaround to add IPv6 support to getifaddrs

ifreq only contains sockaddr structures, which are not big enough for
IPv6 addresses. This takes another, ugly, approach, by parsing fsysopts
/servers/socket/2 options...

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-ifaddrs_v6.diff

Fix sbrk in PIE binaries

In PIE mode, the program gets loaded at very low address, and thus _end is very
low, just before libraries, and thus initializing the brk to it does not make
sense, since there is no room left there. Hardcode the brk to 0x8000000 for now
as a workaround.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-pie-sbrk.diff

tg-setitimer

    Fix Hurd implementation of setitimer.

    The preemptor sigcode doesn't match since the POSIX sigcode SI_TIMER is
    used when SIGALRM is sent. In addition, The inline version of
    hurd_preempt_signals doesn't update _hurdsig_preempted_set. For these
    reasons, the preemptor would be skipped by post_signal.

    * sysdeps/mach/hurd/setitimer.c (setitimer_locked): Fix preemptor setup.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-setitimer.diff

tg-single-select-timeout

commit fffcbbb0b5ff8864086f7b58fdbb8aa4c71943d0
Author: Richard Braun <rbraun@sceen.net>
Date:   Mon Dec 17 09:57:41 2012 +0000

    Fix _hurd_select for single fd sets

    The function attempts to optimize this case by performing one IPC system call
    with the timeout included among the parameters, but in the absence of a reply,
    it will call mach_msg again with the same timeout later, effectively doubling
    the total timeout of the select/poll call.

    Remove this optimization for the time being.

    * hurd/hurdselect.c (_hurd_select): Always call __io_select with no timeout.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-single-select-timeout.diff

tg-hurdsig-boot-fix

2012-09-05  Richard Braun  <rbraun@sceen.net>

       * hurd/hurdsig.c (sigstate_is_global_rcv): Do not return true
       if _HURD_GLOBAL_SIGSTATE is null.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-hurdsig-boot-fix.diff

unsubmitted-gnumach.defs

Build stubs for gnumach.defs

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name unsubmitted-gnumach.defs.diff

Use the new file_exec_paths RPC

Pass the file paths of executable to the exec server, both relative and
absolute, which exec needs to properly execute and avertise #!-scripts.
Previously, the exec server tried to guess the name from argv[0] but argv[0]
only contains the executable name by convention.

2010-08-04  Emilio Pozuelo Monfort  <pochu27@gmail.com>
2017-09-26  Svante Signell  <svante.signell@gmail.com>

* hurd/hurdexec.c (_hurd_exec): Deprecate function.
(_hurd_exec_paths): New function.
* hurd/hurd.h (_hurd_exec): Deprecate function.
(_hurd_exec_paths): Declare function.
* hurd/Versions: Export _hurd_exec_paths.
* sysdeps/mach/hurd/execve.c: Include <stdlib.h> and <stdio.h>
(__execve): Use __getcwd to build absolute path, and use
_hurd_exec_paths instead of _hurd_exec.
* sysdeps/mach/hurd/spawni.c: Likewise.
* sysdeps/mach/hurd/fexecve.c: Use _hurd_exec_paths instead of
_hurd_exec.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name git-exec_filename.diff

libpthread_version

This dates when libpthread was moved to glibc and some features were added.

2.13-31/38/39 dates when the global signal disposition call was added.
TODO: _DEBIAN_ in versions however pose problem. Remove the _DEBIAN_ version
once packages are rebuilt against 2.21.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name libpthread_version.diff

t/libpthread_depends

libpthread needs to be built before modules which need it.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-libpthread_depends.diff

libpthread_build

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name libpthread_build.diff

cvs-libpthread

lsdiff -p2 cvs-libpthread.diff > orderfile
git diff empty --patch-with-stat --src-prefix=glibc/libpthread/ --dst-prefix=glibc/libpthread/ -Oorderfile 64727f50e23a744e1607cb6aff72f0813d56c6c9
from git.savannah.gnu.org:/srv/git/hurd/libpthread.git/
thus by construction only libpthread/ files, thus hurd-only
Date: Tue Sep 26 21:29:14 2017 +0200

 ChangeLog                                     |   6 +
 Makeconfig                                    |  13 +
 Makefile                                      | 365 +++++++++++
 TODO                                          | 131 ++++
 Versions                                      | 155 +++++
 configure                                     |   2 +
 configure.in                                  |   4 +
 forward.c                                     | 281 +++++++++
 sysdeps/pthread/pthread.h                     |  22 +
 include/pthread/pthread.h                     | 835 ++++++++++++++++++++++++++
 include/pthread/pthreadtypes.h                | 125 ++++
 libc_pthread_init.c                           |  35 ++
 libpthread.a                                  |  22 +
 libpthread_pic.a                              |  22 +
 lockfile.c                                    |  65 ++
 not-in-libc.h                                 |  12 +
 pthread/Versions                              |  15 +
 pthread/alloca_cutoff.c                       |  27 +
 pthread/cthreads-compat.c                     | 102 ++++
 pthread/pt-alloc.c                            | 219 +++++++
 pthread/pt-cancel.c                           |  63 ++
 pthread/pt-cleanup.c                          |  28 +
 pthread/pt-create.c                           | 254 ++++++++
 pthread/pt-dealloc.c                          |  69 +++
 pthread/pt-detach.c                           |  80 +++
 pthread/pt-exit.c                             | 112 ++++
 pthread/pt-getattr.c                          |  52 ++
 pthread/pt-initialize.c                       |  88 +++
 pthread/pt-internal.h                         | 330 ++++++++++
 pthread/pt-join.c                             |  76 +++
 pthread/pt-self.c                             |  34 ++
 pthread/pt-setcancelstate.c                   |  47 ++
 pthread/pt-setcanceltype.c                    |  47 ++
 pthread/pt-sigmask.c                          |  33 +
 pthread/pt-spin-inlines.c                     |  34 ++
 pthread/pt-testcancel.c                       |  36 ++
 pthread/pt-yield.c                            |  26 +
 pthreadP.h                                    |  27 +
 shlib-versions                                |   1 +
 sysdeps/generic/fork.h                        |  29 +
 sysdeps/generic/old_pt-atfork.c               |  27 +
 sysdeps/generic/pt-atfork.c                   |  34 ++
 sysdeps/generic/pt-attr-destroy.c             |  28 +
 sysdeps/generic/pt-attr-getdetachstate.c      |  31 +
 sysdeps/generic/pt-attr-getguardsize.c        |  29 +
 sysdeps/generic/pt-attr-getinheritsched.c     |  31 +
 sysdeps/generic/pt-attr-getschedparam.c       |  34 ++
 sysdeps/generic/pt-attr-getschedpolicy.c      |  31 +
 sysdeps/generic/pt-attr-getscope.c            |  31 +
 sysdeps/generic/pt-attr-getstack.c            |  32 +
 sysdeps/generic/pt-attr-getstackaddr.c        |  29 +
 sysdeps/generic/pt-attr-getstacksize.c        |  29 +
 sysdeps/generic/pt-attr-init.c                |  29 +
 sysdeps/generic/pt-attr-setdetachstate.c      |  40 ++
 sysdeps/generic/pt-attr-setguardsize.c        |  29 +
 sysdeps/generic/pt-attr-setinheritsched.c     |  40 ++
 sysdeps/generic/pt-attr-setschedparam.c       |  40 ++
 sysdeps/generic/pt-attr-setschedpolicy.c      |  44 ++
 sysdeps/generic/pt-attr-setscope.c            |  43 ++
 sysdeps/generic/pt-attr-setstack.c            |  51 ++
 sysdeps/generic/pt-attr-setstackaddr.c        |  29 +
 sysdeps/generic/pt-attr-setstacksize.c        |  30 +
 sysdeps/generic/pt-attr.c                     |  41 ++
 sysdeps/generic/pt-barrier-destroy.c          |  27 +
 sysdeps/generic/pt-barrier-init.c             |  53 ++
 sysdeps/generic/pt-barrier-wait.c             |  69 +++
 sysdeps/generic/pt-barrier.c                  |  26 +
 sysdeps/generic/pt-barrierattr-destroy.c      |  27 +
 sysdeps/generic/pt-barrierattr-getpshared.c   |  29 +
 sysdeps/generic/pt-barrierattr-init.c         |  28 +
 sysdeps/generic/pt-barrierattr-setpshared.c   |  39 ++
 sysdeps/generic/pt-cond-brdcast.c             |  45 ++
 sysdeps/generic/pt-cond-destroy.c             |  29 +
 sysdeps/generic/pt-cond-init.c                |  47 ++
 sysdeps/generic/pt-cond-signal.c              |  43 ++
 sysdeps/generic/pt-cond-timedwait.c           | 178 ++++++
 sysdeps/generic/pt-cond-wait.c                |  39 ++
 sysdeps/generic/pt-cond.c                     |  29 +
 sysdeps/generic/pt-condattr-destroy.c         |  29 +
 sysdeps/generic/pt-condattr-getclock.c        |  31 +
 sysdeps/generic/pt-condattr-getpshared.c      |  29 +
 sysdeps/generic/pt-condattr-init.c            |  30 +
 sysdeps/generic/pt-condattr-setclock.c        |  52 ++
 sysdeps/generic/pt-condattr-setpshared.c      |  39 ++
 sysdeps/generic/pt-destroy-specific.c         |  28 +
 sysdeps/generic/pt-equal.c                    |  31 +
 sysdeps/generic/pt-getconcurrency.c           |  27 +
 sysdeps/generic/pt-getcpuclockid.c            |  35 ++
 sysdeps/generic/pt-getschedparam.c            |  32 +
 sysdeps/generic/pt-getspecific.c              |  27 +
 sysdeps/generic/pt-init-specific.c            |  27 +
 sysdeps/generic/pt-key-create.c               |  30 +
 sysdeps/generic/pt-key-delete.c               |  29 +
 sysdeps/generic/pt-key.h                      |  22 +
 sysdeps/generic/pt-kill.c                     |  33 +
 sysdeps/generic/pt-mutex-destroy.c            |  39 ++
 sysdeps/generic/pt-mutex-getprioceiling.c     |  30 +
 sysdeps/generic/pt-mutex-init.c               |  50 ++
 sysdeps/generic/pt-mutex-lock.c               |  37 ++
 sysdeps/generic/pt-mutex-setprioceiling.c     |  30 +
 sysdeps/generic/pt-mutex-timedlock.c          | 196 ++++++
 sysdeps/generic/pt-mutex-transfer-np.c        |  66 ++
 sysdeps/generic/pt-mutex-trylock.c            | 112 ++++
 sysdeps/generic/pt-mutex-unlock.c             | 108 ++++
 sysdeps/generic/pt-mutexattr-destroy.c        |  27 +
 sysdeps/generic/pt-mutexattr-getprioceiling.c |  30 +
 sysdeps/generic/pt-mutexattr-getprotocol.c    |  29 +
 sysdeps/generic/pt-mutexattr-getpshared.c     |  29 +
 sysdeps/generic/pt-mutexattr-gettype.c        |  28 +
 sysdeps/generic/pt-mutexattr-init.c           |  28 +
 sysdeps/generic/pt-mutexattr-setprioceiling.c |  30 +
 sysdeps/generic/pt-mutexattr-setprotocol.c    |  42 ++
 sysdeps/generic/pt-mutexattr-setpshared.c     |  39 ++
 sysdeps/generic/pt-mutexattr-settype.c        |  37 ++
 sysdeps/generic/pt-mutexattr.c                |  45 ++
 sysdeps/generic/pt-once.c                     |  45 ++
 sysdeps/generic/pt-rwlock-attr.c              |  26 +
 sysdeps/generic/pt-rwlock-destroy.c           |  29 +
 sysdeps/generic/pt-rwlock-init.c              |  45 ++
 sysdeps/generic/pt-rwlock-rdlock.c            |  33 +
 sysdeps/generic/pt-rwlock-timedrdlock.c       | 121 ++++
 sysdeps/generic/pt-rwlock-timedwrlock.c       | 104 ++++
 sysdeps/generic/pt-rwlock-tryrdlock.c         |  56 ++
 sysdeps/generic/pt-rwlock-trywrlock.c         |  46 ++
 sysdeps/generic/pt-rwlock-unlock.c            |  99 +++
 sysdeps/generic/pt-rwlock-wrlock.c            |  35 ++
 sysdeps/generic/pt-rwlockattr-destroy.c       |  27 +
 sysdeps/generic/pt-rwlockattr-getpshared.c    |  29 +
 sysdeps/generic/pt-rwlockattr-init.c          |  28 +
 sysdeps/generic/pt-rwlockattr-setpshared.c    |  39 ++
 sysdeps/generic/pt-setconcurrency.c           |  34 ++
 sysdeps/generic/pt-setschedparam.c            |  31 +
 sysdeps/generic/pt-setschedprio.c             |  29 +
 sysdeps/generic/pt-setspecific.c              |  27 +
 sysdeps/generic/pt-startup.c                  |  25 +
 sysdeps/generic/raise.c                       |  51 ++
 sysdeps/generic/sem-close.c                   |  32 +
 sysdeps/generic/sem-destroy.c                 |  38 ++
 sysdeps/generic/sem-getvalue.c                |  33 +
 sysdeps/generic/sem-init.c                    |  46 ++
 sysdeps/generic/sem-open.c                    |  32 +
 sysdeps/generic/sem-post.c                    |  62 ++
 sysdeps/generic/sem-timedwait.c               | 100 +++
 sysdeps/generic/sem-trywait.c                 |  42 ++
 sysdeps/generic/sem-unlink.c                  |  32 +
 sysdeps/generic/sem-wait.c                    |  32 +
 sysdeps/generic/shm-directory.h               |  31 +
 sysdeps/hurd/pt-destroy-specific.c            |  77 +++
 sysdeps/hurd/pt-getspecific.c                 |  39 ++
 sysdeps/hurd/pt-init-specific.c               |  31 +
 sysdeps/hurd/pt-key-create.c                  | 110 ++++
 sysdeps/hurd/pt-key-delete.c                  |  64 ++
 sysdeps/hurd/pt-key.h                         |  77 +++
 sysdeps/hurd/pt-kill.c                        |  52 ++
 sysdeps/hurd/pt-setspecific.c                 |  51 ++
 sysdeps/i386/bits/memory.h                    |  40 ++
 sysdeps/i386/bits/pt-atomic.h                 |  66 ++
 sysdeps/i386/bits/pthreadtypes-arch.h         |  21 +
 sysdeps/pthread/bits/thread-shared-types.h    |  24 +
 sysdeps/i386/machine-sp.h                     |  30 +
 sysdeps/i386/pt-machdep.h                     |  29 +
 sysdeps/mach/bits/spin-lock-inline.h          |  90 +++
 sysdeps/mach/bits/spin-lock.h                 |  38 ++
 sysdeps/mach/hurd/Implies                     |   2 +
 sysdeps/mach/hurd/bits/pthread-np.h           |  38 ++
 sysdeps/mach/hurd/i386/pt-machdep.c           |  83 +++
 sysdeps/mach/hurd/i386/pt-setup.c             | 110 ++++
 sysdeps/mach/hurd/pt-attr-setstackaddr.c      |  29 +
 sysdeps/mach/hurd/pt-attr-setstacksize.c      |  29 +
 sysdeps/mach/hurd/pt-docancel.c               |  66 ++
 sysdeps/mach/hurd/pt-hurd-cond-timedwait.c    | 169 ++++++
 sysdeps/mach/hurd/pt-hurd-cond-wait.c         |  41 ++
 sysdeps/mach/hurd/pt-sigstate-destroy.c       |  28 +
 sysdeps/mach/hurd/pt-sigstate-init.c          |  44 ++
 sysdeps/mach/hurd/pt-sigstate.c               |  80 +++
 sysdeps/mach/hurd/pt-sysdep.c                 |  98 +++
 sysdeps/mach/hurd/pt-sysdep.h                 |  67 +++
 sysdeps/mach/i386/bits/spin-lock-inline.h     |  98 +++
 sysdeps/mach/i386/bits/spin-lock.h            |  39 ++
 sysdeps/mach/pt-block.c                       |  39 ++
 sysdeps/mach/pt-spin.c                        |  36 ++
 sysdeps/mach/pt-stack-alloc.c                 |  69 +++
 sysdeps/mach/pt-thread-alloc.c                |  95 +++
 sysdeps/mach/pt-thread-start.c                |  51 ++
 sysdeps/mach/pt-thread-terminate.c            |  85 +++
 sysdeps/mach/pt-timedblock.c                  |  68 +++
 sysdeps/mach/pt-wakeup.c                      |  38 ++
 sysdeps/posix/pt-spin.c                       |  54 ++
 sysdeps/pthread/Makefile                      |   7 +
 sysdeps/pthread/bits/barrier-attr.h           |  32 +
 sysdeps/pthread/bits/barrier.h                |  39 ++
 sysdeps/pthread/bits/cancelation.h            |  51 ++
 sysdeps/pthread/bits/condition-attr.h         |  34 ++
 sysdeps/pthread/bits/condition.h              |  39 ++
 sysdeps/pthread/bits/mutex-attr.h             |  41 ++
 sysdeps/pthread/bits/mutex.h                  |  75 +++
 sysdeps/pthread/bits/once.h                   |  34 ++
 sysdeps/pthread/bits/pthread-np.h             |  27 +
 sysdeps/pthread/bits/pthread.h                |  38 ++
 sysdeps/pthread/bits/pthreadtypes.h           |  30 +
 sysdeps/pthread/bits/rwlock-attr.h            |  32 +
 sysdeps/pthread/bits/rwlock.h                 |  46 ++
 sysdeps/pthread/bits/semaphore.h              |  44 ++
 sysdeps/pthread/bits/thread-attr.h            |  47 ++
 sysdeps/pthread/bits/thread-specific.h        |  25 +
 sysdeps/pthread/flockfile.c                   |  32 +
 sysdeps/pthread/ftrylockfile.c                |  35 ++
 sysdeps/pthread/funlockfile.c                 |  33 +
 sysdeps/pthread/libc-lockP.h                  | 158 +++++
 sysdeps/pthread/pthread-functions.h           | 141 +++++
 sysdeps/pthread/semaphore.h                   |  81 +++
 tests/.cvsignore                              |   1 +
 tests/Makefile                                |  40 ++
 tests/README                                  |   6 +
 tests/test-1.c                                |  50 ++
 tests/test-10.c                               |  46 ++
 tests/test-11.c                               | 143 +++++
 tests/test-12.c                               |  29 +
 tests/test-13.c                               |  66 ++
 tests/test-14.c                               |  44 ++
 tests/test-15.c                               |  87 +++
 tests/test-16.c                               |  71 +++
 tests/test-17.c                               |  57 ++
 tests/test-2.c                                |  39 ++
 tests/test-3.c                                |  55 ++
 tests/test-4.c                                |  86 +++
 tests/test-5.c                                |  75 +++
 tests/test-6.c                                |  96 +++
 tests/test-7.c                                |  70 +++
 tests/test-8.c                                |  60 ++
 tests/test-9.c                                |  88 +++
 tests/test-__pthread_destroy_specific-skip.c  |  83 +++
 232 files changed, 13224 insertions(+)

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name cvs-libpthread.diff

t/hooks

Add link rules to sort hooks, otherwise they are not properly recorded

2012-04-21  Samuel Thibault  <samuel.thibault@ens-lyon.org>

        * Makerules (shlib.lds): Add hurd hooks sorting rules.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-hooks.diff

Small signal fixes

    22e7268 Hurd signals: fix sigwait for pending signals
    da8bf5e Hurd signals: clear the pending mask in fork and spawn
    8e87205 Hurd signals: don't drop blocked ignored signals

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-hurdsig-fixes-2.diff

implement SA_SIGINFO signal handlers.

    52baaca Hurd signals: Copy bits/sigaction.h
    4232c66 Hurd signals: SA_SIGINFO support
    1831cfe Hurd signals: Use POSIX sigcodes

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-hurdsig-SA_SIGINFO.diff

local-hurdsig-global-dispositions-version

jkoenig's work on signals

2.13-19 dates when the global signal disposition call was added.
TODO: _DEBIAN_ in versions however pose problem. Remove the _DEBIAN_ version
once packages are rebuilt against 2.21.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name local-hurdsig-global-dispositions-version.diff

Global signal dispositions.

Although they should not change the
default behaviors of signals for cthread programs, these patches add
new functions which can be used by libpthread to enable
POSIX-conforming behavior of signals on a per-thread basis.

YYYY-MM-DD  Jeremie Koenig  <jk@jk.fr.eu.org>

    e407ae3 Hurd signals: implement global signal dispositions
    38eb4b3 Hurd signals: provide a sigstate destructor
    344dfd6 Hurd signals: fix sigwait() for global signals
    fb055f2 Hurd signals: fix global untraced signals.

YYYY-MM-DD  Thomas Schwinge  <thomas@codesourcery.com>

* sysdeps/mach/hurd/fork.c (__fork): In the child, reinitialize
the global sigstate's lock.

This is work in progress.

This cures an issue that would very rarely cause a deadlock in the child
in fork, tries to unlock ss' critical section lock at the end of fork.
This will typically (always?) be observed in /bin/sh, which is not
surprising as that is the foremost caller of fork.

To reproduce an intermediate state, add an endless loop if
_hurd_global_sigstate is locked after __proc_dostop (cast through
volatile); that is, while still being in the fork's parent process.

When that triggers (use the libtool testsuite), the signal thread has
already locked ss (which is _hurd_global_sigstate), and is stuck at
hurdsig.c:685 in post_signal, trying to lock _hurd_siglock (which the
main thread already has locked and keeps locked until after
__task_create).  This is the case that ss->thread == MACH_PORT_NULL, that
is, a global signal.  In the main thread, between __proc_dostop and
__task_create is the __thread_abort call on the signal thread which would
abort any current kernel operation (but leave ss locked).  Later in fork,
in the parent, when _hurd_siglock is unlocked in fork, the parent's
signal thread can proceed and will unlock eventually the global sigstate.
In the client, _hurd_siglock will likewise be unlocked, but the global
sigstate never will be, as the client's signal thread has been configured
to restart execution from _hurd_msgport_receive.  Thus, when the child
tries to unlock ss' critical section lock at the end of fork, it will
first lock the global sigstate, will spin trying to lock it, which can
never be successful, and we get our deadlock.

Options seem to be:

  * Move the locking of _hurd_siglock earlier in post_signal -- but that
    may generally impact performance, if this locking isn't generally
    needed anyway?

    On the other hand, would it actually make sense to wait here until we
    are not any longer in a critical section (which is meant to disable
    signal delivery anway (but not for preempted signals?))?

  * Clear the global sigstate in the fork's child with the rationale that
    we're anyway restarting the signal thread from a clean state.  This
    has now been implemented.

Why has this problem not been observed before Jérémie's patches?  (Or has
it?  Perhaps even more rarely?)  In _S_msg_sig_post, the signal is now
posted to a *global receiver thread*, whereas previously it was posted to
the *designated signal-receiving thread*.  The latter one was in a
critical section in fork, so didn't try to handle the signal until after
leaving the critical section?  (Not completely analyzed and verified.)

Another question is what the signal is that is being received
during/around the time __proc_dostop executes.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-hurdsig-global-dispositions.diff

Signal code refactoring.

These patches should not change the current
behavior, although they do fix a few minor bugs which were made
apparent in the process. They are unchanged from my previous post
earlier this month.

    34f5960 _hurd_internal_post_signal: Split into more functions
    420eec9 _hurd_internal_post_signal: Scope variables more restrictively
    1f5accd _hurd_internal_post_signal: Split out inner functions
    1bb0a1d Hurd signals: refactor check_pending_signals
    1764465 Hurd signals: reindent
    28473d2 Hurd signals: make sigsuspend POSIX-conformant.
    26d091a Hurd signals: fix uninitialized value.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-hurdsig-fixes.diff

local-madvise_warn

Do not warn about madvise not being implemented. It does not have any real
semantic anyway, and that brings some -Werror FTBFS.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name local-madvise_warn.diff

t/gai_misc

Add a hurd version of threaded gai_misc.h

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-gai_misc.diff

t/posix_thread

Declare that we have a libpthread.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-posix_thread.diff

local-ED

This is a long funny story, but even if it's standard-compliant, it poses too
many problems.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name local-ED.diff

Add -lmachuser -lhurduser to libc.so on GNU/Hurd.

http://lists.gnu.org/archive/html/bug-hurd/2011-03/msg00112.html

2011-03-29  Samuel Thibault  <samuel.thibault@ens-lyon.org>

        * Makerules ($(inst_libdir)/libc.so): Add -lmachuser -lhurduser to
        libc.so on GNU/Hurd.

It's still unclear what we want to aim for.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-mach-hurd-link.diff

t/grantpt

    /media/Stalin/tmp/glibc.hurd.gnu-2/libc_pic.os: In function `pts_name':
    /media/data/home/thomas/tmp/source/glibc/git/glibc.hurd/login/../sysdeps/unix/grantpt.c:52: undefined reference to `__ptsname_internal'
    collect2: ld returned 1 exit status
    make[1]: *** [/media/Stalin/tmp/glibc.hurd.gnu-2/libc.so] Error 1

Cf. aa9890239a2aef81e64f3f22a31c7e01b6501f69

Need any of the checks (stat, etc.) that Linux' __ptsname_internal does?

Implement close_all_fds as in 139ee080b6b428240bf49f3e6361f3ac729f891a?

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-grantpt.diff

hurd: SCM_CREDS support

Svante Signell  <svante.signell@gmail.com>
Samuel Thibault  <samuel.thibault@ens-lyon.org>

* sysdeps/mach/hurd/sendmsg.c (__libc_sendmsg): On SCM_CREDS
control messages, record uids, pass a rendez-vous port in the
control message, and call __auth_user_authenticate_request to
make auth send credentials on that port.  Do not wait for a
reply.
* sysdeps/mach/hurd/recvmsg.c (contains_uid, contains_gid,
check_auth): New functions.
(__libc_recvmsg): On SCM_CREDS control messages, call check_auth
to check the passed credentials thanks to the answer from the
auth server.
* hurd/Makefile (user-interfaces): Add auth_request and
auth_reply.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-sendmsg-SCM_CREDS.diff

Add support to send file descriptors over Unix sockets

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-sendmsg-SCM_RIGHTS.diff

No hp timing

2010-08-21  Samuel Thibault  <samuel.thibault@ens-lyon.org>

We don't have support for hp timing for now, even the i686 variant, which needs
to know the CPU speed.
Copied from sysdeps/generic/hp-timing.h

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-no-hp-timing.diff

Fix installation of locales-all

Dirty hack to fix installation of locales-all: instead of just locking the
archive extension (which is not supported on hurd-i386), lock it all.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-locarchive.diff

local-disable-ioctls

struct ortentry and struct ifalias req are actually not defined
struct arpreq is defined, but can not be passed to an ioctl on the Hurd.
so let's make packages not believe these are available.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name local-disable-ioctls.diff

Allow the kernel to start earlier than VM_MAX_ADDRESS

A bit hackish.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-bigmem.diff

The critical section lock _can_ be held in these place.

At least since hurd_thread_cancel can be called by another thread and lock our
critical lock.

http://bugs.debian.org/46859

“
Thomas suggested that there is no need to take the critical section
lock.  I believe that taking the critical section lock is necessary to
prevent the target thread from entering a signal handler.  Roland will
look into the problem.
”

Taking the critical section lock makes these assertions bogus.

It happens that hurd_thread_cancel is only called from libports and inside
/hurd/term so this is rare in practice.

A reproducer can be found here:

http://lists.gnu.org/archive/html/bug-hurd/2014-05/msg00025.html

2006-08-05  Samuel Thibault  <samuel.thibault@ens-lyon.org>

       * hurd/thread-cancel.c (hurd_thread_cancel): Do not assert that
       `&ss->critical_section_lock' is unlocked.
       * sysdeps/mach/hurd/jmp-unwind.c (_longjmp_unwind): Likewise, and take
       critical section lock before taking the sigstate lock.
       * sysdeps/mach/hurd/spawni.c (__spawni): Likewise.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-thread-cancel.diff

_dl_random

_dl_random

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-_dl_random.diff

Implement SysV shared memory for GNU/Hurd.

2005-07-11  Marcus Brinkmann  <marcus@gnu.org>

        * hurd/Makefile (routines): Add sysvshm.
        (distribute): Add sysvshm.h.
        * hurd/sysvshm.h: New file.
        * hurd/sysvshm.c: New file.
        * sysdeps/mach/hurd/bits/stat.h (S_IMMAP0): New macro.
        (S_ISPARE): Unset the S_IMMAP0 flag.
        * sysdeps/mach/hurd/ftok.c: New file.
        * sysdeps/mach/hurd/shmat.c: New file.
        * sysdeps/mach/hurd/shmctl.c: New file.
        * sysdeps/mach/hurd/shmdt.c: New file.
        * sysdeps/mach/hurd/bits/posix_opt.h: Define _XOPEN_SHM to 1.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-sysvshm.diff

tlsdesc.sym

i386 version of tlsdesc.sym for GNU Hurd.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-tlsdesc.sym.diff

tls-threadvar

replace the custom threadvar mechanism with generic TLS.
That will fix sigaltstack.

Note: the added reply_port and _hurd_sigstate fields should be kept last.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-tls-threadvar.diff

tls

TLS support.

All by Samuel Thibault.

glibc-2.8/debian/patches/hurd-i386/local-tls-support.diff 3151

2009-07-30  Samuel Thibault  <samuel.thibault@gnu.org>

Align up includes as on Linux, to fix build.
* sysdeps/mach/hurd/tls.h: Include <sysdep.h>.

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-tls.diff

getcontext, makecontext, setcontext, swapcontext

Adapted from the Linux x86 functions.

Not thoroughly tested, but manual testing as well as glibc tests look fine, and
manual -lpthread testing also looks fine (within the given bounds for a new
stack to be used with makecontext).

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name tg-context_functions.diff

local-enable-ldconfig

# DP: Description: Enable ldconfig and such on hurd-i386
# DP: Author: Jeff Bailey <jbailey@nisa.net>
# DP: Related Bugs: #309489
# DP: Upstream status: Not submitted
# DP: Status Details: Upstream disagrees with this patch, but I'm
# DP: putting it in so that we have expected Debian behaviour on the
# DP: Hurd.  We should review this when the ELF standard supports runpath.
# DP: Date: 08 Apr 2003

Gbp-Pq: Topic hurd-i386
Gbp-Pq: Name local-enable-ldconfig.diff

submitted-gmon-start

2015-11-15  Helge Deller  <deller@gmx.de>
    John David Anglin  <danglin@gcc.gnu.org>

PR libc/19170
* sysdeps/hppa/crti.S (gmon_initializer): New.  Use .init_array support
to call gmon_initializer.  Only call __gmon_start__ when symbol is
defined at runtime.
* sysdeps/hppa/crtn.S (__gmon_start__): Delete.
* sysdeps/hppa/dl-lookupcfg.h (DL_FIXUP_MAKE_VALUE): Provide null
function descriptor when MAP argument is null.
* sysdeps/hppa/dl-trampoline.S (_dl_runtime_resolve): Return directly
to previous function when function is undefined.

Gbp-Pq: Topic hppa
Gbp-Pq: Name submitted-gmon-start.diff

local-elf-make-cflags

===================================================================

Gbp-Pq: Topic hppa
Gbp-Pq: Name local-elf-make-cflags.diff

local-inlining

Increase the maximal overall growth of the compilation unit caused
by inlining for dl-reloc.c on hppa. This remove some warnings and
strangely it reduces the size of the final binary.

Gbp-Pq: Topic hppa
Gbp-Pq: Name local-inlining.diff

git-arm64-memcmp

commit 922369032c604b4dcfd535e1bcddd4687e7126a5
Author: Wilco Dijkstra <wdijkstr@arm.com>
Date:   Thu Aug 10 17:00:38 2017 +0100

    [AArch64] Optimized memcmp.

    This is an optimized memcmp for AArch64.  This is a complete rewrite
    using a different algorithm.  The previous version split into cases
    where both inputs were aligned, the inputs were mutually aligned and
    unaligned using a byte loop.  The new version combines all these cases,
    while small inputs of less than 8 bytes are handled separately.

    This allows the main code to be sped up using unaligned loads since
    there are now at least 8 bytes to be compared.  After the first 8 bytes,
    align the first input.  This ensures each iteration does at most one
    unaligned access and mutually aligned inputs behave as aligned.
    After the main loop, process the last 8 bytes using unaligned accesses.

    This improves performance of (mutually) aligned cases by 25% and
    unaligned by >500% (yes >6 times faster) on large inputs.

            * sysdeps/aarch64/memcmp.S (memcmp):
            Rewrite of optimized memcmp.

Gbp-Pq: Topic arm
Gbp-Pq: Name git-arm64-memcmp.diff

local-arm-futex

Gbp-Pq: Topic arm
Gbp-Pq: Name local-arm-futex.diff

unsubmitted-ldso-multilib

Gbp-Pq: Topic arm
Gbp-Pq: Name unsubmitted-ldso-multilib.diff

local-vfp-sysdeps

Gbp-Pq: Topic arm
Gbp-Pq: Name local-vfp-sysdeps.diff

local-soname-hack

For backward compatibility with armhf binaries built with the
old linker SONAME, we need to fake out the linker to believe
the new is the old, until such a point as everything is rebuilt

Gbp-Pq: Topic arm
Gbp-Pq: Name local-soname-hack.diff

unsubmitted-ldso-abi-check

Gbp-Pq: Topic arm
Gbp-Pq: Name unsubmitted-ldso-abi-check.diff

unsubmitted-ldconfig-cache-abi

Gbp-Pq: Topic arm
Gbp-Pq: Name unsubmitted-ldconfig-cache-abi.diff

local-sigaction

Gbp-Pq: Topic arm
Gbp-Pq: Name local-sigaction.diff

submitted-fminmax

2017-12-31  Aurelien Jarno  <aurelien@aurel32.net>

[BZ #20947]
* sysdeps/alpha/fpu/s_fmax.S: Remove file.
* sysdeps/alpha/fpu/s_fmaxf.S: Likewise.
* sysdeps/alpha/fpu/s_fmin.S: Likewise.
* sysdeps/alpha/fpu/s_fminf.S: Likewise.

Gbp-Pq: Topic alpha
Gbp-Pq: Name submitted-fminmax.diff

submitted-rlimit

2017-12-30  Aurelien Jarno <aurelien@aurel32.net>

[BZ #22648]
* sysdeps/unix/sysv/linux/alpha/getrlimit64.c: New file.
* sysdeps/unix/sysv/linux/alpha/setrlimit64.c: Ditto.

Gbp-Pq: Topic alpha
Gbp-Pq: Name submitted-rlimit.diff

submitted-fts64

2016-03-22  Aurelien Jarno  <aurelien@aurel32.net>

* sysdeps/unix/sysv/linux/alpha/fts.c: New file.
* sysdeps/unix/sysv/linux/alpha/fts64.c: New file.

Gbp-Pq: Topic alpha
Gbp-Pq: Name submitted-fts64.diff

local-string-functions

The alpha assembly version of various string functions do not work
correctly on some corner cases, and thus doesn't pass the testsuite.
This patch removes them.

Gbp-Pq: Topic alpha
Gbp-Pq: Name local-string-functions.diff

submitted-dl-support

older versions of glibc would build dl-sysdep as shared-only and dl-support as
static-only.  alpha hooks in a cache variable via dl-auxv.h.  newer versions of
glibc build dl-sysdep as both shared and static which means we now have symbol
duplication for static builds with dl-sysdep and dl-support.  since dl-sysdep
is both shared/static, there is no point in hooking dl-support anymore, so we
can punt it.

Gbp-Pq: Topic alpha
Gbp-Pq: Name submitted-dl-support.diff

local-gcc4.1

2006-05-30  Falk Hueffner <falk@debian.org>

* sysdeps/unix/sysv/linux/alpha/ioperm.c: force the architecture
  to ev6 in assembly code.

{standard input}: Assembler messages:
{standard input}:341: Error: macro requires $at register while noat in effect
{standard input}:374: Error: macro requires $at register while noat in effect
{standard input}:438: Error: macro requires $at register while noat in effect
{standard input}:471: Error: macro requires $at register while noat in effect
make[3]: *** [/tmp/buildd/glibc-2.3.6/build-tree/alpha-libc/misc/ioperm.o] Error 1

Hrm. gcc puts .arch ev4 into the .s, and this overrides -mev6 for as.
I cannot really think of anything better than

Gbp-Pq: Topic alpha
Gbp-Pq: Name local-gcc4.1.diff

submitted-it_IT-thousands_sep

2014-05-05  Aurelien Jarno  <aurelien@aurel32.net>

* locales/it_IT: Use same grouping in LC_NUMERIC as in LC_MONETARY.

Gbp-Pq: Topic localedata
Gbp-Pq: Name submitted-it_IT-thousands_sep.diff

submitted-es_MX-decimal_point

2012-06-06  Aurelien Jarno  <aurelien@aurel32.net>

* locales/es_MX (LC_MONETARY): Set mon_thousands_sep to space
(<U0020>). (LC_NUMERIC): Set decimal_point to dot (<U002E>)
and thousands_sep to space (<U0020>).

Gbp-Pq: Topic localedata
Gbp-Pq: Name submitted-es_MX-decimal_point.diff

submitted-en_AU-date_fmt

2012-04-29  Aurelien Jarno  <aurelien@aurel32.net>

        * locales/en_AU: Change date_fmt to match Australian standards.

Gbp-Pq: Topic localedata
Gbp-Pq: Name submitted-en_AU-date_fmt.diff