Package: emacs;
Reported by: Paul Eggert <eggert <at> cs.ucla.edu>
Date: Sat, 1 Sep 2012 22:41:01 UTC
Severity: normal
Tags: patch
Done: Eli Zaretskii <eliz <at> gnu.org>
Bug is archived. No further changes may be made.
To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 12327 in the body.
You can then email your comments to 12327 AT debbugs.gnu.org in the normal way.
Toggle the display of automated, internal messages from the tracker.
View this report as an mbox folder, status mbox, maintainer mbox
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Sat, 01 Sep 2012 22:41:02 GMT) Full text and rfc822 format available.Paul Eggert <eggert <at> cs.ucla.edu>:bug-gnu-emacs <at> gnu.org.
(Sat, 01 Sep 2012 22:41:02 GMT) Full text and rfc822 format available.Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):
From: Paul Eggert <eggert <at> cs.ucla.edu> To: bug-gnu-emacs <at> gnu.org Cc: Juanma Barranquero <lekktu <at> gmail.com>, Eli Zaretskii <eliz <at> gnu.org> Subject: Signal-handler cleanup for Emacs Date: Sat, 01 Sep 2012 15:38:15 -0700
Tags: patch
Here's a patch to clean up Emacs's signal-handling code somewhat.
As described below, it shrinks the size of Emacs's text segment
by 0.6% on my platform. It also removes a few dozen lines from
the source code.
This patch affects only the mainline code. Some tweaks will be needed
to the Windows port, which will need to define the functions
'sighandler' and 'set_sighandler'; these are specialized replacements
for 'sys_signal' that avoid the overhead of setting/getting when it's
not needed.
I'm planning to install this into the trunk and am publishing it now
as a head-up for the Windows side.
=== modified file 'ChangeLog'
--- ChangeLog 2012-09-01 18:54:38 +0000
+++ ChangeLog 2012-09-01 21:51:42 +0000
@@ -1,5 +1,10 @@
2012-09-01 Paul Eggert <eggert <at> cs.ucla.edu>
+ Signal-handler cleanup.
+ * configure.ac (PTY_OPEN, PTY_TTY_NAME_SPRINTF):
+ Adjust to syssignal.h changes.
+ (SIGNAL_H_AB): Remove; no longer needed.
+
* configure.ac (_FORTIFY_SOURCE): Define only when optimizing.
This ports to glibc 2.15 or later, when configured with
--enable-gcc-warnings. See Eric Blake in
=== modified file 'configure.ac'
--- configure.ac 2012-09-01 18:54:38 +0000
+++ configure.ac 2012-09-01 21:37:48 +0000
@@ -3445,7 +3445,7 @@
cygwin )
AC_DEFINE(PTY_ITERATION, [int i; for (i = 0; i < 1; i++)])
dnl multi-line AC_DEFINEs are hard. :(
- AC_DEFINE(PTY_OPEN, [ do { int dummy; SIGMASKTYPE mask; mask = sigblock (sigmask (SIGCHLD)); if (-1 == openpty (&fd, &dummy, pty_name, 0, 0)) fd = -1; sigsetmask (mask); if (fd >= 0) emacs_close (dummy); } while (0)])
+ AC_DEFINE(PTY_OPEN, [ do { int dummy; sigset_t mask; get_sigmask_block_signal (SIGCHLD, &mask); if (-1 == openpty (&fd, &dummy, pty_name, 0, 0)) fd = -1; set_signal_mask (&mask); if (fd >= 0) emacs_close (dummy); } while (0)])
AC_DEFINE(PTY_NAME_SPRINTF, [])
AC_DEFINE(PTY_TTY_NAME_SPRINTF, [])
;;
@@ -3474,7 +3474,7 @@
AC_DEFINE(PTY_ITERATION, [int i; for (i = 0; i < 1; i++)])
dnl Note that grantpt and unlockpt may fork. We must block SIGCHLD
dnl to prevent sigchld_handler from intercepting the child's death.
- AC_DEFINE(PTY_TTY_NAME_SPRINTF, [{ char *ptyname; sigblock (sigmask (SIGCHLD)); if (grantpt (fd) == -1 || unlockpt (fd) == -1 || !(ptyname = ptsname(fd))) { sigunblock (sigmask (SIGCHLD)); close (fd); return -1; } snprintf (pty_name, sizeof pty_name, "%s", ptyname); sigunblock (sigmask (SIGCHLD)); }])
+ AC_DEFINE(PTY_TTY_NAME_SPRINTF, [{ char *ptyname; block_signal (SIGCHLD); if (grantpt (fd) == -1 || unlockpt (fd) == -1 || !(ptyname = ptsname(fd))) { unblock_signal (SIGCHLD); close (fd); return -1; } snprintf (pty_name, sizeof pty_name, "%s", ptyname); unblock_signal (SIGCHLD); }])
dnl if HAVE_POSIX_OPENPT
if test "x$ac_cv_func_posix_openpt" = xyes; then
AC_DEFINE(PTY_OPEN, [fd = posix_openpt (O_RDWR | O_NOCTTY)])
@@ -3525,12 +3525,12 @@
dnl On SysVr4, grantpt(3) forks a subprocess, so keep sigchld_handler()
dnl from intercepting that death. If any child but grantpt's should die
dnl within, it should be caught after sigrelse(2).
- AC_DEFINE(PTY_TTY_NAME_SPRINTF, [{ char *ptsname (int), *ptyname; sigblock (sigmask (SIGCLD)); if (grantpt (fd) == -1) { emacs_close (fd); return -1; } sigunblock (sigmask (SIGCLD)); if (unlockpt (fd) == -1) { emacs_close (fd); return -1; } if (!(ptyname = ptsname (fd))) { emacs_close (fd); return -1; } snprintf (pty_name, sizeof pty_name, "%s", ptyname); }])
+ AC_DEFINE(PTY_TTY_NAME_SPRINTF, [{ char *ptsname (int), *ptyname; block_signal (SIGCLD); if (grantpt (fd) == -1) { emacs_close (fd); return -1; } unblock_signal (SIGCLD); if (unlockpt (fd) == -1) { emacs_close (fd); return -1; } if (!(ptyname = ptsname (fd))) { emacs_close (fd); return -1; } snprintf (pty_name, sizeof pty_name, "%s", ptyname); }])
;;
unixware )
dnl Comments are as per sol2*.
- AC_DEFINE(PTY_TTY_NAME_SPRINTF, [{ char *ptsname (int), *ptyname; sigblock(sigmask(SIGCLD)); if (grantpt(fd) == -1) fatal("could not grant slave pty"); sigunblock(sigmask(SIGCLD)); if (unlockpt(fd) == -1) fatal("could not unlock slave pty"); if (!(ptyname = ptsname(fd))) fatal ("could not enable slave pty"); snprintf (pty_name, sizeof pty_name, "%s", ptyname); }])
+ AC_DEFINE(PTY_TTY_NAME_SPRINTF, [{ char *ptsname (int), *ptyname; block_signal (SIGCLD); if (grantpt(fd) == -1) fatal("could not grant slave pty"); unblock_signal (SIGCLD); if (unlockpt(fd) == -1) fatal("could not unlock slave pty"); if (!(ptyname = ptsname(fd))) fatal ("could not enable slave pty"); snprintf (pty_name, sizeof pty_name, "%s", ptyname); }])
;;
esac
@@ -3806,13 +3806,6 @@
AC_DEFINE(XOS_NEEDS_TIME_H, 1, [Compensate for a bug in Xos.h on
some systems, where it requires time.h.])
;;
-
- netbsd | openbsd )
- dnl Greg A. Woods <woods <at> weird.com> says we must include signal.h
- dnl before syssignal.h is included, to work around interface conflicts
- dnl that are handled with CPP __RENAME() macro in signal.h.
- AC_DEFINE(SIGNAL_H_AHB, 1, [Define if AH_BOTTOM should include signal.h.])
- ;;
esac
=== modified file 'src/ChangeLog'
--- src/ChangeLog 2012-09-01 13:54:27 +0000
+++ src/ChangeLog 2012-09-01 21:50:12 +0000
@@ -1,3 +1,66 @@
+2012-09-01 Paul Eggert <eggert <at> cs.ucla.edu>
+
+ Signal-handler cleanup.
+ Emacs's signal handlers were written in the old 4.2BSD style with
+ sigblock and sigmask and so forth, and this led to some
+ inefficiencies and confusion. Rewrite these to use
+ pthread_sigmask etc. without copying signal sets around. Also,
+ get rid of the confusing macros 'SIGNAL_THREAD_CHECK' and
+ 'signal', and instead use functions that do not attempt to take
+ over the system name space. This patch causes Emacs's text
+ segment to shrink by 0.6% on my platform, Fedora 17 x86-64.
+ * alloc.c, emacsgtkfixed.c, nsfns.m, widget.c, xmenu.c:
+ Do not include <signal.h> or "syssignal.h", as these
+ modules do not use signals.
+ * atimer.c, callproc.c, data.c, dispnew.c, emacs.c, floatfns.c:
+ * gtkutil.c, keyboard.c, process.c, sound.c, sysdep.c, term.c, xterm.c:
+ Do not include <signal.h>, as "syssignal.h" does that for us now.
+ * callproc.c (Fcall_process) [HAVE_WORKING_VFORK]:
+ Use new get_sigmask_block_signal and set_sigmask functions instead
+ of repeating the code inline.
+ * conf_post.h [SIGNAL_H_AHB]: Do not include <signal.h>;
+ no longer needed here.
+ * emacs.c (main): Use new sighandler function, so that there's
+ no need to block and unblock SIGHUP.
+ * sysdep.c (SYSSIGNAL_INLINE): New macro.
+ (empty_mask): Remove. All uses replaced to create their own
+ masks, as this avoids static storage and offers the compiler more
+ optimization opportunities.
+ (sighandler, get_set_sighandler, set_sighandler) [!WINDOWSNT]:
+ New functions. All users of 'signal' modified to use sighandler
+ if they're readonly, to set_sighandler if they're writeonly, and to
+ sys_signal if they're read+write.
+ (sys_signal): Reimplement in terms of get_set_sighandler.
+ (sys_sigmask) [!__GNUC__]: Remove; no longer needed.
+ (sys_sigblock): Remove; replaced by block_signal and
+ get_sigmask_block_signal. All uses changed.
+ (sys_sigunblock): Remove; replaced by unblock_signal. All uses changed.
+ (sys_sigsetmask): Remove; replaced by set_sigmask. All uses changed.
+ * syssignal.h: Include <signal.h>, since we need its types,
+ and <stdbool.h> since we need bool.
+ (SIGMASKTYPE): Remove. All uses replaced by its definiens, sigset_t.
+ (SIGEMPTYMASK, empty_mask): Remove as described above.
+ (sigmask, sys_sigmask): Remove; no longer needed.
+ (sigpause): Remove. All uses replaced by its definiens, sigsuspend.
+ (sigblock): Remove. All uses replaced by block_signal.
+ (sigunblock): Remove. All uses replaced by unblock_signal.
+ (sigsetmask) [!defined sigsetmask]: Remove. All uses replaced by
+ set_sigmask.
+ (signal): Remove. All uses replaced by sighandler, set_sighandler,
+ or sys_signal as described above.
+ (init_signals): Now ATTRIBUTE_CONST in the usual case.
+ (get_sigmask_block_signal, block_signal, unblock_signal, set_sigmask):
+ New functions.
+ (sigfree): Reimplement in terms of set_sigmask.
+ (sighandler, set_sighandler): New decls.
+ (sys_sigblock, sys_sigunblock, sys_sigsetmask): Remove decls.
+ (sys_sigdel): Remove; unused.
+ (NSIG): Remove a FIXME; the code's fine. Remove an unnecessary ifdef.
+ (forwarded_signal): Rename from SIGNAL_THREAD_CHECK, and turn into
+ a boolean function. This is cleaner, because the caller can now
+ see that the check might immediately return from the caller.
+ All uses changed.
+
2012-09-01 Eli Zaretskii <eliz <at> gnu.org>
* w32uniscribe.c (uniscribe_shape): Handle correctly the case of
=== modified file 'src/alloc.c'
--- src/alloc.c 2012-08-31 10:53:19 +0000
+++ src/alloc.c 2012-09-01 21:37:07 +0000
@@ -26,8 +26,6 @@
#include <limits.h> /* For CHAR_BIT. */
#include <setjmp.h>
-#include <signal.h>
-
#ifdef HAVE_PTHREAD
#include <pthread.h>
#endif
@@ -42,7 +40,6 @@
#include "keyboard.h"
#include "frame.h"
#include "blockinput.h"
-#include "syssignal.h"
#include "termhooks.h" /* For struct terminal. */
#include <setjmp.h>
#include <verify.h>
=== modified file 'src/atimer.c'
--- src/atimer.c 2012-08-23 08:27:08 +0000
+++ src/atimer.c 2012-09-01 21:37:07 +0000
@@ -17,7 +17,6 @@
along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>. */
#include <config.h>
-#include <signal.h>
#include <stdio.h>
#include <setjmp.h>
#include "lisp.h"
@@ -51,8 +50,8 @@
/* Block/unblock SIGALRM. */
-#define BLOCK_ATIMERS sigblock (sigmask (SIGALRM))
-#define UNBLOCK_ATIMERS sigunblock (sigmask (SIGALRM))
+#define BLOCK_ATIMERS block_signal (SIGALRM)
+#define UNBLOCK_ATIMERS unblock_signal (SIGALRM)
/* Function prototypes. */
@@ -378,7 +377,8 @@
alarm_signal_handler (int signo)
{
#ifndef SYNC_INPUT
- SIGNAL_THREAD_CHECK (signo);
+ if (forwarded_signal (signo))
+ return;
#endif
pending_atimers = 1;
@@ -412,7 +412,7 @@
{
if (on)
{
- signal (SIGALRM, alarm_signal_handler);
+ set_sighandler (SIGALRM, alarm_signal_handler);
set_alarm ();
}
else
@@ -426,5 +426,5 @@
free_atimers = stopped_atimers = atimers = NULL;
pending_atimers = 0;
/* pending_signals is initialized in init_keyboard.*/
- signal (SIGALRM, alarm_signal_handler);
+ set_sighandler (SIGALRM, alarm_signal_handler);
}
=== modified file 'src/callproc.c'
--- src/callproc.c 2012-08-25 03:11:12 +0000
+++ src/callproc.c 2012-09-01 21:37:07 +0000
@@ -19,7 +19,6 @@
#include <config.h>
-#include <signal.h>
#include <errno.h>
#include <stdio.h>
#include <setjmp.h>
@@ -500,15 +499,11 @@
int fd_error = fd1;
#ifdef HAVE_WORKING_VFORK
sigset_t procmask;
- sigset_t blocked;
struct sigaction sigpipe_action;
#endif
if (fd_output >= 0)
fd1 = fd_output;
-#if 0 /* Some systems don't have sigblock. */
- mask = sigblock (sigmask (SIGCHLD));
-#endif
/* Record that we're about to create a synchronous process. */
synch_process_alive = 1;
@@ -597,10 +592,8 @@
this sets the parent's signal handlers as well as the child's.
So delay all interrupts whose handlers the child might munge,
and record the current handlers so they can be restored later. */
- sigemptyset (&blocked);
- sigaddset (&blocked, SIGPIPE);
sigaction (SIGPIPE, 0, &sigpipe_action);
- pthread_sigmask (SIG_BLOCK, &blocked, &procmask);
+ get_sigmask_block_signal (&procmask, SIGPIPE);
#endif
BLOCK_INPUT;
@@ -649,9 +642,9 @@
/* GConf causes us to ignore SIGPIPE, make sure it is restored
in the child. */
- signal (SIGPIPE, SIG_DFL);
+ set_sighandler (SIGPIPE, SIG_DFL);
#ifdef HAVE_WORKING_VFORK
- pthread_sigmask (SIG_SETMASK, &procmask, 0);
+ set_sigmask (&procmask);
#endif
child_setup (filefd, fd1, fd_error, (char **) new_argv,
@@ -663,7 +656,7 @@
#ifdef HAVE_WORKING_VFORK
/* Restore the signal state. */
sigaction (SIGPIPE, &sigpipe_action, 0);
- pthread_sigmask (SIG_SETMASK, &procmask, 0);
+ set_sigmask (&procmask);
#endif
#endif /* not WINDOWSNT */
=== modified file 'src/conf_post.h'
--- src/conf_post.h 2012-08-20 16:48:10 +0000
+++ src/conf_post.h 2012-09-01 21:37:07 +0000
@@ -40,11 +40,6 @@
#endif
#endif
-#ifdef SIGNAL_H_AHB
-#undef SIGNAL_H_AHB
-#include <signal.h>
-#endif
-
/* This silences a few compilation warnings on FreeBSD. */
#ifdef BSD_SYSTEM_AHB
#undef BSD_SYSTEM_AHB
=== modified file 'src/data.c'
--- src/data.c 2012-08-27 17:23:48 +0000
+++ src/data.c 2012-09-01 21:37:07 +0000
@@ -19,7 +19,6 @@
#include <config.h>
-#include <signal.h>
#include <stdio.h>
#include <setjmp.h>
@@ -3213,9 +3212,9 @@
static void
arith_error (int signo)
{
- sigsetmask (SIGEMPTYMASK);
-
- SIGNAL_THREAD_CHECK (signo);
+ sigfree ();
+ if (forwarded_signal (signo))
+ return;
xsignal0 (Qarith_error);
}
@@ -3230,5 +3229,5 @@
if (!initialized)
return;
#endif /* CANNOT_DUMP */
- signal (SIGFPE, arith_error);
+ set_sighandler (SIGFPE, arith_error);
}
=== modified file 'src/dispnew.c'
--- src/dispnew.c 2012-09-01 06:38:52 +0000
+++ src/dispnew.c 2012-09-01 21:37:07 +0000
@@ -21,7 +21,6 @@
#define DISPEXTERN_INLINE EXTERN_INLINE
-#include <signal.h>
#include <stdio.h>
#include <setjmp.h>
#include <unistd.h>
@@ -5561,8 +5560,9 @@
struct tty_display_info *tty;
- signal (SIGWINCH, window_change_signal);
- SIGNAL_THREAD_CHECK (signalnum);
+ set_sighandler (SIGWINCH, window_change_signal);
+ if (forwarded_signal (signalnum))
+ return;
/* The frame size change obviously applies to a single
termcap-controlled terminal, but we can't decide which.
@@ -6173,7 +6173,7 @@
#ifndef CANNOT_DUMP
if (initialized)
#endif /* CANNOT_DUMP */
- signal (SIGWINCH, window_change_signal);
+ set_sighandler (SIGWINCH, window_change_signal);
#endif /* SIGWINCH */
/* If running as a daemon, no need to initialize any frames/terminal. */
=== modified file 'src/emacs.c'
--- src/emacs.c 2012-09-01 08:01:36 +0000
+++ src/emacs.c 2012-09-01 21:37:07 +0000
@@ -20,7 +20,6 @@
#include <config.h>
-#include <signal.h>
#include <errno.h>
#include <stdio.h>
@@ -297,9 +296,10 @@
void
fatal_error_signal (int sig)
{
- SIGNAL_THREAD_CHECK (sig);
+ if (forwarded_signal (sig))
+ return;
fatal_error_code = sig;
- signal (sig, SIG_DFL);
+ set_sighandler (sig, SIG_DFL);
TOTALLY_UNBLOCK_INPUT;
@@ -319,7 +319,7 @@
going to send is probably blocked, so we have to unblock it if we
want to really receive it. */
#ifndef MSDOS
- sigunblock (sigmask (fatal_error_code));
+ unblock_signal (fatal_error_code);
#endif
kill (getpid (), fatal_error_code);
@@ -331,8 +331,9 @@
void
memory_warning_signal (int sig)
{
- signal (sig, memory_warning_signal);
- SIGNAL_THREAD_CHECK (sig);
+ set_sighandler (sig, memory_warning_signal);
+ if (forwarded_signal (sig))
+ return;
malloc_warning ("Operating system warns that virtual memory is running low.\n");
@@ -1119,13 +1120,10 @@
#endif
)
{
- sigblock (sigmask (SIGHUP));
/* In --batch mode, don't catch SIGHUP if already ignored.
That makes nohup work. */
- if (! noninteractive
- || signal (SIGHUP, SIG_IGN) != SIG_IGN)
- signal (SIGHUP, fatal_error_signal);
- sigunblock (sigmask (SIGHUP));
+ if (! (noninteractive && sighandler (SIGHUP) == SIG_IGN))
+ set_sighandler (SIGHUP, fatal_error_signal);
}
if (
@@ -1139,9 +1137,9 @@
/* Don't catch these signals in batch mode if dumping.
On some machines, this sets static data that would make
signal fail to work right when the dumped Emacs is run. */
- signal (SIGQUIT, fatal_error_signal);
- signal (SIGILL, fatal_error_signal);
- signal (SIGTRAP, fatal_error_signal);
+ set_sighandler (SIGQUIT, fatal_error_signal);
+ set_sighandler (SIGILL, fatal_error_signal);
+ set_sighandler (SIGTRAP, fatal_error_signal);
#ifdef SIGUSR1
add_user_signal (SIGUSR1, "sigusr1");
#endif
@@ -1149,68 +1147,69 @@
add_user_signal (SIGUSR2, "sigusr2");
#endif
#ifdef SIGABRT
- signal (SIGABRT, fatal_error_signal);
+ set_sighandler (SIGABRT, fatal_error_signal);
#endif
#ifdef SIGHWE
- signal (SIGHWE, fatal_error_signal);
+ set_sighandler (SIGHWE, fatal_error_signal);
#endif
#ifdef SIGPRE
- signal (SIGPRE, fatal_error_signal);
+ set_sighandler (SIGPRE, fatal_error_signal);
#endif
#ifdef SIGORE
- signal (SIGORE, fatal_error_signal);
+ set_sighandler (SIGORE, fatal_error_signal);
#endif
#ifdef SIGUME
- signal (SIGUME, fatal_error_signal);
+ set_sighandler (SIGUME, fatal_error_signal);
#endif
#ifdef SIGDLK
- signal (SIGDLK, fatal_error_signal);
+ set_sighandler (SIGDLK, fatal_error_signal);
#endif
#ifdef SIGCPULIM
- signal (SIGCPULIM, fatal_error_signal);
+ set_sighandler (SIGCPULIM, fatal_error_signal);
#endif
#ifdef SIGIOT
/* This is missing on some systems - OS/2, for example. */
- signal (SIGIOT, fatal_error_signal);
+ set_sighandler (SIGIOT, fatal_error_signal);
#endif
#ifdef SIGEMT
- signal (SIGEMT, fatal_error_signal);
+ set_sighandler (SIGEMT, fatal_error_signal);
#endif
- signal (SIGFPE, fatal_error_signal);
+ set_sighandler (SIGFPE, fatal_error_signal);
#ifdef SIGBUS
- signal (SIGBUS, fatal_error_signal);
+ set_sighandler (SIGBUS, fatal_error_signal);
#endif
- signal (SIGSEGV, fatal_error_signal);
+ set_sighandler (SIGSEGV, fatal_error_signal);
#ifdef SIGSYS
- signal (SIGSYS, fatal_error_signal);
+ set_sighandler (SIGSYS, fatal_error_signal);
#endif
/* May need special treatment on MS-Windows. See
http://lists.gnu.org/archive/html/emacs-devel/2010-09/msg01062.html
Please update the doc of kill-emacs, kill-emacs-hook, and
NEWS if you change this.
*/
- if (noninteractive) signal (SIGINT, fatal_error_signal);
- signal (SIGTERM, fatal_error_signal);
+ if (noninteractive)
+ set_sighandler (SIGINT, fatal_error_signal);
+ set_sighandler (SIGTERM, fatal_error_signal);
#ifdef SIGXCPU
- signal (SIGXCPU, fatal_error_signal);
+ set_sighandler (SIGXCPU, fatal_error_signal);
#endif
#ifdef SIGXFSZ
- signal (SIGXFSZ, fatal_error_signal);
+ set_sighandler (SIGXFSZ, fatal_error_signal);
#endif /* SIGXFSZ */
#ifdef SIGDANGER
/* This just means available memory is getting low. */
- signal (SIGDANGER, memory_warning_signal);
+ set_sighandler (SIGDANGER, memory_warning_signal);
#endif
#ifdef AIX
/* 20 is SIGCHLD, 21 is SIGTTIN, 22 is SIGTTOU. */
- signal (SIGXCPU, fatal_error_signal);
- signal (SIGIOINT, fatal_error_signal);
- signal (SIGGRANT, fatal_error_signal);
- signal (SIGRETRACT, fatal_error_signal);
- signal (SIGSOUND, fatal_error_signal);
- signal (SIGMSG, fatal_error_signal);
+ set_sighandler (SIGXCPU, fatal_error_signal);
+ set_sighandler (SIGIOINT, fatal_error_signal);
+ set_sighandler (SIGGRANT, fatal_error_signal);
+ set_sighandler (SIGRETRACT, fatal_error_signal);
+ set_sighandler (SIGSOUND, fatal_error_signal);
+ set_sighandler (SIGMSG, fatal_error_signal);
#endif /* AIX */
}
@@ -2041,7 +2040,7 @@
/* There is a tendency for a SIGIO signal to arrive within exit,
and cause a SIGHUP because the input descriptor is already closed. */
unrequest_sigio ();
- signal (SIGIO, SIG_IGN);
+ set_sighandler (SIGIO, SIG_IGN);
#endif
#ifdef WINDOWSNT
=== modified file 'src/emacsgtkfixed.c'
--- src/emacsgtkfixed.c 2012-04-23 07:34:29 +0000
+++ src/emacsgtkfixed.c 2012-09-01 21:37:07 +0000
@@ -21,7 +21,6 @@
#include <config.h>
#include "emacsgtkfixed.h"
-#include <signal.h>
#include <stdio.h>
#include <setjmp.h>
#include "lisp.h"
=== modified file 'src/floatfns.c'
--- src/floatfns.c 2012-07-17 02:56:00 +0000
+++ src/floatfns.c 2012-09-01 21:37:07 +0000
@@ -48,7 +48,6 @@
*/
#include <config.h>
-#include <signal.h>
#include <setjmp.h>
#include "lisp.h"
#include "syssignal.h"
@@ -955,13 +954,14 @@
fatal_error_signal (signo);
#ifdef BSD_SYSTEM
- sigsetmask (SIGEMPTYMASK);
+ sigfree ();
#else
/* Must reestablish handler each time it is called. */
- signal (SIGILL, float_error);
+ set_sighandler (SIGILL, float_error);
#endif /* BSD_SYSTEM */
- SIGNAL_THREAD_CHECK (signo);
+ if (forwarded_signal (signo))
+ return;
in_float = 0;
xsignal1 (Qarith_error, float_error_arg);
@@ -1007,7 +1007,7 @@
init_floatfns (void)
{
#ifdef FLOAT_CATCH_SIGILL
- signal (SIGILL, float_error);
+ set_sighandler (SIGILL, float_error);
#endif
in_float = 0;
}
=== modified file 'src/gtkutil.c'
--- src/gtkutil.c 2012-08-30 16:07:44 +0000
+++ src/gtkutil.c 2012-09-01 21:37:07 +0000
@@ -21,7 +21,6 @@
#ifdef USE_GTK
#include <float.h>
-#include <signal.h>
#include <stdio.h>
#include <setjmp.h>
@@ -1979,7 +1978,7 @@
/* I really don't know why this is needed, but without this the GLIBC add on
library linuxthreads hangs when the Gnome file chooser backend creates
threads. */
- sigblock (sigmask (__SIGRTMIN));
+ block_signal (__SIGRTMIN);
#endif /* HAVE_PTHREAD */
#ifdef HAVE_GTK_FILE_SELECTION_NEW
@@ -2001,7 +2000,7 @@
filesel_done = xg_dialog_run (f, w);
#if defined (HAVE_PTHREAD) && defined (__SIGRTMIN)
- sigunblock (sigmask (__SIGRTMIN));
+ unblock_signal (__SIGRTMIN);
#endif
if (filesel_done == GTK_RESPONSE_OK)
@@ -2061,7 +2060,7 @@
Lisp_Object font = Qnil;
#if defined (HAVE_PTHREAD) && defined (__SIGRTMIN)
- sigblock (sigmask (__SIGRTMIN));
+ block_signal (__SIGRTMIN);
#endif /* HAVE_PTHREAD */
w = gtk_font_chooser_dialog_new
@@ -2090,7 +2089,7 @@
done = xg_dialog_run (f, w);
#if defined (HAVE_PTHREAD) && defined (__SIGRTMIN)
- sigunblock (sigmask (__SIGRTMIN));
+ unblock_signal (__SIGRTMIN);
#endif
if (done == GTK_RESPONSE_OK)
=== modified file 'src/keyboard.c'
--- src/keyboard.c 2012-09-01 06:38:52 +0000
+++ src/keyboard.c 2012-09-01 21:37:07 +0000
@@ -21,7 +21,6 @@
#define KEYBOARD_INLINE EXTERN_INLINE
-#include <signal.h>
#include <stdio.h>
#include <setjmp.h>
#include "lisp.h"
@@ -3661,7 +3660,7 @@
hold_keyboard_input ();
#ifdef SIGIO
if (!noninteractive)
- signal (SIGIO, SIG_IGN);
+ set_sighandler (SIGIO, SIG_IGN);
#endif
stop_polling ();
}
@@ -3833,7 +3832,7 @@
unhold_keyboard_input ();
#ifdef SIGIO
if (!noninteractive)
- signal (SIGIO, input_available_signal);
+ set_sighandler (SIGIO, input_available_signal);
#endif /* SIGIO */
start_polling ();
}
@@ -6782,10 +6781,10 @@
#ifdef SIGIO
if (interrupt_input)
{
- SIGMASKTYPE mask;
- mask = sigblock (sigmask (SIGIO));
+ sigset_t mask;
+ get_sigmask_block_signal (&mask, SIGIO);
read_avail_input (expected);
- sigsetmask (mask);
+ set_sigmask (&mask);
}
else
#ifdef POLL_FOR_INPUT
@@ -6794,10 +6793,10 @@
it's always set. */
if (!interrupt_input && poll_suppress_count == 0)
{
- SIGMASKTYPE mask;
- mask = sigblock (sigmask (SIGALRM));
+ sigset_t mask;
+ get_sigmask_block_signal (&mask, SIGALRM);
read_avail_input (expected);
- sigsetmask (mask);
+ set_sigmask (&mask);
}
else
#endif
@@ -6833,10 +6832,10 @@
#ifdef SIGIO
if (interrupt_input)
{
- SIGMASKTYPE mask;
- mask = sigblock (sigmask (SIGIO));
+ sigset_t mask;
+ get_sigmask_block_signal (&mask, SIGIO);
kbd_buffer_store_event (&event);
- sigsetmask (mask);
+ set_sigmask (&mask);
}
else
#endif
@@ -7241,7 +7240,8 @@
{
/* Must preserve main program's value of errno. */
int old_errno = errno;
- SIGNAL_THREAD_CHECK (signo);
+ if (forwarded_signal (signo))
+ return;
#ifdef SYNC_INPUT
interrupt_input_pending = 1;
@@ -7311,7 +7311,7 @@
p->next = user_signals;
user_signals = p;
- signal (sig, handle_user_signal);
+ set_sighandler (sig, handle_user_signal);
}
static void
@@ -7321,7 +7321,8 @@
struct user_signal_info *p;
const char *special_event_name = NULL;
- SIGNAL_THREAD_CHECK (sig);
+ if (forwarded_signal (sig))
+ return;
if (SYMBOLP (Vdebug_on_event))
special_event_name = SSDATA (SYMBOL_NAME (Vdebug_on_event));
@@ -7382,7 +7383,7 @@
for (p = user_signals; p; p = p->next)
if (p->npending > 0)
{
- SIGMASKTYPE mask;
+ sigset_t mask;
if (nstored == 0)
{
@@ -7392,7 +7393,7 @@
}
nstored += p->npending;
- mask = sigblock (sigmask (p->sig));
+ get_sigmask_block_signal (&mask, p->sig);
do
{
buf.code = p->sig;
@@ -7400,7 +7401,7 @@
p->npending--;
}
while (p->npending > 0);
- sigsetmask (mask);
+ set_sigmask (&mask);
}
return nstored;
@@ -8445,7 +8446,7 @@
/* Append entries from tool_bar_item_properties to the end of
tool_bar_items_vector. */
- vcopy (tool_bar_items_vector, ntool_bar_items,
+ vcopy (tool_bar_items_vector, ntool_bar_items,
XVECTOR (tool_bar_item_properties)->contents, TOOL_BAR_ITEM_NSLOTS);
ntool_bar_items += TOOL_BAR_ITEM_NSLOTS;
}
@@ -10784,7 +10785,8 @@
int old_errno = errno;
struct terminal *terminal;
- SIGNAL_THREAD_CHECK (signalnum);
+ if (forwarded_signal (signalnum))
+ return;
/* See if we have an active terminal on our controlling tty. */
terminal = get_named_tty ("/dev/tty");
@@ -10841,7 +10843,7 @@
/* If SIGINT isn't blocked, don't let us be interrupted by
another SIGINT, it might be harmful due to non-reentrancy
in I/O functions. */
- sigblock (sigmask (SIGINT));
+ block_signal (SIGINT);
fflush (stdout);
reset_all_sys_modes ();
@@ -11405,17 +11407,17 @@
SIGINT. There is special code in interrupt_signal to exit
Emacs on SIGINT when there are no termcap frames on the
controlling terminal. */
- signal (SIGINT, interrupt_signal);
+ set_sighandler (SIGINT, interrupt_signal);
#ifndef DOS_NT
/* For systems with SysV TERMIO, C-g is set up for both SIGINT and
SIGQUIT and we can't tell which one it will give us. */
- signal (SIGQUIT, interrupt_signal);
+ set_sighandler (SIGQUIT, interrupt_signal);
#endif /* not DOS_NT */
}
/* Note SIGIO has been undef'd if FIONREAD is missing. */
#ifdef SIGIO
if (!noninteractive)
- signal (SIGIO, input_available_signal);
+ set_sighandler (SIGIO, input_available_signal);
#endif /* SIGIO */
/* Use interrupt input by default, if it works and noninterrupt input
=== modified file 'src/nsfns.m'
--- src/nsfns.m 2012-08-17 23:38:43 +0000
+++ src/nsfns.m 2012-09-01 21:37:07 +0000
@@ -30,7 +30,6 @@
interpretation of even the system includes. */
#include <config.h>
-#include <signal.h>
#include <math.h>
#include <setjmp.h>
#include <c-strcase.h>
=== modified file 'src/process.c'
--- src/process.c 2012-09-01 06:38:52 +0000
+++ src/process.c 2012-09-01 21:37:07 +0000
@@ -23,7 +23,6 @@
#define PROCESS_INLINE EXTERN_INLINE
-#include <signal.h>
#include <stdio.h>
#include <errno.h>
#include <setjmp.h>
@@ -1766,12 +1765,6 @@
int xforkin = forkin;
int xforkout = forkout;
-#if 0 /* This was probably a mistake--it duplicates code later on,
- but fails to handle all the cases. */
- /* Make sure SIGCHLD is not blocked in the child. */
- sigsetmask (SIGEMPTYMASK);
-#endif
-
/* Make the pty be the controlling terminal of the process. */
#ifdef HAVE_PTYS
/* First, disconnect its current controlling terminal. */
@@ -1880,18 +1873,18 @@
/* On AIX, we've disabled SIGHUP above once we start a child on a pty.
Now reenable it in the child, so it will die when we want it to. */
if (pty_flag)
- signal (SIGHUP, SIG_DFL);
+ set_sighandler (SIGHUP, SIG_DFL);
#endif
#endif /* HAVE_PTYS */
- signal (SIGINT, SIG_DFL);
- signal (SIGQUIT, SIG_DFL);
+ set_sighandler (SIGINT, SIG_DFL);
+ set_sighandler (SIGQUIT, SIG_DFL);
/* GConf causes us to ignore SIGPIPE, make sure it is restored
in the child. */
- signal (SIGPIPE, SIG_DFL);
+ set_sighandler (SIGPIPE, SIG_DFL);
/* Stop blocking signals in the child. */
- pthread_sigmask (SIG_SETMASK, &procmask, 0);
+ set_sigmask (&procmask);
if (pty_flag)
child_setup_tty (xforkout);
@@ -1982,7 +1975,7 @@
#endif
#endif /* HAVE_WORKING_VFORK */
/* Stop blocking signals in the parent. */
- pthread_sigmask (SIG_SETMASK, &procmask, 0);
+ set_sigmask (&procmask);
/* Now generate the error if vfork failed. */
if (pid < 0)
@@ -5439,8 +5432,9 @@
static void
send_process_trap (int ignore)
{
- SIGNAL_THREAD_CHECK (SIGPIPE);
- sigunblock (sigmask (SIGPIPE));
+ if (forwarded_signal (SIGPIPE))
+ return;
+ unblock_signal (SIGPIPE);
longjmp (send_process_frame, 1);
}
@@ -5534,7 +5528,7 @@
struct Lisp_Process *p = XPROCESS (proc);
ssize_t rv;
struct coding_system *coding;
- void (*volatile old_sigpipe) (int);
+ signal_handler_t volatile old_sigpipe;
if (p->raw_status_new)
update_status (p);
@@ -5673,7 +5667,7 @@
/* Send this batch, using one or more write calls. */
ptrdiff_t written = 0;
int outfd = p->outfd;
- old_sigpipe = (void (*) (int)) signal (SIGPIPE, send_process_trap);
+ old_sigpipe = sys_signal (SIGPIPE, send_process_trap);
#ifdef DATAGRAM_SOCKETS
if (DATAGRAM_CHAN_P (outfd))
{
@@ -5684,7 +5678,7 @@
written = rv;
else if (errno == EMSGSIZE)
{
- signal (SIGPIPE, old_sigpipe);
+ set_sighandler (SIGPIPE, old_sigpipe);
report_file_error ("sending datagram",
Fcons (proc, Qnil));
}
@@ -5709,7 +5703,7 @@
}
#endif
}
- signal (SIGPIPE, old_sigpipe);
+ set_sighandler (SIGPIPE, old_sigpipe);
if (rv < 0)
{
@@ -5769,7 +5763,7 @@
}
else
{
- signal (SIGPIPE, old_sigpipe);
+ set_sighandler (SIGPIPE, old_sigpipe);
proc = process_sent_to;
p = XPROCESS (proc);
p->raw_status_new = 0;
@@ -6414,7 +6408,8 @@
Lisp_Object proc;
struct Lisp_Process *p;
- SIGNAL_THREAD_CHECK (signo);
+ if (forwarded_signal (signo))
+ return;
while (1)
{
@@ -7397,7 +7392,7 @@
#ifndef CANNOT_DUMP
if (! noninteractive || initialized)
#endif
- signal (SIGCHLD, sigchld_handler);
+ set_sighandler (SIGCHLD, sigchld_handler);
#endif
FD_ZERO (&input_wait_mask);
=== modified file 'src/sound.c'
--- src/sound.c 2012-07-29 08:18:29 +0000
+++ src/sound.c 2012-09-01 21:37:07 +0000
@@ -48,7 +48,6 @@
#include "lisp.h"
#include "dispextern.h"
#include "atimer.h"
-#include <signal.h>
#include "syssignal.h"
/* END: Common Includes */
@@ -316,7 +315,7 @@
turn_on_atimers (1);
#ifdef SIGIO
- sigunblock (sigmask (SIGIO));
+ unblock_signal (SIGIO);
#endif
if (saved_errno != 0)
error ("%s: %s", msg, strerror (saved_errno));
@@ -736,7 +735,7 @@
troubles. */
turn_on_atimers (0);
#ifdef SIGIO
- sigblock (sigmask (SIGIO));
+ block_signal (SIGIO);
#endif
val = sd->format;
@@ -770,7 +769,7 @@
turn_on_atimers (1);
#ifdef SIGIO
- sigunblock (sigmask (SIGIO));
+ unblock_signal (SIGIO);
#endif
}
@@ -786,7 +785,7 @@
be interrupted by a signal. Block the ones we know to cause
troubles. */
#ifdef SIGIO
- sigblock (sigmask (SIGIO));
+ block_signal (SIGIO);
#endif
turn_on_atimers (0);
@@ -795,7 +794,7 @@
turn_on_atimers (1);
#ifdef SIGIO
- sigunblock (sigmask (SIGIO));
+ unblock_signal (SIGIO);
#endif
/* Close the device. */
=== modified file 'src/sysdep.c'
--- src/sysdep.c 2012-09-01 01:13:50 +0000
+++ src/sysdep.c 2012-09-01 21:37:07 +0000
@@ -19,9 +19,9 @@
#include <config.h>
+#define SYSSIGNAL_INLINE EXTERN_INLINE
#define SYSTIME_INLINE EXTERN_INLINE
-#include <signal.h>
#include <stdio.h>
#include <setjmp.h>
#ifdef HAVE_PWD_H
@@ -302,31 +302,43 @@
termination of subprocesses, perhaps involving a kernel bug too,
but no idea what it is. Just as a hunch we signal SIGCHLD to see
if that causes the problem to go away or get worse. */
- sigsetmask (sigmask (SIGCHLD));
+ sigset_t sigchild_mask;
+ sigemptyset (&sigchild_mask);
+ sigaddset (&sigchild_mask, SIGCHLD);
+ set_sigmask (&sigchild_mask);
+
if (0 > kill (pid, 0))
{
- sigsetmask (SIGEMPTYMASK);
+ sigfree ();
kill (getpid (), SIGCHLD);
break;
}
if (wait_debugging)
sleep (1);
else
- sigpause (SIGEMPTYMASK);
+ {
+ sigset_t empty_mask;
+ sigemptyset (&empty_mask);
+ sigsuspend (&empty_mask);
+ }
#else /* not BSD_SYSTEM, and not HPUX version >= 6 */
#ifdef WINDOWSNT
wait (0);
break;
#else /* not WINDOWSNT */
- sigblock (sigmask (SIGCHLD));
+ block_signal (SIGCHLD);
errno = 0;
if (kill (pid, 0) == -1 && errno == ESRCH)
{
- sigunblock (sigmask (SIGCHLD));
+ unblock_signal (SIGCHLD);
break;
}
- sigsuspend (&empty_mask);
+ {
+ sigset_t empty_mask;
+ sigemptyset (&empty_mask);
+ sigsuspend (&empty_mask);
+ }
#endif /* not WINDOWSNT */
#endif /* not BSD_SYSTEM, and not HPUX version >= 6 */
if (interruptible)
@@ -460,7 +472,7 @@
struct save_signal
{
int code;
- void (*handler) (int);
+ signal_handler_t handler;
};
static void save_signal_handlers (struct save_signal *);
@@ -618,8 +630,7 @@
{
while (saved_handlers->code)
{
- saved_handlers->handler
- = (void (*) (int)) signal (saved_handlers->code, SIG_IGN);
+ saved_handlers->handler = sys_signal (saved_handlers->code, SIG_IGN);
saved_handlers++;
}
}
@@ -629,7 +640,7 @@
{
while (saved_handlers->code)
{
- signal (saved_handlers->code, saved_handlers->handler);
+ set_sighandler (saved_handlers->code, saved_handlers->handler);
saved_handlers++;
}
}
@@ -690,9 +701,9 @@
return;
#ifdef SIGWINCH
- sigunblock (sigmask (SIGWINCH));
+ unblock_signal (SIGWINCH);
#endif
- sigunblock (sigmask (SIGIO));
+ unblock_signal (SIGIO);
interrupts_deferred = 0;
}
@@ -709,9 +720,9 @@
#endif
#ifdef SIGWINCH
- sigblock (sigmask (SIGWINCH));
+ block_signal (SIGWINCH);
#endif
- sigblock (sigmask (SIGIO));
+ block_signal (SIGIO);
interrupts_deferred = 1;
}
@@ -1473,16 +1484,27 @@
/* POSIX signals support - DJB */
/* Anyone with POSIX signals should have ANSI C declarations */
-sigset_t empty_mask;
-
#ifndef WINDOWSNT
+/* Return the current signal handler for signal number SIGNO. */
signal_handler_t
-sys_signal (int signal_number, signal_handler_t action)
-{
- struct sigaction new_action, old_action;
+sighandler (int signo)
+{
+ struct sigaction old_action;
+ sigaction (signo, NULL, &old_action);
+ return old_action.sa_handler;
+}
+
+/* Copy the old signal action for SIGNO into *OLD_ACTION if OLD_ACTION
+ is non-null. Then change the signal action to be one whose handler
+ is HANDLER. */
+static void
+get_set_sighandler (int signo, struct sigaction *old_action,
+ signal_handler_t handler)
+{
+ struct sigaction new_action;
sigemptyset (&new_action.sa_mask);
- new_action.sa_handler = action;
+ new_action.sa_handler = handler;
new_action.sa_flags = 0;
#if defined (SA_RESTART)
/* Emacs mostly works better with restartable system services. If this
@@ -1502,54 +1524,26 @@
# endif
new_action.sa_flags = SA_RESTART;
#endif
- sigaction (signal_number, &new_action, &old_action);
- return (old_action.sa_handler);
+ sigaction (signo, &new_action, old_action);
+}
+
+/* Change SIGNO's handler to be HANDLER. */
+void
+set_sighandler (int signo, signal_handler_t handler)
+{
+ get_set_sighandler (signo, NULL, handler);
+}
+
+/* Change SIGNO's handler to be HANDLER. Return the old handler. */
+signal_handler_t
+sys_signal (int signo, signal_handler_t handler)
+{
+ struct sigaction old_action;
+ get_set_sighandler (signo, &old_action, handler);
+ return old_action.sa_handler;
}
#endif /* WINDOWSNT */
-
-#ifndef __GNUC__
-/* If we're compiling with GCC, we don't need this function, since it
- can be written as a macro. */
-sigset_t
-sys_sigmask (int sig)
-{
- sigset_t mask;
- sigemptyset (&mask);
- sigaddset (&mask, sig);
- return mask;
-}
-#endif
-
-/* I'd like to have these guys return pointers to the mask storage in here,
- but there'd be trouble if the code was saving multiple masks. I'll be
- safe and pass the structure. It normally won't be more than 2 bytes
- anyhow. - DJB */
-
-sigset_t
-sys_sigblock (sigset_t new_mask)
-{
- sigset_t old_mask;
- pthread_sigmask (SIG_BLOCK, &new_mask, &old_mask);
- return (old_mask);
-}
-
-sigset_t
-sys_sigunblock (sigset_t new_mask)
-{
- sigset_t old_mask;
- pthread_sigmask (SIG_UNBLOCK, &new_mask, &old_mask);
- return (old_mask);
-}
-
-sigset_t
-sys_sigsetmask (sigset_t new_mask)
-{
- sigset_t old_mask;
- pthread_sigmask (SIG_SETMASK, &new_mask, &old_mask);
- return (old_mask);
-}
-
#if !defined HAVE_STRSIGNAL && !HAVE_DECL_SYS_SIGLIST
static char *my_sys_siglist[NSIG];
@@ -1562,8 +1556,6 @@
void
init_signals (void)
{
- sigemptyset (&empty_mask);
-
#if !defined HAVE_STRSIGNAL && !HAVE_DECL_SYS_SIGLIST
if (! initialized)
{
=== modified file 'src/syssignal.h'
--- src/syssignal.h 2012-07-13 01:19:06 +0000
+++ src/syssignal.h 2012-09-01 21:37:07 +0000
@@ -17,7 +17,8 @@
You should have received a copy of the GNU General Public License
along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>. */
-extern void init_signals (void);
+#include <signal.h>
+#include <stdbool.h>
#ifdef HAVE_PTHREAD
#include <pthread.h>
@@ -26,63 +27,65 @@
#define FORWARD_SIGNAL_TO_MAIN_THREAD
#endif
-/* Don't #include <signal.h>. That header should always be #included
- before "config.h", because some configuration files (like s/hpux.h)
- indicate that SIGIO doesn't work by #undef-ing SIGIO. If this file
- #includes <signal.h>, then that will re-#define SIGIO and confuse
- things. */
-/* XXX This is not correct anymore, there is a BROKEN_SIGIO macro. */
-
-#define SIGMASKTYPE sigset_t
-
-#define SIGEMPTYMASK (empty_mask)
-extern sigset_t empty_mask;
-
-/* POSIX pretty much destroys any possibility of writing sigmask as a
- macro in standard C. We always define our own version because the
- predefined macro in Glibc 2.1 is only provided for compatibility for old
- programs that use int as signal mask type. */
-#undef sigmask
-#ifdef __GNUC__
-#define sigmask(SIG) \
- ({ \
- sigset_t _mask; \
- sigemptyset (&_mask); \
- sigaddset (&_mask, SIG); \
- _mask; \
- })
-#else /* ! defined (__GNUC__) */
-extern sigset_t sys_sigmask ();
-#define sigmask(SIG) (sys_sigmask (SIG))
-#endif /* ! defined (__GNUC__) */
-
-#undef sigpause
-#define sigpause(MASK) sigsuspend (&(MASK))
-
-#define sigblock(SIG) sys_sigblock (SIG)
-#define sigunblock(SIG) sys_sigunblock (SIG)
-#ifndef sigsetmask
-#define sigsetmask(SIG) sys_sigsetmask (SIG)
-#endif
-#undef signal
-#define signal(SIG,ACT) sys_signal(SIG,ACT)
-
-/* Whether this is what all systems want or not, this is what
- appears to be assumed in the source, for example data.c:arith_error. */
+INLINE_HEADER_BEGIN
+#ifndef SYSSIGNAL_INLINE
+# define SYSSIGNAL_INLINE INLINE
+#endif
+
+extern void init_signals (void)
+#if defined HAVE_STRSIGNAL || HAVE_DECL_SYS_SIGLIST
+ ATTRIBUTE_CONST
+#endif
+ ;
+
+SYSSIGNAL_INLINE void
+get_sigmask_block_signal (sigset_t *old_mask, int sig)
+{
+ sigset_t mask;
+ sigemptyset (&mask);
+ sigaddset (&mask, sig);
+ pthread_sigmask (SIG_BLOCK, &mask, old_mask);
+}
+
+SYSSIGNAL_INLINE void
+block_signal (int sig)
+{
+ get_sigmask_block_signal (NULL, sig);
+}
+
+SYSSIGNAL_INLINE void
+unblock_signal (int sig)
+{
+ sigset_t mask;
+ sigemptyset (&mask);
+ sigaddset (&mask, sig);
+ pthread_sigmask (SIG_UNBLOCK, &mask, NULL);
+}
+
+SYSSIGNAL_INLINE void
+set_sigmask (sigset_t *mask)
+{
+ pthread_sigmask (SIG_SETMASK, mask, NULL);
+}
+
+SYSSIGNAL_INLINE void
+sigfree (void)
+{
+ sigset_t empty_mask;
+ sigemptyset (&empty_mask);
+ set_sigmask (&empty_mask);
+}
+
typedef void (*signal_handler_t) (int);
-signal_handler_t sys_signal (int signal_number, signal_handler_t action);
-sigset_t sys_sigblock (sigset_t new_mask);
-sigset_t sys_sigunblock (sigset_t new_mask);
-sigset_t sys_sigsetmask (sigset_t new_mask);
+signal_handler_t sighandler (int);
+void set_sighandler (int, signal_handler_t);
+signal_handler_t sys_signal (int, signal_handler_t);
+
#if ! (defined TIOCNOTTY || defined USG5 || defined CYGWIN)
_Noreturn void croak (char *);
#endif
-#define sys_sigdel(MASK,SIG) sigdelset (&MASK,SIG)
-
-#define sigfree() sigsetmask (SIGEMPTYMASK)
-
#if defined (SIGIO) && defined (BROKEN_SIGIO)
# undef SIGIO
#endif
@@ -97,12 +100,8 @@
#undef SIGPTY
#endif
-
-/* FIXME? Emacs only defines NSIG_MINIMUM on some platforms? */
#if NSIG < NSIG_MINIMUM
-# ifdef NSIG
-# undef NSIG
-# endif
+# undef NSIG
# define NSIG NSIG_MINIMUM
#endif
@@ -133,24 +132,20 @@
#ifdef FORWARD_SIGNAL_TO_MAIN_THREAD
extern pthread_t main_thread;
-#define SIGNAL_THREAD_CHECK(signo) \
- do { \
- if (!pthread_equal (pthread_self (), main_thread)) \
- { \
- /* POSIX says any thread can receive the signal. On GNU/Linux \
- that is not true, but for other systems (FreeBSD at least) \
- it is. So direct the signal to the correct thread and block \
- it from this thread. */ \
- sigset_t new_mask; \
- \
- sigemptyset (&new_mask); \
- sigaddset (&new_mask, signo); \
- pthread_sigmask (SIG_BLOCK, &new_mask, 0); \
- pthread_kill (main_thread, signo); \
- return; \
- } \
- } while (0)
-
-#else /* not FORWARD_SIGNAL_TO_MAIN_THREAD */
-#define SIGNAL_THREAD_CHECK(signo)
-#endif /* not FORWARD_SIGNAL_TO_MAIN_THREAD */
+#endif
+
+SYSSIGNAL_INLINE bool
+forwarded_signal (int signo)
+{
+#ifdef FORWARD_SIGNAL_TO_MAIN_THREAD
+ if (! pthread_equal (pthread_self (), main_thread))
+ {
+ block_signal (signo);
+ pthread_kill (main_thread, signo);
+ return 1;
+ }
+#endif
+ return 0;
+}
+
+INLINE_HEADER_END
=== modified file 'src/term.c'
--- src/term.c 2012-08-31 10:53:19 +0000
+++ src/term.c 2012-09-01 21:37:07 +0000
@@ -25,7 +25,6 @@
#include <sys/file.h>
#include <sys/time.h>
#include <unistd.h>
-#include <signal.h>
#include <setjmp.h>
#include "lisp.h"
@@ -2932,7 +2931,7 @@
no_controlling_tty = 1;
#else
#ifdef TIOCNOTTY /* Try BSD ioctls. */
- sigblock (sigmask (SIGTTOU));
+ block_signal (SIGTTOU);
fd = emacs_open (DEV_TTY, O_RDWR, 0);
if (fd != -1 && ioctl (fd, TIOCNOTTY, 0) != -1)
{
@@ -2940,7 +2939,7 @@
}
if (fd != -1)
emacs_close (fd);
- sigunblock (sigmask (SIGTTOU));
+ unblock_signal (SIGTTOU);
#else
/* Unknown system. */
croak ();
@@ -3074,9 +3073,9 @@
/* On some systems, tgetent tries to access the controlling
terminal. */
- sigblock (sigmask (SIGTTOU));
+ block_signal (SIGTTOU);
status = tgetent (tty->termcap_term_buffer, terminal_type);
- sigunblock (sigmask (SIGTTOU));
+ unblock_signal (SIGTTOU);
if (status < 0)
{
=== modified file 'src/widget.c'
--- src/widget.c 2012-07-10 21:48:34 +0000
+++ src/widget.c 2012-09-01 21:37:07 +0000
@@ -50,9 +50,6 @@
#include <X11/ShellP.h>
#include "../lwlib/lwlib.h"
-#include <signal.h>
-#include "syssignal.h"
-
#include "character.h"
#include "font.h"
=== modified file 'src/xmenu.c'
--- src/xmenu.c 2012-08-17 21:52:15 +0000
+++ src/xmenu.c 2012-09-01 21:37:07 +0000
@@ -32,11 +32,6 @@
#include <config.h>
-#if 0 /* Why was this included? And without syssignal.h? */
-/* On 4.3 this loses if it comes after xterm.h. */
-#include <signal.h>
-#endif
-
#include <stdio.h>
#include <setjmp.h>
=== modified file 'src/xterm.c'
--- src/xterm.c 2012-08-25 20:31:04 +0000
+++ src/xterm.c 2012-09-01 21:37:07 +0000
@@ -21,7 +21,6 @@
/* Xt features made by Fred Pierresteguy. */
#include <config.h>
-#include <signal.h>
#include <stdio.h>
#include <setjmp.h>
@@ -29,9 +28,6 @@
#include "lisp.h"
#include "blockinput.h"
-
-/* Need syssignal.h for various externs and definitions that may be required
- by some configurations for calls to signal later in this source file. */
#include "syssignal.h"
/* This may include sys/types.h, and that somehow loses
@@ -7773,7 +7769,7 @@
#ifdef USG
/* USG systems forget handlers when they are used;
must reestablish each time */
- signal (signalnum, x_connection_signal);
+ set_sighandler (signalnum, x_connection_signal);
#endif /* USG */
}
@@ -7884,9 +7880,9 @@
/* Ordinary stack unwind doesn't deal with these. */
#ifdef SIGIO
- sigunblock (sigmask (SIGIO));
+ unblock_signal (SIGIO);
#endif
- sigunblock (sigmask (SIGALRM));
+ unblock_signal (SIGALRM);
TOTALLY_UNBLOCK_INPUT;
unbind_to (idx, Qnil);
@@ -10814,7 +10810,7 @@
XSetErrorHandler (x_error_handler);
XSetIOErrorHandler (x_io_error_quitter);
- signal (SIGPIPE, x_connection_signal);
+ set_sighandler (SIGPIPE, x_connection_signal);
}
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Sun, 02 Sep 2012 17:54:02 GMT) Full text and rfc822 format available.Message #8 received at submit <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: Paul Eggert <eggert <at> cs.ucla.edu> Cc: lekktu <at> gmail.com, bug-gnu-emacs <at> gnu.org Subject: Re: Signal-handler cleanup for Emacs Date: Sun, 02 Sep 2012 20:51:51 +0300
> Date: Sat, 01 Sep 2012 15:38:15 -0700 > From: Paul Eggert <eggert <at> cs.ucla.edu> > CC: Eli Zaretskii <eliz <at> gnu.org>, Juanma Barranquero <lekktu <at> gmail.com> > > Here's a patch to clean up Emacs's signal-handling code somewhat. > As described below, it shrinks the size of Emacs's text segment > by 0.6% on my platform. It also removes a few dozen lines from > the source code. FWIW, I cannot see any advantage in these changes. They just replace one system of macros by another. The advantage of the current code is that it invokes widely known standard APIs, such as 'signal', 'sigemptyset', and 'sigblock' (never mind that we redefine some of them to call their more modern equivalents: that doesn't impede code reading and understanding in any way). OTOH, the changes you propose replace that with APIs that AFAIK are not known to anyone and, with a couple of exceptions, not used anywhere else. I could go with replacing 'signal' etc. with their modern Posix replacements, such as 'sigaction', directly in the code. That will clean up old APIs while still maintaining clarity for anyone who has ever written signal handlers. But I fail to see any good reasons for changes that, e.g., hide a pair of calls to well-known library functions, such as 'sigemptyset' and 'sigaddset', behind 'sigsetmask' (which AFAIK is a BSD-ism), or behind newly-invented functions such as 'get_sigmask_block_signal'. As for 0.6% reduction in the size of .text: what kind of humongous .text size do you have that makes 0.6% a significant value? On 2 different platforms to which I have access, one of them a x86_64 GNU/Linux, I see ~3.5MB in an unoptimized build and about half that much in an optimized one, which makes the savings around 2KB, too small to justify any change whatsoever. What am I missing?
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Sun, 02 Sep 2012 18:39:02 GMT) Full text and rfc822 format available.Message #11 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: eggert <at> cs.ucla.edu, lekktu <at> gmail.com Cc: 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Sun, 02 Sep 2012 21:37:21 +0300
> Date: Sun, 02 Sep 2012 20:51:51 +0300 > From: Eli Zaretskii <eliz <at> gnu.org> > Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org > > As for 0.6% reduction in the size of .text: what kind of humongous > .text size do you have that makes 0.6% a significant value? On 2 > different platforms to which I have access, one of them a x86_64 > GNU/Linux, I see ~3.5MB in an unoptimized build and about half that > much in an optimized one, which makes the savings around 2KB, too > small to justify any change whatsoever. ^^^ Err, make that 21KB, still too small.
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Sun, 02 Sep 2012 19:03:02 GMT) Full text and rfc822 format available.Message #14 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Paul Eggert <eggert <at> cs.ucla.edu> To: Eli Zaretskii <eliz <at> gnu.org> Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: Signal-handler cleanup for Emacs Date: Sun, 02 Sep 2012 12:01:11 -0700
On 09/02/2012 10:51 AM, Eli Zaretskii wrote: > that doesn't impede code reading and understanding in any way Sure it does, because Emacs defines symbols incompatibly with their normal meanings. E.g., Emacs redefines 'sigblock' to be a macro whose type signature is incompatible with the 'sigblock' that is found on GNU/Linux. This sort of thing makes Emacs's signal-handling code more confusing than it needs to be, and is the main motivation for the patch. > I could go with replacing 'signal' etc. with their modern Posix > replacements, such as 'sigaction', directly in the code. That would make the mainline code significantly less readable. Instead of this, for example: unblock_signal (SIGPIPE); send_process_trap would have to do something like this: sigset_t mask; ... sigemptyset (&mask); sigaddset (&mask, SIGPIPE); pthread_sigmask (SIG_UNBLOCK, &mask, NULL); which is not an improvement. I suspect the main reason that Emacs currently uses macros reminiscent but incompatible with the obsolete 4.2BSD interface, rather than the standard POSIX interface, is because the 4.2BSD interface was simpler. Unfortunately, the 4.2BSD interface does not scale well to modern systems with lots of signals. > I fail to see any good reasons for changes that, e.g., hide a > pair of calls to well-known library functions, such as 'sigemptyset' > and 'sigaddset', behind 'sigsetmask' (which AFAIK is a BSD-ism) There must be some confusion here. The new code does not use sigsetmask. > As for 0.6% reduction in the size of .text: what kind of humongous > .text size do you have that makes 0.6% a significant value? Every little bit helps, no? But the main point of measuring text size is as a rough gauge of the efficiency implications. If the text segment had grown, that would have been a bad sign. The fact that it shrank is a good sign. Fewer instructions mean faster CPU performance, partly due to lower pressure on the instruction cache. I didn't bother to measure CPU performance earlier, since that takes more work, but I just now did that with an artificial benchmark that simply blocks and then unblocks SIGCHLD, and on my platform the proposed patch speeds up this benchmark by 15%.
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Sun, 02 Sep 2012 21:23:02 GMT) Full text and rfc822 format available.Message #17 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: Paul Eggert <eggert <at> cs.ucla.edu> Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: Signal-handler cleanup for Emacs Date: Mon, 03 Sep 2012 00:20:42 +0300
> Date: Sun, 02 Sep 2012 12:01:11 -0700 > From: Paul Eggert <eggert <at> cs.ucla.edu> > CC: 12327 <at> debbugs.gnu.org, lekktu <at> gmail.com > > On 09/02/2012 10:51 AM, Eli Zaretskii wrote: > > > that doesn't impede code reading and understanding in any way > > Sure it does, because Emacs defines symbols incompatibly with their > normal meanings. E.g., Emacs redefines 'sigblock' to be a macro > whose type signature is incompatible with the 'sigblock' > that is found on GNU/Linux. If that's the problem, then fix that, by changing the type. Why do more than is strictly needed? Gratuitous changes are BAAAD! > > I could go with replacing 'signal' etc. with their modern Posix > > replacements, such as 'sigaction', directly in the code. > > That would make the mainline code significantly less readable. > Instead of this, for example: > > unblock_signal (SIGPIPE); > > send_process_trap would have to do something like this: > > sigset_t mask; > ... > sigemptyset (&mask); > sigaddset (&mask, SIGPIPE); > pthread_sigmask (SIG_UNBLOCK, &mask, NULL); > > which is not an improvement. Of course, it's an improvement! That kind of code is a pattern programmers are used to see when signals are being masked. By contrast, when I see a call to unblock_signal, I need to look it up. > I suspect the main reason that Emacs currently uses macros reminiscent > but incompatible with the obsolete 4.2BSD interface, rather than the > standard POSIX interface, is because the 4.2BSD interface was simpler. No, it's just because the BSD interface was there first. > Unfortunately, the 4.2BSD interface does not scale well to modern > systems with lots of signals. I suggested to use sigaction and friends, not go back to the old APIs. > > I fail to see any good reasons for changes that, e.g., hide a > > pair of calls to well-known library functions, such as 'sigemptyset' > > and 'sigaddset', behind 'sigsetmask' (which AFAIK is a BSD-ism) > > There must be some confusion here. Yeah, did I say the new code is confusing? > > As for 0.6% reduction in the size of .text: what kind of humongous > > .text size do you have that makes 0.6% a significant value? > > Every little bit helps, no? Not every little bit can be justified by such non-trivial changes. Changes come at a cost, which should be justified. > But the main point of measuring text size > is as a rough gauge of the efficiency implications. If the text > segment had grown, that would have been a bad sign. Not necessarily. And certainly not at fractions of a percent. > Fewer instructions mean faster CPU performance, partly due to lower > pressure on the instruction cache. This all smells premature optimization to me. Text segment size has no direct relation to pressure on the instruction cache. What matters is the structure of the code in terms of branches and loops, and its locality. The size alone is almost meaningless. I'm sure you know that. > I didn't bother to measure CPU performance earlier, since that takes > more work, but I just now did that with an artificial benchmark that > simply blocks and then unblocks SIGCHLD, and on my platform the > proposed patch speeds up this benchmark by 15%. A meaningless benchmark, IMO. You need to measure an Emacs loop that does these block and unblock operations (if there is such a loop), and see the effect on that. _That_ would be interesting.
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Mon, 03 Sep 2012 08:18:02 GMT) Full text and rfc822 format available.Message #20 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Paul Eggert <eggert <at> cs.ucla.edu> To: Eli Zaretskii <eliz <at> gnu.org> Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: Signal-handler cleanup for Emacs Date: Mon, 03 Sep 2012 01:16:08 -0700
[Message part 1 (text/plain, inline)]
On 09/02/2012 02:20 PM, Eli Zaretskii wrote: > If that's the problem, then fix that, by changing the type. If I understand this suggestion correctly, it's not practical. GNU/Linux sigblock returns 'int', but its signal masks require 1024 bits, too wide for 'int'. This is why sigblock has been removed from POSIX and is documented as obsolete in GNU/Linux. > Changes come at a cost, which should be justified. My point was that the new version is smaller and faster. How much smaller and faster doesn't matter much, as the primary justification for the change is to reduce the confusion caused by the current setup. > I could go with replacing 'signal' etc. with their modern Posix > replacements OK, thanks, I reworked the patch to implement that suggestion. I also modified it a bit in other ways, to boost the performance a bit (as that suggestion caused a small performance hit). The result is attached. I expect the Windows port will need a few tweaks.
[syssignal.txt (text/plain, attachment)]
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Mon, 03 Sep 2012 08:52:01 GMT) Full text and rfc822 format available.Message #23 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Andreas Schwab <schwab <at> linux-m68k.org> To: Paul Eggert <eggert <at> cs.ucla.edu> Cc: lekktu <at> gmail.com, Eli Zaretskii <eliz <at> gnu.org>, 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Mon, 03 Sep 2012 10:49:22 +0200
Paul Eggert <eggert <at> cs.ucla.edu> writes: > /* Must reestablish handler each time it is called. */ That's not true for POSIX signals. Andreas. -- Andreas Schwab, schwab <at> linux-m68k.org GPG Key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5 214B 8276 4ED5 "And now for something completely different."
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Mon, 03 Sep 2012 09:04:01 GMT) Full text and rfc822 format available.Message #26 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Paul Eggert <eggert <at> cs.ucla.edu> To: Andreas Schwab <schwab <at> linux-m68k.org> Cc: lekktu <at> gmail.com, Eli Zaretskii <eliz <at> gnu.org>, 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Mon, 03 Sep 2012 02:02:03 -0700
On 09/03/2012 01:49 AM, Andreas Schwab wrote: > That's not true for POSIX signals. Yes, thanks, that's right. Further cleanup is needed, both there and elsewhere. I've avoided these further cleanups in order to keep the size of the patch down.
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Mon, 03 Sep 2012 09:19:02 GMT) Full text and rfc822 format available.Message #29 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Andreas Schwab <schwab <at> linux-m68k.org> To: Paul Eggert <eggert <at> cs.ucla.edu> Cc: lekktu <at> gmail.com, Eli Zaretskii <eliz <at> gnu.org>, 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Mon, 03 Sep 2012 11:17:02 +0200
Paul Eggert <eggert <at> cs.ucla.edu> writes: > On 09/03/2012 01:49 AM, Andreas Schwab wrote: >> That's not true for POSIX signals. > > Yes, thanks, that's right. Further cleanup is needed, > both there and elsewhere. I've avoided these further > cleanups in order to keep the size of the patch down. But that's really part of the signal->sigaction move. Andreas. -- Andreas Schwab, schwab <at> linux-m68k.org GPG Key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5 214B 8276 4ED5 "And now for something completely different."
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Mon, 03 Sep 2012 15:35:02 GMT) Full text and rfc822 format available.Message #32 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: Paul Eggert <eggert <at> cs.ucla.edu> Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: Signal-handler cleanup for Emacs Date: Mon, 03 Sep 2012 18:32:06 +0300
> Date: Mon, 03 Sep 2012 01:16:08 -0700 > From: Paul Eggert <eggert <at> cs.ucla.edu> > CC: 12327 <at> debbugs.gnu.org, lekktu <at> gmail.com > > > I could go with replacing 'signal' etc. with their modern Posix > > replacements > > OK, thanks, I reworked the patch to implement that suggestion. Thanks.
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Tue, 04 Sep 2012 08:13:01 GMT) Full text and rfc822 format available.Message #35 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Paul Eggert <eggert <at> cs.ucla.edu> To: Andreas Schwab <schwab <at> linux-m68k.org> Cc: lekktu <at> gmail.com, Eli Zaretskii <eliz <at> gnu.org>, 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Tue, 04 Sep 2012 01:12:17 -0700
[Message part 1 (text/plain, inline)]
On 09/03/2012 02:17 AM, Andreas Schwab wrote: > But that's really part of the signal->sigaction move. OK, I looked into that, and several other places are also affected by the move. Attached is a patch to fix the issue that you found, and the other related issues that I saw. This assumes the patch I already sent in <http://bugs.gnu.org/12327#20>.
[syssignao.txt (text/plain, attachment)]
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Thu, 06 Sep 2012 12:02:01 GMT) Full text and rfc822 format available.Message #38 received at submit <at> debbugs.gnu.org (full text, mbox):
From: Andy Moreton <andrewjmoreton <at> gmail.com> To: bug-gnu-emacs <at> gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Thu, 06 Sep 2012 12:59:47 +0100
On Tue 04 Sep 2012, Paul Eggert wrote:
> On 09/03/2012 02:17 AM, Andreas Schwab wrote:
>> But that's really part of the signal->sigaction move.
>
> OK, I looked into that, and several other places are also
> affected by the move. Attached is a patch to fix the issue
> that you found, and the other related issues that I saw. This
> assumes the patch I already sent in <http://bugs.gnu.org/12327#20>.
This patch set (or something similar) was committed in r109893, and
causes a crash in the Windows build (r109892 is ok).
Crash is in r109893 built on Windows XP using MingGW gcc 4.6.2.
To reproduce it, run "M-x rgrep" using Cygwin grep. The grep
output is all displayed in the *grep* buffer, and then emacs aborts.
Attaching gdb gives the output below.
(gdb) thread apply all bt full
Thread 6 (Thread 2072.0x9fc):
#0 0x7c90e514 in ntdll!LdrAccessResource () from C:\WINDOWS\system32\ntdll.dll
No symbol table info available.
#1 0x7c90d9da in ntdll!ZwReadFile () from C:\WINDOWS\system32\ntdll.dll
No symbol table info available.
#2 0x7c801879 in ReadFile () from C:\WINDOWS\system32\kernel32.dll
No symbol table info available.
#3 0x00000628 in ?? ()
No symbol table info available.
#4 0x00000000 in ?? ()
No symbol table info available.
Thread 5 (Thread 2072.0xdfc):
#0 0x7c90e514 in ntdll!LdrAccessResource () from C:\WINDOWS\system32\ntdll.dll
No symbol table info available.
#1 0x7c90da4a in ntdll!ZwRemoveIoCompletion () from C:\WINDOWS\system32\ntdll.dll
No symbol table info available.
#2 0x7c80a7e6 in KERNEL32!GetQueuedCompletionStatus () from C:\WINDOWS\system32\kernel32.dll
No symbol table info available.
#3 0x77e7715c in RPCRT4!I_RpcBindingCopy () from C:\WINDOWS\system32\rpcrt4.dll
No symbol table info available.
#4 0x77e772a0 in RPCRT4!I_RpcBindingCopy () from C:\WINDOWS\system32\rpcrt4.dll
No symbol table info available.
#5 0x77e77328 in RPCRT4!I_RpcBindingCopy () from C:\WINDOWS\system32\rpcrt4.dll
No symbol table info available.
#6 0x77e76ad1 in RPCRT4!I_RpcBindingCopy () from C:\WINDOWS\system32\rpcrt4.dll
No symbol table info available.
#7 0x77e76c97 in RPCRT4!I_RpcBindingCopy () from C:\WINDOWS\system32\rpcrt4.dll
No symbol table info available.
#8 0x7c80b729 in KERNEL32!GetModuleFileNameA () from C:\WINDOWS\system32\kernel32.dll
No symbol table info available.
#9 0x00000000 in ?? ()
No symbol table info available.
Thread 4 (Thread 2072.0x358):
#0 0x7c90e514 in ntdll!LdrAccessResource () from C:\WINDOWS\system32\ntdll.dll
No symbol table info available.
#1 0x7c90da4a in ntdll!ZwRemoveIoCompletion () from C:\WINDOWS\system32\ntdll.dll
No symbol table info available.
#2 0x7c80a7e6 in KERNEL32!GetQueuedCompletionStatus () from C:\WINDOWS\system32\kernel32.dll
No symbol table info available.
#3 0x77e7715c in RPCRT4!I_RpcBindingCopy () from C:\WINDOWS\system32\rpcrt4.dll
No symbol table info available.
#4 0x77e772a0 in RPCRT4!I_RpcBindingCopy () from C:\WINDOWS\system32\rpcrt4.dll
No symbol table info available.
#5 0x77e77328 in RPCRT4!I_RpcBindingCopy () from C:\WINDOWS\system32\rpcrt4.dll
No symbol table info available.
#6 0x77e76ad1 in RPCRT4!I_RpcBindingCopy () from C:\WINDOWS\system32\rpcrt4.dll
No symbol table info available.
#7 0x77e76c97 in RPCRT4!I_RpcBindingCopy () from C:\WINDOWS\system32\rpcrt4.dll
No symbol table info available.
#8 0x7c80b729 in KERNEL32!GetModuleFileNameA () from C:\WINDOWS\system32\kernel32.dll
No symbol table info available.
#9 0x00000000 in ?? ()
No symbol table info available.
Thread 3 (Thread 2072.0x1dc):
#0 0x7c90e514 in ntdll!LdrAccessResource () from C:\WINDOWS\system32\ntdll.dll
No symbol table info available.
#1 0x7c90df5a in ntdll!ZwWaitForSingleObject () from C:\WINDOWS\system32\ntdll.dll
No symbol table info available.
#2 0x7c8025db in WaitForSingleObjectEx () from C:\WINDOWS\system32\kernel32.dll
No symbol table info available.
#3 0x00000620 in ?? ()
No symbol table info available.
#4 0x00000000 in ?? ()
No symbol table info available.
Thread 2 (Thread 2072.0x410):
#0 0x7c90e514 in ntdll!LdrAccessResource () from C:\WINDOWS\system32\ntdll.dll
No symbol table info available.
#1 0x77f16bf2 in GdiDrawStream () from C:\WINDOWS\system32\gdi32.dll
No symbol table info available.
#2 0x00fa2b1b in ?? () from C:\WINDOWS\system32\uxtheme.dll
No symbol table info available.
#3 0x00fa289b in ?? () from C:\WINDOWS\system32\uxtheme.dll
No symbol table info available.
#4 0x00fa2753 in ?? () from C:\WINDOWS\system32\uxtheme.dll
No symbol table info available.
#5 0x00fa2c9f in UxTheme!DrawThemeParentBackground () from C:\WINDOWS\system32\uxtheme.dll
No symbol table info available.
#6 0x00fa5f42 in UxTheme!SetThemeAppProperties () from C:\WINDOWS\system32\uxtheme.dll
No symbol table info available.
#7 0x00fa6485 in UxTheme!SetThemeAppProperties () from C:\WINDOWS\system32\uxtheme.dll
No symbol table info available.
#8 0x00fbc4b9 in UxTheme!GetThemeTextMetrics () from C:\WINDOWS\system32\uxtheme.dll
No symbol table info available.
#9 0x00fa1ac7 in ?? () from C:\WINDOWS\system32\uxtheme.dll
No symbol table info available.
#10 0x00fa1b3d in ?? () from C:\WINDOWS\system32\uxtheme.dll
No symbol table info available.
#11 0x7e4294ed in USER32!GetPropW () from C:\WINDOWS\system32\user32.dll
No symbol table info available.
#12 0x0114b954 in w32_wnd_proc (hwnd=0xa036c, msg=0x85, wParam=0x870409db, lParam=0x0) at w32fns.c:3832
f = 0xa45740
dpyinfo = 0x16626c0
wmsg = {
msg = {
hwnd = 0x12,
message = 0x7240f7c8,
wParam = 0xfc505d,
lParam = 0xfbe152,
time = 0xfbe535,
pt = {
x = 0x7240f838,
y = 0x6
}
},
dwModifiers = 0x6,
rect = {
left = 0x11,
top = 0x11,
right = 0x7240f87c,
bottom = 0x11
}
}
windows_translate = 0xfa39d6
key = 0xa45740
#13 0x7e418734 in USER32!GetDC () from C:\WINDOWS\system32\user32.dll
No symbol table info available.
#14 0x000a036c in ?? ()
No symbol table info available.
#15 0x00000085 in ?? ()
No symbol table info available.
#16 0x870409db in ?? ()
No symbol table info available.
#17 0x00000000 in ?? ()
No symbol table info available.
Thread 1 (Thread 2072.0x564):
#0 0x7c90120f in ntdll!DbgUiConnectToDbg () from C:\WINDOWS\system32\ntdll.dll
No symbol table info available.
#1 0x0115317e in emacs_abort () at w32fns.c:7215
button = 0x6
#2 0x010539f2 in sys_wait (status=0x82f264) at w32proc.c:462
active = 0x0
retval = 0x0
nh = 0x0
pid = 0xc38
cp = 0x165f430
cps = {0x165f430, 0x82f1d8, 0x82f490, 0x82fcdc, 0x1, 0x8, 0x353c81a, 0x0, 0x82f23c, 0x132c256, 0x60, 0x0, 0x82f99c, 0x82f3e8, 0x10e02e4, 0x5, 0x7c90df4a, 0x7c809590, 0x7ffdf000, 0x82f280, 0x7c8095d2, 0x82f22c, 0x7c8095c0, 0x0, 0x8, 0x0, 0x714, 0x710, 0x68c, 0x644, 0x5fc, 0x600}
wait_hnd = {0x0, 0x82f168, 0x1028d66, 0x3555ad2, 0x353c832, 0x82f1d8, 0x1036669, 0x13f203d, 0x3, 0x132c24e, 0x353c81a, 0x163a5e8, 0x82f15c, 0x82fcdc, 0x3555ad0, 0x353c81a, 0x82fcf0, 0x82f1d8, 0x10380e2, 0x3555ad2, 0x353c832, 0x0, 0x82f12c, 0x10320ff, 0x82ffe0, 0x0, 0x3555ad2, 0x353c832, 0x0, 0x0, 0x3, 0x35554aa}
#3 0x010508a0 in record_child_status_change () at process.c:6422
proc = 0x534cee5
p = 0x534cee0
pid = 0xc38
w = 0x0
tail = 0x536717e
#4 0x01050ee0 in handle_child_signal (sig=0x12) at process.c:6529
No locals.
#5 0x01143cfe in handle_on_main_thread (sig=0x12, handler=0x1050ed1 <handle_child_signal>) at sysdep.c:1584
old_errno = 0x16
on_main_thread = 0x1
#6 0x01050eff in deliver_child_signal (sig=0x12) at process.c:6536
No locals.
#7 0x01054d67 in sys_select (nfds=0xa, rfds=0x82f6e4, wfds=0x0, efds=0x0, timeout=0x82f6d4, ignored=0x0) at w32proc.c:1325
orfds = {
bits = {0x309, 0x0}
}
timeout_ms = 0xfa
start_time = 0x5e2323e
i = 0xa
nh = 0x5
nc = 0x2
nr = 0x1
active = 0x5
cp = 0x165f430
cps = {0x165f430, 0x165f3d8, 0x82f5bc, 0xee6b280, 0x0, 0xee6b280, 0x30, 0x0 <repeats 17 times>, 0xa, 0x200, 0x82f628, 0x1014eea, 0x0, 0x0, 0x0, 0x353c81a}
wait_hnd = {0x714, 0x710, 0x68c, 0x644, 0x5fc, 0x600, 0x624, 0x82f5bc, 0x1, 0x82f5e8, 0x82f598, 0x12c0b6b, 0x82f4e8, 0x0, 0x850000, 0x0, 0x1645970, 0x86a4d0, 0x850000, 0x1028d66, 0x15ad2, 0x8, 0x82f3d4, 0x850000, 0x82f4c8, 0xe11f, 0x82f4b8, 0x105917e, 0x50488ca1, 0x36d61600, 0x82f4d8, 0x1103595, 0x50488ca1, 0x36d61600, 0x29e, 0x1, 0x82f99c, 0x3c8c90c, 0x82f4e8, 0x5048, 0x0, 0xe09c0, 0x8ca1, 0x5048, 0x7c80a0e9, 0x0, 0x82f518, 0x100dd13, 0x0, 0xee6b280, 0xee6b280, 0xee6b280, 0x50488ca1, 0xe11f, 0x82f528, 0x105917e, 0x50488ca1, 0x36d61600, 0x82f548, 0x1103595, 0x50488ca1, 0x36d61600, 0x29e, 0xce70001, 0xfbe96c7a, 0x1cd8c24, 0x82f558, 0x5048, 0x0, 0xe09c0, 0x8ca1, 0x5048, 0x0, 0x0, 0x82f588, 0x100dd13, 0x0, 0xee6b280, 0xee6b280, 0x1662720, 0x16626c0, 0x0, 0x82f588, 0x1142a8d, 0x50488ca1, 0x36d61600, 0x0, 0x0, 0xa, 0x1, 0x0, 0x3811de0, 0x1, 0x82f5e8, 0x82f608, 0x10150ed}
fdindex = {0xffffffff, 0x0, 0x3, 0x8, 0x9, 0x353c81a, 0x82f378, 0x1028c3b, 0x163a414, 0x353c81a, 0x0, 0x0, 0x163a414, 0x82f380, 0x82f378, 0x1028a1c, 0x3546de2, 0x457c7c0, 0x353c81a, 0x38e4226, 0x391e1ca, 0x0, 0x82f418, 0x10362be, 0x3546de2, 0x353c81a, 0x392284a, 0x39229ca, 0x4bbf405, 0x54184be, 0x82f438, 0x36f1e8a, 0x353c802, 0x82f4a4, 0x26aaec7a, 0x77c49210, 0x26aaec7a, 0x2fda46, 0x989680, 0x0, 0x200, 0x1, 0x77c4921d, 0xce7e2a0, 0xfbe96c7a, 0x1cd8c24, 0x86a508, 0x7c90cfea, 0x7c80a0e9, 0x710, 0x82f438, 0x7c90cfea, 0x7c80a0e9, 0x710, 0x82f448, 0x12e5c9c, 0x16627bc, 0x353c81a, 0x4405800, 0x353c81a, 0x36f1e8a, 0x82f514, 0x26aaec7a, 0x77c49210}
#8 0x0104ca1f in wait_reading_process_output (time_limit=0x23, nsecs=0x0, read_kbd=0xffffffff, do_display=0x1, wait_for_cell=0x353c81a, wait_proc=0x0, just_wait_proc=0x0) at process.c:4682
timeout_reduced_for_timers = 0x1
channel = 0xa
nfds = 0x1
Available = {
bits = {0x200, 0x0}
}
Writeok = {
bits = {0x0, 0x0}
}
check_write = 0x0
check_delay = 0x0
no_avail = 0x0
xerrno = 0x16
proc = 0x534cee5
timeout = {
tv_sec = 0x0,
tv_nsec = 0x0
}
end_time = {
tv_sec = 0x50488cc4,
tv_nsec = 0x26fb3f80
}
wait_channel = 0xffffffff
got_some_input = 0x1
count = 0x2
#9 0x010faab4 in sit_for (timeout=0x8c, reading=0x1, do_display=0x1) at dispnew.c:5972
sec = 0x23
nsec = 0x0
#10 0x01009478 in read_char (commandflag=0x1, nmaps=0x5, maps=0x82f990, prev_event=0x353c81a, used_mouse_menu=0x82fa78, end_time=0x0) at keyboard.c:2742
tem0 = 0x82f938
timeout = 0x23
delay_level = 0x5
buffer_size = 0x3f
c = 0x353c81a
jmpcount = 0x2
local_getcjmp = {0x82f938, 0x3c8c900, 0x82f99c, 0x3c8c90c, 0x82f7bc, 0x1008d63, 0x82ffe0, 0x0, 0x82f8e8, 0x378e1ca, 0x3565f02, 0x3c8c900, 0x82f8c8, 0x1027b11, 0x379e540, 0x379e540}
save_jump = {0x0 <repeats 16 times>}
key_already_recorded = 0x0
tem = 0x3cf79ee
save = 0xdb15
previous_echo_area_message = 0x353c81a
also_record = 0x353c81a
reread = 0x0
gcpro1 = {
next = 0x12e3542,
var = 0x51e2276,
nvars = 0x354377a
}
gcpro2 = {
next = 0x82f830,
var = 0x354377a,
nvars = 0x82f828
}
polling_stopped_here = 0x0
orig_kboard = 0x353f500
#11 0x0101c9ae in read_key_sequence (keybuf=0x82fc00, bufsize=0x1e, prompt=0x353c81a, dont_downcase_last=0x0, can_return_switch_frame=0x1, fix_current_buffer=0x1) at keyboard.c:9356
interrupted_kboard = 0x353f500
interrupted_frame = 0x3811de0
key = 0x4405805
used_mouse_menu = 0x0
echo_local_start = 0x0
last_real_key_start = 0x0
keys_local_start = 0x0
local_first_binding = 0x0
from_string = 0x353c81a
count = 0x2
t = 0x0
echo_start = 0x0
keys_start = 0x0
nmaps = 0x5
nmaps_allocated = 0x5
defs = 0x82f960
submaps = 0x82f990
orig_local_map = 0x4247c76
orig_keymap = 0x353c81a
localized_local_map = 0x0
first_binding = 0x0
first_unbound = 0x1f
mock_input = 0x0
fkey = {
parent = 0x38f7456,
map = 0x38f7456,
start = 0x0,
end = 0x0
}
keytran = {
parent = 0x3539ec6,
map = 0x3539ec6,
start = 0x0,
end = 0x0
}
indec = {
parent = 0x38f745e,
map = 0x38f745e,
start = 0x0,
end = 0x0
}
shift_translated = 0x0
delayed_switch_frame = 0x353c81a
original_uppercase = 0x36c50
original_uppercase_position = 0xffffffff
dummyflag = 0x0
starting_buffer = 0x4405800
fake_prefixed_keys = 0x353c81a
gcpro1 = {
next = 0x4405800,
var = 0x44c0668,
nvars = 0x82fac8
}
#12 0x01005e1b in command_loop_1 () at keyboard.c:1499
cmd = 0x3565f02
keybuf = {0x200001e0, 0x353c81a, 0x82fb00, 0x0, 0x163a630, 0x353c81a, 0x3745dfa, 0x35554aa, 0x353c81a, 0x0, 0x0, 0x35554a8, 0x0, 0x0, 0x82fc68, 0x1037cfb, 0x35554aa, 0x353c81a, 0x44f6b7e, 0x353c81a, 0xc, 0x353c81a, 0x2, 0x7ffd6000, 0x0, 0x0, 0x82fc98, 0x1033c91, 0x2, 0x44f6b7e}
i = 0x1
prev_modiff = 0x1dd
prev_buffer = 0x4405800
already_adjusted = 0x0
#13 0x01032285 in internal_condition_case (bfun=0x1005929 <command_loop_1>, handlers=0x3548dca, hfun=0x10050fe <cmd_error>) at eval.c:1319
val = 0x44f6b7e
c = {
tag = 0x353c81a,
val = 0x353c81a,
next = 0x82fda4,
gcpro = 0x0,
jmp = {0x82fd68, 0x7ffd6000, 0x0, 0x0, 0x82fcbc, 0x1032232, 0x82ffe0, 0x0, 0x2, 0x0, 0x82fdd0, 0x2, 0x2, 0x0, 0x1, 0x1},
backlist = 0x0,
handlerlist = 0x0,
lisp_eval_depth = 0x0,
pdlcount = 0x2,
poll_suppress_count = 0x0,
interrupt_input_blocked = 0x0,
byte_stack = 0x0
}
h = {
handler = 0x3548dca,
var = 0x353c81a,
chosen_clause = 0x2020e99,
tag = 0x82fcf0,
next = 0x0
}
#14 0x01005556 in command_loop_2 (ignore=0x353c81a) at keyboard.c:1194
val = 0x7ffd6000
#15 0x01031c96 in internal_catch (tag=0x3546612, func=0x1005532 <command_loop_2>, arg=0x353c81a) at eval.c:1076
c = {
tag = 0x3546612,
val = 0x353c81a,
next = 0x0,
gcpro = 0x0,
jmp = {0x82fe18, 0x7ffd6000, 0x0, 0x0, 0x82fd8c, 0x1031c87, 0x82ffe0, 0x0, 0x353c81a, 0x3542e00, 0x740053, 0x6e0061, 0x7ffd6000, 0x82fe18, 0x1027b01, 0x165c4fc},
backlist = 0x0,
handlerlist = 0x0,
lisp_eval_depth = 0x0,
pdlcount = 0x2,
poll_suppress_count = 0x0,
interrupt_input_blocked = 0x0,
byte_stack = 0x0
}
#16 0x01005512 in command_loop () at keyboard.c:1173
No locals.
#17 0x01004abb in recursive_edit_1 () at keyboard.c:794
count = 0x1
val = 0x77c4a190
#18 0x01004de8 in Frecursive_edit () at keyboard.c:858
count = 0x0
buffer = 0x353c81a
#19 0x01002abc in main (argc=0x2, argv=0xa425b0) at emacs.c:1643
dummy = 0x82ffe0
stack_bottom_variable = 0x0
do_initial_setlocale = 0x1
skip_args = 0x0
no_loadup = 0x0
junk = 0x0
dname_arg = 0x0
ch_to_dir = 0x0
(gdb) quit
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Thu, 06 Sep 2012 14:42:01 GMT) Full text and rfc822 format available.Message #41 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: martin rudalics <rudalics <at> gmx.at> To: Andy Moreton <andrewjmoreton <at> gmail.com> Cc: 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Thu, 06 Sep 2012 16:41:27 +0200
> Crash is in r109893 built on Windows XP using MingGW gcc 4.6.2. > > To reproduce it, run "M-x rgrep" using Cygwin grep. The grep > output is all displayed in the *grep* buffer, and then emacs aborts. With r109905 doing M-x shell and then exit gets me Breakpoint 1, emacs_abort () at w32fns.c:7201 7201 button = MessageBox (NULL, (gdb) bt #0 emacs_abort () at w32fns.c:7201 #1 0x010345e2 in sys_wait (status=0x82f284) at w32proc.c:462 #2 0x01031ff0 in record_child_status_change () at process.c:6422 #3 0x01032315 in handle_child_signal (sig=18) at process.c:6529 #4 0x010dd5fa in handle_on_main_thread (sig=18, handler=0x1032306 <handle_child_signal>) at sysdep.c:1584 #5 0x01032334 in deliver_child_signal (sig=18) at process.c:6536 #6 0x010358ec in sys_select (nfds=4, rfds=0x82f704, wfds=0x0, efds=0x0, timeout=0x82f6f4, ignored=0x0) at w32proc.c:1325 #7 0x0102f943 in wait_reading_process_output (time_limit=900, nsecs=0, read_kbd=-1, do_display=1, wait_for_cell=52672538, wait_proc=0x0, just_wait_proc=0) at process.c:4682 #8 0x010a4b7e in sit_for (timeout=3600, reading=true, do_display=1) at dispnew.c:5972 #9 0x010071dc in read_char (commandflag=1, nmaps=3, maps=0x82f9a0, prev_event=52672538, used_mouse_menu=0x82fa78, end_time=0x0) at keyboard.c:2742 #10 0x01011081 in read_key_sequence (keybuf=0x82fc00, bufsize=30, prompt=52672538, dont_downcase_last=0, can_return_switch_frame=1, fix_current_buffer=1) at keyboard.c:9356 #11 0x01004fad in command_loop_1 () at keyboard.c:1499 #12 0x0101e737 in internal_condition_case (bfun=0x1004c3e <command_loop_1>, handlers=52723146, hfun=0x10045e8 <cmd_error>) at eval.c:1319 #13 0x0100496f in command_loop_2 (ignore=52672538) at keyboard.c:1194 #14 0x0101e26b in internal_catch (tag=52712978, func=0x100494b <command_loop_2>, arg=52672538) at eval.c:1076 #15 0x0100492b in command_loop () at keyboard.c:1173 #16 0x01004225 in recursive_edit_1 () at keyboard.c:794 #17 0x01004374 in Frecursive_edit () at keyboard.c:858 #18 0x010028e6 in main (argc=1, argv=0xa32808) at emacs.c:1643 martin, on GNU Emacs 24.2.50.1 (i386-mingw-nt5.1.2600) of 2012-09-06 on MACHNO
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Thu, 06 Sep 2012 16:47:01 GMT) Full text and rfc822 format available.Message #44 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: martin rudalics <rudalics <at> gmx.at> Cc: andrewjmoreton <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Thu, 06 Sep 2012 19:46:43 +0300
> Date: Thu, 06 Sep 2012 16:41:27 +0200 > From: martin rudalics <rudalics <at> gmx.at> > Cc: 12327 <at> debbugs.gnu.org > > > Crash is in r109893 built on Windows XP using MingGW gcc 4.6.2. > > > > To reproduce it, run "M-x rgrep" using Cygwin grep. The grep > > output is all displayed in the *grep* buffer, and then emacs aborts. > > With r109905 doing M-x shell and then exit gets me > > Breakpoint 1, emacs_abort () at w32fns.c:7201 > 7201 button = MessageBox (NULL, > (gdb) bt > #0 emacs_abort () at w32fns.c:7201 > #1 0x010345e2 in sys_wait (status=0x82f284) at w32proc.c:462 > #2 0x01031ff0 in record_child_status_change () at process.c:6422 > #3 0x01032315 in handle_child_signal (sig=18) at process.c:6529 > #4 0x010dd5fa in handle_on_main_thread (sig=18, handler=0x1032306 <handle_child_signal>) at sysdep.c:1584 > #5 0x01032334 in deliver_child_signal (sig=18) at process.c:6536 > #6 0x010358ec in sys_select (nfds=4, rfds=0x82f704, wfds=0x0, efds=0x0, timeout=0x82f6f4, ignored=0x0) at w32proc.c:1325 I think I fixed this in revision 109907 on the trunk, please test. At least "M-x shell" followed by "exit RET" no longer winds up in emacs_abort for me. I cannot test Andrew's recipe, as I don't have Cygwin installed.
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Thu, 06 Sep 2012 16:56:01 GMT) Full text and rfc822 format available.Message #47 received at submit <at> debbugs.gnu.org (full text, mbox):
From: Andy Moreton <andrewjmoreton <at> gmail.com> To: bug-gnu-emacs <at> gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Thu, 06 Sep 2012 17:54:29 +0100
On Thu 06 Sep 2012, Eli Zaretskii wrote:
>> Date: Thu, 06 Sep 2012 16:41:27 +0200
>> From: martin rudalics <rudalics <at> gmx.at>
>> Cc: 12327 <at> debbugs.gnu.org
>>
>> > Crash is in r109893 built on Windows XP using MingGW gcc 4.6.2.
>> >
>> > To reproduce it, run "M-x rgrep" using Cygwin grep. The grep
>> > output is all displayed in the *grep* buffer, and then emacs aborts.
>>
>> With r109905 doing M-x shell and then exit gets me
>>
>> Breakpoint 1, emacs_abort () at w32fns.c:7201
>> 7201 button = MessageBox (NULL,
>> (gdb) bt
>> #0 emacs_abort () at w32fns.c:7201
>> #1 0x010345e2 in sys_wait (status=0x82f284) at w32proc.c:462
>> #2 0x01031ff0 in record_child_status_change () at process.c:6422
>> #3 0x01032315 in handle_child_signal (sig=18) at process.c:6529
>> #4 0x010dd5fa in handle_on_main_thread (sig=18, handler=0x1032306 <handle_child_signal>) at sysdep.c:1584
>> #5 0x01032334 in deliver_child_signal (sig=18) at process.c:6536
>> #6 0x010358ec in sys_select (nfds=4, rfds=0x82f704, wfds=0x0, efds=0x0,
>> timeout=0x82f6f4, ignored=0x0) at w32proc.c:1325
>
> I think I fixed this in revision 109907 on the trunk, please test. At
> least "M-x shell" followed by "exit RET" no longer winds up in
> emacs_abort for me. I cannot test Andrew's recipe, as I don't have
> Cygwin installed.
Thanks Eli, it all seems to happy again.
AndyM
Eli Zaretskii <eliz <at> gnu.org>:Paul Eggert <eggert <at> cs.ucla.edu>:Message #52 received at 12327-done <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: Andy Moreton <andrewjmoreton <at> gmail.com> Cc: 12327-done <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Thu, 06 Sep 2012 20:20:09 +0300
> From: Andy Moreton <andrewjmoreton <at> gmail.com> > Date: Thu, 06 Sep 2012 17:54:29 +0100 > > > I think I fixed this in revision 109907 on the trunk, please test. At > > least "M-x shell" followed by "exit RET" no longer winds up in > > emacs_abort for me. I cannot test Andrew's recipe, as I don't have > > Cygwin installed. > > Thanks Eli, it all seems to happy again. Thanks for testing, I'm closing the bug.
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Thu, 06 Sep 2012 17:37:01 GMT) Full text and rfc822 format available.Message #55 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: martin rudalics <rudalics <at> gmx.at> To: Eli Zaretskii <eliz <at> gnu.org> Cc: andrewjmoreton <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Thu, 06 Sep 2012 19:37:01 +0200
> I think I fixed this in revision 109907 on the trunk, please test. At > least "M-x shell" followed by "exit RET" no longer winds up in > emacs_abort for me. I cannot test Andrew's recipe, as I don't have > Cygwin installed. Works now. Thanks, martin
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Fri, 07 Sep 2012 01:37:02 GMT) Full text and rfc822 format available.Message #58 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Paul Eggert <eggert <at> cs.ucla.edu> To: 12327 <at> debbugs.gnu.org Cc: Juanma Barranquero <lekktu <at> gmail.com>, Eli Zaretskii <eliz <at> gnu.org> Subject: Signal-handler cleanup for Emacs Date: Thu, 06 Sep 2012 18:35:59 -0700
I installed the reworked patch <http://bugs.gnu.org/12327#20> into the trunk as bzr 109909. I also installed the first chunk of the later cleanup <http://bugs.gnu.org/12327#35> as trunk bzr 109893. This had a typo that prevented it working on Windows hosts -- sorry about that, and thanks to Andy and Martin for reporting it and Eli for fixing it. Other parts of the later cleanup still need installing, to fix the problem Andreas mentioned, plus some other race conditions I've found during analysis and testing on GNU/Linux. Since Bug#12327 is now marked as done and is getting a bit hard-to-follow anyway, I'll submit these separately, as a different bug#.
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Fri, 07 Sep 2012 06:03:01 GMT) Full text and rfc822 format available.Message #61 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: Paul Eggert <eggert <at> cs.ucla.edu> Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: Signal-handler cleanup for Emacs Date: Fri, 07 Sep 2012 09:02:04 +0300
> Date: Thu, 06 Sep 2012 18:35:59 -0700 > From: Paul Eggert <eggert <at> cs.ucla.edu> > CC: Eli Zaretskii <eliz <at> gnu.org>, Juanma Barranquero <lekktu <at> gmail.com> > > I installed the reworked patch <http://bugs.gnu.org/12327#20> > into the trunk as bzr 109909. This doesn't compile on Windows: gcc -I. -c -gdwarf-2 -g3 -DEMACSDEBUG -fno-crossjumping -Id:/usr/include/libxml2 -DGLYPH_DEBUG=1 -Demacs=1 -I../lib -I../nt/inc -DHAVE_NTGUI=1 -DUSE_CRT_DLL=1 -o oo/i386/editfns.o editfns.c alloc.c: In function `die': alloc.c:6716: error: `SIGABRT' undeclared (first use in this function) alloc.c:6716: error: (Each undeclared identifier is reported only once alloc.c:6716: error: for each function it appears in.) makefile:322: recipe for target `oo/i386/alloc.o' failed make[1]: *** [oo/i386/alloc.o] Error 1 make[1]: *** Waiting for unfinished jobs.... make[1]: Leaving directory `D:/gnu/bzr/emacs/trunk/src' makefile:432: recipe for target `all-other-dirs-gmake' failed make: *** [all-other-dirs-gmake] Error 2
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Fri, 07 Sep 2012 07:27:02 GMT) Full text and rfc822 format available.Message #64 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: eggert <at> cs.ucla.edu Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Fri, 07 Sep 2012 10:26:17 +0300
> Date: Fri, 07 Sep 2012 09:02:04 +0300 > From: Eli Zaretskii <eliz <at> gnu.org> > Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org > > This doesn't compile on Windows: > > gcc -I. -c -gdwarf-2 -g3 -DEMACSDEBUG -fno-crossjumping -Id:/usr/include/libxml2 -DGLYPH_DEBUG=1 -Demacs=1 -I../lib -I../nt/inc -DHAVE_NTGUI=1 -DUSE_CRT_DLL=1 -o oo/i386/editfns.o editfns.c > alloc.c: In function `die': > alloc.c:6716: error: `SIGABRT' undeclared (first use in this function) > alloc.c:6716: error: (Each undeclared identifier is reported only once > alloc.c:6716: error: for each function it appears in.) > makefile:322: recipe for target `oo/i386/alloc.o' failed > make[1]: *** [oo/i386/alloc.o] Error 1 > make[1]: *** Waiting for unfinished jobs.... > make[1]: Leaving directory `D:/gnu/bzr/emacs/trunk/src' > makefile:432: recipe for target `all-other-dirs-gmake' failed > make: *** [all-other-dirs-gmake] Error 2 Actually, this isn't Windows-specific, it happens in any build that specifies --enable-checking. Fixed by revision 109918 on the trunk. The following link failure _is_ Windows-specific, stand by for a fix. gcc -o oo/i386/temacs.bin -gdwarf-2 -g3 -Wl,-stack,0x00800000 -Wl,-heap,0x00100000 -Wl,-image-base,0x01000000 -Wl,-subsystem,console -Wl,-entry,__start -Wl,-Map,oo/i386/temacs.map oo/i386/firstfile.o oo/i386/emacs.res oo/i386/temacs0.a oo/i386/temacs1.a oo/i386/temacs2.a oo/i386/lastfile.a ../lib/oo/i386/libgnu.a -lwinmm -ladvapi32 -lgdi32 -lcomdlg32 -luser32 -lmpr -lshell32 -lwinspool -lole32 -lcomctl32 -lusp10 oo/i386/temacs0.a(emacs.o)(.text+0x1066): In function `main': D:\gnu\bzr\emacs\trunk\src/emacs.c:1129: undefined reference to `sigaction' oo/i386/temacs0.a(emacs.o)(.text+0x1091):D:\gnu\bzr\emacs\trunk\src/emacs.c:1133: undefined reference to `sigaction' oo/i386/temacs0.a(emacs.o)(.text+0x10ff):D:\gnu\bzr\emacs\trunk\src/emacs.c:1157: undefined reference to `sigaction' oo/i386/temacs0.a(emacs.o)(.text+0x111a):D:\gnu\bzr\emacs\trunk\src/emacs.c:1184: undefined reference to `sigaction' oo/i386/temacs0.a(emacs.o)(.text+0x1135):D:\gnu\bzr\emacs\trunk\src/emacs.c:1188: undefined reference to `sigaction' oo/i386/temacs0.a(emacs.o)(.text+0x1159):D:\gnu\bzr\emacs\trunk\src/emacs.c:1198: more undefined references to `sigaction' follow collect2: ld returned 1 exit status makefile:501: recipe for target `oo/i386/temacs.exe' failed make[1]: *** [oo/i386/temacs.exe] Error 1 make[1]: Leaving directory `D:/gnu/bzr/emacs/trunk/src' makefile:432: recipe for target `all-other-dirs-gmake' failed make: *** [all-other-dirs-gmake] Error 2
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Fri, 07 Sep 2012 08:28:01 GMT) Full text and rfc822 format available.Message #67 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: eggert <at> cs.ucla.edu Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Fri, 07 Sep 2012 11:27:06 +0300
> Date: Fri, 07 Sep 2012 10:26:17 +0300 > From: Eli Zaretskii <eliz <at> gnu.org> > Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org > > The following link failure _is_ Windows-specific, stand by for a fix. > > gcc -o oo/i386/temacs.bin -gdwarf-2 -g3 -Wl,-stack,0x00800000 -Wl,-heap,0x00100000 -Wl,-image-base,0x01000000 -Wl,-subsystem,console -Wl,-entry,__start -Wl,-Map,oo/i386/temacs.map oo/i386/firstfile.o oo/i386/emacs.res oo/i386/temacs0.a oo/i386/temacs1.a oo/i386/temacs2.a oo/i386/lastfile.a ../lib/oo/i386/libgnu.a -lwinmm -ladvapi32 -lgdi32 -lcomdlg32 -luser32 -lmpr -lshell32 -lwinspool -lole32 -lcomctl32 -lusp10 > oo/i386/temacs0.a(emacs.o)(.text+0x1066): In function `main': > D:\gnu\bzr\emacs\trunk\src/emacs.c:1129: undefined reference to `sigaction' > oo/i386/temacs0.a(emacs.o)(.text+0x1091):D:\gnu\bzr\emacs\trunk\src/emacs.c:1133: undefined reference to `sigaction' > oo/i386/temacs0.a(emacs.o)(.text+0x10ff):D:\gnu\bzr\emacs\trunk\src/emacs.c:1157: undefined reference to `sigaction' > oo/i386/temacs0.a(emacs.o)(.text+0x111a):D:\gnu\bzr\emacs\trunk\src/emacs.c:1184: undefined reference to `sigaction' > oo/i386/temacs0.a(emacs.o)(.text+0x1135):D:\gnu\bzr\emacs\trunk\src/emacs.c:1188: undefined reference to `sigaction' > oo/i386/temacs0.a(emacs.o)(.text+0x1159):D:\gnu\bzr\emacs\trunk\src/emacs.c:1198: more undefined references to `sigaction' follow > collect2: ld returned 1 exit status > makefile:501: recipe for target `oo/i386/temacs.exe' failed > make[1]: *** [oo/i386/temacs.exe] Error 1 > make[1]: Leaving directory `D:/gnu/bzr/emacs/trunk/src' > makefile:432: recipe for target `all-other-dirs-gmake' failed > make: *** [all-other-dirs-gmake] Error 2 Should be fixed with trunk revision 109919. (Why are there still calls to 'signal', as opposed to 'sigaction', in some parts of Emacs?)
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Fri, 07 Sep 2012 09:00:01 GMT) Full text and rfc822 format available.Message #70 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Paul Eggert <eggert <at> cs.ucla.edu> To: Eli Zaretskii <eliz <at> gnu.org> Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Fri, 07 Sep 2012 01:59:20 -0700
On 09/07/2012 01:27 AM, Eli Zaretskii wrote: > Should be fixed with trunk revision 109919. Thanks, I looked for other problems of that ilk and fixed them in trunk bzr 109921. > (Why are there still calls to 'signal', as opposed to 'sigaction', in > some parts of Emacs?) Thanks for catching that. Three of them were wrong -- a glitch when merging patches. Fixed in trunk bzr 109922. The other calls are setting signal handlers to SIG_DFL or SIG_IGN. For these, signal and sigaction do the same thing, so I left the calls alone, to minimize changes. If it's better, for regularity purposes, to replace these 'signal' calls with 'sigaction' calls, that's an easy and simple change that I can do.
bug-gnu-emacs <at> gnu.org:bug#12327; Package emacs.
(Fri, 07 Sep 2012 10:22:01 GMT) Full text and rfc822 format available.Message #73 received at 12327 <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: Paul Eggert <eggert <at> cs.ucla.edu> Cc: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org Subject: Re: bug#12327: Signal-handler cleanup for Emacs Date: Fri, 07 Sep 2012 13:21:34 +0300
> Date: Fri, 07 Sep 2012 01:59:20 -0700 > From: Paul Eggert <eggert <at> cs.ucla.edu> > CC: lekktu <at> gmail.com, 12327 <at> debbugs.gnu.org > > > (Why are there still calls to 'signal', as opposed to 'sigaction', in > > some parts of Emacs?) > > Thanks for catching that. Three of them were wrong -- a glitch when > merging patches. Fixed in trunk bzr 109922. > > The other calls are setting signal handlers to SIG_DFL or SIG_IGN. > For these, signal and sigaction do the same thing, so I left the > calls alone, to minimize changes. If it's better, for regularity > purposes, to replace these 'signal' calls with 'sigaction' calls, > that's an easy and simple change that I can do. I think it's better to use a single family of functions, it makes the code slightly easier to comprehend. But I'm okay with leaving those calls to 'signal', if no one else cares. It just means that 'sys_signal' in the Windows sources will have to be rewritten to call 'sigaction' internally (I left it alone in revision 109919 because I thought it was on its way out).
Debbugs Internal Request <help-debbugs <at> gnu.org>
to internal_control <at> debbugs.gnu.org.
(Fri, 05 Oct 2012 11:24:03 GMT) Full text and rfc822 format available.
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.