src/w32proc.c (waitpid): Avoid busy-waiting when called with WNOHANG
if the child process is still running. Instead, exit the wait
loop and return zero.
* callproc.c, process.h (record_kill_process):
New function, containing part of the old call_process_kill.
(call_process_kill): Use it.
This does not change call_process_kill's behavior.
* process.c (Fdelete_process): Use record_kill_process to fix a
race condition that could cause Emacs to lose track of a child.
* callproc.c (Fcall_process): Do record-unwind-protect on MSDOS
at the same time as other platforms, to simplify analysis.
No need for fd0_volatile since we have synch_process_fd.
Avoid needless emacs_close; arg is always negative.
src/fileio.c (file_name_as_directory, directory_file_name) [DOS_NT]:
Encode the file name before passing it to dostounix_filename, in
case it will downcase it (under w32-downcase-file-names).
Fixing this bug required some cleanup of the signal-handling code.
As a side effect, this change also fixes a longstanding rare race
condition whereby Emacs could mistakenly kill unrelated processes,
and it fixes a bug where a second C-g does not kill a recalcitrant
synchronous process in GNU/Linux and similar platforms.
The patch should also fix the last vestiges of Bug#9488,
a bug which has mostly been fixed on the trunk by other changes.
* callproc.c, process.h (synch_process_alive, synch_process_death)
(synch_process_termsig, sync_process_retcode):
Remove. All uses removed, to simplify analysis and so that
less consing is done inside critical sections.
* callproc.c (call_process_exited): Remove. All uses replaced
with !synch_process_pid.
* callproc.c (synch_process_pid, synch_process_fd): New static vars.
These take the role of what used to be in unwind-protect arg.
All uses changed.
(block_child_signal, unblock_child_signal):
New functions, to avoid races that could kill innocent-victim processes.
(call_process_kill, call_process_cleanup, Fcall_process): Use them.
(call_process_kill): Record killed processes as deleted, so that
zombies do not clutter up the system. Do this inside a critical
section, to avoid a race that would allow the clutter.
(call_process_cleanup): Fix code so that the second C-g works again
on common platforms such as GNU/Linux.
(Fcall_process): Create the child process in a critical section,
to fix a race condition. If creating an asynchronous process,
record it as deleted so that zombies do not clutter up the system.
Do unwind-protect for WINDOWSNT too, as that's simpler in the
light of these changes. Omit unnecessary call to emacs_close
before failure, as the unwind-protect code does that.
* callproc.c (call_process_cleanup):
* w32proc.c (waitpid): Simplify now that synch_process_alive is gone.
* process.c (record_deleted_pid): New function, containing
code refactored out of Fdelete_process.
(Fdelete_process): Use it.
(process_status_retrieved): Remove. All callers changed to use
child_status_change.
(record_child_status_change): Remove, folding its contents into ...
(handle_child_signal): ... this signal handler. Now, this
function is purely a handler for SIGCHLD, and is not called after
a synchronous waitpid returns; the synchronous code is moved to
wait_for_termination. There is no need to worry about reaping
more than one child now.
* sysdep.c (get_child_status, child_status_changed): New functions.
(wait_for_termination): Now takes int * status and bool
interruptible arguments, too. Do not record child status change;
that's now the caller's responsibility. All callers changed.
Reimplement in terms of get_child_status.
(wait_for_termination_1, interruptible_wait_for_termination):
Remove. All callers changed to use wait_for_termination.
* syswait.h: Include <stdbool.h>, for bool.
(record_child_status_change, interruptible_wait_for_termination):
Remove decls.
(record_deleted_pid, child_status_changed): New decls.
(wait_for_termination): Adjust to API changes noted above.
Fixes: debbugs:12980
(modify_region_1): ...new prototype.
* textprop.c (modify_region): Now static. Adjust users.
* insdel.c (modify_region): Rename to...
(modify_region_1): ...new function to work with current buffer.
Adjust comment and users. Use true and false for boolean arg.
(safe_alloca_unwind): Use it.
* lisp.h (free_save_value): New prototype.
* editfns.c (save_excursion_save): Use Lisp_Misc_Save_Value.
Add comment.
(save_excursion_restore): Adjust to match saved data structure.
Use free_save_value to offload some work from GC. Drop obsolete
#if 0 code.
* alloc.c (xpalloc): Comment fix.
* charset.c (Fdefine_charset_internal): If xpalloc exhausts memory
and signals an error, do not clear charset_table_size, as
charset_table is still valid.
* doprnt.c (evxprintf): Clear *BUF after freeing it.
* callproc.c (Fcall_process):
* process.c (create_process):
Don't save and restore environ; no longer needed.
* callproc.c (child_setup):
Use execve, not execvp, to preserve environ.
Fixes: debbugs:13054
src/w32fns.c: Remove prototype of atof.
(syspage_mask): Declared DWORD_PTR, for compatibility with 64-bit
builds.
(file_dialog_callback): Declared UINT_PTR.
src/w32common.h (syspage_mask): Declare DWORD_PTR, for compatibility
with 64-bit builds.
src/w32.c (FILE_DEVICE_FILE_SYSTEM, METHOD_BUFFERED)
(FILE_ANY_ACCESS, CTL_CODE) [_MSC_VER]: Define only if not already
defined.
some suspicious corner cases (see Bug#13007 and Bug#13012).
(mode_line_update_needed): New function.
(redisplay_internal, redisplay_window): Use it.
(ensure_selected_frame): New function.
(redisplay_internal, unwind_redisplay): Use it.
(redisplay_internal): Move comment about buffer_shared...
(buffer_shared_and_changed): ...near to its real use.
Do not assume that fd[0], count, filefd, and save_environ survive
vfork. Fix bug whereby wrong errno value could be reported for
pipe failure. Some minor cleanups, too, as follows. Move buf and
bufsize to the context where they're needed. Change new_argv to
be of type char **, as this is more convenient and avoids casts.
(CALLPROC_BUFFER_SIZE_MIN, CALLPROC_BUFFER_SIZE_MAX):
Now local constants, not macros.
src/xdisp.c (draw_glyphs): Don't draw in mouse face if mouse
highlighting on the frame was cleared. Prevents assertion
violations when repeatedly clicking on the "Top" link of the
"bread-crumbs" in Info buffers.
This is a backport from the trunk, consisting of:
2012-11-17 Eli Zaretskii <eliz@gnu.org>
* nt/inc/sys/wait.h: New file, with prototype of waitpid and
definitions of macros it needs.
* nt/inc/ms-w32.h (wait): Don't define, 'wait' is not used anymore.
(sys_wait): Remove prototype.
* nt/config.nt (HAVE_SYS_WAIT_H): Define to 1.
* src/w32proc.c (create_child): Don't clip the PID of the child
process to fit into an Emacs integer, as this is no longer a
restriction.
(waitpid): Rename from sys_wait. Emulate a Posix 'waitpid' by
reaping only the process specified by PID argument, if that is
positive. Use PID instead of dead_child to know which process to
reap. Wait for the child to die only if WNOHANG is not in
OPTIONS.
(sys_select): Don't set dead_child.
* src/sysdep.c (wait_for_termination_1): Remove the WINDOWSNT portion,
as it is no longer needed.
* src/process.c (waitpid, WUNTRACED) [!WNOHANG]: Remove definitions,
no longer needed.
(record_child_status_change): Remove the setting of
record_at_most_one_child for the !WNOHANG case.
2012-11-03 Paul Eggert <eggert@cs.ucla.edu>
Fix a race condition that causes Emacs to mess up glib (Bug#8855).
This is a backport from the trunk.
The symptom is a diagnostic "GLib-WARNING **: In call to
g_spawn_sync(), exit status of a child process was requested but
SIGCHLD action was set to SIG_IGN and ECHILD was received by
waitpid(), so exit status can't be returned." The diagnostic
is partly wrong, as the SIGCHLD action is not set to SIG_IGN.
The real bug is a race condition between Emacs and glib: Emacs
does a waitpid (-1, ...) and reaps glib's subprocess by mistake,
so that glib can't find it. Work around the bug by invoking
waitpid only on subprocesses that Emacs itself creates.
* src/process.c (create_process, record_child_status_change):
Don't use special value -1 in pid field, as the caller now must
know the pid rather than having the callee infer it. The
inference was sometimes incorrect anyway, due to another race.
(create_process): Set new 'alive' member if child is created.
(process_status_retrieved): New function.
(record_child_status_change): Use it.
Accept negative 1st argument, which means to wait for the
processes that Emacs already knows about. Move special-case code
for DOS_NT (which lacks WNOHANG) here, from caller. Keep track of
processes that have already been waited for, by testing and
clearing new 'alive' member.
(CAN_HANDLE_MULTIPLE_CHILDREN): Remove, as record_child_status_change
now does this internally.
(handle_child_signal): Let record_child_status_change do all
the work, since we do not want to reap all exited child processes,
only the child processes that Emacs itself created.
* src/process.h (Lisp_Process): New boolean member 'alive'.
since all users are dead long ago.
* nsterm.h (struct x_output): Use the only dummy member.
* w32menu.c (pending_menu_activation): Remove since not
really used.
(set_frame_menubar): Adjust user.
* w32term.h (struct x_output): Drop outdated #if 0 code.
(struct w32_output): Use bitfields for explicit_parent,
asked_for_visible and menubar_active members. Drop
unused pending_menu_activation member.
* xterm.h (struct x_output): Drop outdated #if 0 code.
Use bitfields for explicit_parent, asked_for_visible,
has_been_visible and net_wm_state_hidden_seen members.
src/xdisp.c (set_cursor_from_row): Skip step 2 only if point is not
between bpos_covered and bpos_max. This fixes cursor display when
several display strings follow each other.
src/.gdbinit (pgx): If the glyph's object is a string, display the
pointer to string data, rather than the value of the string object
itself (which barfs under CHECK_LISP_OBJECT_TYPE).
* admin/CPP-DEFINES (HAVE_CLOSEDIR, HAVE_DIRENT_H): Remove.
* admin/notes/copyright: Adjust to src/ndir.h -> nt/inc/dirent.h renaming.
* configure.ac: Do not check for dirent.h or closdir.
* nt/inc/dirent.h: Rename from ../src/ndir.h, with these changes:
(struct dirent): Rename from struct direct. All uses changed.
* nt/inc/sys/dir.h: Remove.
* src/dired.c: Assume HAVE_DIRENT_H.
(NAMLEN): Remove, replacing with ...
(dirent_namelen): New function. All uses changed. Use the GNU macro
_D_EXACT_NAMELEN if available, as it's faster than strlen.
(DIRENTRY): Remove, replacing all uses with 'struct dirent'.
(DIRENTRY_NONEMPTY): Remove. All callers now assume it's nonzero.
* src/makefile.w32-in (DIR_H): Remove. All uses replaced with
$(NT_INC)/dirent.h.
($(BLD)/w32.$(O)): Do not depend on $(SRC)/ndir.h.
* src/ndir.h: Rename to ../nt/inc/dirent.h.
* src/sysdep.h (closedir) [!HAVE_CLOSEDIR]: Remove.
Do not include <dirent.h>; no longer needed.
* src/w32.c: Include <dirent.h> rather than "ndir.h".
Fixes: debbugs:12958
