Package: emacs;
Reported by: Paul Eggert <eggert <at> cs.ucla.edu>
Date: Thu, 29 Nov 2012 06:55:01 UTC
Severity: normal
Tags: patch
Done: Paul Eggert <eggert <at> cs.ucla.edu>
Bug is archived. No further changes may be made.
Message #8 received at 13026 <at> debbugs.gnu.org (full text, mbox):
From: Eli Zaretskii <eliz <at> gnu.org> To: Paul Eggert <eggert <at> cs.ucla.edu> Cc: 13026 <at> debbugs.gnu.org Subject: Re: bug#13026: Assume POSIX 1003.1-1988 or later for signal.h. Date: Thu, 29 Nov 2012 19:33:56 +0200
> Date: Wed, 28 Nov 2012 22:52:31 -0800
> From: Paul Eggert <eggert <at> cs.ucla.edu>
>
> Attached is a proposed patch to simplify Emacs by assuming
> POSIX 1003.1-1988 or later for signal.h, which I'd
> like to install soon. This patch is relative to
> trunk bzr 111026. Tested on Fedora 17.
This will break the Windows build in more than one way, unless
accompanied by suitable Windows-only changes. Is this the entire
changeset, or did you omit some more changes?
Here are the problems I found:
> === modified file 'lib-src/emacsclient.c'
> --- lib-src/emacsclient.c 2012-11-21 21:06:52 +0000
> +++ lib-src/emacsclient.c 2012-11-29 06:46:35 +0000
> @@ -1801,7 +1801,6 @@
> needlf = str[0] == '\0' ? needlf : str[strlen (str) - 1] != '\n';
> exit_status = EXIT_FAILURE;
> }
> -#ifdef SIGSTOP
> else if (strprefix ("-suspend ", p))
> {
> /* -suspend: Suspend this terminal, i.e., stop the process. */
> @@ -1810,7 +1809,6 @@
> needlf = 0;
> kill (0, SIGSTOP);
> }
> -#endif
> else
> {
> /* Unknown command. */
This will break compilation of emacsclient.c, because SIGSTOP is not
defined on MS-Windows, and therefore the -suspend option cannot be
supported. And even if we define SIGSTOP to some value, having
"-suspend" pretend to work, but do nothing is not a good idea. (We
could emulate this option, but that would be a non-trivial change,
certainly not belonging in this changeset.)
Also, there's no 'kill' on Windows, and the emulation in w32proc.c is
only compiled into Emacs itself, not into lib-src programs.
> --- src/keyboard.c 2012-11-27 05:17:07 +0000
> +++ src/keyboard.c 2012-11-29 06:46:35 +0000
> @@ -10552,8 +10552,7 @@
> void
> stuff_buffered_input (Lisp_Object stuffstring)
> {
> -#ifdef SIGTSTP /* stuff_char is defined if SIGTSTP. */
> - register unsigned char *p;
> + unsigned char *p;
>
> if (STRINGP (stuffstring))
> {
> @@ -10584,7 +10583,6 @@
> }
>
> input_pending = 0;
> -#endif /* SIGTSTP */
> }
This will break suspend-emacs on MS-Windows (which does work now by
launching a sub-shell), because stuff_char, called by
stuff_buffered_input, will now signal an error.
> -#ifdef SIGTSTP /* Support possible in later USG versions */
> /*
> - * On systems which can suspend the current process and return to the original
> - * shell, this command causes the user to end up back at the shell.
> + * Cause the user to end up back at the shell.
> * The "Auto-save" and "Abort" questions are not asked until
> * the user elects to return to emacs, at which point he can save the current
> * job and either dump core or continue.
> */
> sys_suspend ();
> -#else
> - /* Perhaps should really fork an inferior shell?
> - But that would not provide any way to get back
> - to the original shell, ever. */
> - printf ("No support for stopping a process on this operating system;\n");
> - printf ("you can continue or abort.\n");
> -#endif /* not SIGTSTP */
> #ifdef MSDOS
> /* We must remain inside the screen area when the internal terminal
> is used. Note that [Enter] is not echoed by dos. */
This will also break Windows, because it cannot_suspend, so calling
sys_suspend for it does nothing useful (the Windows 'kill' emulation
only works on processes other than Emacs, and again SIGTSTP is not
going to suspend anything).
> @@ -5791,7 +5777,6 @@
>
> switch (signo)
> {
> -#ifdef SIGCONT
> case SIGCONT:
> p->raw_status_new = 0;
> pset_status (p, Qrun);
> @@ -5802,7 +5787,7 @@
> redisplay_preserve_echo_area (13);
> }
> break;
> -#endif /* ! defined (SIGCONT) */
This won't compile, since SIGCONT is not defined on Windows.
> @@ -5929,11 +5911,8 @@
> pset_command (p, Qnil);
> return process;
> }
> -#ifdef SIGCONT
> - process_send_signal (process, SIGCONT, current_group, 0);
> -#else
> - error ("No SIGCONT support");
> -#endif
> +
> + process_send_signal (process, SIGCONT, current_group, 0);
> return process;
> }
Likewise.
> @@ -5994,28 +5973,18 @@
>
> if (0)
> ;
> -#ifdef SIGUSR1
> parse_signal ("usr1", SIGUSR1);
> -#endif
> -#ifdef SIGUSR2
> parse_signal ("usr2", SIGUSR2);
> -#endif
Likewise: SIGUSR1 and SIGUSR2 are not defined.
> -#ifdef SIGSTOP
> parse_signal ("stop", SIGSTOP);
> -#endif
> -#ifdef SIGTSTP
> parse_signal ("tstp", SIGTSTP);
> -#endif
> -#ifdef SIGCONT
> parse_signal ("cont", SIGCONT);
> -#endif
> -#ifdef SIGCHLD
> parse_signal ("chld", SIGCHLD);
> -#endif
> -#ifdef SIGTTIN
> parse_signal ("ttin", SIGTTIN);
> -#endif
> -#ifdef SIGTTOU
> parse_signal ("ttou", SIGTTOU);
> -#endif
And the same for all these, except SIGCHLD.
> --- src/sysdep.c 2012-11-27 03:10:32 +0000
> +++ src/sysdep.c 2012-11-29 06:46:35 +0000
> @@ -213,8 +213,6 @@
> }
>
>
> -#ifdef SIGTSTP
> -
> /* Arrange for character C to be read as the next input from
> the terminal.
> XXX What if we have multiple ttys?
> @@ -233,8 +231,6 @@
> error ("Cannot stuff terminal input characters in this version of Unix");
> #endif /* no TIOCSTI */
> }
> -
> -#endif /* SIGTSTP */
See above: this will now signal an error on Windows.
> @@ -428,20 +424,15 @@
> void
> sys_suspend (void)
> {
> -#if defined (SIGTSTP) && !defined (MSDOS)
> -
> - {
> - pid_t pgrp = getpgrp ();
> - EMACS_KILLPG (pgrp, SIGTSTP);
> - }
> -
> -#else /* No SIGTSTP */
> +#ifndef MSDOS
> + kill (0, SIGTSTP);
> +#else
> /* On a system where suspending is not implemented,
> instead fork a subshell and let it talk directly to the terminal
> while we wait. */
> sys_subshell ();
>
> -#endif /* no SIGTSTP */
> +#endif
> }
Windows should also call sys_subshell here.
> @@ -702,14 +693,12 @@
> static void
> tcsetpgrp_without_stopping (int fd, pid_t pgid)
> {
> -#ifdef SIGTTOU
> signal_handler_t handler;
> block_input ();
> handler = signal (SIGTTOU, SIG_IGN);
> tcsetpgrp (fd, pgid);
> signal (SIGTTOU, handler);
> unblock_input ();
> -#endif
> }
This won't compile because SIGTTOU is not defined. Also, there's no
tcsetpgrp on Windows, so this won't link unless we provide an
emulation.
> -# ifdef SIGCONT
> sys_siglist[SIGCONT] = "Continued";
> -# endif
Won't compile.
> -# ifdef SIGSTOP
> sys_siglist[SIGSTOP] = "Stopped (signal)";
> -# endif
Likewise.
> -# ifdef SIGTSTP
> sys_siglist[SIGTSTP] = "Stopped (user)";
> -# endif
> -# ifdef SIGTTIN
> sys_siglist[SIGTTIN] = "Stopped (tty input)";
> -# endif
> -# ifdef SIGTTOU
> sys_siglist[SIGTTOU] = "Stopped (tty output)";
> -# endif
Likewise.
> -# ifdef SIGUSR1
> sys_siglist[SIGUSR1] = "User defined signal 1";
> -# endif
> -# ifdef SIGUSR2
> sys_siglist[SIGUSR2] = "User defined signal 2";
> -# endif
Likewise.
Please also note that NSIG is 23 on Windows, so the sys_siglist array
cannot be much larger than it already is (I think most of the 23 slots
are already used up).
> @@ -1861,12 +1819,8 @@
> sigaction (SIGFPE, &action, 0);
> }
>
> -#ifdef SIGUSR1
> add_user_signal (SIGUSR1, "sigusr1");
> -#endif
> -#ifdef SIGUSR2
> add_user_signal (SIGUSR2, "sigusr2");
> -#endif
Won't compile.
GNU bug tracking system
Copyright (C) 1999 Darren O. Benham,
1997,2003 nCipher Corporation Ltd,
1994-97 Ian Jackson.