GNU bug report logs - #12316
Simplify redefinition of 'abort'.

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 12316 in the body.
You can then email your comments to 12316 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

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: Simplify redefinition of 'abort'.
Date: Thu, 30 Aug 2012 20:14:51 -0700

Tags: patch

Here's a patch I'd like to install after a bit more testing.
It simplifies how 'abort' is configured, on POSIX hosts.
This affects the Windows port but I haven't tested that part.
I'm CC'ing this to Eli and Juanma to give them a heads-up.

=== modified file 'ChangeLog'
--- ChangeLog	2012-08-28 16:01:59 +0000
+++ ChangeLog	2012-08-31 03:06:13 +0000
@@ -1,3 +1,8 @@
+2012-08-31  Paul Eggert  <eggert <at> cs.ucla.edu>
+
+	Simplify redefinition of 'abort'.
+	* configure.ac (NO_ABRT): Remove.
+
 2012-08-26  Paul Eggert  <eggert <at> cs.ucla.edu>
 
 	* configure.ac (CFLAGS): Prefer -g3 to -g if -g3 works

=== modified file 'admin/CPP-DEFINES'
--- admin/CPP-DEFINES	2012-08-18 21:36:13 +0000
+++ admin/CPP-DEFINES	2012-08-31 03:06:13 +0000
@@ -428,7 +428,6 @@
 MAIL_USE_SYSTEM_LOCK
 MAXPATHLEN
 NLIST_STRUCT
-NO_ABORT
 NO_EDITRES
 NO_MATHERR
 NO_TERMIO

=== modified file 'admin/ChangeLog'
--- admin/ChangeLog	2012-08-28 19:26:41 +0000
+++ admin/ChangeLog	2012-08-31 03:06:13 +0000
@@ -1,3 +1,8 @@
+2012-08-31  Paul Eggert  <eggert <at> cs.ucla.edu>
+
+	Simplify redefinition of 'abort'.
+	* CPP-DEFINES (NO_ABORT): Remove.
+
 2012-08-28  Glenn Morris  <rgm <at> gnu.org>
 
 	* bzrmerge.el (bzrmerge-merges): Allow unversioned files in the tree.

=== modified file 'configure.ac'
--- configure.ac	2012-08-28 16:01:59 +0000
+++ configure.ac	2012-08-31 03:06:13 +0000
@@ -3321,12 +3321,6 @@
     AC_DEFINE(BROKEN_PTY_READ_AFTER_EAGAIN, 1, [Define on FreeBSD to
       work around an issue when reading from a PTY.])
     ;;
-
-  dnl Define the following so emacs symbols will not conflict with those
-  dnl in the System framework.  Otherwise -prebind will not work.
-  darwin)
-    AC_DEFINE(NO_ABORT, 1, [Do not define abort in emacs.c.])
-    ;;
 esac
 
 case $opsys in

=== modified file 'nt/ChangeLog'
--- nt/ChangeLog	2012-08-28 16:01:59 +0000
+++ nt/ChangeLog	2012-08-31 03:06:13 +0000
@@ -1,3 +1,9 @@
+2012-08-31  Paul Eggert  <eggert <at> cs.ucla.edu>
+
+	Simplify redefinition of 'abort'.
+	* inc/ms-w32.h (w32_abort): Remove.
+	* inc/unistd.h: Include <stdlib.h>, for 'environ'.
+
 2012-08-22  Juanma Barranquero  <lekktu <at> gmail.com>
 
 	* config.nt: Sync with autogen/config.in.

=== modified file 'nt/inc/ms-w32.h'
--- nt/inc/ms-w32.h	2012-08-07 11:03:48 +0000
+++ nt/inc/ms-w32.h	2012-08-31 03:06:13 +0000
@@ -334,16 +334,7 @@
 #include <malloc.h>
 #endif
 
-/* stdlib.h must be included after redefining malloc & friends, but
-   before redefining abort.  Isn't library redefinition funny?  */
 #include <stdlib.h>
-
-/* Redefine abort.  */
-#ifdef HAVE_NTGUI
-#define abort	w32_abort
-extern _Noreturn void w32_abort (void);
-#endif
-
 #include <sys/stat.h>
 
 /* Define for those source files that do not include enough NT system files.  */

=== modified file 'nt/inc/unistd.h'
--- nt/inc/unistd.h	2011-02-27 19:48:31 +0000
+++ nt/inc/unistd.h	2012-08-31 03:06:13 +0000
@@ -3,8 +3,9 @@
 #ifndef _UNISTD_H
 #define _UNISTD_H
 
+#include <stdlib.h> /* for 'environ' */
+
 extern ssize_t readlink (const char *, char *, size_t);
 extern int symlink (char const *, char const *);
 
 #endif	/* _UNISTD_H */
-

=== modified file 'src/.gdbinit'
--- src/.gdbinit	2012-08-20 17:32:31 +0000
+++ src/.gdbinit	2012-08-31 03:06:13 +0000
@@ -1222,14 +1222,9 @@
   set $tem = (struct Lisp_String *) $ptr
   set $tem = (char *) $tem->data
 
-  # Don't let abort actually run, as it will make stdio stop working and
-  # therefore the `pr' command above as well.
-  if $tem[0] == 'w' && $tem[1] == 'i' && $tem[2] == 'n' && $tem[3] == 'd'
-    # The windows-nt build replaces abort with its own function.
-    break w32_abort
-  else
-    break abort
-  end
+  # Don't let emacs_abort actually run, as it will make stdio stop
+  # working and therefore the 'pr' command above as well.
+  break emacs_abort
 end
 
 # x_error_quitter is defined only on X.  But window-system is set up

=== modified file 'src/ChangeLog'
--- src/ChangeLog	2012-08-30 16:07:44 +0000
+++ src/ChangeLog	2012-08-31 03:06:13 +0000
@@ -1,3 +1,23 @@
+2012-08-31  Paul Eggert  <eggert <at> cs.ucla.edu>
+
+	Simplify redefinition of 'abort'.
+	Do not try to redefine the 'abort' function.  Instead, redo
+	'emacs' so that it calls 'emacs_abort' rather than 'abort'.
+	This removes the need for the NO_ABORT configure-time macro
+	and makes it easier to change the abort code to do a backtrace.
+	* .gdbinit: Just stop at emacs_abort, not at w32_abort or abort.
+	* conf_post.h: Include <string.h> and <stdlib.h> toward the end,
+	so that we can use NO_INLINE here and can define 'abort' cleanly.
+	(emacs_abort) [emacs]: New decl.
+	(abort) [emacs]: New macro.
+	* emacs.c (abort) [!DOS_NT && !NO_ABORT]:
+	Remove; sysdep.c's emacs_abort now takes its place.
+	* msdos.c (dos_abort) [defined abort]: Remove; not used.
+	(abort) [!defined abort]: Rename to ...
+	(emacs_abort) [HAVE_NTGUI]: ... new name.
+	* sysdep.c (emacs_abort) [!HAVE_NTGUI]: New function, taking
+	the place of the old 'abort' in emacs.c.
+
 2012-08-30  Jan Djärv  <jan.h.d <at> swipnet.se>
 
 	* gtkutil.c (x_wm_set_size_hint): Use 1 col for base_width so it

=== modified file 'src/conf_post.h'
--- src/conf_post.h	2012-08-20 16:48:10 +0000
+++ src/conf_post.h	2012-08-31 03:06:13 +0000
@@ -177,12 +177,6 @@
 #endif
 #endif
 
-#include <string.h>
-/* If you think about removing the line below, note that the
-   MS-Windows build relies on it for declaration of 'environ' needed
-   by a few source files.  */
-#include <stdlib.h>
-
 #if __GNUC__ >= 3  /* On GCC 3.0 we might get a warning.  */
 #define NO_INLINE __attribute__((noinline))
 #else
@@ -233,4 +227,14 @@
 # define lint_assume(cond) ((void) (0 && (cond)))
 #endif
 
+#include <string.h>
+#include <stdlib.h>
+/* Under Emacs, arrange for 'abort ()' to call a function that is
+   breakpointable via GDB.  */
+#ifdef emacs
+extern _Noreturn void emacs_abort (void) NO_INLINE;
+# undef abort
+# define abort() emacs_abort ()
+#endif
+
 /* conf_post.h ends here */

=== modified file 'src/emacs.c'
--- src/emacs.c	2012-08-25 06:38:43 +0000
+++ src/emacs.c	2012-08-31 03:10:04 +0000
@@ -344,22 +344,6 @@
   force_auto_save_soon ();
 }
 #endif
-
-/* We define abort, rather than using it from the library,
-   so that GDB can return from a breakpoint here.
-   MSDOS has its own definition in msdos.c.  */
-
-#if ! defined (DOS_NT) && ! defined (NO_ABORT)
-
-void
-abort (void)
-{
-  kill (getpid (), SIGABRT);
-  /* This shouldn't be executed, but it prevents a warning.  */
-  exit (1);
-}
-#endif
-
 
 /* Code for dealing with Lisp access to the Unix command line.  */
 

=== modified file 'src/msdos.c'
--- src/msdos.c	2012-08-21 10:21:04 +0000
+++ src/msdos.c	2012-08-31 03:06:13 +0000
@@ -4215,26 +4215,9 @@
 }
 #endif
 
-#ifdef abort
-#undef abort
-void
-dos_abort (char *file, int line)
-{
-  char buffer1[200], buffer2[400];
-  int i, j;
-
-  sprintf (buffer1, "<EMACS FATAL ERROR IN %s LINE %d>", file, line);
-  for (i = j = 0; buffer1[i]; i++) {
-    buffer2[j++] = buffer1[i];
-    buffer2[j++] = 0x70;
-  }
-  dosmemput (buffer2, j, (int)ScreenPrimary);
-  ScreenSetCursor (2, 0);
-  abort ();
-}
-#else
-void
-abort (void)
+#ifdef HAVE_NTGUI
+void
+emacs_abort (void)
 {
   dos_ttcooked ();
   ScreenSetCursor (10, 0);

=== modified file 'src/sysdep.c'
--- src/sysdep.c	2012-08-18 02:49:24 +0000
+++ src/sysdep.c	2012-08-31 03:06:13 +0000
@@ -1838,6 +1838,15 @@
 }
 #endif
 
+#ifndef HAVE_NTGUI
+/* Using emacs_abort lets GDB return from a breakpoint here.  */
+void
+emacs_abort (void)
+{
+  (abort) ();
+}
+#endif
+
 int
 emacs_open (const char *path, int oflag, int mode)
 {

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: Simplify redefinition of 'abort'.
Date: Fri, 31 Aug 2012 14:39:43 +0300

> Date: Thu, 30 Aug 2012 20:14:51 -0700
> From: Paul Eggert <eggert <at> cs.ucla.edu>
> CC: Eli Zaretskii <eliz <at> gnu.org>, Juanma Barranquero <lekktu <at> gmail.com>
> 
> Tags: patch
> 
> Here's a patch I'd like to install after a bit more testing.
> It simplifies how 'abort' is configured, on POSIX hosts.
> This affects the Windows port but I haven't tested that part.
> I'm CC'ing this to Eli and Juanma to give them a heads-up.

I don't understand what you want to happen after this change on
MS-Windows.  Please explain more.

Message #11 received at 12316 <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, 12316 <at> debbugs.gnu.org
Subject: Re: Simplify redefinition of 'abort'.
Date: Fri, 31 Aug 2012 05:53:14 -0700

On 08/31/2012 04:39 AM, Eli Zaretskii wrote:
> I don't understand what you want to happen after this change on
> MS-Windows.  Please explain more.

Currently, MS-Windows defines a function w32_abort that implements the
equivalent of 'abort' for MS-Windows.  So .gdbinit calls w32_abort on
MS-Windows, 'abort' on POSIX platforms, and nt/inc/ms-w32.h #defines
abort to w32_abort and declares w32_abort.

Due to the change, POSIX platforms define a function
emacs_abort rather than replacing the system 'abort'.
To keep in sync with that, the MS-Windows part of the patch
renames w32_abort to emacs_abort.  Hence .gdbinit
can now call emacs_abort on all platforms, without a special
case for MS-Windows, and nt/inc/ms-w32.h can remove its
special declarations of w32_abort and of 'abort'.

Also, this redefinition of 'abort' provides another
reason why conf_post.h must include <stdlib.h>, which makes
its comment about 'environ' confusing.  To help remove this
confusion I modified nt/inc/unistd.h to include <stdlib.h>;
this documents the fact that on POSIX platforms unistd.h
declares 'environ', moving that documentation to the
platform-specific area which is a better location for it.

Message #14 received at 12316 <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, 12316 <at> debbugs.gnu.org
Subject: Re: Simplify redefinition of 'abort'.
Date: Fri, 31 Aug 2012 17:29:02 +0300

> Date: Fri, 31 Aug 2012 05:53:14 -0700
> From: Paul Eggert <eggert <at> cs.ucla.edu>
> CC: 12316 <at> debbugs.gnu.org, lekktu <at> gmail.com
> 
> Due to the change, POSIX platforms define a function
> emacs_abort rather than replacing the system 'abort'.

But this is not equivalent to the old code: with your changes, only
those calls to 'abort' that originate within Emacs sources will wind
up at 'emacs_abort'.  E.g., if a function from one of the libraries
linked into Emacs, like the image libraries, calls 'abort', that will
bypass our function.  Why is it a good idea to lose this
functionality?

And if we do want to lose this, why not simply replace all the calls
to 'abort' with calls to 'emacs_abort'?  Isn't that cleaner?

> To keep in sync with that, the MS-Windows part of the patch
> renames w32_abort to emacs_abort.

I don't see this part in your patches.  Where is it?  Without this, I
cannot see how this will possibly work.

> Also, this redefinition of 'abort' provides another
> reason why conf_post.h must include <stdlib.h>, which makes
> its comment about 'environ' confusing.  To help remove this
> confusion I modified nt/inc/unistd.h to include <stdlib.h>;

Please don't do that.  A standard header should not include another
standard header; I don't want our private emulations violate this
rule, to avoid conflicts.  Surely, there are better and more
straight-forward ways of dealing with outdated comments.

Message #17 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: Simplify redefinition of 'abort'.
Date: Fri, 31 Aug 2012 17:41:23 +0300

> Date: Thu, 30 Aug 2012 20:14:51 -0700
> From: Paul Eggert <eggert <at> cs.ucla.edu>
> CC: Eli Zaretskii <eliz <at> gnu.org>, Juanma Barranquero <lekktu <at> gmail.com>
> 
> === modified file 'src/msdos.c'
> --- src/msdos.c	2012-08-21 10:21:04 +0000
> +++ src/msdos.c	2012-08-31 03:06:13 +0000
> @@ -4215,26 +4215,9 @@
>  }
>  #endif
>  
> -#ifdef abort
> -#undef abort
> -void
> -dos_abort (char *file, int line)
> -{
> -  char buffer1[200], buffer2[400];
> -  int i, j;
> -
> -  sprintf (buffer1, "<EMACS FATAL ERROR IN %s LINE %d>", file, line);
> -  for (i = j = 0; buffer1[i]; i++) {
> -    buffer2[j++] = buffer1[i];
> -    buffer2[j++] = 0x70;
> -  }
> -  dosmemput (buffer2, j, (int)ScreenPrimary);
> -  ScreenSetCursor (2, 0);
> -  abort ();
> -}
> -#else
> -void
> -abort (void)
> +#ifdef HAVE_NTGUI       <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
> +void
> +emacs_abort (void)
>  {
>    dos_ttcooked ();
>    ScreenSetCursor (10, 0);

The line marked above is wrong: code on msdos.c has no relation
whatsoever to the Windows build.

Message #20 received at 12316 <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, 12316 <at> debbugs.gnu.org
Subject: Re: Simplify redefinition of 'abort'.
Date: Fri, 31 Aug 2012 10:55:56 -0700

[Message part 1 (text/plain, inline)]

On 08/31/2012 07:29 AM, Eli Zaretskii wrote:

> why not simply replace all the calls
> to 'abort' with calls to 'emacs_abort'?  Isn't that cleaner?

Yes, I prefer that, too.  I avoided that only because
it bloated the patch.  But it's easy enough to add, and
the attached patch does that.

>> To keep in sync with that, the MS-Windows part of the patch
>> renames w32_abort to emacs_abort.
>
> I don't see this part in your patches.  Where is it?

Sorry, I missed one place.  Fixed in attached patch.
This also should fix the HAVE_NTGUI issue you reported in
a separate email.

> with your changes, only those calls to 'abort' that
> originate within Emacs sources will wind up at 'emacs_abort'.

Yes, that's part of the idea.  Intercepting other libraries'
call to 'abort' tends to break things.  We've caught some of
the breakages and have disabled them with NO_ABORT but it's
safer to avoid the problem in the first place.  Doing this
will also make it a bit easier for me to add a new feature,
namely backtraces on production aborts.

> A standard header should not include another
> standard header

That guideline is needed for glibc for conformance reasons,
and it used to be important long ago when multiple inclusion
of standard headers was a problem, but these days it's not
necessary and is even somewhat counterproductive in
portability emulations.  We violate that guideline all the
time in gnulib (and therefore in Emacs) and it's not a
problem.  However, if you prefer using the glibc style
for the Microsoft port that's easy to do I've revised
the patch to do that.

The attached patch is fairly long because of the abort ->
emacs_abort thing, so here's a precis that skips the boring
parts.

=== modified file 'ChangeLog'
--- ChangeLog	2012-08-28 16:01:59 +0000
+++ ChangeLog	2012-08-31 03:18:42 +0000
@@ -1,3 +1,8 @@
+2012-08-31  Paul Eggert  <eggert <at> cs.ucla.edu>
+
+	Simplify redefinition of 'abort' (Bug#12316).
+	* configure.ac (NO_ABRT): Remove.
+
 2012-08-26  Paul Eggert  <eggert <at> cs.ucla.edu>

 	* configure.ac (CFLAGS): Prefer -g3 to -g if -g3 works

=== modified file 'configure.ac'
--- configure.ac	2012-08-28 16:01:59 +0000
+++ configure.ac	2012-08-30 18:23:15 +0000
@@ -3321,12 +3321,6 @@
     AC_DEFINE(BROKEN_PTY_READ_AFTER_EAGAIN, 1, [Define on FreeBSD to
       work around an issue when reading from a PTY.])
     ;;
-
-  dnl Define the following so emacs symbols will not conflict with those
-  dnl in the System framework.  Otherwise -prebind will not work.
-  darwin)
-    AC_DEFINE(NO_ABORT, 1, [Do not define abort in emacs.c.])
-    ;;
 esac

 case $opsys in

=== modified file 'nt/ChangeLog'
--- nt/ChangeLog	2012-08-28 16:01:59 +0000
+++ nt/ChangeLog	2012-08-31 16:26:56 +0000
@@ -1,3 +1,8 @@
+2012-08-31  Paul Eggert  <eggert <at> cs.ucla.edu>
+
+	Simplify redefinition of 'abort' (Bug#12316).
+	* inc/ms-w32.h (w32_abort) [HAVE_NTGUI]: Remove.
+
 2012-08-22  Juanma Barranquero  <lekktu <at> gmail.com>

 	* config.nt: Sync with autogen/config.in.

=== modified file 'nt/inc/ms-w32.h'
--- nt/inc/ms-w32.h	2012-08-07 11:03:48 +0000
+++ nt/inc/ms-w32.h	2012-08-30 22:25:36 +0000
@@ -334,16 +334,7 @@
 #include <malloc.h>
 #endif

-/* stdlib.h must be included after redefining malloc & friends, but
-   before redefining abort.  Isn't library redefinition funny?  */
 #include <stdlib.h>
-
-/* Redefine abort.  */
-#ifdef HAVE_NTGUI
-#define abort	w32_abort
-extern _Noreturn void w32_abort (void);
-#endif
-
 #include <sys/stat.h>

 /* Define for those source files that do not include enough NT system files.  */

=== modified file 'nt/inc/unistd.h'
--- nt/inc/unistd.h	2011-02-27 19:48:31 +0000
+++ nt/inc/unistd.h	2012-08-31 16:31:23 +0000
@@ -3,8 +3,12 @@
 #ifndef _UNISTD_H
 #define _UNISTD_H

+/* On Microsoft platforms, <stdlib.h> declares 'environ'; on POSIX
+   platforms, <unistd.h> does.  Every file in Emacs that includes
+   <unistd.h> also includes <stdlib.h>, so there's no need to declare
+   'environ' here.  */
+
 extern ssize_t readlink (const char *, char *, size_t);
 extern int symlink (char const *, char const *);

 #endif	/* _UNISTD_H */
-

=== modified file 'src/.gdbinit'
--- src/.gdbinit	2012-08-20 17:32:31 +0000
+++ src/.gdbinit	2012-08-30 19:23:45 +0000
@@ -1222,14 +1222,9 @@
   set $tem = (struct Lisp_String *) $ptr
   set $tem = (char *) $tem->data

-  # Don't let abort actually run, as it will make stdio stop working and
-  # therefore the `pr' command above as well.
-  if $tem[0] == 'w' && $tem[1] == 'i' && $tem[2] == 'n' && $tem[3] == 'd'
-    # The windows-nt build replaces abort with its own function.
-    break w32_abort
-  else
-    break abort
-  end
+  # Don't let emacs_abort actually run, as it will make stdio stop
+  # working and therefore the 'pr' command above as well.
+  break emacs_abort
 end

 # x_error_quitter is defined only on X.  But window-system is set up

=== modified file 'src/ChangeLog'
--- src/ChangeLog	2012-08-31 10:53:19 +0000
+++ src/ChangeLog	2012-08-31 16:34:18 +0000
@@ -1,3 +1,23 @@
+2012-08-31  Paul Eggert  <eggert <at> cs.ucla.edu>
+
+	Simplify redefinition of 'abort' (Bug#12316).
+	Do not try to redefine the 'abort' function.  Instead, redo
+	the code so that it calls 'emacs_abort' rather than 'abort'.
+	This removes the need for the NO_ABORT configure-time macro
+	and makes it easier to change the abort code to do a backtrace.
+	* .gdbinit: Just stop at emacs_abort, not at w32_abort or abort.
+	* emacs.c (abort) [!DOS_NT && !NO_ABORT]:
+	Remove; sysdep.c's emacs_abort now takes its place.
+	* lisp.h (emacs_abort): New decl.  All calls from Emacs code to
+	'abort' changed to use 'emacs_abort'.
+	* msdos.c (dos_abort) [defined abort]: Remove; not used.
+	(abort) [!defined abort]: Rename to ...
+	(emacs_abort): ... new name.
+	* sysdep.c (emacs_abort) [!HAVE_NTGUI]: New function, taking
+	the place of the old 'abort' in emacs.c.
+	* w32.c, w32fns.c (abort): Do not #undef.
+	* w32.c (emacs_abort): Rename from w32_abort.
+
 2012-08-31  Dmitry Antipov  <dmantipov <at> yandex.ru>

 	Remove mark_ttys function and fix tty_display_info initialization.

=== modified file 'src/conf_post.h'
--- src/conf_post.h	2012-08-20 16:48:10 +0000
+++ src/conf_post.h	2012-08-31 14:47:20 +0000
@@ -178,9 +178,6 @@
 #endif

 #include <string.h>
-/* If you think about removing the line below, note that the
-   MS-Windows build relies on it for declaration of 'environ' needed
-   by a few source files.  */
 #include <stdlib.h>

 #if __GNUC__ >= 3  /* On GCC 3.0 we might get a warning.  */

=== modified file 'src/emacs.c'
--- src/emacs.c	2012-08-25 06:38:43 +0000
+++ src/emacs.c	2012-08-31 14:58:40 +0000
@@ -345,22 +345,6 @@
 }
 #endif

-/* We define abort, rather than using it from the library,
-   so that GDB can return from a breakpoint here.
-   MSDOS has its own definition in msdos.c.  */
-
-#if ! defined (DOS_NT) && ! defined (NO_ABORT)
-
-void
-abort (void)
-{
-  kill (getpid (), SIGABRT);
-  /* This shouldn't be executed, but it prevents a warning.  */
-  exit (1);
-}
-#endif
-
-
 /* Code for dealing with Lisp access to the Unix command line.  */

 static void

=== modified file 'src/msdos.c'
--- src/msdos.c	2012-08-21 10:21:04 +0000
+++ src/msdos.c	2012-08-31 14:56:42 +0000
@@ -4215,26 +4215,8 @@
 }
 #endif

-#ifdef abort
-#undef abort
 void
-dos_abort (char *file, int line)
-{
-  char buffer1[200], buffer2[400];
-  int i, j;
-
-  sprintf (buffer1, "<EMACS FATAL ERROR IN %s LINE %d>", file, line);
-  for (i = j = 0; buffer1[i]; i++) {
-    buffer2[j++] = buffer1[i];
-    buffer2[j++] = 0x70;
-  }
-  dosmemput (buffer2, j, (int)ScreenPrimary);
-  ScreenSetCursor (2, 0);
-  abort ();
-}
-#else
-void
-abort (void)
+emacs_abort (void)
 {
   dos_ttcooked ();
   ScreenSetCursor (10, 0);
@@ -4250,7 +4232,6 @@
 #endif /* __DJGPP_MINOR__ >= 2 */
   exit (2);
 }
-#endif

 void
 syms_of_msdos (void)

=== modified file 'src/sysdep.c'
--- src/sysdep.c	2012-08-18 02:49:24 +0000
+++ src/sysdep.c	2012-08-30 22:25:36 +0000
@@ -1838,6 +1838,15 @@
 }
 #endif
 
+#ifndef HAVE_NTGUI
+/* Using emacs_abort lets GDB return from a breakpoint here.  */
+void
+emacs_abort (void)
+{
+  (abort) ();
+}
+#endif
+
 int
 emacs_open (const char *path, int oflag, int mode)
 {

=== modified file 'src/w32.c'
--- src/w32.c	2012-08-26 10:29:37 +0000
+++ src/w32.c	2012-08-31 15:03:25 +0000
@@ -6640,8 +6640,7 @@
 		      buffer,
 		      "Emacs Abort Dialog",
 		      MB_OK | MB_ICONEXCLAMATION | MB_TASKMODAL);
-      /* Use the low-level Emacs abort. */
-#undef abort
+	  /* Use the low-level system abort. */
 	  abort ();
 	}
       else

=== modified file 'src/w32fns.c'
--- src/w32fns.c	2012-08-18 06:06:39 +0000
+++ src/w32fns.c	2012-08-31 15:04:03 +0000
@@ -7194,10 +7194,8 @@
   syms_of_w32uniscribe ();
 }

-#undef abort
-
 void
-w32_abort (void)
+emacs_abort (void)
 {
   int button;
   button = MessageBox (NULL,

[no-abort.txt (text/plain, attachment)]

Message #25 received at 12316-done <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, 12316-done <at> debbugs.gnu.org
Subject: Re: Simplify redefinition of 'abort'.
Date: Tue, 04 Sep 2012 10:36:17 -0700

I've installed that patch and am marking this bug as done.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.