bug#13968: 24.3.50; emacs_backtrace.txt

bug#13968: 24.3.50; emacs_backtrace.txt

[Drew Adams]

Backtrace:
0x01158AA0
0x01158B12
0x01001459
0x01021E80
0x0114E571
0x7E418730
0x7E418812
0x7E4189C9
0x7E418A0C
0x0114CB8E
0x0114CE2D
0x7C80B725
 

In GNU Emacs 24.3.50.1 (i386-mingw-nt5.1.2600)
 of 2013-03-04 on ODIEONE
Bzr revision: 111935 yamaoka@jpl.org-20130304102733-4qy111z41qwoh2as
Windowing system distributor `Microsoft Corp.', version 5.1.2600
Configured using:
 `configure --with-gcc (4.7) --no-opt --enable-checking --cflags
 -IC:/Devel/emacs/build/include --ldflags -LC:/Devel/emacs/build/lib'
 

bug#13968: 24.3.50; emacs_backtrace.txt

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Date: Fri, 15 Mar 2013 13:59:09 -0700
> 
> Backtrace:
> 0x01158AA0
> 0x01158B12
> 0x01001459
> 0x01021E80
> 0x0114E571
> 0x7E418730
> 0x7E418812
> 0x7E4189C9
> 0x7E418A0C
> 0x0114CB8E
> 0x0114CE2D
> 0x7C80B725

  w32_backtrace at C:\Devel\emacs\repo\build\src/w32fns.c:7727
  emacs_abort at C:\Devel\emacs\repo\build\src/w32fns.c:7759
  terminate_due_to_signal at C:\Devel\emacs\repo\build\src/emacs.c:343
  die at C:\Devel\emacs\repo\build\src/alloc.c:6556
  w32_wnd_proc at C:\Devel\emacs\repo\build\src/w32fns.c:3163
  ??
  ??:0
  ??
  ??:0
  ??
  ??:0
  ??
  ??:0
  w32_msg_pump at C:\Devel\emacs\repo\build\src/w32fns.c:2493
  w32_msg_worker@4 at C:\Devel\emacs\repo\build\src/w32fns.c:2619

Useless without a recipe.

bug#13968: 24.3.50; emacs_backtrace.txt

[Eli Zaretskii]

> Date: Fri, 15 Mar 2013 23:06:53 +0200
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: 13968@debbugs.gnu.org
> 
>   w32_backtrace at C:\Devel\emacs\repo\build\src/w32fns.c:7727
>   emacs_abort at C:\Devel\emacs\repo\build\src/w32fns.c:7759
>   terminate_due_to_signal at C:\Devel\emacs\repo\build\src/emacs.c:343
>   die at C:\Devel\emacs\repo\build\src/alloc.c:6556
>   w32_wnd_proc at C:\Devel\emacs\repo\build\src/w32fns.c:3163
>   ??
>   ??:0
>   ??
>   ??:0
>   ??
>   ??:0
>   ??
>   ??:0
>   w32_msg_pump at C:\Devel\emacs\repo\build\src/w32fns.c:2493
>   w32_msg_worker@4 at C:\Devel\emacs\repo\build\src/w32fns.c:2619
> 
> Useless without a recipe.

Forgot to show where it aborts:

	  form.rcArea.bottom = (WINDOW_BOTTOM_EDGE_Y (w)
				- WINDOW_MODE_LINE_HEIGHT (w));

It's an assertion violation, but the report doesn't even say what was
the text of the assertion message.  The above 2 macros could abort in
XFRAME, XWINDOW, or XBUFFER.

bug#13968: 24.3.50; emacs_backtrace.txt

[Drew Adams]

> > Useless without a recipe.
> 
> Forgot to show where it aborts:
> 
> 	  form.rcArea.bottom = (WINDOW_BOTTOM_EDGE_Y (w)
> 				- WINDOW_MODE_LINE_HEIGHT (w));
> 
> It's an assertion violation, but the report doesn't even say what was
> the text of the assertion message.  The above 2 macros could abort in
> XFRAME, XWINDOW, or XBUFFER.

Yes, well there never is a recipe for this kind of thing (but thanks for taking
a look)...  The crashes I get seem to happen any old time, and each time I send
in emacs_backtrace.txt you say it is useless (and I believe you).

And I've gotten plenty of crashes, seemingly ever since Emacs Dev starting
adding more assertions here and there to the C code.

What's the point of adding such assertions and providing backtraces that are
apparently useless?  That doesn't seem to have accomplished much that is
constructive, but it does seem to have increased the number of crashes (just a
guess).

Anyway, multiple reports of crashes (by me and others) have not prevented Emacs
24.2 or 24.3 from being released.  FWIW, before Emacs 24 I rarely had an Emacs
crash.

The C code seems to be a bit flaky now.  There seems to be a lot of C
development since Emacs 24.1, and perhaps there is not as much in the way of
fixing problems that get introduced along the way.  Not an informed judgment,
just one, naive impression.

I'm not faulting you, Eli, and I do appreciate your taking a look at the bug
reports.  Just mentioning that in my naive view, Emacs seems less stable than
before, on the C side.

Presumably such assertions are turned off when Emacs is released (?), so
releases should crash less, at least.  I tend to use trunk builds, so perhaps I
underestimate the release stability in this regard.

As you hint above, perhaps the assertions can be refined, so the backtraces
produced become more informative.  I will continue to report them, unless you
say it's not worth the trouble in general.

bug#13968: 24.3.50; emacs_backtrace.txt

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Cc: <13968@debbugs.gnu.org>
> Date: Fri, 15 Mar 2013 14:34:09 -0700
> 
> > > Useless without a recipe.
> > 
> > Forgot to show where it aborts:
> > 
> > 	  form.rcArea.bottom = (WINDOW_BOTTOM_EDGE_Y (w)
> > 				- WINDOW_MODE_LINE_HEIGHT (w));
> > 
> > It's an assertion violation, but the report doesn't even say what was
> > the text of the assertion message.  The above 2 macros could abort in
> > XFRAME, XWINDOW, or XBUFFER.
> 
> Yes, well there never is a recipe for this kind of thing

As I probably already said, they are probably related to your heavy
use of separate and minibuffer-less frames.  But this conclusion is
not instrumental to fixing those aborts, without a clear recipe to
reproduce a similar problem and/or without the ability to look around
when the abort does happen.  I urge you (again) to run Emacs from GDB
and leave the GDB session running long enough for us to ask you to
investigate after the abort.

> What's the point of adding such assertions and providing backtraces that are
> apparently useless?

The assertions are not specific to the aborts you experience.  They
are part of C macros, such as XWINDOW and XBUFFER, which extract C
data structures from a Lisp object.  These macros are used all over
the code.

> Anyway, multiple reports of crashes (by me and others) have not prevented Emacs
> 24.2 or 24.3 from being released.

In a release version, these assertions are not compiled into the
binary.

> FWIW, before Emacs 24 I rarely had an Emacs crash.

Probably because pretest and snapshot binaries before that were not
compiled with these assertions.

> Presumably such assertions are turned off when Emacs is released (?)

Yes.

> As you hint above, perhaps the assertions can be refined, so the backtraces
> produced become more informative.  I will continue to report them, unless you
> say it's not worth the trouble in general.

Please continue reporting them.  However, it will be much more useful
if you could let us investigate by using GDB.