bug#1295: Segfault with `M-- M-r' when `scroll-margin' > 0

bug#1295: Segfault with `M-- M-r' when `scroll-margin' > 0

[Chong Yidong]

> Emacs crashes with segmentation fault when I type command `M-- M-r'
> while variable `scroll-margin' has value 1 more more.
>
>  1. $ emacs -Q
>
>  2. C-h r
>
>  3. Don't move the cursor. Set the `scroll-margin' variable:
>
>         M-x set-variable RET scroll-margin RET 4 RET
>
>  4. M-- M-r
>
>         Fatal error (11)Segmentation fault

Hi Martin,

I think this may be a consequence of your 2008-11-02 checkin to src (I
didn't pinpoint the exact change that caused it, though).

Could you check?

bug#1295: Segfault with `M-- M-r' when `scroll-margin' > 0

[martin rudalics]

 >> Emacs crashes with segmentation fault when I type command `M-- M-r'
 >> while variable `scroll-margin' has value 1 more more.
 >>
 >>  1. $ emacs -Q
 >>
 >>  2. C-h r
 >>
 >>  3. Don't move the cursor. Set the `scroll-margin' variable:
 >>
 >>         M-x set-variable RET scroll-margin RET 4 RET
 >>
 >>  4. M-- M-r
 >>
 >>         Fatal error (11)Segmentation fault
 >
 > Hi Martin,
 >
 > I think this may be a consequence of your 2008-11-02 checkin to src (I
 > didn't pinpoint the exact change that caused it, though).
 >
 > Could you check?

The report was for Emacsen built as

     GNU Emacs 22.2.1 (i486-pc-linux-gnu, GTK+ Version 2.12.11) of
     2008-10-17 on raven, modified by Debian

and development snapshot version

     GNU Emacs 23.0.60.1 (i486-pc-linux-gnu, GTK+ Version 2.12.11) of
     2008-10-18 on elegiac, modified by Debian

and I verified here that it crashes with a negative argument for all
versions I keep personally.

BTW, Richard apparently once tried to make `move-to-window-line' handle
scroll margins but failed somehow.

martin

bug#1295: Segfault with `M-- M-r' when `scroll-margin' > 0

[Chong Yidong]

martin rudalics <rudalics@gmx.at> writes:

> The report was for Emacsen built as
>
>     GNU Emacs 22.2.1 (i486-pc-linux-gnu, GTK+ Version 2.12.11) of
>     2008-10-17 on raven, modified by Debian
>
> and development snapshot version
>
>     GNU Emacs 23.0.60.1 (i486-pc-linux-gnu, GTK+ Version 2.12.11) of
>     2008-10-18 on elegiac, modified by Debian
>
> and I verified here that it crashes with a negative argument for all
> versions I keep personally.

Sorry, my bad.  For some reason, reproducing this bug needs a full
bootstrap after reversion.  A simple make recompile/make doesn't track
the bug accurately, which is why it mistakenly pointed to your checkin.

After some more investigation, it looks to me like the bug was
introduced between 2008-10-06 (no crash) and 2008-10-07 (crash).
Weird---there weren't any redisplay changes made at this time.

bug#1295: Segfault with `M-- M-r' when `scroll-margin' > 0

[Chong Yidong]

Interesting.  This bug appears to be due to an interaction between Emacs
and the window manager's size hints.  At least, I can't reproduce it
with the latest CVS changes from today, in which a correction has been
made to how Emacs sends size hints to the window manager:

2008-11-03  Chong Yidong  <cyd@stupidchicken.com>

	* term/x-win.el (x-initialize-window-system): Call
	x-wm-set-size-hint after initialization.

Could you see whether the problem is fixed for you?

bug#1295: Segfault with `M-- M-r' when `scroll-margin' > 0

[Teemu Likonen]

Chong Yidong (2008-11-03 14:54 -0500) wrote:

> Interesting.  This bug appears to be due to an interaction between Emacs
> and the window manager's size hints.  At least, I can't reproduce it
> with the latest CVS changes from today, in which a correction has been
> made to how Emacs sends size hints to the window manager:
>
> 2008-11-03  Chong Yidong  <cyd@stupidchicken.com>
>
> 	* term/x-win.el (x-initialize-window-system): Call
> 	x-wm-set-size-hint after initialization.
>
> Could you see whether the problem is fixed for you?

I'm not able to compile Emacs myself. For development versions I depend
on prepackaged snapshots[1]. Sorry. I'll try it when a new snapshot becomes
available.

----------
 1. http://emacs.orebokech.com/

bug#1295: Segfault with `M-- M-r' when `scroll-margin' > 0

[martin rudalics]

 > Interesting.  This bug appears to be due to an interaction between Emacs
 > and the window manager's size hints.  At least, I can't reproduce it
 > with the latest CVS changes from today, in which a correction has been
 > made to how Emacs sends size hints to the window manager:
 >
 > 2008-11-03  Chong Yidong  <cyd@stupidchicken.com>
 >
 > 	* term/x-win.el (x-initialize-window-system): Call
 > 	x-wm-set-size-hint after initialization.
 >
 > Could you see whether the problem is fixed for you?

Still crashes here.  But just before crashing, display of the center line
of the window gets garbled now.  So it appears related to these size hints.

martin

bug#1295: Segfault with `M-- M-r' when `scroll-margin' > 0

[Richard M. Stallman]

     > 2008-11-03  Chong Yidong  <cyd@stupidchicken.com>
     >
     > 	* term/x-win.el (x-initialize-window-system): Call
     > 	x-wm-set-size-hint after initialization.
     >
     > Could you see whether the problem is fixed for you?

    Still crashes here.  But just before crashing, display of the center line
    of the window gets garbled now.  So it appears related to these size hints.

If we think that a bug in Lisp code causes a crash,
we should think about whether there's a bug in the C code
which it is triggering.

bug#1295: Segfault with `M-- M-r' when `scroll-margin' > 0

[martin rudalics]

 > If we think that a bug in Lisp code causes a crash,
 > we should think about whether there's a bug in the C code
 > which it is triggering.

It's a bug in C.  With the OP's recipe I get:

Program received signal SIGSEGV, Segmentation fault.
0x01028412 in try_window_reusing_current_matrix (w=0x3784600) at xdisp.c:14471
(gdb) bt
#0  0x01028412 in try_window_reusing_current_matrix (w=0x3784600) at xdisp.c:14471
#1  0x0102c328 in redisplay_window (window=58213892, just_this_one_p=1) at xdisp.c:13764
#2  0x0102dd61 in redisplay_window_1 (window=58213892) at xdisp.c:12249
#3  0x01009d1b in internal_condition_case_1 (bfun=0x102dd34 <redisplay_window_1>, arg=58213892, handlers=43976013, hfun=0x101d96a <redisplay_window_error>) at eval.c:1559
#4  0x0102f775 in redisplay_internal (preserve_echo_area=2732) at xdisp.c:11867
#5  0x0105946f in read_char (commandflag=1, nmaps=2, maps=0x82fb20, prev_event=43993089, used_mouse_menu=0x82fb68, end_time=0x0) at keyboard.c:2649
#6  0x0105c47d in read_key_sequence (keybuf=0x82fcb0, bufsize=30, prompt=43993089, dont_downcase_last=0, can_return_switch_frame=1, fix_current_buffer=1) at keyboard.c:9344
#7  0x0105e342 in command_loop_1 () at keyboard.c:1621
#8  0x01009fae in internal_condition_case (bfun=0x105e190 <command_loop_1>, handlers=44056841, hfun=0x105773c <cmd_error>) at eval.c:1511
#9  0x01051cca in command_loop_2 () at keyboard.c:1338
#10 0x01009ee3 in internal_catch (tag=2732, func=0x1051ca7 <command_loop_2>, arg=43993089) at eval.c:1247
#11 0x01051ad7 in command_loop () at keyboard.c:1317
#12 0x01051b70 in recursive_edit_1 () at keyboard.c:942
#13 0x01051c91 in Frecursive_edit () at keyboard.c:1004
#14 0x01002e46 in main (argc=2, argv=0xa327e0) at emacs.c:1777
(gdb)

martin