Emacs 22.0.50 and suspending while logged into another system

Emacs 22.0.50 and suspending while logged into another system

[Quanah Gibson-Mount]

Hello,

One of our users reported the following problem with Emacs 22.0.50 (CVS CO 
9/22/2006):

If one logs into another system, and then runs emacs -nw, then does the 
following 2-3 times:

suspend emacs
foreground emacs

Control characters stop being interpreted, and instead show up in the edit 
buffer text.  In addition, foreground emacs does not redraw the screen.

It is important to note that I can only reproduce this behavior if I've 
logged in from system A to system B.  Doing it only on the system I'm 
directly connected to does not trigger the bug.

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITS/Shared Application Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html

Re: Emacs 22.0.50 and suspending while logged into another system

[Kim F. Storm]

Quanah Gibson-Mount <quanah@stanford.edu> writes:

> Hello,
>
> One of our users reported the following problem with Emacs 22.0.50
> (CVS CO 9/22/2006):

What systems are involved here?

How do you log into the remote system?

What is the exact emacs version on the target system?
(pls. use C-u M-x emacs-version RET)

Does this happen with emacs 21.3 (or 21.4) ?

If not, do you know when this started to happen with CVS emacs?

>
> If one logs into another system, and then runs emacs -nw, then does
> the following 2-3 times:
>
> suspend emacs
> foreground emacs
>
> Control characters stop being interpreted, and instead show up in the
> edit buffer text.  In addition, foreground emacs does not redraw the
> screen.
>
> It is important to note that I can only reproduce this behavior if
> I've logged in from system A to system B.  Doing it only on the system
> I'm directly connected to does not trigger the bug.
>
> --Quanah
>
> --
> Quanah Gibson-Mount
> Principal Software Developer
> ITS/Shared Application Services
> Stanford University
> GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html

-- 
Kim F. Storm  http://www.cua.dk

Re: Emacs 22.0.50 and suspending while logged into another system

[Quanah Gibson-Mount]

--On Wednesday, October 11, 2006 12:51 AM +0200 "Kim F. Storm" 
<no-spam@cua.dk> wrote:

> Quanah Gibson-Mount <quanah@stanford.edu> writes:
>
>> Hello,
>>
>> One of our users reported the following problem with Emacs 22.0.50
>> (CVS CO 9/22/2006):
>
> What systems are involved here?

Solaris 8

> How do you log into the remote system?

I can reproduce it with both klogin and ssh

> What is the exact emacs version on the target system?
> (pls. use C-u M-x emacs-version RET)

GNU Emacs 22.0.50.1 (sparc-sun-solaris2.8, GTK+ Version 2.8.17) of 
2006-09-22 on solaris8-build.Stanford.EDU


> Does this happen with emacs 21.3 (or 21.4) ?

No idea, the last version of emacs I could get to build and work on Solaris 
was 21.2.

> If not, do you know when this started to happen with CVS emacs?

Nope.


Here's a simple cut and paste after I've managed to trigger it:
;; This buffer is for notes you don't want to save, and for Lisp evaluation.
;; If you want to create a file, visit that file with C-x C-f,
;; then enter the text in that file's own buffer.

Test 1
Test 2
Test 3
Test 4
Test 5
Test 6
Test 7
Test 8
Test 9
Test 10
^X^C well this sucks I can't ^X^C out of emacs
^P^P^N^N^N^N^N^P^N^P^N^P^F


--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITS/Shared Application Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html

Re: Emacs 22.0.50 and suspending while logged into another system

[Kim F. Storm]

Quanah Gibson-Mount <quanah@stanford.edu> writes:

> --On Wednesday, October 11, 2006 12:51 AM +0200 "Kim F. Storm"
> <no-spam@cua.dk> wrote:
>
>> Quanah Gibson-Mount <quanah@stanford.edu> writes:
>>
>>> Hello,
>>>
>>> One of our users reported the following problem with Emacs 22.0.50
>>> (CVS CO 9/22/2006):
>>
>> What systems are involved here?
>
> Solaris 8

I don't have access to such systems, so I don't know how to debug this.
Can anyone help to debug this?

Can anyone reproduce it on other platforms (eg. GNU/Linux).

>
>> How do you log into the remote system?
>
> I can reproduce it with both klogin and ssh
>
>> What is the exact emacs version on the target system?
>> (pls. use C-u M-x emacs-version RET)
>
> GNU Emacs 22.0.50.1 (sparc-sun-solaris2.8, GTK+ Version 2.8.17) of
> 2006-09-22 on solaris8-build.Stanford.EDU
>
>
>> Does this happen with emacs 21.3 (or 21.4) ?
>
> No idea, the last version of emacs I could get to build and work on
> Solaris was 21.2.
>
>> If not, do you know when this started to happen with CVS emacs?
>
> Nope.
>
>
> Here's a simple cut and paste after I've managed to trigger it:
> ;; This buffer is for notes you don't want to save, and for Lisp evaluation.
> ;; If you want to create a file, visit that file with C-x C-f,
> ;; then enter the text in that file's own buffer.
>
> Test 1
> Test 2
> Test 3
> Test 4
> Test 5
> Test 6
> Test 7
> Test 8
> Test 9
> Test 10
> ^X^C well this sucks I can't ^X^C out of emacs
> ^P^P^N^N^N^N^N^P^N^P^N^P^F
>
>
> --Quanah
>
> --
> Quanah Gibson-Mount
> Principal Software Developer
> ITS/Shared Application Services
> Stanford University
> GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html

-- 
Kim F. Storm  http://www.cua.dk

Re: Emacs 22.0.50 and suspending while logged into another system

[Richard Stallman]

    Control characters stop being interpreted, and instead show up in the edit 
    buffer text.  In addition, foreground emacs does not redraw the screen.

I think wrong terminal modes are being installed.
Can you investigate using GDB what happens during init_sys_modes?

Re: Emacs 22.0.50 and suspending while logged into another system

[Quanah Gibson-Mount]

--On Wednesday, October 11, 2006 2:50 PM -0400 Richard Stallman 
<rms@gnu.org> wrote:

>     Control characters stop being interpreted, and instead show up in the
> edit      buffer text.  In addition, foreground emacs does not redraw the
> screen.
>
> I think wrong terminal modes are being installed.
> Can you investigate using GDB what happens during init_sys_modes?

I'd be happy to, if you can provide some explicit steps to take with gdb to 
gather the information you have requested.  Debugging under gdb is still 
something I'm learning.

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITS/Shared Application Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html

Re: Emacs 22.0.50 and suspending while logged into another system

[Richard Stallman]

    I'd be happy to, if you can provide some explicit steps to take with gdb to 
    gather the information you have requested.

I am too overloaded to do this myself, but how to use GDB is
not a secret.  You could order the GDB manual on paper from fsf.org
or else download its source code.

Re: Emacs 22.0.50 and suspending while logged into another system

[Quanah Gibson-Mount]

--On Thursday, October 12, 2006 11:02 AM -0400 Richard Stallman 
<rms@gnu.org> wrote:

>     I'd be happy to, if you can provide some explicit steps to take with
> gdb to      gather the information you have requested.
>
> I am too overloaded to do this myself, but how to use GDB is
> not a secret.  You could order the GDB manual on paper from fsf.org
> or else download its source code.

I understand the basics of using GDB (Setting break points, printing 
variables, etc).  What I'm not clear on, is what specific data from the 
init_sys_modes funtion you want.

Here's some data, which may or may not be useful to you:

When entering the function:

(gdb) print tty
$1 = {main = {c_iflag = 777708, c_oflag = 0, c_cflag = 0, c_lflag = 0, c_cc 
= '\0' <repeats 18 times>}}

(gdb) print tty.main
$2 = {c_iflag = 777708, c_oflag = 0, c_cflag = 0, c_lflag = 0, c_cc = '\0' 
<repeats 18 times>}

(gdb) print Vtty_erase_char
$11 = 3139585


After exiting emacs_set_tty:

(gdb) print tty
$30 = {main = {c_iflag = 10243, c_oflag = 1, c_cflag = 189, c_lflag = 2609, 
c_cc = "\a\a\177\025\001", '\0' <repeats 13 times>}}
(gdb) print tty.main
$31 = {c_iflag = 10243, c_oflag = 1, c_cflag = 189, c_lflag = 2609, c_cc = 
"\a\a\177\025\001", '\0' <repeats 13 times>}



and my TERM setting is:

TERM=vt100

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITS/Shared Application Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html

Re: Emacs 22.0.50 and suspending while logged into another system

[Richard Stallman]

    I understand the basics of using GDB (Setting break points, printing 
    variables, etc).  What I'm not clear on, is what specific data from the 
    init_sys_modes funtion you want.

I'm not asking for specific data; I just said I suspect that the
problem is there.  But more specifically, if my suspicion is right,
there would be bad data in `tty' when EMACS_SET_TTY is called.

Re: Emacs 22.0.50 and suspending while logged into another system

[Quanah Gibson-Mount]

--On Friday, October 13, 2006 7:19 AM -0400 Richard Stallman <rms@gnu.org> 
wrote:

>     I understand the basics of using GDB (Setting break points, printing
>     variables, etc).  What I'm not clear on, is what specific data from
> the      init_sys_modes funtion you want.
>
> I'm not asking for specific data; I just said I suspect that the
> problem is there.  But more specifically, if my suspicion is right,
> there would be bad data in `tty' when EMACS_SET_TTY is called.

When you said:

"I think wrong terminal modes are being installed.
Can you investigate using GDB what happens during init_sys_modes?"



I assumed this was a request for specific data, so I apologize for my 
misunderstanding of your request, and sending the information it appeared 
you were asking about to the list about an ongoing and valid bug.

In any case, you now have the information on what "tty" contained after it 
exited from EMACS_SET_TTY, and I'll leave it up to you and the other 
developers to decide whether or not that data was 'bad' or not.

If there *is* any data that the developers are interested in me collecting 
on this bug, please let me know.

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITS/Shared Application Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html

Re: Emacs 22.0.50 and suspending while logged into another system

[Richard Stallman]

    In any case, you now have the information on what "tty" contained after it 
    exited from EMACS_SET_TTY, and I'll leave it up to you and the other 
    developers to decide whether or not that data was 'bad' or not.

Did you examine that value just before the failure?
That is, did it fail just after this call to init_sys_modes?
init_sys_modes will be called each time you resume Emacs after
suspending; the time that we need to see the value
is the time that it fails.

It would be good to see the value of tty after EMACS_SET_TTY,
and the value of old_tty too.

Re: Emacs 22.0.50 and suspending while logged into another system

[Quanah Gibson-Mount]

--On Sunday, October 15, 2006 12:20 AM -0400 Richard Stallman <rms@gnu.org> 
wrote:

>     In any case, you now have the information on what "tty" contained
> after it      exited from EMACS_SET_TTY, and I'll leave it up to you and
> the other      developers to decide whether or not that data was 'bad' or
> not.
>
> Did you examine that value just before the failure?
> That is, did it fail just after this call to init_sys_modes?
> init_sys_modes will be called each time you resume Emacs after
> suspending; the time that we need to see the value
> is the time that it fails.
>
> It would be good to see the value of tty after EMACS_SET_TTY,
> and the value of old_tty too.

Hi Richard,


I set up the following in gdb:

l sysdep.c:1637
break 1637

(this is the line after EMACS_SET_TTY is called in the source I have).

Then, I

run -nw

Then use control-Z to suspend.

This *usually* then breaks at line 1637.  If this is where the break 
occurs, everything continues to work as desired.

However, sometimes it breaks in _libc_kill instead, completely bypassing 
EMACS_SET_TTY.
After this, things don't work correctly.  So it appears that the same logic 
is not always being followed when a SIGTSTP is caught.


The usual path I get is:

(control-Z):
Breakpoint 1, init_sys_modes () at sysdep.c:1642
1642          if (!flow_control) ioctl (input_fd, TCXONC, 1);
(gdb) cont
Program received signal SIGTSTP, Stopped (user).
0xfe6990cc in _poll () from /usr/lib/libc.so.1
(gdb)


Then I issue "signal SIGCONT" to resume.

On the other path:

(control-Z):
Program received signal SIGTSTP, Stopped (user).
0xfe69b638 in _libc_kill () from /usr/lib/libc.so.1

I tried "step"ing from here, but it just hangs forever, without entering 
init_sys_modes.

I can reproduce this very easily.  If I give a "signal SIGCONT" here, then 
it'll let me back into emacs.

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITS/Shared Application Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html

Re: Emacs 22.0.50 and suspending while logged into another system

[Quanah Gibson-Mount]

--On Monday, October 16, 2006 3:30 PM -0700 Quanah Gibson-Mount 
<quanah@stanford.edu> wrote:


Oh, here are the back traces of the different behaviors:

> The usual path I get is:
> (control-Z):
> Breakpoint 1, init_sys_modes () at sysdep.c:1642
> 1642          if (!flow_control) ioctl (input_fd, TCXONC, 1);

(gdb) bt
#0  init_sys_modes () at sysdep.c:1642
#1  0x00126a98 in unbind_to (count=Variable "count" is not available.
) at eval.c:3348
#2  0x000c515c in Fsuspend_emacs (stuffstring=3139585) at keyboard.c:10333
#3  0x00128340 in Ffuncall (nargs=Variable "nargs" is not available.
) at eval.c:2993
#4  0x0012555c in Fcall_interactively (function=3188649, 
record_flag=3139585, keys=3198980) at callint.c:880
#5  0x000c4264 in Fcommand_execute (cmd=3188649, record_flag=3139585, 
keys=3139585, special=Variable "special" is not available.
) at keyboard.c:9854
#6  0x000cc4e8 in command_loop_1 () at keyboard.c:1856
#7  0x00126548 in internal_condition_case (bfun=0xcc0e4 <command_loop_1>, 
handlers=3203745, hfun=0xc4dc0 <cmd_error>) at eval.c:1477
#8  0x000bea34 in command_loop_2 () at keyboard.c:1326
#9  0x00126168 in internal_catch (tag=3197977, func=0xbea10 
<command_loop_2>, arg=3139585) at eval.c:1218
#10 0x000be814 in command_loop () at keyboard.c:1305
#11 0x000be8ac in recursive_edit_1 () at keyboard.c:1003
#12 0x000bea00 in Frecursive_edit () at keyboard.c:1064
#13 0x000bd850 in main (argc=Variable "argc" is not available.
) at emacs.c:1794


> On the other path:
>
> (control-Z):
> Program received signal SIGTSTP, Stopped (user).
> 0xfe69b638 in _libc_kill () from /usr/lib/libc.so.1

(gdb) bt
#0  0xfe69b638 in _libc_kill () from /usr/lib/libc.so.1
#1  0x000c5150 in Fsuspend_emacs (stuffstring=3139585) at keyboard.c:10332
#2  0x00128340 in Ffuncall (nargs=Variable "nargs" is not available.
) at eval.c:2993
#3  0x0012555c in Fcall_interactively (function=3188649, 
record_flag=3139585, keys=3198980) at callint.c:880
#4  0x000c4264 in Fcommand_execute (cmd=3188649, record_flag=3139585, 
keys=3139585, special=Variable "special" is not available.
) at keyboard.c:9854
#5  0x000cc4e8 in command_loop_1 () at keyboard.c:1856
#6  0x00126548 in internal_condition_case (bfun=0xcc0e4 <command_loop_1>, 
handlers=3203745, hfun=0xc4dc0 <cmd_error>) at eval.c:1477
#7  0x000bea34 in command_loop_2 () at keyboard.c:1326
#8  0x00126168 in internal_catch (tag=3197977, func=0xbea10 
<command_loop_2>, arg=3139585) at eval.c:1218
#9  0x000be814 in command_loop () at keyboard.c:1305
#10 0x000be8ac in recursive_edit_1 () at keyboard.c:1003
#11 0x000bea00 in Frecursive_edit () at keyboard.c:1064
#12 0x000bd850 in main (argc=Variable "argc" is not available.
) at emacs.c:1794


--
Quanah Gibson-Mount
Principal Software Developer
ITS/Shared Application Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html

Re: Emacs 22.0.50 and suspending while logged into another system

[Richard Stallman]

    #0  0xfe69b638 in _libc_kill () from /usr/lib/libc.so.1
    #1  0x000c5150 in Fsuspend_emacs (stuffstring=3139585) at keyboard.c:10332
    #2  0x00128340 in Ffuncall (nargs=Variable "nargs" is not available.
    ) at eval.c:2993

That is very strange.
I think you need to consult a wizard for your system
to figure out what is happening here.

Re: Emacs 22.0.50 and suspending while logged into another system

[Quanah Gibson-Mount]

--On Tuesday, October 17, 2006 5:45 AM -0400 Richard Stallman <rms@gnu.org> 
wrote:

>     #0  0xfe69b638 in _libc_kill () from /usr/lib/libc.so.1
>     #1  0x000c5150 in Fsuspend_emacs (stuffstring=3139585) at
> keyboard.c:10332     #2  0x00128340 in Ffuncall (nargs=Variable "nargs"
> is not available.     ) at eval.c:2993
>
> That is very strange.
> I think you need to consult a wizard for your system
> to figure out what is happening here.

Okay, would that wizard be a gnu emacs developer specializing in Solaris, 
or someone else? ;)  I.e., where do I pursue this next?

thanks,

Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITS/Shared Application Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html

Re: Emacs 22.0.50 and suspending while logged into another system

[Richard Stallman]

    Okay, would that wizard be a gnu emacs developer specializing in Solaris, 
    or someone else? ;)  I.e., where do I pursue this next?

He has to be a Solaris wizard, but he doesn't need to know anything
about Emacs.  It is a small part of Emacs that he would need to
investigate.