Re: How to help debug numerous crashes of 21.4.4 on Windows?

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Jody Klymak]

dmkarr@earthlink.net (David M. Karr) writes:

> I have Cygwin-XEmacs 21.4.4 installed on my win2k system.  Every couple of
> days, it crashes at random times, usually while doing things in GNUS.  Just
> before it goes away, I see a DOS window come up and print something, but it
> goes away before I can read it.  Just now, it happened again, and I guess it
> had to do some swapping right after the DOS window came up, so I was able to
> read the message, which said something like: "Windows procedure called during
> GC?".

Hello David,

Just so you don't feel that you are going insane, I get similar
behaviour on my win2k machine, but using the native build.

XEmacs 21.4 (patch 3) "Academic Rigor" (win32) of Thu May 17 2001 on
SHALOM

gnus version ognus-0.04, though it does in on ognus-0.03 and gnus-5.8.8
as well.  

XEmacs sometimes crashes when entering a group that is not on my
machine for the first time (nnimap groups in my case), but not when
entering groups that are on my machine (i.e. nndiary, nndraft).  So
I'm guessing it is a difficulty in calling external processes (maybe
IMAP?).  Note that by "first time" I mean the first time entering
*any* group not on my machine.

As you say, it is intermittent.  Starting and stopping Xemacs a few
times makes it go away.  Some days its worse than others.  

If it does not crash on the first group entered, it continues to work
until I exit gnus.  But a new gnus session either works or it doesn't.
If it doesn't I need to restart Xemacs until I get one that works.

Umm, I didn't check if nntp groups fail or not.  

Rebooting does not seem to help.

By "crash", I mean the behaviour you describe above.  An error window
pops up, a DOS console with a bunch of debugging messages pops up.
The XEmacs window disappears.

I submitted a bug report about this a month or so ago.

Cheers,  Jody 

-- 
Jody M. Klymak               mailto:jklymak@oce.orst.edu
College of Oceanic and Atmospheric Sicences
Oregon State University, Corvallis, OR, 97331  

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Kai Großjohann]

Jody Klymak <jklymak@OCE.ORST.EDU> writes:

> Just so you don't feel that you are going insane, I get similar
> behaviour on my win2k machine, but using the native build.  [...]
>
> I submitted a bug report about this a month or so ago.

Knowing that it happens on the Cygwin and the native build might help
the XEmacs maintainers find the problem.  David, do you see a
backtrace that you could submit?  Do you see the same as Jody?

kai
-- 
Symbol's function definition is void: signature

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[David M. Karr]

>>>>> "Kai" == Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:

    Kai> Jody Klymak <jklymak@OCE.ORST.EDU> writes:
    >> Just so you don't feel that you are going insane, I get similar
    >> behaviour on my win2k machine, but using the native build.  [...]
    >> 
    >> I submitted a bug report about this a month or so ago.

    Kai> Knowing that it happens on the Cygwin and the native build might help
    Kai> the XEmacs maintainers find the problem.  David, do you see a
    Kai> backtrace that you could submit?  Do you see the same as Jody?

Do you mean a lisp backtrace?  I don't get any diagnostics except for the DOS
window with the message, and then it's gone.  At this point, I have to manually
kill the gnuserv process, as I can't bring up XEmacs again without killing the
old gnuserv process.

I would gladly configure my XEmacs to give me more diagnostics when this
happens, if I knew how to do that.

I'm not certain if I'm getting the same symptoms as Jody.  I haven't done a
very good job of remembering what I was doing when it bombs, except for doing
things in GNUS.

-- 
===================================================================
David M. Karr          ; Best Consulting
dmkarr@earthlink.net   ; Java/Unix/XML/C++/X ; BrainBench CJ12P (#12004)

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Jody Klymak]

dmkarr@earthlink.net (David M. Karr) writes:

> Do you mean a lisp backtrace?  I don't get any diagnostics except for the DOS
> window with the message, and then it's gone.  At this point, I have to manually
> kill the gnuserv process, as I can't bring up XEmacs again without killing the
> old gnuserv process.

Same here.  The backtrace, if that is what it is, is in that DOS
window.  The window self-destructs when XEmacs is done dying.

However, I can see what it says before it dies.  It changes each time.
How about that?

> I'm not certain if I'm getting the same symptoms as Jody.  I haven't done a
> very good job of remembering what I was doing when it bombs, except for doing
> things in GNUS.

Hmmm, doing things?  Mine is pretty "reproducible" on *first* entering
my nnimap groups.  Once I get an XEmacs that can read one nnimap group
it can read them all. 

Do you use nnimap?  

What version of Gnus?  Surely you are not using GNUS.

If anyone comes up with a way for David and I to debug this, I'll be
happy to do it (short of buying VC++)

Cheers,  Jody

-- 
Jody M. Klymak               mailto:jklymak@oce.orst.edu
College of Oceanic and Atmospheric Sicences
Oregon State University, Corvallis, OR, 97331  

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Stephen J. Turnbull]

>>>>> "David" == David M Karr <dmkarr@earthlink.net> writes:

    David> I have Cygwin-XEmacs 21.4.4 installed on my win2k system.

21.4.4 is deprecated on Windows.  Please read the announcements.

>>>>> "Kai" == Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:

    Kai> Jody Klymak <jklymak@OCE.ORST.EDU> writes:

    >> Just so you don't feel that you are going insane, I get similar
    >> behaviour on my win2k machine, but using the native build.
    >> [...]

    >> I submitted a bug report about this a month or so ago.

    Kai> Knowing that it happens on the Cygwin and the native build
    Kai> might help the XEmacs maintainers find the problem.  David,
    Kai> do you see a backtrace that you could submit?  Do you see the
    Kai> same as Jody?

This is a known bug, it's the reason that using 21.4.4 is deprecated
on Windows.  Ben and Adrian are working on it, the trial solutions are
not in 21.4 and will not be until I have time to verify that they
don't break other platforms (my impression is that they do involve
cross-platform code, but I don't even have time to confirm or
disconfirm that right now---I know it's a pretty big change).  Or we
get some experience with them in 21.5.

-- 
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
_________________  _________________  _________________  _________________
What are those straight lines for?  "XEmacs rules."

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Kai Großjohann]

Jody Klymak <jklymak@OCE.ORST.EDU> writes:

> Same here.  The backtrace, if that is what it is, is in that DOS
> window.  The window self-destructs when XEmacs is done dying.
>
> However, I can see what it says before it dies.  It changes each time.
> How about that?

Can you start XEmacs from a DOS window?  Then maybe the DOS window
stays around, and hopefully the error messages go in that window.

But I'm not an XEmacs expert, especially not on DOS.

kai
-- 
Symbol's function definition is void: signature

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Kai Großjohann]

"Stephen J. Turnbull" <turnbull@sk.tsukuba.ac.jp> writes:

> This is a known bug, it's the reason that using 21.4.4 is deprecated
> on Windows.

Ah!  Well, everything is clear, then, I think.

kai
-- 
Symbol's function definition is void: signature

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Jody Klymak]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> "Stephen J. Turnbull" <turnbull@sk.tsukuba.ac.jp> writes:
>
>> This is a known bug, it's the reason that using 21.4.4 is deprecated
>> on Windows.
>
> Ah!  Well, everything is clear, then, I think.

Yes, except I'm using 21.4.3...

Cheers,  Jody

-- 
Jody M. Klymak               mailto:jklymak@oce.orst.edu
College of Oceanic and Atmospheric Sicences
Oregon State University, Corvallis, OR, 97331  

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Andy Piper]

At 01:33 PM 9/2/01 -0700, Jody Klymak wrote:
>XEmacs sometimes crashes when entering a group that is not on my
>machine for the first time (nnimap groups in my case), but not when
>entering groups that are on my machine (i.e. nndiary, nndraft).  So
>I'm guessing it is a difficulty in calling external processes (maybe
>IMAP?).  Note that by "first time" I mean the first time entering
>*any* group not on my machine.
>
>As you say, it is intermittent.  Starting and stopping Xemacs a few
>times makes it go away.  Some days its worse than others.
>
>If it does not crash on the first group entered, it continues to work
>until I exit gnus.  But a new gnus session either works or it doesn't.
>If it doesn't I need to restart Xemacs until I get one that works.
>
>Umm, I didn't check if nntp groups fail or not.
>
>Rebooting does not seem to help.
>
>By "crash", I mean the behaviour you describe above.  An error window
>pops up, a DOS console with a bunch of debugging messages pops up.
>The XEmacs window disappears.
>
>I submitted a bug report about this a month or so ago.

The reasons for this are known, we just need to fix it :)

andy

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[David M. Karr]

>>>>> "Stephen" == Stephen J Turnbull <turnbull@sk.tsukuba.ac.jp> writes:

>>>>> "David" == David M Karr <dmkarr@earthlink.net> writes:
    David> I have Cygwin-XEmacs 21.4.4 installed on my win2k system.

    Stephen> 21.4.4 is deprecated on Windows.  Please read the announcements.

Yes, I missed that announcement.  I knew it had problems on Windows, but I
didn't know it was "deprecated".  I still have 21.4.3 installed (I'm running it
now).

-- 
===================================================================
David M. Karr          ; Best Consulting
dmkarr@earthlink.net   ; Java/Unix/XML/C++/X ; BrainBench CJ12P (#12004)

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Stephen J. Turnbull]

>>>>> "JMK" == Jody Klymak <jklymak@OCE.ORST.EDU> writes:

    JMK> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
    >> "Stephen J. Turnbull" <turnbull@sk.tsukuba.ac.jp> writes:
    >> 
    >>> This is a known bug, it's the reason that using 21.4.4 is
    >>> deprecated on Windows.
    >> Ah!  Well, everything is clear, then, I think.

    JMK> Yes, except I'm using 21.4.3...

Right.  It's not just in 21.4.4, but something about 21.4.4 means it
gets tickled far more often.  Apparently Jody always gets clobbered in
21.4.3 but most other people didn't.

See the thread "Call Stack for Window..." starting around July 24 for
more details, but basically it's a problem that a structure containing
live Lisp pointers was getting collected too early.  Messing with GC
is inherently destabilizing; I'm not going to apply patches
willy-nilly.

-- 
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
_________________  _________________  _________________  _________________
What are those straight lines for?  "XEmacs rules."

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Andy Piper]

Can you point me at the 21.5 change that is supposed to deal with this?

Thanks

andy

At 10:28 AM 9/4/01 +0900, Stephen J. Turnbull wrote:
> >>>>> "JMK" == Jody Klymak <jklymak@OCE.ORST.EDU> writes:
>
>     JMK> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>     >> "Stephen J. Turnbull" <turnbull@sk.tsukuba.ac.jp> writes:
>     >>
>     >>> This is a known bug, it's the reason that using 21.4.4 is
>     >>> deprecated on Windows.
>     >> Ah!  Well, everything is clear, then, I think.
>
>     JMK> Yes, except I'm using 21.4.3...
>
>Right.  It's not just in 21.4.4, but something about 21.4.4 means it
>gets tickled far more often.  Apparently Jody always gets clobbered in
>21.4.3 but most other people didn't.
>
>See the thread "Call Stack for Window..." starting around July 24 for
>more details, but basically it's a problem that a structure containing
>live Lisp pointers was getting collected too early.  Messing with GC
>is inherently destabilizing; I'm not going to apply patches
>willy-nilly.
>
>--
>University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
>Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
>_________________  _________________  _________________  _________________
>What are those straight lines for?  "XEmacs rules."

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Stephen J. Turnbull]

Cleaned ccs; leaving ding.

    Andy> Can you point me at the 21.5 change that is supposed to deal
    Andy> with this?  Thanks

http://list-archive.xemacs.org/xemacs-patches/200106/msg00029.html

Here's the relevant part of the ChangeLog.  Look right to you?  Is it
safe for 21.4.5?

2001-06-08  Ben Wing  <ben@xemacs.org>

------ gc-in-window-procedure fixes ------
	
	* alloc.c:
	* alloc.c (struct post_gc_action):
	* alloc.c (register_post_gc_action):
	* alloc.c (run_post_gc_actions):
	* alloc.c (garbage_collect_1):
	Create "post-gc actions", to avoid those dreaded "GC during window
	procedure" problems.

	* event-msw.c:
	Abort, clean and simple, when GC in window procedure.  We want
	to flush these puppies out.

	* glyphs-msw.c:
	* glyphs-msw.c (finalize_destroy_window):
	* glyphs-msw.c (mswindows_finalize_image_instance):
	Use a post-gc action when destroying subwindows.

	* lisp.h:
	Declare register_post_gc_action().

	* scrollbar-msw.c:
	* scrollbar-msw.c (unshow_that_mofo):
	Use a post-gc action when unshowing scrollbar windows, if in gc.

	* redisplay.c (mark_redisplay):
	Add comment about the utter evilness of what's going down here.



-- 
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
_________________  _________________  _________________  _________________
What are those straight lines for?  "XEmacs rules."

Re: How to help debug numerous crashes of 21.4.4 on Windows?

[Andy Piper]

At 12:08 PM 9/4/01 +0900, Stephen J. Turnbull wrote:
>Cleaned ccs; leaving ding.
>
>     Andy> Can you point me at the 21.5 change that is supposed to deal
>     Andy> with this?  Thanks
>
>http://list-archive.xemacs.org/xemacs-patches/200106/msg00029.html
>
>Here's the relevant part of the ChangeLog.  Look right to you?

Yes, although I object to the function named unshow_that_mofo. I'm assuming 
that we switch off assertions in production, the comments seem to indicate so.

>Is it
>safe for 21.4.5?

Yes. I've reviewed the patch and it may not fix all the occurrences but we 
should get it in the hands of users and see whether there are anymore. 
Especially since we have users with easily reproducible test cases.

andy

RE: How to help debug numerous crashes of 21.4.4 on Windows?

[Andy Piper]

I'm going to apply this in my local 21.4 workspace and run with it for a
while, since I see these crashes fairly often.

andy

-----Original Message-----
From: xemacs-beta-admin@xemacs.org
[mailto:xemacs-beta-admin@xemacs.org]On Behalf Of Stephen J. Turnbull
Sent: Monday, September 03, 2001 8:08 PM
To: Andy Piper
Cc: xemacs-beta@xemacs.org; ding@gnus.org
Subject: Re: How to help debug numerous crashes of 21.4.4 on Windows?


Cleaned ccs; leaving ding.

    Andy> Can you point me at the 21.5 change that is supposed to deal
    Andy> with this?  Thanks

http://list-archive.xemacs.org/xemacs-patches/200106/msg00029.html

Here's the relevant part of the ChangeLog.  Look right to you?  Is it
safe for 21.4.5?

2001-06-08  Ben Wing  <ben@xemacs.org>

------ gc-in-window-procedure fixes ------

	* alloc.c:
	* alloc.c (struct post_gc_action):
	* alloc.c (register_post_gc_action):
	* alloc.c (run_post_gc_actions):
	* alloc.c (garbage_collect_1):
	Create "post-gc actions", to avoid those dreaded "GC during window
	procedure" problems.

	* event-msw.c:
	Abort, clean and simple, when GC in window procedure.  We want
	to flush these puppies out.

	* glyphs-msw.c:
	* glyphs-msw.c (finalize_destroy_window):
	* glyphs-msw.c (mswindows_finalize_image_instance):
	Use a post-gc action when destroying subwindows.

	* lisp.h:
	Declare register_post_gc_action().

	* scrollbar-msw.c:
	* scrollbar-msw.c (unshow_that_mofo):
	Use a post-gc action when unshowing scrollbar windows, if in gc.

	* redisplay.c (mark_redisplay):
	Add comment about the utter evilness of what's going down here.



--
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
_________________  _________________  _________________  _________________
What are those straight lines for?  "XEmacs rules."