Bug 1565 - Nautilus sometimes stop working and will not finish
: Nautilus sometimes stop working and will not finish
Status: CLOSED FIXINBUILD
Product: opensolaris
desktop
: unspecified
: Other Solaris
: P2 major (vote)
: in-200811
Assigned To: fujiwara
: Watcher account for desktop bugs
:
: rn3
:
:
:
  Show dependency treegraph
 
Reported: 2008-04-25 03:43 UTC by Fuyuki Hasegawa
Modified: 2009-03-20 11:38 UTC (History)
3 users (show)

See Also:


Attachments
Patch for eel-enumeration.c (761 bytes, patch)
2008-04-30 08:36 UTC, fujiwara
no flags Details | Diff


Note

You need to log in before you can comment on or make changes to this bug.


Description Fuyuki Hasegawa 2008-04-25 03:43:14 UTC
When log in gnome, background is grayed out and mouse bottom doesn't
work on the background while mouse bottom work on panel.

When this happened, nautilus-debug-log.txt presents under $home
and it's repeatedly updated with the following message.

819c228 2008/04/25 10:59:12.0948 (USER): debug log dumped due to signal 11
819c228 2008/04/25 10:59:12.1202 (USER): debug log dumped due to signal 11
819c228 2008/04/25 10:59:12.1456 (USER): debug log dumped due to signal 11
    :

Once this happens, re-login doesn't help.
The odd nautilus process will not be finished by logout
and the number of the odd process will increase every login.
To escape from this phenomenon, manually kill the odd
nautilus processes or bonobo-activation-server process.

We, G11N team, sometimes encountered this problem both during
live boot and after install from earlier images. Today I also
encountered it in the RC2 installed system.
We tried to find out the reproducible case, but still don't know.
FYI, not always, but we sometimes encountered the problem when
changed login locale and set it as default through gdm lang session.
Comment 1 Lin Ma 2008-04-28 03:07:25 UTC
Could you remote log in the machine and verify whether gam_server is running.
If gam_server isn't living you may check if /tmp/fam-${LOGUSER}/fam- is
existing. If it is, it's a gamin bug and you may replace the SUNWgamin package
with the one in the Vermillion 88 or above.
Comment 2 jack.a.schwartz 2008-04-28 17:30:22 UTC
I have hit this problem as well.  Thru the following sequence, the problem will
usually reproduce, sometimes when I login changing the locale for "just this
session", sometimes when changing it permanently.

1) Install RC3.  Select Korean as the default locale.

2) login without changing the locale.  (May or may not contribute to the
reproduction, but this is the sequence I followed.)

3) logout.

4) At login screen, click lower left button and select American English, and
then say yes to restarting the login screen with chosen language.

5) login, then select to make American English the default for just this
session.
(Sometimes I'll see the problem at this point, sometimes not.)  If I don't see
the problem, I continue...

6) logout

7) At login screen, select American English again, and then say yes again to
restarting the login screen with chosen language.

8) login, then select to make American English the default for future sessions.
 (Sometimes I'll see the problem at this point.)

When I see the problem:
- Takes a long time for the screen to initialize to the point where I have
drop-downs menus at the top of the screen.
- I see the same nautilus-debug-log.txt messages.  The messages are continuous,
and the file gets truncated at ~3Mb and builds up again.
- No dark blue (wallpaper?) screen, only blue-gray solid color background.
- Mouse right-click on blank part of screen does not bring up a menu like it
should.


To reply to comment (1):
- gam_server is running
- /tmp/fam-schwartz/fam- exists. (my username is schwartz).  echo $LOGUSER
shows nothing, however.

I've seen this with both the regular RC3 and the global RC3 images.
Comment 3 Fuyuki Hasegawa 2008-04-30 04:53:22 UTC
To be release noted for 2008.5.

Description:

When you log in gnome, you may encounter the phenomenon that
desktop background is blue-gray solid color and mouse right-click
on the background does not bring up a menu. If this occurred,
nautilus is no longer working correctly. Nautilus would consume
50% CPU time and nautilus-debug-log.txt presents under your
home directory.

Workaround:

Manually kill the broken nautilus process.

1. Launch a gnome-terminal through Applications->System Tool->Terminal.
2. Kill the oldest nautilus process in the gnome-terminal. For example:

bash$ ps -ef | grep naut
    test 13425 13374   0 19:59:30 ?           0:00 nautilus --no-default-window
--sm-client-id default2
    test 13227     1  50 17:39:39 ?          39:36 nautilus --no-default-window
--sm-client-id default2

bash$ kill 13227

3. logout
Comment 4 Fuyuki Hasegawa 2008-04-30 05:16:56 UTC
I encountered the problem again.
I logged in Japanese locale and left the session a while, around 1 hour(?). 
Then I logout and login with American English (for just this session).
The problem was reproduced. I'm not sure when nautilus stopped working,
before logout or during logout... 

gam_server was running at that time.

There were two nautilus processes.

test@rc3:~$ ps -ef | grep naut
    test 13425 13374   0 19:59:30 ?           0:00 nautilus --no-default-window
--sm-client-id default2
    test 13227     1  50 17:39:39 ?          39:36 nautilus --no-default-window
--sm-client-id default2

The new nautilus process (13425) seemed normal status (pollsys).
The old nautilus process (13227) consumed 50% of CPU.

test@rc3:~$ pstack 13227
13227:  nautilus --no-default-window --sm-client-id default2
 fef05e35 __write  (c, 91c234c, 20000) + 15
 feed8ee9 _xflsbuf (8199ea0) + a9
 feed9b22 fputs    (906e6e0, 8199ea0) + f2
 0810c1c0 write_string (84a9758, 8199ea0, 906e6e0, 0) + 20
 0810c2b9 nautilus_debug_log_dump (84a9758, 0) + b5
 080b653e dump_debug_log (fef98000, 3, 400, 804754c, fef0146f, b) + 26
 080b6608 sigfatal_handler (b, 0, 80475e4) + 24
 fef0146f __sighndlr (b, 0, 80475e4, 80b65e4) + f
 feef6312 call_user_handler (b, 0, 80475e4) + 2bf
 feef6546 sigacthandler (b, 0, 80475e4) + d0
 --- called from signal handler with signal 11 (SIGSEGV) ---
 fe4c1a3e magazine_cache_push_magazine (1, 81e2ca0, 33) + 56
 fe4c1ee6 thread_memory_magazine2_unload (819c870, 1) + 2a
 fe4c23ec g_slice_free1 (10, 81ceb00) + 1c0
 fe49afc5 g_hash_nodes_destroy (81ceb00, 0, 0) + 55
 fe49a60c g_hash_table_remove_all (819e400, 81cecd0, fee40a54, 80478ec,
fedef236
, 819e400) + 34
 fe499ef4 g_hash_table_destroy (819e400) + 28
 fedef236 enumeration_table_free (0, 81c6398, 81cc600, 0, 8047b3c, feffde48) +
3
a
 fede7da4 eel_debug_shut_down (8047bf4, 8047b20, feffb7cc, 8047b0c, 804795c,
816
72dc) + 3c
 080b6f70 main     (4, 8047b64, 8047b78) + 8c0
 080a5c3a _start   (4, 8047c5c, 8047c65, 8047c79, 8047c88, 0) + 7a
Comment 5 fujiwara 2008-04-30 08:36:35 UTC
Created an attachment (id=272) [details]
Patch for eel-enumeration.c

It's caused by the memory allocation.
The strdup() is needed since enumeration_table_free() is called.
Comment 6 fujiwara 2008-09-03 04:01:21 UTC
GNOME 2.24 will be Indiana 2008.11.