Fri 27 May 2005 10:03:59 PM UTC, comment #26:
transformed into feature request for later work
|
Thu 26 May 2005 10:17:26 AM UTC, comment #25:
See patch #4045
|
Thu 26 May 2005 10:15:55 AM UTC, comment #24:
"- the bug does not exist in 2.5.16w1"
I don´t think this is true. The patch below shows that also
this core prints "Core stopped" in its logfile when started
with "-daemon".
Yet the bug looks different, from the logfile:
===============================================================
26/ 5, 12:11:04: Core stopped
26/ 5, 12:11:04: Saving shared files and sources
26/ 5, 12:11:04: Options correctly saved
26/ 5, 12:13:35: Core stopped
26/ 5, 12:13:35: Saving shared files and sources
26/ 5, 12:13:35: Options correctly saved
===============================================================
--- ../cvs/knocker-2-5-16w1/src/daemon/driver/driverMain.ml 2005-05-19 00:13:15.000000000 +0200
+++ ./src/daemon/driver/driverMain.ml 2005-05-26 12:08:06.000000000 +0200
@@ -505,10 +505,10 @@
CommonGlobals.print_localtime ();
lprintf (_b "Core started\n");
core_included := true;
-(*
+
CommonGlobals.do_at_exit (fun _ -> lprintf "Core stopped\n");
CommonGlobals.do_at_exit (fun _ -> CommonGlobals.print_localtime ());
-*)
+
if not !keep_console_output then begin
lprintf (_b "Disabling output to console, to enable: stdout true\n");
===============================================================
This is the result of the patch from 2004/04/06 which solved
another bug, so I still think removing "-daemon" is the way to
go as we have an easy replacement.
|
Thu 19 May 2005 02:46:16 AM UTC, comment #23:
Some new thoughts about that one:
- the bug does not exist in 2.5.16w1
- I think it was introduced in 2.5.18:
2004/04/06: Fabrice
- CommonGlobals: only execute 'do_at_exit' in the main process, not
in the forked ones (should solve bug #7738).
|
Wed 18 May 2005 09:37:08 PM UTC, comment #22:
Updated patch so it prints infos when "mlnet -daemon" was executed:
"Option -daemon was removed.
Use 'mlnet > /dev/null 2>&1 &' instead, exiting..."
The core will exit so users are forced to update their scripts.
Most users won´t notice as most common distros don´t use that function anyway.
|
Wed 18 May 2005 09:08:51 PM UTC, comment #21:
Here is a summary how MLDonkey is started on some distros:
Gentoo (mldonkey.initd): /usr/bin/mlnet &>${LOG} &
Debian: uses /usr/bin/mldonkey-server which daemonizes mlnet itself with "--daemon" option
Fedora(ATrpms): http://dl.atrpms.net/production/sources/fedora-4-i386/atrpms/mldonkey-2.5.28.1-43.rhfc3.92.at.src.rpm
exec /usr/@lib@/mldonkey/mlnet >> /mldonkey.log 2>&1 < /dev/null" &
PLD: ftp://ftp.pld-linux.org/dists/2.0/PLD/SRPMS/SRPMS/mldonkey-2.5.30.1-1.src.rpm
exec $MLDONKEY_PATH > ${LOG:-/dev/null} 2>&1
|
Wed 18 May 2005 09:34:34 AM UTC, comment #20:
As I wrote in comment #7 starting the core with
mlnet > /dev/null 2>&1 &
works without problems. Having a function which causes data loss
is also not a good deal. I spent already two hours in debugging
but did not succeed. If no one comes up with a better solution
package maintainers should change to the other way starting the core.
|
Wed 18 May 2005 07:21:53 AM UTC, comment #19:
removing functionality used by those who dont use/have init scripts for mldonkey isnt very user friendly and is a wontfix and not a fix
|
Tue 17 May 2005 08:17:38 PM UTC, comment #18:
I think we should drop "-daemon", patch attached.
|
Tue 17 May 2005 08:07:33 PM UTC, comment #17:
It seems the core goes through the stop process once it is started.
gentoo a # ps axHu | grep mlnet
root 2970 0.7 2.8 31896 18112 ? Ssl 21:58 0:03 ./mlnet -daemon
root 2970 0.0 2.8 31896 18112 ? SNsl 21:58 0:00 ./mlnet -daemon
Process 2969 does not exist anymore, so I think the main thread
is dropped immediatly after core start because when "MlUnix.detach_daemon ()"
is called the core already is threaded. See the comment in the code:
" (* Question: is-it not to late to go in background ? Is-it possible that
we have started some threads already ? What happens then ? *)"
The answer to one is yes, two also, three? Core stops main thread and produces this bug.
|
Tue 17 May 2005 08:05:07 PM UTC, comment #16:
From logfile:
Welcome to MLdonkey
Starting with pid 2969
Options correctly saved on localtime: 17/ 5, 21:58:10
Core started on localtime: 17/ 5, 21:58:10
Disabling output to console, to enable: stdout true
threads are enabled!!
Options correctly saved on localtime: 17/ 5, 21:58:10
on localtime: 17/ 5, 21:58:10
Core stopped
Core stopped on localtime: 17/ 5, 21:58:10
|
Tue 17 May 2005 08:00:57 PM UTC, comment #15:
It seems daemon´izing the core does not work properly. For debugging I tested this patch:
--- ./src/daemon/driver/driverMain.ml 2005-05-17 21:34:42.000000000 +0200
+++ ./src/daemon/driver/driverMain.ml 2005-05-17 21:39:11.000000000 +0200
@@ -659,10 +659,10 @@
()
end else
close_log ();
-
+
+ if BasicSocket.has_threads () then lprintf "\n\nthreads are enabled!!\n\n";
(* Question: is-it not to late to go in background ? Is-it possible that
we have started some threads already ? What happens then ? *)
if !daemon then MlUnix.detach_daemon ()
end
-
|
Mon 09 May 2005 12:58:54 PM UTC, comment #14:
Comment #13 is solved by patch #3993
|
Mon 09 May 2005 11:42:35 AM UTC, comment #13:
With MinGW core CTRL+C does not save ini files and config_files_space.tmp is not deleted.
Of course it does not because in ./src/config/mingw/mlUnix.ml
let set_signal signal f = ()
|
Thu 14 Apr 2005 10:45:49 PM UTC, comment #12:
the problem is still in 2-5-30-5
I started with
#./mlnet -log_file log -deamon
end of log file:
>Core started on localtime: 15/ 4, 00:39:06
>Disabling output to console, to enable: stdout true
>Options correctly saved on localtime: 15/ 4, 00:39:08
> on localtime: 15/ 4, 00:39:08
>Core stopped
>Core stopped on localtime: 15/ 4, 00:39:08
the temp file "config_files_space.tmp" is gone
BUT the core is running
|
Fri 01 Apr 2005 05:45:58 PM UTC, comment #11:
kempston and whoever had/has this problem: could you please check if the problem, that you ini files are not saved, is fixed in 2.5.30ac ?
|
Wed 16 Mar 2005 09:08:11 PM UTC, comment #10:
oh sorry i meant revision 231:
https://opensvn.csie.org/viewcvs.cgi?root=mlnet&rev=231&view=rev
|
Wed 16 Mar 2005 09:04:32 PM UTC, comment #9:
perhaps my revision 228 fixes some of these problems.
https://opensvn.csie.org/viewcvs.cgi?root=mlnet&rev=228&view=rev
|
Thu 03 Mar 2005 01:19:42 PM UTC, comment #8:
"from driverMain.ml:
CommonGlobals.do_at_exit (fun _ ->
BasicSocket.close_all ();
Sys.remove security_space_filename;
DriverInteractive.save_config ();
CommonGlobals.print_localtime ();
lprintf (_b "Core stopped\n"));
These commands should be executed every time the core is stopped, regardless if through process signals (SIGTERM etc.), the "kill"-command or CTRL+C
"
These commands are probably not executed everytime the core is stopped ...
Currently I am lacking time, to take a closer look at it and debug it ... however:
The first command should -better must- fail with an exception (permission denied), if mldonkey is stared with the -run_as_user or -run_as_uid options.
Remember, a setuid can only be done with root-privileges and the sockets are bound prior the setuid with root-privileges. Socket-binding with root priviledges while not running the core with root privileges is the purpose of the above options, right?
... I bet due to an exception the game is over for the remaining commands then ...
Not sure about a plain -daemon option, without an additional setuid, a plain -daemon should save the ini-files.
One can check the neccessary permissions to close the sockets in different startup-configurations with:
ls -al /proc/$(($(ps -C mlnet -o pid --no-header f|head -n1)))/fd|grep socket
|
Thu 10 Feb 2005 12:23:15 PM UTC, comment #7:
It seems that "-daemon" is the reason why ini files are not saved. Here I normally start mlnet like this:
/usr/local/bin/nice -n 10 /usr/local/bin/chroot /mnt/mlnet /mlnet > /dev/null 2>&1 &
so, no "-daemon", thats the reason I never experienced that bug...
|
Thu 10 Feb 2005 06:41:24 AM UTC, comment #6:
config_files_space.tmp is gone after kill
starting: mlnet -daemon
--> *.ini not saved & no "Core stopped" in logfile
starting: mlnet -stdout
--> *.ini saved & and "Core stopped" on console
|
Tue 08 Feb 2005 07:48:21 PM UTC, comment #5:
from driverMain.ml:
CommonGlobals.do_at_exit (fun _ ->
BasicSocket.close_all ();
Sys.remove security_space_filename;
DriverInteractive.save_config ();
CommonGlobals.print_localtime ();
lprintf (_b "Core stopped\n"));
These commands should be executed every time the core is stopped, regardless if through process signals (SIGTERM etc.), the "kill"-command or CTRL+C
Do you get any messages in logfile (try starting with ./mlnet -stdout) when you issue the kill command in telnet interface?
|
Tue 08 Feb 2005 07:37:06 PM UTC, comment #4:
I tried it with 2-5-28r ten times or more and each time the inis
were saved wken "kill" was issued through telnet, HTML or sancho.
What about config_files_space.tmp?
Is this file deleted when you kill the core?
|
Tue 08 Feb 2005 10:40:29 AM UTC, comment #3:
I can vertify this bug, tried it several times.
version 2-5-28-r
|
Wed 19 Jan 2005 03:37:27 PM UTC, comment #2:
I tried it several times before posting this bug.
My stats are posted here:
http://mldonkey.org/portal/modules.php?op=modload&name=phpBB2&file=viewtopic&t=5166
And Surround confirmed it.
|
Mon 10 Jan 2005 06:24:21 PM UTC, comment #1:
Can´t confirm that bug with 2-5-28m. Actually I never saw this bug.
Two files compared:
- before killing the core
-rw-r--r-- 1 root root 2653089 Jan 10 19:13 file_sources.ini
-rw-r--r-- 1 root root 2632173 Jan 10 18:53 file_sources.ini.old
- after killing the core on 7:21pm
-rw-r--r-- 1 root root 2662071 Jan 10 19:21 file_sources.ini
-rw-r--r-- 1 root root 2653089 Jan 10 19:13 file_sources.ini.old
Here the core directory before killing the core via telnet:
drwxr-xr-x 2 root 65534 1024 Jun 28 2004 .mldonkey
drwxr-xr-x 2 mlnet mlnet 1024 Dec 14 13:41 backup
-rw-r--r-- 1 root root 1349 Jan 10 19:13 bittorrent.ini
-rw-r--r-- 1 root root 1349 Jan 10 18:53 bittorrent.ini.old
-rw-r--r-- 1 root root 1 Nov 16 15:13 comments.met
-rw-r--r-- 1 root root 10485760 Jan 10 12:33 config_files_space.tmp
drwxr-xr-x 2 root root 1024 Aug 13 09:31 dev
-rw-r--r-- 1 root 65534 1685 Dec 4 16:47 directconnect.ini
-rw-r--r-- 1 root root 11102 Jan 10 19:13 donkey.ini
-rw-r--r-- 1 root root 11102 Jan 10 18:53 donkey.ini.old
-rw-r--r-- 1 root root 22377 Jan 10 19:13 downloads.ini
-rw-r--r-- 1 root root 22377 Jan 10 18:53 downloads.ini.old
drwxr-xr-x 2 mlnet mlnet 1024 Feb 10 2004 etc
-rw-r--r-- 1 root root 1303 Jan 10 19:13 fasttrack.ini
-rw-r--r-- 1 root root 1303 Jan 10 18:53 fasttrack.ini.old
-rw-r--r-- 1 root root 951 Jan 10 19:13 fileTP.ini
-rw-r--r-- 1 root root 951 Jan 10 18:53 fileTP.ini.old
-rw-r--r-- 1 root root 2653089 Jan 10 19:13 file_sources.ini
-rw-r--r-- 1 root root 2632173 Jan 10 18:53 file_sources.ini.old
-rw-r--r-- 1 root root 607369 Jan 10 19:13 files.ini
-rw-r--r-- 1 root root 607310 Jan 10 18:53 files.ini.old
-rw-r--r-- 1 root root 116 Jan 10 19:13 friends.ini
-rw-r--r-- 1 root root 116 Jan 10 18:53 friends.ini.old
-rw-r--r-- 1 root root 1793 Jan 10 19:13 gnutella.ini
-rw-r--r-- 1 root root 1793 Jan 10 18:53 gnutella.ini.old
-rw-r--r-- 1 root root 2197 Jan 10 19:13 gnutella2.ini
-rw-r--r-- 1 root root 2197 Jan 10 18:53 gnutella2.ini.old
-rw-r--r-- 1 root root 2520205 Jan 9 06:45 guarding.p2p
drwxr-xr-x 2 mlnet mlnet 1024 May 30 2004 lib
-rw------- 1 root root 3370 Jan 4 18:50 messages.log
-rwxr-xr-x 1 root 65534 916 Nov 16 15:13 mldonkey_submit
-rwxr-xr-x 1 root root 4042984 Jan 10 12:32 mlnet
-rwxr--r-- 1 root 65534 291 Nov 16 15:13 mlsubmit.reg
-rw-r--r-- 1 root 65534 48 Jan 10 19:13 onlinesig.dat
-rw-r--r-- 1 root 65534 1260 Dec 4 16:47 opennap.ini
drwxr-xr-x 2 root root 1024 Dec 4 16:48 searches
-rw-r--r-- 1 root root 2095 Jan 10 19:13 searches.ini
-rw-r--r-- 1 root root 2095 Jan 10 18:53 searches.ini.old
-rw-r--r-- 1 root root 60413 Jan 10 19:13 servers.ini
-rw-r--r-- 1 root root 60478 Jan 10 18:53 servers.ini.old
-rw-r--r-- 1 root root 61025 Jan 10 19:13 shared_files_new.ini
-rw-r--r-- 1 root root 61025 Jan 10 18:53 shared_files_new.ini.old
drwxr-xr-x 8 mlnet mlnet 1024 Aug 11 01:24 shares
-rw-r--r-- 1 root 65534 1270 Dec 4 16:47 soulseek.ini
-rw-r--r-- 1 root root 1521 Jan 10 19:13 stats.ini
-rw-r--r-- 1 root root 1521 Jan 10 18:53 stats.ini.old
-rw-r--r-- 1 root root 10113 Jan 10 19:13 stats_mod.ini
-rw-r--r-- 1 root root 10113 Jan 10 18:53 stats_mod.ini.old
drwxr-xr-x 5 mlnet mlnet 2048 Jun 21 2004 torrents
Here the directory after killing the core:
drwxr-xr-x 2 root 65534 1024 Jun 28 2004 .mldonkey
drwxr-xr-x 2 mlnet mlnet 1024 Dec 14 13:41 backup
-rw-r--r-- 1 root root 1349 Jan 10 19:21 bittorrent.ini
-rw-r--r-- 1 root root 1349 Jan 10 19:13 bittorrent.ini.old
-rw-r--r-- 1 root root 1 Nov 16 15:13 comments.met
drwxr-xr-x 2 root root 1024 Aug 13 09:31 dev
-rw-r--r-- 1 root 65534 1685 Dec 4 16:47 directconnect.ini
-rw-r--r-- 1 root root 11102 Jan 10 19:21 donkey.ini
-rw-r--r-- 1 root root 11102 Jan 10 19:13 donkey.ini.old
-rw-r--r-- 1 root root 22377 Jan 10 19:20 downloads.ini
-rw-r--r-- 1 root root 22377 Jan 10 19:13 downloads.ini.old
drwxr-xr-x 2 mlnet mlnet 1024 Feb 10 2004 etc
-rw-r--r-- 1 root root 1303 Jan 10 19:21 fasttrack.ini
-rw-r--r-- 1 root root 1303 Jan 10 19:13 fasttrack.ini.old
-rw-r--r-- 1 root root 951 Jan 10 19:21 fileTP.ini
-rw-r--r-- 1 root root 951 Jan 10 19:13 fileTP.ini.old
-rw-r--r-- 1 root root 2662071 Jan 10 19:21 file_sources.ini
-rw-r--r-- 1 root root 2653089 Jan 10 19:13 file_sources.ini.old
-rw-r--r-- 1 root root 607311 Jan 10 19:21 files.ini
-rw-r--r-- 1 root root 607369 Jan 10 19:13 files.ini.old
-rw-r--r-- 1 root root 116 Jan 10 19:21 friends.ini
-rw-r--r-- 1 root root 116 Jan 10 19:13 friends.ini.old
-rw-r--r-- 1 root root 1793 Jan 10 19:21 gnutella.ini
-rw-r--r-- 1 root root 1793 Jan 10 19:13 gnutella.ini.old
-rw-r--r-- 1 root root 2197 Jan 10 19:21 gnutella2.ini
-rw-r--r-- 1 root root 2197 Jan 10 19:13 gnutella2.ini.old
-rw-r--r-- 1 root root 2520205 Jan 9 06:45 guarding.p2p
drwxr-xr-x 2 mlnet mlnet 1024 May 30 2004 lib
-rw------- 1 root root 3370 Jan 4 18:50 messages.log
-rwxr-xr-x 1 root 65534 916 Nov 16 15:13 mldonkey_submit
-rwxr-xr-x 1 root root 4042984 Jan 10 12:32 mlnet
-rwxr--r-- 1 root 65534 291 Nov 16 15:13 mlsubmit.reg
-rw-r--r-- 1 root 65534 48 Jan 10 19:20 onlinesig.dat
-rw-r--r-- 1 root 65534 1260 Dec 4 16:47 opennap.ini
drwxr-xr-x 2 root root 1024 Dec 4 16:48 searches
-rw-r--r-- 1 root root 2095 Jan 10 19:21 searches.ini
-rw-r--r-- 1 root root 2095 Jan 10 19:13 searches.ini.old
-rw-r--r-- 1 root root 60438 Jan 10 19:21 servers.ini
-rw-r--r-- 1 root root 60413 Jan 10 19:13 servers.ini.old
-rw-r--r-- 1 root root 61025 Jan 10 19:20 shared_files_new.ini
-rw-r--r-- 1 root root 61025 Jan 10 19:13 shared_files_new.ini.old
drwxr-xr-x 8 mlnet mlnet 1024 Aug 11 01:24 shares
-rw-r--r-- 1 root 65534 1270 Dec 4 16:47 soulseek.ini
-rw-r--r-- 1 root root 1521 Jan 10 19:20 stats.ini
-rw-r--r-- 1 root root 1521 Jan 10 19:13 stats.ini.old
-rw-r--r-- 1 root root 10113 Jan 10 19:20 stats_mod.ini
-rw-r--r-- 1 root root 10113 Jan 10 19:13 stats_mod.ini.old
drwxr-xr-x 5 mlnet mlnet 2048 Jun 21 2004 torrents
|
Fri 07 Jan 2005 08:16:35 AM UTC, original submission:
If you quit mldonkey with "kill" from telnet-Interface it don't saves the *.ini files.
Stopping mldonkey by Ctrl-C or "kill mlnet" has not this bug.
|