With cordite in the air, splintered steel, shell casings and powder burns, there’s only one explanation...
WQ3 pre B2.2 archive. Links may not work.

Postby hika » Thu Jun 08, 2006 10:01 pm

sig11 wrote:The log files so far are not very helpful ... I hope I can fix this :|

Perhaps, a segfault is causing this crash. I 'm sure you know what I'm talking about ;-)
SG name: Manchot
SG fansite: http://western.bsdmon.com
Image
userbar originally created by Caffeine
User avatar
hika
SG Team
 
Posts: 703
Joined: Thu May 25, 2006 2:32 pm
Location: Trévoux, France



Postby hika » Tue Jun 20, 2006 12:01 am

Hmm, after running a Western Quake 3 server with the b2.10 testpatch01 for a while, I have noticed several "crash", I mean a stopped q3ded, after few days.

In the logfile, I only see that it was just after a "ShutdownGame" by a timelimit hit.
It seems that it did not launch the next map, and stopped.

Its map rotation is the same as the Wasteland - Deathmatch.
The difference is the bot_minplayers 2. So more chances to reach a timelimit.

I don't know if it is just a coincidence, but my others two q3ded (Wasteland - Deathmatch and Wasteland - Bank robbery) in b2.10, do not crash for a quite while.

I don't notice any segfault, in my system log file.
Quite hard to do a bug report (if bug there is).
SG name: Manchot
SG fansite: http://western.bsdmon.com
Image
userbar originally created by Caffeine
User avatar
hika
SG Team
 
Posts: 703
Joined: Thu May 25, 2006 2:32 pm
Location: Trévoux, France



Postby Pardner » Tue Jun 20, 2006 2:46 am

mmmm.... i have not had this problem. i ran a "dedicated 1" (LAN) server for 3 or 4 hours. i cant remember what or how many maps we played, but we had 12 people playing br. (fun times 8) )

dumb question, but did you spell the map names correctly? thats the only time ive had a server die on me; granted i dont run them too often.

Image
Image
User avatar
Pardner
SG Team
 
Posts: 1786
Joined: Fri Nov 18, 2005 5:48 am
Location: MD, USA



Postby hika » Tue Jun 20, 2006 9:06 am

I have rechecked the logfile, just to see any map name mispells, but they are corrects.

However, I have noticed something else.
Currently, the map rotation is "normal", but before the "crash", the server kept loading the wq_coyotebluff map.
The only difference between the time before the "crash" and now is that I have done a rcon command, before, to change map to wq_coyotebluff.

I have played one round, entirely, hit the fraglimit, and disconnect.
Then, the server has reloaded wq_coyotebluff and the timelimit has been hit 4 times, until the "crash".
The map was loaded 5 times, successively.

That lets me think that there is some code, doing the server shutdown if a same map is loaded many times.

Hope someone could correct me (if i'm wrong) or confirm.
I can provide the logfile, if needed.
SG name: Manchot
SG fansite: http://western.bsdmon.com
Image
userbar originally created by Caffeine
User avatar
hika
SG Team
 
Posts: 703
Joined: Thu May 25, 2006 2:32 pm
Location: Trévoux, France



Postby sig11 » Tue Jun 20, 2006 9:43 pm

Actually, I've seen different crashes and simple shutdowns for no reason, too. I could not make up any systematics so far :-( ... and it also happened on my "normal" servers which don't have coyote_bluff in their rotation.

I'm losely following the changes in the icculus repository. They fix stuff here and there but I didn't notice a red flag saying "this bug is causing these shutdowns" ... OTOH I may not recognize such things ;-)

@hika: if you can reproduce crashes, try observing the cvars "nextmap" and "mapname". There was a bug concerning these cvars and restarting maps. Unfortunately, I can't find the diff in icculus right now.
In-game name: =SG= Django (yes, it's cliché :-) )
User avatar
sig11
SG Team
 
Posts: 616
Joined: Sun Mar 30, 2003 1:00 pm
Location: Germany



Another crash

Postby hika » Sat Jul 15, 2006 9:43 am

Hmm, my "testing" server crashed last night (GMT+2), one more time.
According to the log file, it crashed at the middle of a DM round, this time.
I'm quite sure it only occurred after a change map (by vote or rcon), but not immediately after ....

sig11 wrote:@hika: if you can reproduce crashes, try observing the cvars "nextmap" and "mapname". There was a bug concerning these cvars and restarting maps

Quite hard to reproduce crashes immediately.
I have looked into these cvars :
"mapname" is correct and reflects the correct map loaded.
However, I cannot see "nextmap" var in the "InitGame" line, in the logfile.
(Don't know If there is another way to dump all cvars)
I only guess it is not set, because the last map kept restarting after each round end.

Any clues ?
SG name: Manchot
SG fansite: http://western.bsdmon.com
Image
userbar originally created by Caffeine
User avatar
hika
SG Team
 
Posts: 703
Joined: Thu May 25, 2006 2:32 pm
Location: Trévoux, France



Postby sig11 » Sat Jul 15, 2006 2:31 pm

I don't see "nextmap" in the log file either. OTOH, I use screen (ports/sysutils/screen) to run my servers and there I can recall all cvars (you may already very well know all that).
In case you want to test if that bug hit you : this is what I meant (a really small patch), but the symptons sound differently anyways :? ...

In the long run, I can also produce a standalone version for FreeBSD. Lately, my servers did not crash, so maybe one of the fixed bugs is responsible for that, and a standalone version would help (probably, now that I wrote that, one of my servers will crash ;-) ...)
In-game name: =SG= Django (yes, it's cliché :-) )
User avatar
sig11
SG Team
 
Posts: 616
Joined: Sun Mar 30, 2003 1:00 pm
Location: Germany



Postby hika » Sat Jul 15, 2006 8:39 pm

sig11 wrote:OTOH, I use screen (ports/sysutils/screen) to run my servers and there I can recall all cvars (you may already very well know all that).

Actually, I'm running the servers from a shell command that put q3ded in background
Code: Select all
q3ded +set fs_game westernq3 +set dedicated 2 +set net_port 27962 +set sv_strictAuth 0 +set com_hunkmeg 25 +sets gamestartup "`date +"%D %T"`" +exec dm_server.cfg > /dev/null 2>&1 &

By the meantime, I will just do 2 things :
1) I will try the ports/game/cleanq3, for the test server, which provide a cleanq3ded binary.
(I'm using this port at the client side, and it runs well)
It will probably not correct the shutdown problem, but at least, I will preserve the current ports/game/quake3 for the others (Wasteland Deathmatch and Bank robbery), and can do patch stuff on the cleanq3 one.
2) I will replace /dev/null by a regular file, in order to log all the Com_Printf() message (same result than screen ;))

sig11 wrote:In case you want to test if that bug hit you : this is what I meant (a really small patch), but the symptons sound differently anyways :? ...

Thanks for the link :)
If the shutdown still occurs, perhaps messages in my new log file will bring me some new clues ?
It not, I will consider using the small patch
SG name: Manchot
SG fansite: http://western.bsdmon.com
Image
userbar originally created by Caffeine
User avatar
hika
SG Team
 
Posts: 703
Joined: Thu May 25, 2006 2:32 pm
Location: Trévoux, France



Previous

Return to WQ3 Beta 2.2 Testing

Show Sidebar
Show Sidebar

User Control Panel

cron