PDA

View Full Version : Seg Fault since latest VI Patch



Spaz
01-23-2002, 09:14 PM
FYI, on test:

"(MOB)(MOB#)Has already been removed from zone before we processed it.
Segmentation Fault"

Could just be this zone, but I've been here for a while and it just started today.

Not changed any configs and this is the first seg fault I've seen since going to SEQ 4

sequser
01-29-2002, 07:54 AM
Seems to happen much more often in crowded zones with a lot of mobs (like Sebilis). I'm guessing this is just what it seems? ie, someone is killing a mob right when it spawns, before we have got around to processing that it exists yet.

fluke
01-29-2002, 11:29 PM
was in and out of the The Grey alot last night and couldn't get SEQ to stay up for more than a few seconds after zoning into The Grey. Most of the time I got the

"(MOB)(MOB#)Has already been removed from zone before we processed it.
Segmentation Fault"

but at least twice there was simply the Seg Fault nothing else.

Last update I did was this afternoon (about 3:30 PM CST on the 29th of Jan).

will see if I can get any better data than this, will try a few more zones and see if there is any pattern.

fee
01-30-2002, 11:45 AM
I have experienced this once so far. Kinda smells like some array is getting blown to hell somewhere in the packet processing code... Would be nice to catch this in gdb or generate a core. I promise as soon as this decides to occur for me again, it will get taken care of.

fee

Azerael
01-30-2002, 12:38 PM
I'd reccomend goiong to a mob heavy zone like Chardok when there is a crew doing AOE slaughter/XP grinds.... Last night I was seeing 10 to 15 at a time there, and it coincided exactly with the group of casters that was AOE blasting up near the fort.

Come to think of it, another place I've seen it occur regularly, and in the same scenario, is Dragon Necropolis when we had 3 groups down by the chetari city practicing AOE Tactics.

Rarely do I find it happening on individual mobs being slain. In my experience, it seems to be occuring when a whole bunch die at once and are looted quickly...

I will say that for me, only once has it actually taken ShowEQ down, I just see the messages while i'm monitoring STDOUT as they fly by.

Anyway, take it FWIW, YMMV.

fluke
02-01-2002, 02:11 PM
Ok,

Still getting the seg faults pretty regularly in some zones (The Grey, Maidens Eye, etc...).

To provide more useful info I wanted to post the results of core files

I set the ulimit to get a core file

$ ulimit -c unlimited

now getting the

Seg fault error (core dumped)

message. Before ulimit no (core dumped) message was seen

Problem is I can't find the core file. I have looked everywhere I can think and done

$ find / -name core -print

and nothing.

I wanted to post the results of a few of these to see if anyone has any ideas on fixing them but I can't seem to generate one.


PS: I'm a Solaris sysadmin so if my syntax is not the quickest or shortest versions Linux supports please let me know.

Cryonic
02-01-2002, 03:07 PM
try:

find /* -name core -print

don't recall if core is all lower case.

sequser
02-13-2002, 01:06 PM
If you loot a corpse (usually one that has nothing on it), and then stand up (making corpse poof), before seq has got around to processing the corpse packet, it seems to make this happen pretty reliably (assuming a crowded zone). Just fight a mob that likely will have no loot (froglok ghoul knight in seb crypt a good choice), and be spam right clicking the mob as it is about to die. As soon as you are on the corpse hit ESC real fast.

fee
02-15-2002, 03:36 AM
To folks experiencing this segfault, please try the following.

In your showeq.conf file, under section [Network] look for the "ArqSeqGiveUp" setting. I think by default it is commented out. Set it to 128 and see if this lessens the amount of segfaults you recieve.

The quick and dirty: ArqSeqGiveUp sets the internal packet cache size of showeq. We all have seen our Lag-O-Meters show packet loss in-game, well ShowEQ's packet cache is designed to cope with this loss. Setting the size larger allows showeq to store more out-of-sequence packets in hopes of eventually recieving the one it needs. By default the setting is 32, this is also the hardcoded min.

Dialup users may find setting between 128 and 256 optimal while high speed connections with little to no loss will be fine with 128 or less.

If you want to get an idea of how showeq is doing, define PACKET_CACHE_DIAG and PACKET_PROCESS_DIAG in packet.cpp
have fun!

fee

fluke
02-21-2002, 11:55 PM
Fee,

tried this,

I have a cable modem and don't normally experience much lag, but add that line to the network section of my showeq.conf file (wasn't there at all to start with), and tried both 128 and 256 and still get seg fault with the

"Has already been removed from zone before we processed it."

error

fee
02-22-2002, 01:47 AM
Yeah I finally captured one of the segfaults in an debugger. It looks like the stl map that is going bOOm is in the spawnlist code. I am not very familiar with this section but will start looking at it over the next few weeks. Maybe someone will pick up on it and submit a patch.

The other segfault being seen where the map code blows up looks to be caused by having 2 pets... I will also try to look into it.


fee

fluke
02-22-2002, 04:46 PM
agreed, a problem exists with two pets. You probably saw my post about the seg fault with stalking probe eye up and pet up at same time. Think I left a BT on that one, if not will try and get one to you all.

Studace
03-27-2002, 05:00 PM
I recently installed SEQ4 and seemed to be running fine. I did the update through instructions I found in posts. I now get segmentation faults when I zone, or bring the character in. It may even do a fault after running for some time. I tried increasing the ArqSeqGiveUp to 256 (it was set for 96), still the same thing. The only messages I get are something like
Clearing Cache: Count =0
Resetting sequence cache
then segmentation fault

One time I got Clearing Cache: Count = 3

Would appreciate any help.