Core bug with events. Result = Crash

[RusseL]

We have two scripts with events listeners.
1. 1st script is for example NPC script, but could be anything - http://pastebin.com/C0mw8k61
Create NPC with this script, and start it with "all come" command.
2. 2nd script is another listener, for example characters backpack - http://pastebin.com/yQ0Q7gwa (that is a command)

Start your cmd with .testcycle, be sure that npc script is running (prints "NPC LOOP" in console), and start to flood with message "test" near your npc. In 5-10 seconds it'll crash.

Solution:
add wait_for_event(0); after while( me ) in NPC script

But it must be fixed in core, because this bug prints really NOTHING in pol console, nothing in dump, no backtrace, that is terrible.
And it could appear everywhere - boats, npc and another listeners.

For example today i found it in my boat script. And that was a great piece of luck, that i've found it, otherwise it's impossible to catch it...

ps: as i understand only 2 listeners and 1 cycle are necessary

[Yukiko]

Sounds like it is a buffer overflow possibly due to a full event queue maybe?

[RusseL]

DiscardOldEvents=1 doesn't help

[Turley]

Sounds like fun, cannot promise but I try to look into it this weekend.

[Turley]

Should be fixed in Rev 573.
Was a problem with speechtokens member in the event struct.

[RusseL]

Should be fixed in Rev 573.
Was a problem with speechtokens member in the event struct.

thanks:)
i already thought that problem was with tokens, because that was only one thing that i've found in changelog:)
thanks, i'll check it soon

[RusseL]

ooops Turley, u've broken something. Segfault when i logIn..

Code: Select all

[09/29 00:52:54] Account admin selecting character Coffee Effect
Caught SIGSEGV (Segfault).  Please post the following on http://forums.polserver.com/tracker.php :
=== CUT ===
Build: POL099-2011-05-02 Break Everything Even Rudder (ubuntu)
Last Script: scripts/ai/spellkillpcs.ecl PC: 965
Stack Backtrace:
[0x82b099c]
[0x110420]
[0x80dd3ac]
[0x809262e]
[0x809290c]
[0x814720a]
[0x81475aa]
[0x81484c0]
[0x814b15e]
[0x8216e5d]
[0x819ae62]
[0x82b7b32]
[0x82b845b]
[0x831b418]
[0x8364a9e]
=== CUT ===

[Turley]

mmmh damn...
Can you give me the source lines from the backtrace?
gdb pol and the l *0xaddresses

[RusseL]

my live server starts, but i can not login.
my test server doesn't start:)

here, Turley http://pastebin.com/cwQC31LP

[RusseL]

Update:
no it doesn't work
crash on startup

info here:
http://pastebin.com/qLi64jAm

[Turley]

Weird I will try to analyze it tomorrow. Based on my latest changes. Maybe I have to send you a test build.

[RusseL]

solved.
you didn's said that i must recompile all
Thank you, turley

[AsYlum]

solved.
you didn's said that i must recompile all
Thank you, turley

RusseL ... http://www.youtube.com/watch?v=BNsrK6P9QvI

[RusseL]

nope. crash again
i must update fedora on my server, because i can't start dynamic bin there to get a gdb output

[RusseL]

i don't know, but my server is completly unstable since last updates
http://pastebin.com/RpuyT1Wy

[Turley]

Do you use the new item methods splitstackat or splitstackinto? They have bugs, tomi is working on a sollution.

[Terciob]

[10/06 14:22:02] [pkg/systems/worldSaver/worldSaver.ecl]: 06.10/14:22 Saving World State
[10/06 14:22:08] Exception during world save! (basic_ios::clear)

Crash while saving can be shadow realm issue http://forums.polserver.com/viewtopic.php?f=20&t=3483. I had this problem months ago.

[RusseL]

last rev. repsys is completely broken.
i see the innocent character, but he has a green/alliance name, and then when i try to attack him - crash
http://pastebin.com/Fk0H5dNF

as i understood something is wrong when repsys:OnAttack is called

[Turley]

The problem with shadow realm deleting and leftover storage items should be fixed in the latest revision.
As always I cannot reproduze your problem with repsys ...

[RusseL]

As always I cannot reproduze your problem with repsys ...

Try to use distro repsys hook

[Turley]

again no problem for me