POL097-2006-08-30

[Xandros]

I took a look at the new core and I'm really getting desperate....
The stat window still doesn't work for us, only that now it doesn't just
hang the client, it crashes the whole server.

I haven't been able to make any serious testing of POL since the
first POL097 version because of these problems. I'd really
like to give feedback about all those other changes, but I can't let
our players test it with those issues right now.

I'll include some dump files, there is no helpful output in pol.log.
Please tell me if you need anything else. And please tell me if you
were able to reproduce our client hanging problems in the previous
POL097 versions, or if you need help to reproduce them, to that you
have exactly the same situation that we have and can verify that
a future release really fixes this issue.

Thanks,
Xandros

[MuadDib]

I'll get with shini or austin to check the dumps for me (slow ass internet delays me downloading what i need to check them).

I couldn't reproduce the hanging at all myself. i did notice the packets was incorrectly padded (the length was right, but amount of data sent was wrong, etc). So I fixed that problem. I set my account to T2A, fine, AOS, fine, SE, fine, etc. And the packets was the correct size per expansion base.

Can you email me your scriptbase so that I can test live with it myself? You can email the inc/src files to sroyalty@gmail.com so I can work with them.

[MuadDib]

Did some research and it's still all working fine for me. Strange. It would really help if you sent me the scriptbase you use to try to work this out. If i mimick what you are doing, there should not be any problem. But, for some reason, there is.

So, soon as you can get that to me, it would help expedite this and get it resolved it is on our end. If not, then it helps me get you straight either way

[Xandros]

The zip file has about 23 Megs, with compiled scripts (I'd prefer not to
send sources unless you _really_ can't find the problem this way), without
realms and with an empty worldfile. Are you prepared for a mail of that
size?

By the way, while preparing the empty worldfile I noticed that even
when I try to create a character, the shard crashes now.

Xandros

[MuadDib]

Only thing i need is cfg files, inc files, and src files. No data directory, no realm directory, no log directory, no executables.

And yes, I'd prefer source for debugging myself. It's safe to say no one else will get the source files from me. I've done this with several shards to resolve issues and deleted their scripts soon as we got it straight.

[Austin]

Muad and I are discussing it right now. Its a crash we see a lot that occurs in STL-Port's library in the threadex.c file

Can you give us information on your system specifications?
processor type, memory amount, and operting system at the least.

[Austin]

Also add to your pol.cfg

LogLevel=20
DebugLevel=20

We've tracked down what happened, it started with a client status thread.
What we cant figure out is why you get the crash and we dont. Do you have any sort of packet hooks?

Nonetheless, the console will atleast give you some thread checkpoint reports to help us.

[Xandros]

[Xandros]

Oh yes and we have quite a number of packet hooks. They worked fine
with POL96. I can deactivate them for a test to see if it changes
anything.

Xandros

[Austin]

Okay next crash, let us know what you get in the logs and console around the report of it happening.

Also if you can, provide a list of packets youre hooking? MuadDib may get some extra insight from that.

[Xandros]

[Xandros]

I did some tests, the results are interesting...

Stat Window Crash:
-----------------------
If I set LogLevel and DebugLevel to 20, POL doesn't crash when
I open my stat window. The client still hangs an doesn't show any
data in the stat window though, as it did with the previous POL097
versions. Deactivating all the packet hooks doesn't change this
behaviour.
Unfortunately that also means I can't give you any more log output
about this crash. But it's interesting that these settings seem to
influence the server stability.

Char Creation Crash:
------------------------
This crash happens no matter if LogLevel and DebugLevel are set
to 20 or not. Deactivating all the packet hooks also doesn't change
anything.
The log output for this crash:

Account Xandros logged in from 127.0.0.1
[08/31 20:28:04] Account Xandros logged in from 127.0.0.1
[08/31 20:28:08] Unhandled Exception! Writing Minidump file.
This file with explanation and last lines from log files should be given to the development team.
Saved dump file to 'POL097-2006-08-30 Coregina-20060831202113-0.dmp'
Last Script: pkg/misc/syshooks/regen.ecl PC: 26

I'll include some dump files of the char creation crash below.
Our regen.src file contains the RegenRate functions and
maximum functions for all the vitals. Please tell me if you need
this script.

Thanks,
Xandros

[MuadDib]

Do me a favor. Try it with loglevel changes anyway. That creates "checkpoints" that are logged to file. That will help us.

Second, try setting multithread support in pol.cfg to 0. I keep seeing threading issues in the crash dumps and wonder if that might affect this.

Shini is also going to upgrade his shard to use the 097 core this weekend if he gets time. He uses a similar setup as you. T2A expansion with ML support in logon for elves and statbar without tooltips.

[Xandros]

I tried Multithread=0, it seems there is no stat window crash with this
setting (even with LogLevel and DebugLevel=0). The client still doesn't
get any data in the stat window and hangs though.
The char creation crash still happens with Multithread=0.
By the way, this setting wreaks havok with CPU time consumption on
my machine. Sysload permanently stays at 100% as long as pol is
running, when normally it goes back to a few percent a minute
after starting POL with Multithread=1.

I found out that with Multithread=1, I can go up to LogLevel and
DebugLevel=10, and the stat window crash still happens, if I go further
up to 11, it doesn't happen anymore.

Dump files with maximum Log and Debug settings below...

Xandros

[MuadDib]

What client version are you using?

What is the state of XP Service packs?

Anything special for the 64 cpu you are using? Any special settings?

What checkpoints are being written to you pol.log with the loglevels? try setting them higher, like 20 or 30, see if you get checkpoints written to the pol.log file.

And last, still haven't gotten an email with the cfg/src/inc files yet to try to reproduce and test this

Client 5.0.1h is not giving me any crashes, or hangs, with any of the 097 cores (but it did help me find the padding issue that was fixed with last beta).