Announcements > Development
2015/11/03 - Blockland r1972
Kyuande:
Yeah, seems to have a higher risk when player count is higher than 18
Heedy:
--- Quote from: TheBlackParrot on December 05, 2015, 01:54:02 AM ---also now dealing with R6025
had ~21 players on at the time, download freezing was also rampant
--- End quote ---
This is similar to the console logs I get. It just happened again with only one player in my server. He was loading, he disconnected, and it crashed (without displaying the disconnect message). Last lines of the console are identical in format.
--- Code: ---Client 27277 disconnected.
Issuing Disconnect packet.
CDROP: 27277 - Xeidious - [IP]
Posting to master server
--- End code ---
--- Quote from: Badspot on November 26, 2015, 06:43:18 AM ---I'll get right on that as soon as anyone gives me any kind of information to work with whatsoever.
--- End quote ---
Can you tell us if any of this information is helpful at all? Multiple people have reported the same circumstances for which the error occurs (player loading), and someone even provided a potential lead for the function itself.
TheBlackParrot:
I managed to find a server folder I haven't updated yet, it's still on r1967 if anyone wants it to host with for the time being.
Kyuande:
--- Quote from: Heedy on December 05, 2015, 02:27:41 AM ---Can you tell us if any of this information is helpful at all? Multiple people have reported the same circumstances for which the error occurs (player loading), and someone even provided a potential lead for the function itself.
--- End quote ---
He can't find the error just by looking at the code, but, someone did find the cause of it. I'll bring it up again.
--- Quote from: Val on December 03, 2015, 07:13:51 PM ---Zeblote sent me a picture of what the stack looks like when the game crashes with the pure virtual call error. It seems to occur in octTree::castRay_r, and the call to the only virtual function referenced calls on an item from a big container (seems to be octTree::smServerTree)
Maybe it calls on an object in the tree (with the type being an abstract base class) that hasn't quite initialized yet?
--- End quote ---
I hope this helps Badspot.
Val:
It's probably not exclusive to just that spot either. A big chunk of objects with a large amount of pure virtual functions that can be called from anywhere makes a lot of places potentially responsible, but just one location should be enough.
Regardless, it's helpful to report what the stack looks like when you get the error. Open blockland, attach ollydbg, click debug->run to resume execution, and wait for it to happen. Once the error occurs you'll want to navigate to the bottom right window (which is the current state of the stack):
Then right click edit->select all, edit->copy as table, paste into a text file/pastebin and report back here.
The stack is important because it holds a chain of return addresses eventually leading to the culprit function. It's very useful when tracking down exactly what situation caused the crash.