[Fixed] MC-Server crash R530

[NiLSPACE]

MC-Server sometimes just crashes but i can't upload the dmp file

---

i changed the dmp into a txt file so i can upload it

[xoft]

Not enough information. Please follow this thread to provide as much as you can:
https://forum.cuberite.org/showthread.php?tid=400

If unable to upload files to the forum, feel free to use a filesharing web service, such as www.ulozto.net

[NiLSPACE]

log:
  LOG_1338483192.txt (Size: 26.09 KB / Downloads: 219)
crashdump: you have it already
detailed crashdump: don't have one
Reproduce: it is hard to reproduce becouse it just happens sometimes
Plugins/Settings:

[Server]
Port=25565
MaxPlayers=42
Description=MCServer - Slightly more custom!

[Worlds]
DefaultWorld=world

[Plugins]
NewPlugin=Core
NewPlugin=ChatLog
NewPlugin=sTick
NewPlugin=ChunkWorx

[HelpPlugin]
ShowPluginNames=1

[Physics]
Water=0

[Redstone]
SimulateRedstone=0

[Monsters]
AnimalsOn=0
AnimalSpawnInterval=10
Types=Spider,Chicken,Cow,Pig,Sheep,Squid,Enderman,Zombiepigman,Cavespider,Creeper,Ghast,Silverfish,Skeleton,Slime,Spider,Zombie

[Authentication]
Server=session.minecraft.net
Address=/game/checkserver.jsp?user=%USERNAME%&serverId=%SERVERID%
Authenticate=0

OS: Windows 8 Consumer Preview
Location MC-Server: G:\MCServer\MCServer_Win_530
number of players: 1 (myself)

[xoft]

well, the detailed log would come really handy in this case. I can see the line of failure, but I'd really need to know the variables' values, which only the detailed crashdump provides.

[NiLSPACE]

im working on it

---

ok i got the dmp file it is 124 mb and it took verry long to let MC-Server crash :S
http://www.mediafire.com/?17w1dsutncw4890

[xoft]

The crashdump file compresses pretty nicely, actually, using 7z or rar on it could have reduced it easily to a third of that size. But whatever

I am officially crazy. I solved this bug yesterday at night. While sleeping. In a dream. Totally weird. But when I woke up and went through the sources, the solution that I came up with in the dream seems right. The detailed crashdump confirms it. Unbelievable.
Your hint that the crash doesn't happen too often was the first lead. The crashing line was the second. It's rather simple - the random generator generated too large a number, and when we added an existing coord to it, it overflowed, causing a (very) negative coord to be used, thus accessing memory outside of the chunk data.

Fixed in Rev 533.

[tigerw]

That's two things that you've solved whilst sleeping

(One was the client crash on kick, and the other this)

Awesome