There is a bug inside GetWorldHeight... in some cirumstances it will return an error instead of expected z-value (no Core crash).

affected releases:
pol-linuxbin-097-2006-10-18-gentoo.tar.gz
pol-linuxbin-097-2006-10-29-RC-Coregina-gentoo.tar.gz

To be correct, _all_ releases are affected but all other releases just hide this problem. These releases were build on a new compiler release (gcc4.1.1).

Shinigami

Can I ask what are the circumstances of that error? or is it random error?

the values was not initialized, but it was never a problem here. the value was taken from stack and it was greater than 1000 (huge value in real but positive) in any case... in Win, Linux and FreeBSD release if they use 32 bit integer, just by luck.

but now, using newer gcc, 64 bit integers are used (just by default). so the value taken from stack is different. not in includes an older return address and the whole value will become negative.

the algorithm inside needs a larger positive value to init... so it will break here...

Shinigami