[Plib-users] Bug report

[<dva...@cs...>]

Here comes a bunch of bugs that I've found when using PLib on Win9x
platform. I use a PLib version from the cvs approx. a month ago, so if
some of them already has been fixed I appologize.

* ssgaShape. Crashes after regeneration if kidStates has been set. The
removeAllKids deletes all references to the ssgState but the pointer is
still used after that. Works as long as the state is ref:ed before used in
the Shape but I don't believe that's the way it's supposed to work.

* ssgLOS doesn't work. It uses some code that are specific to the HOT case
(i.e. assume that the ray always points along the Z axis).

* sgQuat is broken. The conversion between other rotational types are NOT
done correctly. Conversions like quat -> matrix -> quat etc results in
axes switching and radians/degrees conversion errors.

* The 3ds-loader (and maybe other loaders as well?) crashes if the model
file isn't found since it tries to do a fseek on an invalid file handle.

* The 3ds-loader also have problems with large models. It can load them
ok, but when ypu try to manipulate them PLib will crash. This is because
the loader creates VtxTables with too many vertices in them. The getVertex
function can only handle shorts but the number of vertices variable is an
int, making it possible to load more vertices than some of PLib's other
functions can handle.

I've also found some mem leaks that shouldn't be a problem on a platform
with good memory handling, but since I'm stuck with Windows... Some of
these leaks are things that could (should?) be handled by a ssgExit
function.

* _ssgModelPath/_ssgTexturePath are never deallocated.

* _ssgCurrentContext is never deallocated.

* In ssgLoad3ds the current_branch is never deleted after the parsing of
the file is finished.

CU.

/Fredde

Magic is Life - Technology is Death