On Mon, 31 Jan 2005, Mark Taylor wrote:
> If you can narrow it down to one JNI call then I (or you) could take a
> Very Hard Look at said bit of JNIAST/AST.
Mark & DavidB,
OK, I've been stepping around a disassembled JNIAST DLL using the various
memory addresses and offsets that Java spits out when it crashes and it
may be looking more stable (in terms of where an error might be) running
under Java 1.5.
Case 1, starts in Frame_pickAxes
calls jniastMakeObject
calls astGetC
(must be call that gets the class name)
calls astGetAttrib
calls astStore
(could be line: axis_attrib=astStore(NULL,attrib,len+4))
call memcpy
and down it goes with an EXCEPTION_ACCESS_VIOLATION,
which probably means the memory is unwritable.
Case 2, starts in FrameSet_getFrame
calls jniastMakeObject
then the same as above!
Case 3, occasionally no problems at all.
Of course these same calls have succeeded many times before we get to this
stage. I'll have a closer look at the code of this trace tomorrow (still
no debugger to actually do this when live), but feel free to have a lock
yourselves.
Cheers,
Peter.
|