> > I played a bit with the test program, perhaps these findings help > > you: > > - it doesn't matter how many threads you are starting (it also > > crashes > > with only one thread) > > - the threads crash when they return > > - they don't crash anymore, when you have more than 224 loops: with > > up > > to 223 loops, they will always crash, with 224 and more loops, they > > never crash - 100% reproducible. > That's no real help is it=3F thanks for your efforts, but so far I > can't see > anything that's wrong. It looks like the stack gets corrupted eaarly > on You should know :-) At least it's clear that it doesn't have to do with the thread handling. BTW I tried the (fixed and the bad one) mbuf test with no extra threads up to 100 threads. Works indeed fine now. Do anyone on this list has a dual-CPU machine to test the code=3F Adios... Axel.