On 2008-07-10 at 16:32:02 [+0200], Sergei Dolgov <dolgov@xxxxx> wrote: > http://dev.haiku-os.org/ticket/2049 - closed by bonefish as invalid: > "This was obviously a bug in the Mozilla (BeOS port) code. You can't > expect that object pointers are unique across applications." > > I cannot be so sure as bonefish, as things aren't investigated to > logical end, and all that code works in BeOS and Zeta. > > At least it points to some difference in port or tls methods behaviour > in R5.* and Haiku, and until that difference is totally clear, I > wouldn't close the bug so resolutely. Ticket name is now wrong for sure, > but not the problem. Haiku has a different address space layout and a different userland memory allocator than BeOS/Zeta. That may make it more likely that PRThread objects in question get the same address in different applications. At any rate using an object pointer when you need a globally unique name is a bug. Using the wrong port sounds like a very good explanation for the observed misbehavior and fixing the bug apparently solves it. Not sure what else you want to investigate. CU, Ingo