tst-basic3 failing
Paul Mackerras <[email protected]>
| Newsgroups | gmane.comp.lib.phil |
|---|---|
| Message-ID | <[email protected]> |
I am making some good progress on getting NPTL running on PowerPC. However, I have run into a problem with tst-basic3. The initial thread segfaults after creating the 20 child threads. It turns out that it is doing a longjmp on an uninitialized jmp_buf. It calls pthread_exit, which calls __do_cancel, which calls __libc_longjmp on self->cancelbuf. The cancelbuf field gets initialized in start_thread, but of course the initial thread doesn't call start_thread. Unfortunately most of the thread exit processing logic seems to be in start_thread too. Are other architectures seeing this problem? It looks to me like it will take some restructuring to fix. Paul.