At Wed, 4 May 2005 15:12:34 -0700 (PDT), Roland McGrath wrote: > > > At Wed, 4 May 2005 14:07:41 -0700 (PDT), > > Roland McGrath wrote: > > > > > > There isn't any big issue with _cthread_exit_routine, because it doesn't > > > need to be set so early. Your initializer code (DSO initializer, or your > > > _cthread_init_routine function) can just assign _cthread_exit_routine. > > > > > > > Good point. Does that mean glibc should be to actually call > > _cthread_exit_initializer? Should I file a bug report? > > Huh? I just described why no further libc change is required.
In my second email, I asked two questions: 1) does the same treatment that you have suggested need to be applied to _cthread_exit_routine and 2) _cthread_exit_routine is never called should we fix it. As I understood your reply, you responded to the first but said nothing with respect to the second. Basically, [a]s far as I can tell, _cthread_exit_routine is not currently called when the main thread exits as libc/sysdeps/mach/start.c would seem to suggest: we use libc/sysdeps/generic/libc-start.c (which is called by libc/sysdeps/i386/elf/start.S) to call main. This isn't a problem for libpthread as libpthread doesn't use this call back. libcthreads, on the other hand, does to wait until all other threads have exited. Thanks, Neal _______________________________________________ Bug-hurd mailing list Bug-hurd@gnu.org http://lists.gnu.org/mailman/listinfo/bug-hurd
