https://bugs.kde.org/show_bug.cgi?id=372504
--- Comment #16 from Philippe Waroquiers <philippe.waroqui...@skynet.be> --- (In reply to David Hagood from comment #15) > #1 0x00000000380936a9 in do_syscall_for_client (syscall_mask=0x80ddebdb0, > tst=0x80200c130, syscallno=130) at m_syswrap/syswrap-main.c:339 According to the above, the 2 blocked threads are doing a sigsuspend syscall. Now, to get a thread out of a syscall, valgrind sends a special signal VG_SIGVGKILL (which is signal 64 on my setup. I think it changes depending on the platform, but I guess it will be 64 on Ubuntu also). So, maybe what happens is that the signal VG_SIGVGKILL is blocked ? > > But, funny thing: after attaching to each of the three pending threads, then > detaching, Valgrind exited. So it's like the second 2 threads in that group > needed a kick in the head from the debugger to exit. I guess that gdb is doing a ptrace syscall or something like that that finally gets the thread 2 and 3 out of the syscall, allowing the threads to terminate, and so the main thread detects they are dead. I guess the question now is: why is the signal sent by valgrind to these 2 threads not getting these out of the syscall ? Can you check the value of the mask ? -- You are receiving this mail because: You are watching all bug changes.
