On Thursday 02 June 2011 18:39, Oleg Nesterov wrote:
> On 06/02, Oleg Nesterov wrote:
> >
> > On 06/02, Denys Vlasenko wrote:
> > >
> > > The problem is that right now it seems that if tracer doesn't catch
> > > EVENT_EXIT and detach tracee when it sees it, really weird things
> > > happen.
> >
> > The test-case is wrong afaics...
> >
> > Perhaps this should be considered as a bug in glibc, I dunno.
> >
> > > thread1(void *unused)
> > > {
> > > //	usleep(100*1000);
> > > //	VERBOSE("WINCH\n");
> > > //	raise(SIGWINCH);
> > >
> > > 	usleep(100*1000);
> > > 	VERBOSE("DYING\n");
> > > 	raise(SIGUSR1);
> >
> > This doesn't send a signal. This does tgkill(tgid, 0, SIGUSR1) which
> > fails correctly with -EINVAL.

Yes. After I fixed this, it works as expected. See attached.
The output is:

# ./thread_leader_exit
2876: thread leader
2876: status:0003057f WIFSTOPPED sig:5 (TRAP) event:CLONE eventdata:0xb3d
2877: status:0000137f WIFSTOPPED sig:19 (STOP) event:none eventdata:0x0
Leader exits
2876: status:0006057f WIFSTOPPED sig:5 (TRAP) event:EXIT eventdata:0x7700
Sending WINCH to self (pid,tid:2876,2877)
2877: status:00001c7f WIFSTOPPED sig:28 (WINCH) event:none eventdata:0x0
res:0
Sending USR1 to self (pid,tid:2876,2877)
2877: status:00000a7f WIFSTOPPED sig:10 (USR1) event:none eventdata:0x0
2877: status:0006057f WIFSTOPPED sig:5 (TRAP) event:EXIT eventdata:0xa
2877: status:0000000a WIFSIGNALED sig:10 (USR1)
2876: status:0000000a WIFSIGNALED sig:10 (USR1)



> > > static int
> > > thread_leader(void *unused)
> > > {
> > > 	/* malloc gives sufficiently aligned buffer.
> > > 	 * long buf[] does not! (on ia64).
> > > 	 */
> > > 	clone2(thread1, malloc(16 * 1024), 16 * 1024, 0
> >
> > Probably because of this clone2.

This seems to be not a problem (it is defined to clone()).

-- 
vda