BUG: NTPL: waitpid() doesn't return?

[Roland McGrath]

> NB: if you'd like me to upload 4.5.1 to Debian-unstable, and/or the CVS
> head to Debian-experimental, just say so.

Um, sure.  What I was reasonably used to doing was making a source-only
package and uploading it, and then the buildd's seemed to take care of
things.  They changed the system so it refused a source-only upload,
and that made it just that much more inconvenient for me.  

> On the SMP system it passes about 10% of the time, 40% hangs, and the
> rest crashes with "umoven:no such process" _and_ "[tcb table full]".

Hmm.  The [tcb table full] really shouldn't ever happen, I must have a bug
there.  fork_tcb ought to get called and expand the table before alloctcb
can ever get called.

> On my single-CPU system, half of the time it passes, the rest dies with
> "[tcb table full]" (no "umoven" error).

That is more consistent anyway.  Can you debug that one and at least see
why the table isn't getting expanded.

> I'll report what, if anything, changes with CVS head.

Great.


Thanks,
Roland