PATCH? tracehook_notify_jctl && SIGCONT

[Roland McGrath]

> Now we send SIGCHLD with si_code = CLD_STOPPED twice. By T1 from
> do_signal_stop(), and by T1 or T2 from get_signal_to_deliver() which
> checks SIGNAL_CLD_MASK.

Yes, I considered this problem.  It's just not so big a deal as to worry
overmuch about this corner case in the first version.  What seems to me
will be the obvious and straightforward way to address it is to give
utrace_report_jctl() a return value that tracehook_notify_jctl() uses.
Then we can omit a notification that has been superceded.

Your patch does not seem very straightforward to me.  Moreover, you moved
some ptrace magic out of the tracehook function back into core signals code.
That is going in the wrong direction and we won't have any of that.


Thanks,
Roland