> Roland, I think it better to change tracehook definition more, please > see below.
I don't really object to this in principle. But this touches signal.c a lot more in less obviously-trivial ways than my tracehook patch. That is more of an issue at the outset than some extra fiddling in the utrace code. I think we should consider this for a later clean-up after merging. > BTW. exit_signals() calls tracehook_notify_jctl(why => CLD_STOPPED), > could you confirm this is right? Yes, it's right. I considered passing CLD_EXITED here to distinguish this odd case, but that would make the vanilla tracehook_notify_jctl() definition less simple. Instead, we put the onus on a ->report_jctl hook to check for PF_EXITING to tell if it's really going to stop. Thanks, Roland