Yeah I feel like pidfd_open() is basically just as racey as being careful with waitid(), haha

if we did PORT_SOURCE_PID you could get notification prior to final consumption with a waitid(), which would then free up the pid for reuse

I think if you want that you'd need a variant on FORK_WAITPID or similar.

[illumos-gate] 16785 fcoei: overflow in conversion -- Toomas Soome <tsoome⊙mc>

[illumos-gate] 16786 usba: possible memory leak -- Toomas Soome <tsoome⊙mc>

fenix: 16789

every time

fenix: #16789

BUG 16789: qlc(4D) contains incorrect copies of much of the DDI (New)

↳ illumos.org/issues/16789

tsoome: that's what I was trying to badger you into fixing

but also if someone wanted a really bite-size bug, and also happened to use qlogic storage...

.oO SHELF OVER TEMPERATURE 173 C (343 F) wonder if the sensor was crazy or they had fire on their hands...

rmustacc: Yes, you are correct of course. I was assuming the use of FORK_NOSIGCHLD|FORK_WAITPID but I should have been explicit

Which I think you basically already need to use to get good semantics generally

PORT_SOURCE_PID would mean you could stop either burning threads in waitid() or polling with waitid(WNOHANG) etc

WNOHANG|WNOWAIT that is

well the ombld omnios env fixed most of my build, but I'm seeing errors from it trying to use perl from /usr/perl5/5.28. I see it mentioned in noise-i386.ref, not sure if that's supposed to suppress it getting mentioned in the mail output.

looks like it's stuff in cmd/perl/Sun/Solaris

err cmd/perl/contrib/Sun/Solaris

josephholsten - 5.28 seems rather old.

huh, it's only in the prevous build logs, I wonder if the noise catcher is a bit touchy. I should do a clean build anyway.

Ah right, in the noise differences from the previous build, that makes sense.