finalize-staged: Don't listen to SIGTERM, just let kernel exit us
#2704
Conversation
|
Skipping CI for Draft Pull Request. |
| g_unix_signal_add (SIGTERM, sigterm_cb, &running); | ||
| g_print ("Waiting for SIGTERM\n"); | ||
| while (running) | ||
| while (TRUE) | ||
| g_main_context_iteration (NULL, TRUE); |
There was a problem hiding this comment.
If the loop isn't doing anything, why not just use pause()? That's actually exactly what sleep infinity does. Seems silly to have glib setup an event loop that does nothing.
There was a problem hiding this comment.
Well, we're just optimizing a no-op at this point, but yes, done 😄
I didn't know about the pause system call actually. It looks like it may be x86 only; glibc has a fallback that just uses ppoll() https://github.com/bminor/glibc/blob/02ca25fef2785974011e9c5beecc99b900b69fd7/sysdeps/unix/sysv/linux/pause.c#L31 with NULLs.
There was a problem hiding this comment.
I ran strace on sleep infinity and that's where it blocked. I think it is provided most everywhere on linux since it's in POSIX. pause is defined in most syscall tables in the kernel and there's a generic implementation.
Followup from discussion in ostreedev#2544 (comment) This is more efficient; no need to have the kernel context switch us in at shutdown time just so we can turn around and call `exit()`.
cgwalters
force-pushed
the
finalize-no-sigterm
branch
from
0081bf3 to
683e4ef
Compare
Followup from discussion in
#2544 (comment)
This is more efficient; no need to have the kernel context switch
us in at shutdown time just so we can turn around and call
exit().