Most of the patch I don't understand, but this seems like this might
be causing some leaks:
@@ -457,11 +455,9 @@ pthread_create(pthread_t *thread, const
if (!PTQ_EMPTY(&pthread__deadqueue)) {
pthread_mutex_lock(&pthread__deadqueue_lock);
PTQ_FOREACH(newthread, &pthread__deadqueue, pt_deadq) {
- /* Still running? */
+ /* Still busily exiting, or finished? */
if (newthread->pt_lwpctl->lc_curcpu ==
- LWPCTL_CPU_EXITED ||
- (_lwp_kill(newthread->pt_lid, 0) == -1 &&
- errno == ESRCH))
+ LWPCTL_CPU_EXITED)
break;
}
if (newthread)
Can you try if the leak is gone if you only revert this part of the patch?
Jaromir
Le jeu. 6 juin 2024 à 20:13, Brian Marcotte <marco...@panix.com> a écrit :
>
> > All started with this commit:
> >
> > https://mail-index.netbsd.org/source-changes/2020/01/27/msg113341.html
> >
> > Undoing part of this commit (diff attached) makes the leak disappear
>
> I'm trying this now. I can already see it's helping these:
>
> gpg-agent
> opendmarc
> dkimpy_milter (python3)
>
> I'm also trying it with mariadb, but it may take a few days to see if
> the leak is gone.
>
> Thanks for looking into this.
>
> --
> - Brian
>
