lib/60925: named and cmake dump core in 10.2_STABLE amd64 after pthread_atfork.c changes

["oster%netbsd.org@localhost via gnats" <gnats-admin%NetBSD.org@localhost>]

>Number:         60925
>Category:       lib
>Synopsis:       named and cmake dump core in 10.2_STABLE amd64 after pthread_atfork.c changes
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    lib-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Oct 25 21:25:00 +0000 2026
>Originator:     Greg Oster
>Release:        NetBSD 10.2_STABLE
>Organization:
	
>Environment:
	
	
System: NetBSD pepe 10.1_STABLE NetBSD 10.1_STABLE (GENERIC) #0: Mon Sep 7 09:22:42 CST 2026 oster@gonzo:/u1/builds/build451/src/obj/amd64/u1/builds/build451/src/sys/arch/amd64/compile/GENERIC amd64
Architecture: x86_64
Machine: amd64
>Description:
   With today's NetBSD 10.2_STABLE checkout, cmake dumps core when running the bootstrap version
of cmake when trying to build cmake in pkgsrc.  The stack trace looks, in part, like:

#0  0x000076728c57f79a in _lwp_kill () from /usr/lib/libc.so.12
#1  0x000076728c584fa0 in abort () from /usr/lib/libc.so.12
#2  0x000076728e2220fa in uv__signal_global_init () at src/unix/signal.c:73
#3  0x000076728bc07cb6 in pthread_once () from /usr/lib/libpthread.so.1
#4  0x000076728e21dc0b in uv_once (guard=guard@entry=0x76728e42cb00 <uv.signal_global_init_guard>, 
    callback=callback@entry=0x76728e21a99d <uv__signal_global_init>) at src/unix/thread.c:454
#5  0x000076728e21aa1c in uv__signal_global_once_init () at src/unix/signal.c:111
#6  0x000076728e216d8a in uv_loop_init (loop=0x76728eca8000) at src/unix/loop.c:83
#7  0x00000000e078dcd1 in cmUVProcessChain::InternalData::Prepare(cmUVProcessChainBuilder const*) ()
 ...

  Within the same environment, named also dumps core, with the following stack trace:
Program terminated with signal SIGABRT, Aborted.
#0  0x000071db8af7f79a in _lwp_kill () from /usr/lib/libc.so.12
(gdb) where
#0  0x000071db8af7f79a in _lwp_kill () from /usr/lib/libc.so.12
#1  0x000071db8af84fa0 in abort () from /usr/lib/libc.so.12
#2  0x000071db9046d115 in ?? () from /usr/lib/libisc.so.21
#3  0x000071db8ec07cb6 in pthread_once () from /usr/lib/libpthread.so.1
#4  0x000071db90464e55 in uv_once () from /usr/lib/libisc.so.21
#5  0x000071db9045c26d in uv_loop_init () from /usr/lib/libisc.so.21
#6  0x000071db9043604d in isc.netmgr_create () from /usr/lib/libisc.so.21
#7  0x000071db904229b8 in isc_managers_create () from /usr/lib/libisc.so.21
#8  0x0000000095a666f2 in main ()
(gdb) 

This is replicated both on a full 10.2_STABLE install in userland (for named) on two
separate hosts, and in a 10.2_STABLE chroot (both with 10.1_STABLE and 10.2_STABLE kernels) for cmake.

>How-To-Repeat:
     Upgrade from 10.1_STABLE to today's 10.2_STABLE.  Run '/etc/rc.d/named onestart'.  Watch named fail to start
and dump core (if enabled).  Attempt to compile /usr/pkgsrc/devel/cmake in a 10.2_STABLE virtual environment.
Watch it also fail and dump core.

Note that the problem does not exist in the official 10.2 release.  It is only in 10.2_STABLE.

>Fix:
     Reverting this commit:

cvs rdiff -u -r1.17 -r1.17.2.1 src/lib/libc/gen/pthread_atfork.c

gets things working for me.  But no idea as to the underlying issue.