Reenable RVV code that it compiles
Brad backported an llvm fix for the type mismatch that was plaguing the
riscv vector intrinsics. Also Brad added runtime detection to this port
through elf_aux_info(3) earlier this summer, so we can now enable the RVV code
in highway. But don't drop -DHWY_CMAKE_RVV=OFF, because it would make
unconditional use of RVV through CFLAGS, and OpenBSD/riscv64 doesn't assume
RVV support / RVA23.
rpki-client: rework handling of the expire time
The expire time for certificates was added as a hack for filemode and has
been unused in normal mode. We can use it to track the expiry time along
the validating chain of all objects by setting it when validating CAs.
TAs expire with their not after, intermediate CAs and EE certs expire at
the minimum of their notafter, their CRL's nextupdate and their issuer's
expire time. Signed objects inherit the expire time from their EE cert
(this can be handled more cleanly later on).
This way we do not need to grab a lock to determine the expire time of any
object and we can stop walking up the validation chain and look up the same
CRLs over and over again.
ok job
Randonly, when starting a daemon that sets rc_bg=YES, we may end up with:
/etc/rc.d/foo: kill: 123456: no such process
There's a race in the rc.subr(8) system where we can end up killing a non
existing process (which PID may already be reused by another one).
The reason is that rc_alarm timer sends a SIGALRM to all children of rc_cmd to
detach rc_start. But the timer is also a child... and can then die while
rc_cmd() is about to kill it; that's when you get the error (as it's already
gone).
Re-implement so the timer ignore SIGALRM and let it get killed by rc_cmd.
My tests confirm this fixes the issue, so commit it early in the release
process to make sure this does not introduce any regression.
reported on bugs@ by adfsilva at gmail dot com
ok kn@
