The following reply was made to PR pkg/48555; it has been noted by GNATS.
From: David Holland <dholland-***@netbsd.org>
To: gnats-***@NetBSD.org
Cc: ***@tx.rr.com
Subject: Re: misc/48555: rc.conf handling of program=yes/no
Date: Thu, 16 Oct 2014 04:48:11 +0000
Is this dirmngr from pkgsrc/security/dirmngr? Try changing
line 56 of /etc/rc.d/dirmngr from
eval `${dirmngr_command} ${rc_flags}`
to
eval "${dirmngr_command} ${rc_flags}"
(double quotes instead of backticks). I have not tested this, but
it's possible that it will help.
Backticks are definitely wrong here. Eval is unnecessary. And
required_* stuff is handled by run_rc_command already. check_process
should be used too instead of doing it manually.
So I'd say, recategorize this against pkgsrc to fix the rc.d file.
On digging some, it seems that the backticks are there because
dirmngr, instead of doing something sane like writing out a pid file,
prints "export DIRMNGR_INFO=blahblahblah" when it daemonizes, so
you're supposed to eval this in the parent shell.
So that's not it - or at least not overtly. Also, this isn't run at
shutdown time, although part of the output from this is passed to kill
in dirmngr_stop and who knows what glop might have got into this.
To the original submitter if you're still there: what ends up in
/tmp/dirmngr/dirmngr.info after dirmngr starts (using this rc script)?
On a further matter, it is not clear where the additional wrapper
program in files/runDirmngr.c came from but a quick inspection shows
it has a bug or two...
--
David A. Holland
***@netbsd.org