Re: 8.0 rc.d scripts less verbose?

Please follow up to this post on the freebsd-rc@xxxxxxxxxxx list, thanks.

See below for my comments.

Ian Smith wrote:
On Wed, 16 Sep 2009, Doug Barton wrote:
Cezary Morga wrote:
Dimitry Andric pisze:
Index: etc/rc.subr
===================================================================
--- etc/rc.subr (revision 196888) +++ etc/rc.subr
(working copy) @@ -680,7 +680,7 @@ run_rc_command()

# setup the full command to run # - [ -z
"${rc_quiet}" ] && echo "Starting ${name}." + echo
"Starting ${name}." if [ -n "$_chroot" ]; then _doit="\
${_nice:+nice -n $_nice }\

Wouldn't it be easier to set rc_quiet to zero-length value,
like rc_quiet=""?

That actually would give off too much information because
rc_quiet suppresses error messages about scripts that don't
currently have _enable set to anything useful. That's arguably a
bug, but it's a bug that we've come to depend on.

The attached patch adds a knob you can set in rc.conf to see
"Starting foo" messages again. If people find this useful I'll be
glad to commit it. I'm also open to suggestions for a better
name.

From your patch:
- [ -z "${rc_quiet}" ] && echo "Starting ${name}."
+ [ -z "${rc_quiet}" -o -n "${rc_starting}" ] && echo "Starting
${name}."

Presumably this variable would go in /etc/defaults/rc.conf .. seems
that all the other rc.conf binary choices are either set "YES" or
"NO" rather than something or empty, so perhaps more consistent
with that might be:

[ -z "${rc_quiet}" -o "${rc_starting}" = "YES" ] && echo "Starting
${name}."

No, we don't test boolean flags for specific values. I've attached a
proper patch to this message.

As for name, not fussed, but maybe rc_startmsgs maybe more
indicative?

That sounds fine to me, if anyone else has a better idea follow up on
the freebsd-rc list.

Also, I agree with Oliver; YES should probably be the default
(POLA) so folks annoyed by 'too much information' in console.log
can disable it.

This late in the release cycle I'm comfortable adding a knob, but I'm
not comfortable toggling the default. If there is consensus on the
freebsd-rc list to make the default "yes" then I will go with that
consensus.

Dimitry Andric wrote:
On 2009-09-17 09:17, Oliver Fromme wrote:
My feeling is that hiding all of the "starting" messages is a
regression that needs to be fixed. I cannot think of a good
reason why they should be hidden, but there are certainly good
reasons to display them.

The commit that introduced this, r179946, has this log message:

==== Move the diagnostic output when the rc.subr(8) glue
automatically starts a
service behind $rc_quiet. Instead, output a warning if the
pre-command routine or the command itself failed. Arguably, it's
more useful to know when
a command failed to start than it is to have an endless list of
"Starting ...." lines[1].

[1] - This change actually helped me to discover a bug in
rc.d/{lockd,statd}
(fixed in r179941) that used to fail silently before. ====

It is really a matter of taste, I guess. There is something to be
said for eliminating verbosity, in order to make real errors more
visible.

On the other hand, if one of your startup rc scripts hangs (for
whatever reason), it is nice to have an approximate idea which one
it is. :)

I tend to agree that the current default is a little too un-verbose.
If there is support for adding this knob I'll follow up with re@ to
see about getting it into 8.0-release, although I'm not sure how
they'll feel about a change like this so late in the cycle.


Doug

--

This .signature sanitized for your protection


Index: share/man/man5/rc.conf.5
===================================================================
--- share/man/man5/rc.conf.5 (revision 197259)
+++ share/man/man5/rc.conf.5 (working copy)
@@ -24,7 +24,7 @@
.\"
.\" $FreeBSD$
.\"
-.Dd September 2, 2009
+.Dd September 17, 2009
.Dt RC.CONF 5
.Os
.Sh NAME
@@ -95,6 +95,13 @@
Informational messages are displayed when
a condition that is not serious enough to warrant a warning or
an error occurs.
+.It Va rc_startmsgs
+.Pq Vt bool
+If set to
+.Dq Li YES ,
+show
+.Dq Starting foo:
+when faststart is used (e.g., at boot time).
.It Va early_late_divider
.Pq Vt str
The name of the script that should be used as the
Index: etc/defaults/rc.conf
===================================================================
--- etc/defaults/rc.conf (revision 197259)
+++ etc/defaults/rc.conf (working copy)
@@ -23,6 +23,7 @@

rc_debug="NO" # Set to YES to enable debugging output from rc.d
rc_info="NO" # Enables display of informational messages at boot.
+rc_startmsgs="NO" # Show "Starting foo:" messages at boot
rcshutdown_timeout="30" # Seconds to wait before terminating rc.shutdown
early_late_divider="FILESYSTEMS" # Script that separates early/late
# stages of the boot process. Make sure you know
Index: etc/rc.subr
===================================================================
--- etc/rc.subr (revision 197259)
+++ etc/rc.subr (working copy)
@@ -708,7 +708,13 @@

# setup the full command to run
#
- [ -z "${rc_quiet}" ] && echo "Starting ${name}."
+ _show_startmsgs=1
+ if [ -n "${rc_quiet}" ]; then
+ if ! checkyesno rc_startmsgs; then
+ unset _show_startmsgs
+ fi
+ fi
+ [ -n "$_show_startmsgs" ] && echo "Starting ${name}."
if [ -n "$_chroot" ]; then
_doit="\
${_nice:+nice -n $_nice }\
_______________________________________________
freebsd-stable@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "freebsd-stable-unsubscribe@xxxxxxxxxxx"

Relevant Pages

  • Re: Nexus Programming Language
    ... And no operator precedence? ... The reason this is rarely seen is because ... There certainly seems to be a bug here somewhere;-) ... The UNIX system has a command, ...
    (comp.lang.ruby)
  • Re: [PATCH 1/2] LogFS proper
    ... Do enums have a significant ... Reason enough to use it for the simple cases. ... why are function and line included in BUG() then? ... More than in other filesystems simply because LogFS is younger. ...
    (Linux-Kernel)
  • Wolfpack 4.3.10 Server Release
    ... Plug file descriptor leak in add command. ... This feature was flawed (it encourages players to ... Redesign of synchronization between commands, update and shutdown, ... The bug allowed players to delay ...
    (rec.games.empire)
  • Accesing structures data (getting lengthy!)
    ... > assignment statement just won't stand out amongst all the other ... > meaningless dummy variables that you have to keep track of. ... > comparitively easy to pick out of a long, boring, command line history, ... > original, for some reason, wraps me up in knots. ...
    (comp.soft-sys.matlab)
  • Re: My Weekly Drash - Vayera
    ... you're asking what reason there might be for a ... >> literally speaking in one voice) and concluding that the law had never ... I have problem with that understanding, ... Because when the command _to him_ is to do ...
    (soc.culture.jewish.moderated)
Loading