Bug#683654: lsb-base: some messages should be conditioned by $VERBOSE
Dear OdyX,
On Wednesday 08 August 2012 16:10:34 Didier 'OdyX' Raboud wrote:
> > In Debian the recommendation for all init scripts is to check the
> > variable VERBOSE from /etc/default/rcS and only print a message on
> > console only if this variable is not 0.
>
> From the content of the debian-policy package, I don't think that's true;
> can you point me to where this recommendation comes from ?
This is not in current policy as far as I'm aware.
However /etc/init.d/skeleton use value of VERBOSE variable to decide whenever
show or suppress messages produced during start/stop actions.
VERBOSE variable is briefly documented in rcS(5) man page and most scripts in
/etc/init.d are controlling verbosity according to /etc/rcS:VERBOSE.
Occasionally people report bugs something like "lack of respect for
rcS:VERBOSITY in init.d/some-daemon breaks splash screen....".
At the moment I found the following message which may be relevant:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=208010#233
There might be something else of course....
> As only the init scripts themselves can know which of the messages handed
> to the log_msg_* functions are important or not, I think the
> responsability of handling an eventual VERBOSE variable is upon the
> individual iniscripts and not upon lsb-base.
Good point, but if implemented, documented behaviour of LSB functions in that
regards will help to reach consistent print-out behaviour.
Perhaps it is possible to decide which functions will respect VERBOSITY
settings. This may help developers to adjust respective scripts, eventually
(in long term).
> > Is that correct that [info] messages are only printed by
> > 'log_action_msg'?
>
> Correct.
>
Thank you.
> > It may be used for messages like "service is already running" and I'm not
> > sure if we want to suppress such messages.
>
> Indeed, there's no way to differentiate useful from useless log messages.
>
Very true... I'm not sure how shall we decide the final behaviour.
> I think that the fact that a call to log_*_msg always leads to having a
> message printed is actually a feature and a characteristic of these shell
> functions. Changing this interface to start printing messages conditionally
> is a quite intrusive change IMHO.
You're right, I'm with you. But in the end it comes down to whatever Debian
scripts should respect VERBOSITY setting - if so such support would be much
more effectively implemented in lsb-base rather than hope that every
maintainer will make sure their scripts correctly obey it.
> Although I would have been tempted to directly tag +wontfix or mark the bug
> - done, I'm hereby tagging as +moreinfo to see if I can get convinced.
Thank you, this is much appreciated.
Desregarding of the outcome I hope your decision may help to clarify some
aspects of init scripts' VERBOSITY.
All the best,
Dmitry.
Reply to: