Subject: Re: rc.d boot log [was: Re: Cosmetic changes to rc.d scripts]
To: Mike M. Volokhov <mishka@apk.od.ua>
From: Mike M. Volokhov <mishka@apk.od.ua>
List: tech-userlevel
Date: 10/23/2004 19:44:50
On Tue, 19 Oct 2004 17:40:04 +0300
"Mike M. Volokhov" <mishka@apk.od.ua> wrote:

> On Fri, 15 Oct 2004 18:46:27 +0200 mouss <usebsd@free.fr> wrote:
> 
> > Mike M. Volokhov wrote:
> > 
> > > However, I meant a single log, i.e. when at a boot time the output will
> > > grew too much (for example up to few megabytes). In this case variables
> > > will store huge sets of data. Is it ok?
> > > 
> > 
> > I tested your modifs with a dummy script (that outputs about 20M) and 
> > noticed no special problem (except that _rclogbuf initialization causes 
> > a blank line to be appended but that's not a serious problem). However, 
> > this may be an issue for small systems.
> 
> In any case we must decide what way we're using (tee or var) *before*
> command execution. There is possible situation where we use tee method
> and some rc.d script creates behaviour where /var/log/rc.log becomes
> unavailable. This means that tee will fail. It possible to workaround by
> buffering any output and then flushing buffer after each rc.d script
> when /var/log/rc.log becomes available.
[snip]
> Well, could you test, please, my yet another patch to rc, rc.shutdown,
> and rc.subr (in attachment)?

After some testing I've close many nuances and publish new patch as
bin/27393. It works well for me.

Many thanks all (especially mouss) for you time and replies.

--
Best wishes,
Mishka.