Re: device stuff - cvs

[Jan-Henrik Haukeland]

Christian Hopp <address@hidden> writes:

> On Tue, 10 Jun 2003, Jan-Henrik Haukeland wrote:
> 
> > Christian Hopp <address@hidden> writes:
> 
> (...)
> 
> > > IF CHILD(REN) <operator> <int> [<int>] <action>
> >
> > One question, what does this mean :) Does it mean that if a process
> > has e.g. more than X children then -> action?
> 
> Yep... something like software birth control. (-: 

B-) I understand and I can see how it may be usefull.


BTW if you change the resource statements:

IF MEM(ORY) <operator> <int|real> <unit|%> [<int>] <action>

IF CPU <operator> <int|real> % [<int>] <action>

IF CHILD(REN) <operator> <int> [<int>] <action>

IF TOTALMEM(ORY) <operator> <int|real> <unit|%> [<int>] <action>

Please make *usage* a noise keyword so it's possible to say:

IF MEM(ORY) usage <operator> ..
IF CPU usage <operator> ..

> > > We can leave the old syntaxes till 4.0.
> >
> > The next release, that is, this release _is_ going to be 4.0. At least
> > I think we agreed upon this, didn't we?  Since it contains a whole new
> > functionality branch with device check++ and language
> > changes/extensions[1].
> 
> Maybe I have missed something. (-:  Because I always read 3.3 in the
> docs.

We step up one version number in the CHANGES.txt and configure.ac
during development, but if you look in Martin's latest checkin for
these files they use 4.0. It may be that Martin and I just assumed
that the next release was going to be 4.0, but I think we agreed upon
this before :)


> > Anyway, I think it should be possible to deprecate the current
> > resource syntax for this release already because this particularry
> > change is slight. Simply add the same deprecation code to the parser
> > like we have done with for instance autostart:
> >
> > autostart       : AUTOSTART {
> >                     YERROR("Warning: 'autostart' is deprecated - "
> >                        "use the 'mode' statement instead");
> >                   }
> >                 ;
> 
> That's not a "warning"... warnings are usually non fatal.  You should say
> error!

Errors should be reserved for stopping monit after parsing. I think
that using a statement without any effect is a warning because monit
may and should run. But using the word 'deprecated' is probably
unfortunate since the user may think that the statement has effect but
that he is recomended to use the new statement version. This is how
"deprecated" is used in e.g. Java, i.e. using a deprecated
method/class/whatever still works. In our case we should use the word
'defunct' instead so the user knows that the statement has ceased to
exist. What do you think?


Ps. In case you haven't seen it I have created a new mail about the
language issues you have mentioned, in the general mailinglist.

-- 
Jan-Henrik Haukeland