Re: Defaults for a minimal embedded GNUstep tool

[Pablo Di Noto]

Richard Frith-Macdonald wrote:
> On 6 Sep 2004, at 17:41, Alex Perez wrote:
>
> > On Mon, 6 Sep 2004, Richard Frith-Macdonald wrote:
> >
> > > On 6 Sep 2004, at 16:39, Pablo Di Noto wrote:
> > >
> > > > Can I set default values into the tool (using no external files) to
> > > > avoid those messages? Where should I look for examples?
> > >
> > > Not really ... but you can silence the thing programmatically either by
> > > changing the descriptor that NSLog writes to, or by replacing the
> > > handler function which performs the write.
> > >
> > > See base/Source/NSLog.m for details.  This stuff should be documented
> > > properly, but is pretty hard to find right now I'm afraid.
> > >
> > > However ... if you have no resources ... you have no characterset
> > > information ... I'd be worried about that in almost any program.
> > > The timezone stuff is less critical.
> >
> > It should be falling back on POSIX envvars for this sort of stuff when
> > no defaults file is present and/or these defaults are not set. If it's
> > not, that should probably be
> > considered a bug.
>
> Unless I'm misunderstanding, we are not looking at an issue of selecting
> options etc via defaults or environment variables here ...
> the idea is to have a standalone tool without the GNUstep system ...
> which means without things like timezone data, information on things like
> which unicode characters are considered to be uppercase letters,
> localised message text etc.

Well, both of you are right in a sense.

The tool works in its actual incarnation, which is a very early release. 
 I may be hitting NSString issues later because of lack of character 
set definition. Not sure about that NSBundle complaint either.

On the other hand, for this particular case, I would prefer no external 
files whatsoever, since the tool will be called from init (and may be in 
the future be the init process itself...)

> The program may be intended to operate in a particular set of
> circumstances where none of these things are required ... (indeed, Pablo
> says the program works ... which would seem to confirm that) in which case
> it does not seem unreasonable to want to silence warnings about the
> fact that they are not available.

My idea was to hardcode suitable defaults. I'm right now learning the 
NSDefault concept and GNUstep tool startup sequence in order to check my 
options (and make more specific questions...)

So far I'm a little bit confused about how my main() function is being 
overriden. GS_FAKE_MAIN is 0 into GSConfig.h;

Any insight/pointer to more info will be appreciated!
///Pablo