Re: [bug-hello] Does "Hello, world!" need a configuration file?

[Karl Berry]

       not meant to be saying, "you have to use your longopts list."  

Ah.  Yes, now I think I understand betterwhat's going on.  But tell me
again, just so I'm sure -- configFileLoad just parses whatever is there?
You don't have to give it a description in advance?  So if all we wanted
to do was look for one option, we could just say
        tOptionValue *opt_p = optionGetValue(ov_p, "greeting");
and that's it?  Thi

    But I _hate_ autoconf.  :-)  Attached anyway.

Thanks.  The test per se looks more or less ok, but we really need to
encapsulate this in a macro.  The point here is not that anyone cares
about a config file for hello per se, but to provide an example of
handling config files.  We don't want to recommend that everyone dump 10
lines of shell code into their configure.ac's, that makes updating
painful.

I guess this means we should throw something about this into gnulib.
But before I write that group, I see one problem:

+    LIBS="${LIBS} `autoopts-config ldflags`"

Unfortunately, it's necessary to separate the loader flags (like
-L/whatever, which get added to LDFLAGS) from the actual library calls
(like -lautoopt, which get added to LIBS).  I don't remember
specifically why any more, but I know it's the case :(.  I fear this
means autoopts-config has to change ...

(Also, CPPFLAGS should be used instead of CFLAGS, but that doesn't seem
like a problem.  CFLAGS is reserved for the user and should not contain
any definitions which are required for building ... see the coding
standards and autoconf.)

What sorts of things do autoopts-config cflags and ldflags output?

It's always the overhead stuff that's the killer ...

Thanks,
karl