named warnings & individual warning control

[Ian Lance Taylor]

DJ Delorie <dj@redhat.com> writes:

> >     status_warning (status, "%s does not support the %<%%%c%> %s format",
> >                     WARNTYPE_unsupported_std,
> 
> Yeah, that was basically the compromise Mark and I talked about.
> 
> I slightly favor putting the new WARNTYPE_* option before the string,
> rather than between the string and its parameters, just to keep the
> string and its parameters together.

Yes.

> I also slightly favor avoiding the word "type" in something with such
> an ingrained idea of what a "type" is.  WARNGROUP might be less
> confusing.

OK.

> Alternately, a string representation of the option could be passed,
> but that would slow the lookup down.

On the other hand, a string representation would make it easier to
build a heirarchy.  And the strings would be fixed for a given
instantiation of the compiler, so it would be possible to perfect hash
them.  But I don't feel very strongly either way.

Ian