Re: paramspec value_validate

[Kevin Ryde <user42 zip com au>]

Torsten Schoenfeld <kaffeetisch gmx de> writes:
> There's one thing that still isn't clear to me: what about the
> gunichar issue? Your wrapper of g_param_value_validate uses
> gperl_value_from_sv,

Yes, I basically followed g_object_set doing init_property_value() and
gperl_value_from_sv() once I realized value_validate isn't a general
validity check, only further restrictions on a GValue of already correct
GType.

> so it will not accept single-char strings, only
> plain numbers.  Wouldn't a user expect to be able to pass in
> single-char strings?

Maybe, but I think value_validate and value_cmp will be working on
things that you're soon to pass to or from get_property and
set_property, so I think they might best follow those latter, for better
or worse.

> It does leave me
> worrying about the inconsistent gunichar API we offer, however.
> Manually wrapped functions return and take single-char strings.  But
> everything which goes through the property system needs to be a plain
> integer.  Is there anything we can do about this?

I guess the blame lies with GValue unhelpfully not distinguishing a
unichar.  I suspect trying to fight against that in unichar paramspecs
might only lead to more pain, or only pain of a different kind. :-)

The best I can think is just try to confine the numeric style to things
related to parameters and paramspecs and put it down to hysterical
raisins. :-(

> "svn diff" from the root.

I'll try to do that at least for multiple files.