Re: Official .defs files

[James Henstridge <james daa com au>]

On 25 May 2001, Rob Browning wrote:

> Havoc Pennington <hp redhat com> writes:
>
> > The idea is to avoid including the GTK C headers. While pretty
> > painful to implement, there is a noticeable compile speed advantage
> > for app writers if you can pull it off.
>
> Well, I guess to me it depends on what the goal is.  If the goal is
> fast compilation times, then perhaps you've got a point.  If the goal
> is to define a general "C API definition format" that can be used as
> universally as possible, then I think the compilation times aren't
> relevant when considered against being able to wrap a wider range of
> APIs (whether they're doing things they "shouldn't be" or not), and
> making sure that there's as little chance of wrapper <-> C-API skew as
> possible.
>
> But again, without knowing for sure what the goals are here, it's hard
> to know if my objections have any merit.

Note that the defs file format already has a number of features
specifically designed for wrapping libraries using the glib GType/GObject
system.  At least in my opinion, wrapping other APIs is a secondary goal.

Given that the (define-enum ...) and (define-flags ...) definitions will
most likely be generated by querying the GType system, adding actual enum
values won't be that difficult.

James.

-- 
Email: james daa com au
WWW:   http://www.daa.com.au/~james/