Re: GObject introspection support

[Simón Pena <spenap gmail com>]

On 12/07/10 20:32, Víctor M. Jáquez L. wrote:
> On Mon, Jul 12, 2010 at 08:24:44PM +0200, Juan A. Suarez Romero wrote:
>>
>> IMHO, I would not change the #defines. If they are going to fix this
>> problem in g-ir-scanner, why do we need to change Grilo for something
>> that is not wrong here, but in a different place?
>>
>> Also, IIRC, Victor handled #defines using a custom vala file. I know
>> that this is not as automatic as desirable. But still, I think it is
>> better to go with it until introspection becomes fixed in trunk.

Maybe I wasn't too clear. If we keep the #defines, there would be two
options:

a) Having the g-ir-scanner parse them, use the already existing tags,
and generate the resulting .gir file.

b) Use the 'g-ir-scanner --inject' option to include a custom file with
those missing functions in the resulting .gir file.

As for the first option, they (people at the IRC channel) didn't provide
an answer on whether it is supported or not. I'll take a look at
gstreamer as Víctor suggested, but I'm not too confident.

The problem with the second one is that they had support for injecting
custom files, but that support is broken now and isn't going to be fixed
(by fixing the 'upstream library', they meant us fixing ours, not they
fixing theirs).

However, I have the IRC logs: we can check them tomorrow to see if I
misunderstood them.

> I recommend you to take a look to gstreamer. They build the gir in their
> sources, and the have several macros aliasing functions. But I suspect they
> just ignore them...

Tomorrow I'll take a look at this. If they handle the #defines nicely,
I'll follow their approach. If they ignore them, I guess we'd need to
use functions...

Cheers,

Simón