gecko-embed API additions

[Marco Pesenti Gritti]

Owen Williams wrote:
> On Thu, 2006-09-21 at 01:41 +0200, Marco Pesenti Gritti wrote:
>
>   
>> Hi Owen,
>>
>> this made me rethink a bit about the whole gecko embedding situation. I 
>> will try to post about it later but the quick summary is... I dropped 
>> gecko-embed and made sugar use gtkmozembed directly. Also, I added the 
>> gtkmozembed bindings to sugar-jhbuild.
>>
>> The remaining issue is open-uri being basically broken in the bindings, 
>> right? It should be fixable, we can figure it out tomorrow.
>>
>> Thanks,
>> Marco
>>
>>     
>
>
> Take a look at the code for the Democracy Player.  They have already
> solved the open-uri bug in their own code.  I don't know if their method
> is sound, but whatever they did, it works.
>
> >From MozillaBrowser.pyx:
> "Here's the deal on the open-uri callback hack:
>
> GtkMozEmbed doesn't define the open-uri callback correctly.  The uri
> value is declared as a gpointer, instead of a string.  It's not a
> problem in C but it doesn't work in Python.
>
> To get around this, we declare a C callback to handle things.  Since
> this callback is coming straight from the C code, we need to acquire the
> python GIL or we'll segfault.  Because of this, we can't create any
> local python object because they will be DECREFed at the end of the
> function -- after we've given back the GIL.  To make things less messy,
> the C callback invokes OpenUriCallback to do most of the work."
>   

Have you tried it with latest pygobject? (i.e. in your sugar-jhbuild 
tree). I remember to have seen a similar issue but now I tried to 
reproduce and I couldn't. Anyway please ping me on IRC and we can find a 
solution for this.

Marco