Re: g-io-add-watch, GIOFunc?

[Andreas Rottmann]

Jan Nieuwenhuizen <address@hidden> writes:

> Jan Nieuwenhuizen writes:
>
>> Duh.  I keep falling for this same kind of error.  I'll add a wrapper.
>
> This is harder than I thought.  I made a patch that works reasonably,
> but it crashes here
>
>     +                 scm_c_gtype_instance_to_scm ((GTypeInstance*) source),
>
> where it turns out that source is garbage...  Note that this source
> should be a channel, which is created and called using
>
IIRC, the IO channels of GLib are not GTyped/GObjectized, so you
cannot use scm_c_gtype_instance_to_scm(). GIOChannels are wrapped as
WCT/WCP, so you'll have to use gw_wcp_assimilate_ptr(), e.g.:

gw_wcp_assimilate_ptr (source, 
                       SCM_VARIABLE_REF (scm_c_lookup ("<gio-channel*>")));


> Could this be related to not using (make <g-io-channel>) ?  
>
No, there isn't even a <g-io-channel> class.

> And if so, it seems I really need channel-new-file.
>
> There is something I do not understand, why are some GIOChannel
> related C functions wrapped as GIOChannel methods, and others as
> functions?
>
Something probably causes the h2def scanner to emit "inconsistent"
declarations; dunno why exactly.

Regards, Andy
-- 
Andreas Rottmann         | address@hidden      | address@hidden | address@hidden
http://yi.org/rotty      | GnuPG Key: http://yi.org/rotty/gpg.asc
Fingerprint              | DFB4 4EB4 78A4 5EEE 6219  F228 F92F CFC5 01FD 5B62

Software Patents: Where do you want to stifle inovation today?