Re: Weird problems with Enchant and GTK

For what it's worth: it works fine with enchant 1.3.0 (even valgrind
does not complain), but i don't have the zemberek provider. So my guess
would be an error in the zemberek provider for now.

But surely Dom has better clues than I have :)

  Marc

On Sun, 2008-08-03 at 19:47 +0200, Enrico Tröger wrote:
> Hey all,
>
>
> I just stumbled on a weird problem:
>
> see the attached test case which simply creates a broker object,
> requests a dictionary object and frees BOTH again.
> When running this code once, there is no problem. But I got weird DBus
> warnings and similar code in my app actually causes a segfault when
> the sample code is ran twice (therefore the for loop around it).
>
> I know that Enchant isn't linked to GTK and this is what makes it weird.
>
> Details:
> Compile the sample code this way:
>
> gcc -o "test" -Wall "enchant_test.c" `pkg-config --cflags --libs
> enchant`
>
> And run the binary
> ./enchant_test 1
> ./enchant_test 3
>
> And this should work. But now, compile the code this way (notice the
> added gtk+-2.0 package):
>
> gcc -o "test" -Wall "enchant_test.c" `pkg-config --cflags --libs
> enchant gtk+-2.0`
>
> And running
> ./enchant_test 1
> is still fine but once you keep the loop running more than once, there
> will happen strange things:
>
> enrico@endor ;-): /misc$ ./enchant_test 2
> finished
>
> (process:18884): GLib-GObject-WARNING **: cannot register existing type
> `DBusGProxy'
>
> (process:18884): GLib-GObject-CRITICAL **: g_type_set_qdata: assertion
> `node != NULL' failed
>
> (process:18884): GLib-GObject-CRITICAL **:
> g_boxed_type_register_static: assertion `g_type_from_name (name) == 0'
> failed
>
> (process:18884): GLib-GObject-CRITICAL **: g_type_set_qdata: assertion
> `node != NULL' failed
>
> (process:18884): GLib-GObject-WARNING **: cannot register existing type
> `DBusGProxy'
>
> (process:18884): GLib-GObject-CRITICAL **: g_object_new: assertion
> `G_TYPE_IS_OBJECT (object_type)' failed finished
>
>
> Again, this only happens when the application is linked against GTK.
> And it seems it only happens when the "zemberek" provider is used.
> This is another thing which seems strange to me:
> when I choose a non-sense dictionary name like "nonsense123" I would
> expect enchant gives me an error saying that the dictionary doesn't
> exist. Instead, it returns a valid dictionary object from the
> "zemberek" provider. It seems this provider is used for tr-dictionaries?
>
> When I choose another dictionary like "de" or "en" which both are
> provided by aspell (on my system), the above example code works fine,
> even when called multiple times.
>
> So, am I doing something wrong or might there be a problem in the
> "zemberek" provider code?
>
>
> Regards,
> Enrico
>
Received on Sun Aug 3 21:08:07 2008