Re: cursor refcounts

[Tim Janik <timj gtk org>]

On 30 Jan 2000, Havoc Pennington wrote:

> 
> Tim Janik <timj@gtk.org> writes: 
> > eh? what's broken with the cursor test, it works fine for me with 1.2 and
> > couple of days old 1.3.
> > 
> 
> Segfaults somewhere in the GdkDrawable code, I didn't look that
> closely. Possibly it's accidentally using 1.2 GDK or something else is
> wrong with my setup. If it doesn't happen for you I'd say don't worry
> about it, I don't really have a 1.3 environment.

that's because you code didn't include the assertments we have everywhere
else. in gtk there are a bunch of cases where cursors are allocated like:
  attributes.cursor = entry->cursor = gdk_cursor_new (GDK_XTERM);
  attributes_mask |= GDK_WA_CURSOR;

and later:
      gdk_cursor_destroy (entry->cursor);
      entry->cursor = NULL;

this produces problems with all entry based widgets and also for ones dervied
from clist.

it'd be nice if you could test new code at least to a point where it doesn't
segfault/warn anymore before comitting.

> 
> Havoc
> 

---
ciaoTJ