Re: Patches for the sheet branch

[John Darrington]

On Sat, Jul 07, 2012 at 10:00:40PM -0700, Ben Pfaff wrote:
     John Darrington <address@hidden> writes:
     
     So is this an accurate summary of your take?
     
        - Don't implement GtkObject.destroy in new code.
     
        - Implement unrealize to undo everything done by realize and
               release Gdk resources.
     
        - Implement dispose to just unref objects.
     
        - Implement finalize to do other cleanup.

That would be a reasonable summary.  I don't claim that it's 100%
correct - but you asked for rules of thumb.
     
     > I changed the var_sheet and data_sheet to keep a reference to the uim
     > objects which they return from their _get_ui_manager objects.  One
     > of the comments in the code (which I removed in another commit), 
mentioned
     > that there were problems when the final reference to uim was unreffed.
     > This change fixed that problem, by ensureing that it wasn't finalised,
     > until the object which owned it was disposed.  In fact, I think that a 
     > later commit, made this change unecessary - at least it should be, once
     > all the dispose/finalise methods are in order.. But it can't to a lot of 
     > harm keeping this extra ref.
     
     Would you mind organizing the commits or editing the commit
     messages to better explain the above?

I'll do that.
     
     >      The patch "Reduce the flicker when redrawing the toolbar and
     >      menubar." appears to leak a reference in
     >      psppire_data_editor_split_window().  At least, I see nothing that
     >      will eventually unref de->old_vbox_widget.
     >
     > I think you're right.  But when I add a call to _unref in dipose, it 
     > provokes a Critical.  This needs more investigation. I think the problem
     > is that make_{single,split}_datasheet returns a floating object.  They
     > should probably call g_object_ref_sink.
     
     OK.  Can you add a comment in psppire_data_editor_split_window()
     mentioning the leak?  Otherwise we'll forget about it.

OK.
     

J'

-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://keys.gnupg.net or any PGP keyserver for public key.

signature.asc
Description: Digital signature