On 25 July 2011 15:49, Jeremy O'Donoghue <jeremy.odonoghue@gmail.com> wrote:
Hi all,

On 25 July 2011 07:45, Eric Y. Kow <eric.kow@gmail.com> wrote:
On Mon, Jul 25, 2011 at 01:10:15 +0100, Dave Tapley wrote:
> A very good suggestion.
> I cabal unpacked wxcore- and it appears to have a fix in it¹.
> Do we know why this fix is in the release, but not in darcs?

I suspect the answer lies somewhere in here:

 darcs changes --repo http://code.haskell.org/wxhaskell --from-tag .  --match 'hunk 45999'

The answer is that it comes from my clean-up for wxWidgets 2.9, and is probably a case of being over-zealous. The problem wasn't caught because (most likely) no-one built tip for Gtk. It's not an environment I have readily available to me.

I'm not sure that there is a straightforward way to get similar behaviour on Gtk, so I think the only clean solution is to revert the old 'NULL' implementation. What do you think?

EWXWEXPORT(wxCursor*,Cursor_CreateLoad)(wxString* name,long type,int width,int height)
#if (wxVERSION_NUMBER >= 2900)
    wxBitmapType bm_type = (wxBitmapType) type;
    long bm_type = type;

#ifdef __WXGTK__
 // See http://thread.gmane.org/gmane.comp.lib.wxwidgets.general/45999
    return NULL;
    return new wxCursor(*name, bm_type, width, height);

Well, the 2.9.2 dev release¹ appears to include that constructor (in include/wx/gtk/cursor.h) if wxUSE_IMAGE is set.
So perhaps we should check that?

¹ http://www.wxwidgets.org/downloads/#latest_dev