focus tracking

Jos Poortvliet jospoortvliet at
Tue Jul 13 19:48:35 BST 2010

On Wednesday 07 July 2010 14:23:20 Thomas Lübking wrote:
> Am Wednesday 07 July 2010 schrieb Martin Sandsmark:
> > On Wednesday 07 July 2010 11:21:25 Piñeiro wrote:
> > > As Joseph explained, this "little smartness" are the heuristics
> > > implemented in Orca.
> > 
> > If I understood Thomas right, the "little smartness" is highly widget/app
> > specific, and therefore doesn't make sense to try to have outside of the
> > widget toolkit/application.
> Indeed.
> Think of a mix'd kwrite window. Nearby the entire screen is covered by the 
> focus'd widget, but only the client process (kwrite) has any reliable* chance 
> to determine the exact position of the I-beam.
> Better:
> it even quite knows *why* the focus has been given, allowing you to prevent 
> nasty jumps when sth. just grabs focus programatically.
> A 3rd prcocess could maybe diff the frames and match that against the focus'd 
> widget reagion, what would dramatically fail when kwrite highlights barackets 
> etc.
> *except for doubling headers, inspecting the process memory, looking for class 
> footprints, pray for matching ABI and reinterpret_cast'ing the relevant 
> portion - of ref'd stack, in doubt ;-)
> On Wednesday 07 July 2010 11:21:25 Piñeiro wrote:
> > do you want to bypass at-spi and made a direct comunication
> > between a Qt app and the magnification tool?
> No - I did not intend to suggest bypassing AT-SPI, I didn't mean to talk about 
> protocols at all.
> I just wanted to point out that there's a better way to track focus than 
> relying on (unspecified) heuristics of a foreign process. (and in that not 
> even intended to offend Orca - it might be required for Gtk+ - dunno ;-)
> Cheers,
> Thomas

Thanks everybody for the input. It seems the picture isn't that grim - the Qt-AT-SPI2-bridge 'just' needs updating to the latest D-BUS API and basically Qt apps should be accessible for Orca and other tools, bringing improved accessibility to millions of users. Of course now someone needs to actually go do that.

The meeting at Akademy was interesting, hopefully some comes out of it.

Cheers and again all thanks!


More information about the kde-core-devel mailing list