http://svn.gnome.org/viewvc/at-spi/trunk/registryd/deviceeventcontroller.c?revision=326&view=markupTo me, it kind of seems rather inefficient for the prototypical use case of mouse motion events, which is to drive the region of interest for the magnifier. I logged a bug/rfe a while ago and to just have gnome-mag track the mouse directly rather than require two CORBA actions per mouse movement (one from the bridge to Orca and one from Orca to the magnifier): http://bugzilla.gnome.org/show_bug.cgi?id=522967. I haven't had time to create a patch, and the current maintainer is only doing so very sporadically due to time constraints.
Regardless of making modifications such as those requested in bug #522967, I wonder if we can make modifications to the bridge to only do the polling if someone has requested to be notified of the events?
Will Matthias Clasen wrote:
On Thu, Jul 31, 2008 at 2:33 PM, Luis Villa <luis tieguy org> wrote:I'd love to see hard performance numbers before we reach that conclusion. (I really don't care about memory numbers. Geeks look at top; my fiancee just sits and taps her fingers waiting for GNOME to log in.)Here is one datapoint: https://bugzilla.redhat.com/show_bug.cgi?id=227476 a11y causes frequent wakeups which eat power. Matthias