Re: [g-a-devel] At-spi2-atk performance tests

[Joanmarie Diggs <joanied gnome org>]

Hi Alejo!

Since you have subsequently posted a note saying that you found another
bug, I'll wait until you provide new data before I look at the numbers.
In the meantime....

> Now, to solve the orca consumption problem into the results, 

Well, Orca does a lot of stuff -- some of which it simply needs to stop
doing (inefficient and/or unnecessary calls). Some of which it has no
choice (work required to present the needed details to the user in
speech and refreshable braille). Regardless, I would love to see numbers
comparing an AT listening for events versus Orca.

While that non-Orca "AT" listening for events *could* be a python
script; it could instead be Accerciser with full event monitoring on.
The reason it might be worth using Accerciser with full event monitoring
on is that not only would it be listening to events (like Orca) but it
would also be updating the tree of accessible objects. Or to put it a
different way, I *think* Accerciser the closest we can come to comparing
Orca versus some AT which is not Orca but doing many of the same tasks
that Orca must do. 

In summary, what I would really like to know is: How much of this
problem is crap in Orca versus how much of it is "the cost of being an
AT?" -- something I'm not convinced your modified python script will
show us. *If* it would be appropriate for you to obtain these numbers as
part of your project, I would find it most helpful; if not, no worries.
Again, you're providing us with tools we can use to get this data
ourselves and for that I am already extremely grateful.

Take care, and thank you again for all your efforts in this area!
--joanie

p.s. if you do decide to do such a comparison, please ping me so that I
can give you the list of events to attend to. For instance Orca doesn't
listen to object:bounds-changed events because we don't care and that
event gets spewed out like it's going out of style anyway. (Thus
listening to it in Accerciser would potentially skew your numbers and
not really be a true comparison between the two ATs.)