[BRLTTY] BRLTTY and AT-SPI

[Samuel Thibault]

Aura Kelloniemi, le Sat 29 Aug 2015 01:14:43 +0300, a écrit :
>  > You can tell brltty which widgets it should take control of with
>  > -X type=terminal
> 
> Sometimes this does not work, and Orca takes over the braille display
> entirely, or then Orca's braille output is lost completely.

Indeed, Orca would need to be told not to read terminal widgets. This
was raised by Sébastien at some point.  Another way would be to define
priorities, but this is yet to implement.

> >> or cursor routing in its AT-SPI driver
> 
>  > I'm surprised, IIRC cursor routing is independent from the screen reader
>  > actually, it's the core which tries to move the cursor.
> 
> Yes, but cursor movement and copy/paste require support from the driver - the
> driver must be able to emulate key presses, and this apparently is not
> supported by the AT-SPI2 driver.

Well, there actually is supposed in atspi2 for key simulation, see
Drivers/Screen/AtSpi2/screen.c (insertKey_AtSpi2Screen).

> I also should add to my problem list the following:
> - When using the AT-SPI2 driver there is no cursor tracking. For some reason
>   the display does not follow the cursor, even though cursor tracking is on.

I'm again surprised, I'm pretty sure this was working.

With your .xinitrc I'm however realizing that I've always only tested
with gnome-terminal, since that was the only accessible xterm at the
time.

> >> If there is going to be some sort of AT-SPI2 driver development, I
> >> think that the driver should use the library interface to AT-SPI, instead of
> >> directly calling the D-Bus methods - this way the code will probably look
> >> better and it will be more tolerant to API changes.
> 
>  > Except that there is no such thing any more. With AT-SPI1 we could use
>  > one, but for AT-SPI2 only the server part really has bindings.
> 
> Isn't libatspi just this.

It's supposed to be, yes. Perhaps it has improved since when I wrote the
AtSpi2, but at the time the rationale was that the old at-spi1 libatspi
was not to be rewritten for at-spi2. I can see that there are
atspi_text_get_text helpers now, so I guess it does have improved, so it
might be usable nowadays.

>  > X window is mostly out of the picture actually, Glib shouldn't be
>  > involved much, so it's more about the brltty codebase, AT-SPI and D-Bus,
>  > you're welcome for questions etc. in any case.
> 
> Thank you! I try to grasp the workings of at-spi2, but it is not at all well
> documented.

Make sure to ask for help on the orca mailing list. Joanmarie happily
provides pointers etc.

Samuel