Re: [orca-list] ORCA improvements

[Piñeiro <apinheiro igalia com>]

From: Alejandro Leiva <aleiva emergya es>

> A guy from CIDAT told me about bug #321123 [1], what do you think ?

As you see in the comments I have been working in this bug some years
ago.

I was working on that as I was working in the same issue but with
HildonCalendar. HildonCalendar is a GtkCalendar C&P, modified because
they required to apply hildon theme there. Although originally hildon
tried to use directly GtkCalendar, in the end they required to modify
it directly, as the calendar drawing was made directly on GtkCalendar
and were not really customizable.

The main problem I found was trying to get a complete and accurate
resolution of the different day cells. This is because the calendar is
drawn by GtkCalendar but each cell data is not exposed, they are
private data. In my implementation I required to copy the complex
formulae to compute it, but it is still not accurate (and copy the
formulae are not really a good and maintainable idea).

The ideal would be implement the a11y support directly on gtkcalendar,
so we could access to this private data, as GtkIconView does. But, the
problem is that in my implementation I used GailCell. As I said on bug
490174 [3], this library is not exposed by GAIL, and in the same way,
as Li Yuan said on comment 14 "Ideally gtk should not include any gail
header file".

In the end, on Hail [2] I decided to implement the support on Hail,
not directly on HildonCalendar. GAIL maemo downstream version export
this header, so I could use that. As I said this means that some cell
coordinates are not computed properly, and the way it is computed
other coordinates are really hacky. But other things, like AtkAction
(ie, select a day) works fine, so it was fine from the automatic
testing POV.

I didn't include any GtkCalendar patch due all this problems (in
summary this is a hacky solution), but you can check HailCalendar as
reference.

Anyway, as far as I know this wasn't a really problematic
issue. Usually with a GtkCalendar you also have an entry where anyone
could write on it. This missing a11y support just mean that the user
can't use the graphical calendar to use that, but they are able to
introduce a date. But probably I'm wrong, this is just a quick
thought.

BR

> [1]: https://bugzilla.gnome.org/show_bug.cgi?id=321123
[2] http://maemo.gitorious.org/hail
[3] https://bugzilla.gnome.org/show_bug.cgi?id=490174

===
API (apinheiro igalia com)