[g-a-devel] About the fallbacks related with the new method get_text_for_offset, and deadlines

[Piñeiro <apinheiro igalia com>]

Hi everybody,

as you are aware due the other thread, we are adding a new method,
[atk/atspi]_get_text_for_offset, that plans to replace all
get_text_[at/after/before]_offset, and use a new, and smaller enum,
called Granularity, replacing Boundary.

But as we want to avoid any kind of regression until the new method is
implemented by all the platforms, the current patches include two fallbacks:

  * at-spi2-atk (aka, the bridge): will fallback to the method
get_text_at_offset if calling get_text_for_offset fails.
  * atk: will still use get_text_at_offset in the implementation of
get_bounded_ranges (trying one and fallbacking to the other doesn't
makes too much sense here)

That means that for a while, the new API will be in place, so ATs (like
Orca) could start to use it, but internally and in some specific cases,
the old one will be still used on the implementation. The rationale is
that the old one is the one that all the implementors implement right now.

So about how much time. Taking into account that we are at the end of
this release cycle, my proposal is keeping those fallbacks for 3.10, and
removing them for 3.12. In that sense, this can be summarized as:
  * 3.10: new sensible API, but old API and implementation still around
  * 3.12: we only care about the new API, if you want to maintain the
old (deprecated) API, is your problem

A full release cycle seems a really good time margin to adapt to me.

This sounds ok to you?

BR

-- 
Alejandro Piñeiro Iglesias