Precise pointer events for Wayland platform #1854
Conversation
| std::pair<float, float> const& pos, | ||
| std::pair<float, float> const& scroll) = 0; |
There was a problem hiding this comment.
We really ought to have strong types for pos and scroll. Lists of identical types are easy to get in the wrong order and this continues down that "slippery slope".
I know you've mentioned floating point geometry before (#1829) and this seems like an opportunity to do something towards it.
(I'm not convinced that generalizing the current "geometry" classes is the best way forward, but we could think about that.)
There was a problem hiding this comment.
Indeed, following the style here, why isn't this just tuple<float, float, float, float> const& axes ;)
There was a problem hiding this comment.
why isn't this just
tuple<float, float, float, float> const& axes;)
Because that has substantially more ways to induce bugs? As it currently is, the main way to get is wrong is passing parameters out of order (which you can check because they have meaningful names) or mixing up .first and .second which again are at least obviously wrong once you notice. A 4-part tuple has all of those problems, plus making it ambiguous (unless explained in comments) what order things are supposed to be in. A more convincing alternative to me would be float pos_x, float pos_y, float scroll_horz, float scroll_vert, but I'm not sure if that has any particular advantages over std::pair.
Of course you're right that we need a real solution to #1829, but I'm not convinced this needs to be blocked on that. The in-practice maintenance burden here is low.
There was a problem hiding this comment.
In part I agree: this doesn't touch any API accessible from downstream.
But if we're passing a generic 'tuple around several interfaces we really ought to have a type. And that internal type could be replaced with a "real solution to #1829".
There was a problem hiding this comment.
a generic 'tuple
In this case pair is just a 2-tuple
wmww
force-pushed
the
wayland-platform-precise-input
branch
from
e60ce90
to
1630616
Compare
|
This now uses the new floating point geometry types. Stacked on #1961. |
wmww
force-pushed
the
wayland-platform-precise-input
branch
from
1630616
to
d8b5c99
Compare
|
Build succeeded: |
This adds support for floating point pointer motion and scrolling to the Wayland platform. It appears touches were already precise. Fixes #1852. There's still a notable problem, where scrolling events don't appear to be precise because they are arbitrarily multiplied by 10 by the Wayland frontend. This will be fixed in a different PR.