Bugzilla

(In reply to Hiroyuki Ikezoe (:hiro) from comment #11)
> Maybe [this comment](https://bugzilla.mozilla.org/show_bug.cgi?id=1774519#c0) help?

Yes, this is very helpful -- thanks!

(In reply to Razvan Cojocaru from comment #10)
> It looks like [this is the early exit](https://searchfox.org/mozilla-central/rev/1564027eba41993868973e4a7b93bea942a16b78/gfx/layers/apz/src/APZCTreeManager.cpp#1760)

Based on the comment Hiro linked, I would guess that the key event which takes that early exit has `keyInput.mType == KEY_DOWN`, and the key press also results in a second event with `keyInput.mType == KEY_PRESS`, which takes a different early-exit?

> What should I do to make sure I end up in `AsyncPanZoomController::GetKeyboardDestination()`?

Do **subsequent** key presses (i.e. not the first one, but starting from the second) get to `AsyncPanZoomController::GetKeyboardDestination()`? That would be consistent with the findings in bug 1774519.

Also, are you testing on codepen.io? Are the results any different if you instead copy the HTML/JS/CSS snippets into a standalone HTML file, and test case? Whether we hit the APZ codepath for keyboard scrolling is sensitive to things like the presence of key event listeners, so maybe codepen.io has some key event listeners which interfere.

(In reply to Hiroyuki Ikezoe (:hiro) from comment #11)
> Maybe [this comment](https://bugzilla.mozilla.org/show_bug.cgi?id=1774519#c0) help?

Yes, this is very helpful -- thanks!

(In reply to Razvan Cojocaru from comment #10)
> It looks like [this is the early exit](https://searchfox.org/mozilla-central/rev/1564027eba41993868973e4a7b93bea942a16b78/gfx/layers/apz/src/APZCTreeManager.cpp#1760)

Based on the comment Hiro linked, I would guess that the key event which takes that early exit has `keyInput.mType == KEY_DOWN`, and the key press also results in a second event with `keyInput.mType == KEY_PRESS`, which takes a different early-exit?

> What should I do to make sure I end up in `AsyncPanZoomController::GetKeyboardDestination()`?

Do **subsequent** key presses (i.e. not the first one, but starting from the second) get to `AsyncPanZoomController::GetKeyboardDestination()`? That would be consistent with the findings in bug 1774519.

Also, are you testing on codepen.io? Are the results any different if you instead copy the HTML/JS/CSS snippets into a standalone HTML file, and test that? Whether we hit the APZ codepath for keyboard scrolling is sensitive to things like the presence of key event listeners, so maybe codepen.io has some key event listeners which interfere.