Do not recognise digits as valid first keys if Shift is depressed.

[smblott-github]

Currently, we cannot set passkeys for symbols accessed with Shift depressed on a digit key: e.g. $, *. This affects native keyboard shortcuts in GMail.

@mrmr1993. This would be fixed if we had event.key, right? In the meantime, is this fix correct? Surely there are no keyboards with digits requiring the Shift key?

Fixes #1996.
Fixes #1757.

[mrmr1993]

Haven't tested (I'm away for a few days), but I expect this breaks g$ on google (we leak the $'s keydown).

Adding a shift modifier (eg. <s-1>) and implementing #1368 or similar would be the way I would look to solve the problem.

This would be fixed if we had event.key, right?

Yes, event.key fixes all of the edge cases for our key handling.

[smblott-github]

but I expect this breaks g$ on google (we leak the $'s keydown)

You mean Google's instant predictions? Yup. It breaks g$ in that case.

To summarize:

• We're currently incorrectly blocking keydown events on one important Google site (GMail).
• But the fix - which is correct in principle, I think - breaks another important Google site (if instant prediction is enabled).

(Google instant prediction and Vimium aren't great bedfellows.)

[lydell]

The widely used AZERTY layout has symbols on the number row and numbers in its shift layer. So there are many users who must use shift to type numbers :)

Se programmers also like to swap the regular and shift layers of the number row because they feel that they use symbols more often than numbers, which reduces the need to press shift.

[smblott-github]

Thanks for the input, @lydell. So, I guess we're leaking a lot of keydown events on AZERTY keyboards.

In any case, the status quo wins. In light of the points from @mrmr1993 and @lydell, we probably just stick with the status quo until event.key is available.