Re: [PATCH] grub-core/term/efi/console.c: Use EFI_SIMPLE_TEXT_INPUT_EX_PROTOCOL if available

[Josh Triplett]

On Mon, Dec 30, 2013 at 08:47:25AM +0100, Vladimir 'φ-coder/phcoder' Serbinenko 
wrote:
> On 29.12.2013 08:04, Josh Triplett wrote:
> > +  term->data = grub_efi_locate_protocol(&simple_text_input_ex_guid, NULL);
> This will find the first handle that provides
> EFI_SIMPLE_TEXT_INPUT_EX_PROTOCOL, not necessarily the same as normal
> console and you may inadvertently switch to another console. You have to
> look for protocols on input handle.

Will do in v2.

> > +      if (shift_state & (GRUB_EFI_RIGHT_SHIFT_PRESSED |
> GRUB_EFI_LEFT_SHIFT_PRESSED))
> > +   result |= GRUB_TERM_SHIFT;
> The keys which have natural meaning for shifted version never should
> receive GRUB_TERM_SHIFT. E.g. Shift+d should give 'D', not
> 'd'|GRUB_TERM_SHIFT.
> Easiest way is to add GRUB_TERM_SHIFT only if GRUB_TERM_EXTENDED is
> already present.

The EFI interface already does that; the documentation for
EFI_SIMPLE_TEXT_INPUT_EX_PROTOCOL explicitly states that it will not
return the shift modifiers for shift+character combinations that compose
into a single Unicode character.  So, shift+'d' will indeed show up as
'D' with shift not set, while shift+left will show up as shift+left.  I
did explicitly test that case; it works exactly as it should.

Also, filtering on GRUB_TERM_EXTENDED wouldn't give the right results
for things like shift+space, shift+tab, or shift+backspace, since the
corresponding unshifted characters don't have GRUB_TERM_EXTENDED set.

> What about providing keystatus function?

The protocol doesn't have a function to return the current modifier
state, only to return a character with accompanying modifiers.

- Josh Triplett