Re: Implementing EFI_SIMPLE_TEXT_INPUT_PROTOCOL

["Ni, Ruiyu" <ruiyu.ni@Intel.com>]

On 9/6/2018 11:27 AM, Heinrich Schuchardt wrote:
> Hello Ruiyu,
> 
> currently I am struggling a bit with interpreting the UEFI spec
> concerning the EFI_SIMPLE_TEXT_INPUT_PROTOCOL.
> 
> In UEFI spec 2.7. chapter 12.1.2 ConsoleIn Definition there is this
> sentence:
> 
> "Only the control characters defined in Table 100 have meaning in the
> Unicode input or output streams."
> 
> Table 100:
> 
> U+0000 Null character ignored when received.
> U+0008 Backspace. Moves cursor left one column.
> U+0009 Tab.
> U+000A Linefeed. Moves cursor  to the next line.
> U+000D Carriage Return. Moves cursor to left margin of the current line.
> 
> Rob interpreted this in a patch for U-Boot such that he simply
> suppressed all other Unicode characters in the 0x00-01F range except for
> special treatment of 0x001b as ESC.
> 
> When I look at EDK2 function USBKeyboardReadKeyStroke()
> (MdeModulePkg/Bus/Usb/UsbKbDxe/EfiKey.c:700) it seems that EDK2 would
> pass CTRL+C as Unicode character U+0003 (cf. your patch 608817ad7114
> "Change the SimpleTextInEx implementation to return CTRL+C").
> 
> The same seems to be the case in function KeyboardReadKeyStroke()
> (MdeModulePkg/Bus/Isa/Ps2KeyboardDxe/Ps2KbdTextIn.c:265).
> 
> So do I get it right that for CTRL+A to CTRL+Z we should return a value
> U+0001 - U+001a in Key->UnicodeChar and table 100 is about meaning of
> control characters only and does *not* prescribe a filter?
The reason to return U+0003 for CTRL+C because SimpleTextIn cannot 
return the CTRL state so it uses U+0003 to tell caller that CTRL+C is 
pressed. I think your interpretation to the spec is right.

> 
> But what about CTRL+[ - CTRL+_ ? Why are they suppressed in the EDK2
> keyboard drivers? How do we enter U+001c - U+001f?
Good question. I didn't think about how to translate CTRL+[ and etc.
The patch I made 7 years ago was to meet the purpose how to tell caller 
the CTRL+[ALPHA] is pressed. There was no requirement to tell caller 
whether CTRL+[ (etc.) is pressed at that moment.
I think since SimpleTextInEx is introduced in UEFI Spec, which can 
return more precise key state/toggle information, caller should use the 
Ex version.
I think it's ok for your driver to translate CTRL+[. I doubt any 
consumer code cares about that.

> 
> Best regards
> 
> Heinrich
> 


-- 
Thanks,
Ray