Re: [Eric-develop] We can read scan codes in Win32 :/

[Gavin R. B. <ga...@ga...>]

> Ismail PAZARBASI wrote:
> 
> Guys,
> 
>     I wrote a code in Win32 platform, that can read from keyboard
> microcontroller' s port 0x60 (96), and can write there too. However,
> actually, I did not try to write anything to keyboard...
> 
>     The thing that it makes now is simple. There is a function in a C
> code, that is reading from 0x60, this is the last pressed key's scan
> code. then there is another function to match the scan codes to ascii
> codes is called... We can read scan code, convert it into ascii...
> 
> What I want to do exactly?
> 
> I want to first, check if keyboard is present. Then, check if keyboard
> buffer is full. If it is, what is it? Did kernel processed it? If no,
> it is not full (=empty), keep tracking it. When user presses a key,
> keyboard writes some code (I dont remember it, it is in my
> bookmarks.), and we invoke kernel' s keyread function (I examined test
> directory in CVSROOT, and I realized that io.cpp!! Dammit guys, I have
> been searching for such thing for days, and it was on there, under my
> hands for 3 weeks :p :) Kernel then starts to check when key will be
> released, and is there another key being pressed. Meanwhile, kernel' s
> keyboard driver's asciimatch function matches scancode to ascii
> character, and checks for kernel' s command list for any special
> meaning of the key (e.g. alt+function keys in FreeBSD console changes
> ttys). There is also one more point here: is key pressed alone? or
> pressed with modifier keys (alt, altgr, ctrl, shift)... Keyboard then
> stops reading, when user releases key kernel stops.


Good approach.

Could you research 'defensive programming'? I believe it is mentioned in
'Software Engineering', by Ian Sommerville. We need all the defenses we
can get when doing assembler. (It's just too dangerous to do
otherwise..)

 
> The Mach Thing:
> 
> I tried but failed to download MachOS. I don' t know why, I could not
> do it. I could not find server (I have found address from yahoo! and
> cmu.edu). I want to have it. You guys have any URLs? Is there keyboard
> stuff inside Mach kernel currently? (PS: I need source, there are
> hundereds of sites offering *free* suggestions, knowledge, 2 lines of
> examples...)

Guys, visit this site:

ftp://ftp.gnu.org/pub/gnu/gnumach/gnumach-1.2.tar.gz
 
> After this point, I mean reading from keyboard, we can simply try to
> code in Eric environment. I believe we can use something at least we
> can write our names to screen with keyboard :p Anyway, seriously, we
> need a loader, I think it will be ELF, is this right? If it is so, we
> can use lots of compatible binaries.

Use the FreeBSD loader. 
 
> Finally, guys, I want to be enlightened about kernel' s current
> status, and utility (i.e. can we use it for a few stupid commands?). I
> am so excited right now, and I realized I really wrote this mail very
> untidy....

 
> Summary: We have a Win32 source/binary to read from keyboard, the
> scancode, and convert it into ASCII.. Win32 codes are available in C
> and Assembler (MASM) formats. Note that assembler source can only read
> codes and saves them into stack. We don' t have ANSI C/C++, AT&T
> assembler codes right now.

Use intel2gas. d;-)

> I hope we will :/ Because it did not work
> in my FreeBSD box, and a Red Hat Linux box in my school. It failed
> with "Bus error", at the line where we try to read from 0x60. Kernels
> might be restricting us to read, or my code is wrong. Code is at
> below, take a look.

Ouch. There is a timing chain problem here. 
 
> short port = 0x69;
> unsigned char nKeyScanCode=0;
> 
> // Initialization is cool

agreed d:-D
 
> __asm__ __volatile__("inb %%dx,%%al;outb %%al,$0x80":"=a"
> (nKeyScanCode):"d" (port));

Ahh, no offence, but methinks you have activated the interrupt and been
calling from the keyboard buffer without any form of buffer protection.

It might be a good idea to scrap the entire inline assembly thing, and
just put all the nasty assembler definitions into .S file. This saves on
all the *ugly* inline code...(Pls. dont take this personal, it's just a
feature of AT&T...)
 
> there is another code, i grabbed from somewhere else (I dunno where it
> was, but from a Linux source site)

Grab all the code you can d;-) 

There is simply no point in 'reinventing the wheel'. We have no time for
that.

>          __asm__ __volatile__("in%B0 (%1)": "=a" (nKeyScanCode) : "d"
> (port));

Even more primal, dude. *grin*. I still think you guys need to use:

1) External function defs in .S file (Fully annotated)
2) Buffer protection built in, (perhaps at the C++ level?)
 
All the best,
		Gavin.