Re: [MiNT] key-repeat-bug on bootup

[Vincent Rivière <vincent.riviere@freesbee.fr>]

Helmut Karlowski wrote:
> BTW: The repeat-fix does not work on TT.

Kbdvbase() returns a pointer to an array of IKBD vectors, named KBDVECS 
below. Just before this array, there is an additional vector named kbdvec. 
It is used when a keyboard-only packet is received, and can be used to 
simulate key press/releases without dealing with the ACIA.

From EmuTOS sources, in bios/aciavecs.S:

// In addition to the documented features, it was deemed necessary to add
// the following undocumented features, located in memory immediately before
// and after the documented part of the KBDVECS structure:
//
// struct UNDOCUMENTED {
//   void (*kbdvec)( UBYTE data );   /* KBD Input, TOS >= 2.0 */
//   KBDVECS kbdvecs;
//   char ikbdstate;
//   char kbdlength;
// };
//
//- kbdvec (undocumented feature of TOS >= 2.0) is called with the
//  received data byte in d0.

We must make a list of OS versions where this kbdvec vector exists, and use 
it only when available. Note that the actual routine has always existed as a 
private function in the TOS. When that vector is not available, we could 
even imagine jumping directly to the well-known ROM address determined form 
the OS version (would be very ugly).

We must keep in mind the final goal: its about stopping the key-repeat 
routine triggered by the VBL. Unfortunately the variable disabling the 
active key repeat is undocumented and its address is not the same between 
TOS versions. Thus I got the idea to access that variable indirectly by 
generating a fake key release.

--
Vincent Rivière