Using mliparam at this time could lead to corruption at the
start of the new executed program if BSS is real full and
mliparam is over $BB00.
The fix is to open the file from the loader stub instead of doing
it before the C library shutdown.
The Pascal Firmware Protocol Bytes ID are not enough to differentiate an SSC card from a IIgs serial firmware:
http://www.1000bit.it/support/manuali/apple/technotes/misc/tn.misc.08.html
Loading a2(e).ssc.ser on a IIgs succeeds, then goes to limbo when one tries to use the serial port.
Check first byte on the slot's firmware in addition to the four existing ones, as it's supposed to be $2C (BIT instruction) on an SSC card, and $EF (65C816 SEP instruction) on the IIgs' serial firmware (ROM revisions 0, 1, 3).
There is no need to TryToSend before getting the character. We
send bytes during SER_PUT, and if interrupted during sending, we
still try to do it at the beginning of the next SER_PUT.
Apple2 and Atmos have Index in X, but can still use it for the
best-case path as long as we reload it in the worst-case part
(when we assert flow control).
Also, standardize the free space to trigger flow control to 32
characters left (compare with RecvFreeCnt before decrement)
There's no target with more than one serial driver (and I don't see that change anytime soon) so it's a no-brainer to apply the standard driver concept to serial drivers.
The Receive Data Register and the Transmit Data Register share share a single address. Accessing that address with STA abs,X in order to fill the Transmit Data Register causes a 6502 false read which causes the Receive Data Register to be emptied.
The simplest way to work around that issue - which I chose here - is to move the base address for all ACIA accesses from page $C0 to page $BF. However, that adds an additional cycle to all read accesses. An alternative approach would be to only modify the single line `sta ACIA_DATA,x`.
