printf() and sscanf().

in size (or >4G-1byte depending on how the return value was used).
This is going to trigger a bunch of warnings like this:
conversion from '__int64 ' to 'unsigned long ', possible loss of data
which will be fixed over type (either with typecasts or by supporting 64-bit
file sizes, as the case may be).

1) We know it's not empty
-- and --
2) We want to wait until it is

*stab*

to actually work).

fixed numerous scoring bugs (that have been plagueing me for 2 years)

Clear the outbuf empty event BEFORE writing a byte to prevent clearing the
event if the output thread sends the newly added byte before the event is
cleared:

ie:
Main		Output
RingBufWrite
                RingBufRead
                SetEvent
ClearEvent

and making the flush callback call fflush()

the vsnprintf() call in lprintf().

and stop passing an aboirted callback since this sucks back CPU hard
on some IO methods (such as curses) for no apparent benefit.

This seems to fix the last of the massive CPU usage during transfers issue.

loop.

I don't even know why this is there.

with data_waiting resulted in a single byte being parsed before each keyboard
poll and screen update.

Stop using mystery ch global.

or potentially more with a timeout to at least one byte.