[Coco] Timing problem with the CoCo and the WD1773 ?

[Kenneth Schunk]

Oh, I sooo got hit by that in 1986 when working on DEFT Pascal. I can
confirm that it's not a OS-9 thing, and that it's not a CoCo-3 thing. I lost
a few disk during development, to the point of being so fed up with the
thing that I packed the entire computer up and put it in the attic for a
month.

You've narrowed it down farther than I did - I was able to <eventually>
determine that it was an odd/even address thing, and patched the DEFT disk
library to ensure that it would be on a safe address. I took the TANDY code
from the ROMS (the DEFT code isn't identical) and reassembled it to move
from an odd to even address, and was able to make the Tandy code fail. I
always wondered - did Tandy/Microsoft just get lucky with the address, or
did they run into this during development and came up with the safe address.

It's been twenty years, so I don't remember all that many details, but
that's exactly the error I was seeing, and moving the address by one fixed
it. I was pretty sure that it must be an undocumented timing difference
between odd address fetches and even ones.

Ken Schunk