Re: [hatari-devel] SCC register handling on TT

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On 11/27/2016 10:04 PM, Thomas Huth wrote:
> Am Sun, 27 Nov 2016 19:51:13 +0200
> schrieb Eero Tamminen <oak@xxxxxxxxxxxxxx>:
> > When looking at SCC register handling in relation to:
> > http://www.atari-forum.com/viewtopic.php?f=51&t=29989&p=306208#p306208
> >
> > Hatari isn't yet emulating SCC, but I started to wonder
> > about Hatari's handling of non-implemented SCC registers...
> >
> >  From what I see from ioMem*.c sources on SCC regs access:
> > * ST & STE bus error
> >    -> corresponds to bus error tester results
> > * Deliberately voided for Falcon STE compat mode (ioMem.c)
> >    -> corresponds to bus error tester results
> > * Deliberately voided for normal Falcon mode (ioMemTabFalcon.c)
> >    -> corresponds to bus error tester results
> > * Bus error for TT
> >    -> does NOT correspond to bus error tester results
> > 	$FF8C20 - $FF8C80
> > 	$FF8C90 - $FF8E01
> >
> > Is above interpretation right?
>
> I think so, yes.
>
> > Thomas, does above bus error test mean that $FF8C80
> > generates bus error or not (i.e. is the output range
> > inclusive or exclusive or range end)?
>
> IIRC the right hand side of the entries in the bus error tester output
> do _not_ generate a bus error anymore. So $FF8C80 should not generate a
> bus error.
>
> The TT IO memory region has never been fixed for proper bus error
> accuracy, mainly since quite a bit of emulation is still missing.
>
> For example, only disabling the bus errors in the SCC IO memory region
> without providing real emulation is a kind of double-edged sword - some
> programs might suddenly run, but others that use bus errors for
> detecting the presence of the SCC might stop working, because they
> suddenly think that the SCC is there and depend on the right behavior
> of the registers which is not emulated yet.

I noticed from freemint sources that its HW detection does use
bus errors for this.

In the above linked thread, the bus error loop happens according
to OL before _main, so it seems like some library initialization
code is not doing things correctly.

In any case, I think TT & Falcon emulation should work the same
way in this matter.  Both machines are supposed to have SCC, but
Hatari doesn't emulate it for neither, so both should either
ignore accesses or bus error on them.
  -> Should I change TT or Falcon emulation?


> So the really right way to fix this: Implement proper SCC emulation!
> (and to get this right, you also might need to implement proper
> emulation of the second MFP of the TT). ... can of worms ... ;-)

Thanks for the offer, but I I'll pass for now. :-)

(I don't have docs on how it should work, suitable test-cases
for testing that the implementation is correct, and something
interesting that would start working after it's implemented.
Time's issue too.)


	- Eero