Re: [hatari-devel] Small bug in Hatari Falcon emulation

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On 11/24/20 1:46 AM, Roger Burrows wrote:
> On 23 Nov 2020 at 12:16, Eero Tamminen wrote:
> > On 11/23/20 2:46 AM, Roger Burrows wrote:
> > > A quasi-random sampling of values shows 0xff is returned.  That looks like
> > a
> > > floating bus (but what would I know, I'm a software guy).
> > >
> > > > is the target location value just unmodified?
> > > I'm just doing reads, so I'm not sure what you mean.
> >
> > I meant the read result (e.g. register where you
> > read it).
> I still don't know what you mean.  I'm doing a read from a non-existent RAM
> address.

Hopefully this pseudo-code (without bus error
checking) clarifies what I meant:
-------------------------------------
for (addr = start; addr < end; addr++) {
  old = value;
  value = *addr;
  /* this check is what I'm referring to: */
  if (value == old) {
     value = ^value;
     old = value;
     value = *addr;
     if (value == old) {
        printf("0x%06x = <no value>\n", addr);
        continue;
     }
  }
  printf("0x%06x = 0x02x\n", addr, value);
}
-------------------------------------


> I did write a separate program to do:
> 	read
> 	write new/read
> 	write old/read
> and the value returned isn't always $ff (and rewriting the original value
> doesn't restore it, unsurprising considering there's no RAM there).  I think
> it's just floating bus values, but I'm not a hardware guy.
>
> > You seemed to be doing byte reads.  For IO HW
> > regs, it matters whether read is byte or word
> > sized.  Did you try word access?
> 1. As I said in my original post, the report is NOT about accessing the I/O
> memory area.  Hatari already has some code to handle that & I haven't
> investigated whether that code is correct or not.  The report is strictly about
> accessing the RAM area, both existing & non-existing memory.

You said that value is 0xff, which I assume to
be from a byte read.  If you do word reads from
non-existing RAM area, is the value 0xffff (as
expected), or something else?


> 2. I spent some more of my time and added support for different widths of
> memory access.  Unsurprisingly (to me) the same issue happens with
> byte/word/long widths.
>
> If you want any more answers, I'll send you the code & you can investigate
> further.

Unfortunately I don't have Falcon HW.


	- Eero