Sorry, I didn't notice this earlier as it wasn't
to hatari-devel (I've now changed my mailing list
rules to check also subject line, not just to/cc).
On 12/6/20 8:00 PM, Laurent Sallafranque wrote:
You can try to download this program again.
It looks like the same problem I had with Fun industries demo (it
didn't work with all my versions of hatari, but I was pretty sure it
was a working demo). I've downloaded another version and now it works.
I set all my Atari related files as read-only, unless program doesn't
work when that's done
(GEMDOS HD trace shows file whether files are
being opened as writable).
That way programs can't accidentally corrupt their
files.
I also keep the original zip files around so that
I can verify files against the original files afterwards. None of
Sidetracked demo files have
changed/corrupted.
I just noticed that tracing shows following just
before it panics:
----------------------------
...
XBIOS 0x6D Dsp_ExecProg(0x2BE26, 0x2DD, 0x8001)
GEMDOS opcode invoked outside of cartridge space
Panic: Illegal Instruction
----------------------------
=> cpu/hatari-glue.c OpCode_GemDos() invokes
illegal instruction exception
If I copy the program to floppy disk or HD image,
it does (still) work with EmuTOS (latest snapshot
with DSP support):
$ atari-hd-image 16 16mb.img TEST sidetrac/
$ hatari -s 4 --monitor vga --machine falcon --mmu on --tos
etos512k.img--ide-master 16mb.img --auto "C:SIDETRAK.PRG"
I.e. it's only with GEMDOS HD and TOS4 that it
doesn't work with.
Thomas, any comments on the GEMDOS opcode part?
There shouldn't be any regression with this patch. It changes nothing
except that it tests for a hostport interrupt at every DSP instruction.
The previous way of testing was wrong (only after a crossbar 2 DSP
exchange).
There's no music in lockup for now. I'm working on this, but I have to
add a new interrupt to the DSP (SCI timer), and a timer.
Ok, thanks!
I thought this part of the DSP was not to be emulated (the Falcon
developper doc says : don't use the SCI and don't rely on it), but
Marcusz shows that the doc is wrong (he pushed the limits ;)
So, I'm working on it.
- Eero