Hatari Crash with Spectre 3.0

[jfceklosky]

Any thoughts on the following crash running Spectre?

This runs fine on a real STE, Mega ST, and the ST core on the MiSTer

[Eero Tamminen]

Any thoughts on the following crash running Spectre?

This runs fine on a real STE, Mega ST, and the ST core on the MiSTer

Screenshot shows Hatari running MegaSTE, not one of the machines you listed.

Does Spectre crash under Hatari when it emulates exactly same HW & SW setup (same TOS version, amount of RAM, same disks/HDs emulated) as where you tested it to work on real HW?

[czietz]

Note that this was a bug in Hatari; which is now fixed. If you run Hatari on Windows, you can use the latest snapshot build from here: http://antarctica.no/~hatari/latest/windows/. On other OS, you might have to build Hatari from source from its git repository.

In any case, with this fix, Spectre runs on Hatari:

grab0006.png

[Eero Tamminen]

In any case, with this fix, Spectre runs on Hatari:

Thanks for debugging this, and identifying the exact (bus error handling) problem in Hatari!

(And thanks to Toni for fixing the issue in WinUAE CPU core & Nicolas merging that quickly to Hatari for the forthcoming 2.5 release.)

[vanfanel]

@Eero Tamminen

I'm also trying to run Spectre on Hatari, built from latest github code.

My problem is that Spectre always gets stuck at the "Welcome to Macintosh" banner.
I am trying to use the same System 6.0.8 HDD image (that image works with MAME with no problems), and I have also tried to use the example 100 MB HDD image available at https://obsolescenceguaranteed.blogspot ... e-mac.html

Using either of the HDD images as ACSI disks in HATARI, Spectre can detect their partition, and even boot, but stays there on the "Welcome to Macintosh" banner seemingly forever.

Maybe there's a recommended Atari machine/TOS combination/RAM amount to use Spectre on Hatari?

I also see this on the TTY when running Spectre on Hatari:

Code: Select all

WARN : Bus Error writing at address $0, PC=$195f0 addr_e3=195f0 op_e3=2ac7
WARN : Bus Error writing at address $efe1fe, PC=$4442ba addr_e3=4442ba op_e3=13fc

In case it would be relevant...

I have tried both ST and MegaST models with their corresponding TOS versions, but Spectre is always stuck there on the "Welcome to Macintosh" banner.

Also, I have tried run Spectre from a GEM disk, or from a floppy .ST disk, but the results are the same...

[vanfanel]

Answering myself: Configuring HATARI to emulate an ST/STE/MSTE machine with 2MB of RAM makes Spectre boot System 6 completely.

But I must say Spectre's compatibility isn't what I had hoped: Both Prince of Persia and Arkanoid fail miserably

[Eero Tamminen]

Bus errors on writing to $0 (or some other single digit address value) i.e. program basically doing NULL pointer access, are always suspicious and likely meaning that the program crashed. "--trace os_base" option would show whether program terminated with Pterm() after that.

[jfceklosky]

I am glad to see that version 2.5.0 fixed the issue with Spectre 3.0.
Thanks for all of the hard work!

[jfceklosky]

Hatari 2.5.0 works great on as an ST, STE, MSTE, but I got the following crash running as a TT.

Anyone else try Hatari 2.5.0 as a TT and Spectre 3.0?

[jfceklosky]

Hatari 2.5.0 works great on as an ST, STE, MSTE, but I got the following crash running as a TT.

Anyone else try Hatari 2.5.0 as a TT and Spectre 3.0?

[Eero Tamminen]

Hatari 2.5.0 works great on as an ST, STE, MSTE, but I got the following crash running as a TT.

Anyone else try Hatari 2.5.0 as a TT and Spectre 3.0?

68000 programs do not necessarily work on TT/Falcon...

Could you try Hatari Git version and disabling data cache ("--data-cache off")?

Or if you're limited to Hatari release version, disabling cycle exact & prefetch emulation ("--cpu-exact off --compatible off"), which means also disabling data cache?

What if you try Falcon instead of TT?

[npomarede]

Also note that most importantly maybe in that case, 68000 and 68030 don't have the same stack frame for all exceptions. And as Spectre is doing some "low level" emulation, intercepting exception vectors I guess (but I haven't checked) this could be the reason for crashing.

[jfceklosky]

I tried 2.5.0 with --cpu-exact off --compatible off and got the same crash as above.

Spectre 3.0 never ran on the Falcon. David Small stopped working on Spectre before the Falcon came out I think.