Civilization instability...WTF?!?

[rhester72]

Please note this applies *ONLY* to the _original_ Civilization, NOT the AGA release, and _NOT_ to the WHDLoad version.

I'm at a complete loss with this game. No matter what I do, sooner or later, one of the following WILL happen (in rough order of likelihood):

- Guru 0000 0003 or 8000 0003 (the latter being FAR more likely)
- Freeze on the city production management screen when trying to present the suggestion from the domestic advisor
- Freeze during computer opponent movement calculation (flashing white/gray cursor in the bottom left)
- Soft reset of Amiga

Problems can occur literally during game start (i.e. before you even select an input method) or can take hours of play to manifest, but again, they WILL happen...it is impossible to complete a full game (unless you lose).

I've tried both original and cracked versions of the 855.01 and 855.04 English game executables, all with the same ultimate result.

The specs are VERY modest - 1M required, that's it. A hard drive is needed for practical play of the game, but can be played without.

Here are things I've tried without success:

- Changing the hard drive controller type. I've gone through A590 XT, A2091 SCSI, AdIDE, and plain old UAE hardfile controller configurations in RDB mode (with every firmware version of each and trying 1.3 and 2.0 FastFileSystems in RDB)...the only interesting thing I learned from all this is that the A590/2091 with DMAC-01 ends up with bizarre colors during actual gameplay but DMAC-02 is OK (and the same happens with Star Control), but I also can't explain what the DMAC version would possibly have to do with that
- Playing entirely from floppy
- Using an emulated A2620 accelerator (as seen in the A2500, listed on the game box as compatible)
- Having 2MB of fast RAM on the emulated motherboard, on the accelerator card, and on the hard drive controller (and both enabling and disabling DMA with each configuration)
- Having only 512KB of chip and 512KB of slow RAM
- Having only 1MB of chip RAM
- Trying Cycle-exact and non CE (with and without cycle-exact DMA)
- Trying OCS, ECS Agnus, full ECS and AGA chipsets
- Trying Kickstart 1.3 and 2.04
- Trying an emulated Quickstart A1200 with native IDE controller HD attached
- Trying with and without SetPatch loaded (either 1.3 or 2.0 version)

As you might have guessed...I've played a *hell of a lot* of this game trying all the various combinations, so I'm quite sure that all of the above _do not work_. =)

Ordinarily, I'd say the game itself is just bugged, except I played MANY hours of this in college on an A500 w/1M chip, 2.04, AdRAM 540 and an AdIDE without issue whatsoever. I don't recall anything like this ever happening...which _suggests_ it's a problem with emulation, except...

- The WHDLoad version works great (the only indication of why possibly being "DMA wait fixed")
- The AGA version works flawlessly on an emulated A1200HD (and I used the *exact same configuration*, simply swapping the drive image, to test the behavior of the OCS version on an AGA machine and it crashed just like it did on a Kickstart 1.3 machine)

I'm completely at a loss. I don't even know how to file a competent bug report because it takes potentially hours to trigger the problem (and also knowing how to actually play the game while having patience to do so) and it's so random that it's impossible to capture a savestate just beforehand.

If anyone has any viable suggestions at all (besides "just play the WHDLoad and/or AGA version" *laughs*), I'd be most grateful.

[Matt_H]

There's this, but I don't know whether it's one of the versions you've already tried. Or whether it's even an official update. (Ah, the good old days when stuff was posted to BBSes with no documentation whatsoever.)

I'd get rid of the 2620. Try 1MB chip, 2091 w/ 2MB, fresh install of 2.04 or 2.1, no UAE ROM. Basically as close to a stock (later-model) A2000HD as you can get.

[rhester72]

Originally Posted by Matt_H

There's this, but I don't know whether it's one of the versions you've already tried. Or whether it's even an official update. (Ah, the good old days when stuff was posted to BBSes with no documentation whatsoever.)

I'd get rid of the 2620. Try 1MB chip, 2091 w/ 2MB, fresh install of 2.04 or 2.1, no UAE ROM. Basically as close to a stock (later-model) A2000HD as you can get.

That update is the official Microprose game engine update to version 855.04 - so yes, it's been tried (cracked and uncracked). =) [And amen on the way patches were distributed back in the day! lol]

I've tried the config you've described...in terms of raw configs, I've covered the following:

- A500 w/1.2, OCS, no hard drive
- A500 w/1.3, OCS, A590 XT [with and without RAM]
- A500 w/1.3, OCS, A590 SCSI [with and without RAM]
- A500 w/2.0, ECS, A590 [XT and SCSI, with and without RAM]
- A500 w/2.0, ECS, AdIDE and AdRAM 540
- A1200 w/3.0 and 40MB IDE hard drive [with and without fast RAM]

with various permutations of 512KB chip/512KB slow and 1MB chip thrown in.

Ironically, that's what I thought the root of the problem was initially - that I'd tried *too* hard to simulate actual hardware and created some sort of unexpected incompatibility.

[Matt_H]

Strange! You're sure it was stable on real hardware? I remember the AGA version crashing on me a number of times. Never played the ECS version, though.

Is there a reference card/insert for the Amiga version that might have some hints? Nothing's made it onto HOL and my AGA copy doesn't have one.

You might have to get into the advanced monitoring tools of WinUAE to figure this out. One more idea: have you tried a different version of UAE? Maybe a bug was accidentally introduced in a recent version...?

[TCD]

Originally Posted by Matt_H

Is there a reference card/insert for the Amiga version that might have some hints? Nothing's made it onto HOL and my AGA copy doesn't have one.

Yep, would be good to have the 'technical supplement' for Amiga that the manual mentions on HOL.

[jotd]

i don't want to troll but that's exactly why whdload can run emulated kickstart games.

[rhester72]

Originally Posted by Matt_H

Strange! You're sure it was stable on real hardware? I remember the AGA version crashing on me a number of times. Never played the ECS version, though.

Very. I'd guess I probably went through 3-4 dozen campaigns back then with zero crashes.

Originally Posted by Matt_H

Is there a reference card/insert for the Amiga version that might have some hints? Nothing's made it onto HOL and my AGA copy doesn't have one.

Possible. My actual box is in storage, but I'll check. I don't recall anything specific about that - back then, I just installed and played.

Originally Posted by Matt_H

You might have to get into the advanced monitoring tools of WinUAE to figure this out. One more idea: have you tried a different version of UAE? Maybe a bug was accidentally introduced in a recent version...?

It's possible - I've been using the 5.0 beta series. I'll go back a couple of releases and see if anything's different.

[rhester72]

Originally Posted by jotd

i don't want to troll but that's exactly why whdload can run emulated kickstart games.

tbh, I fully agree on real hardware - but on emulation, it's kind of pointless, especially when the game itself directly supports hard drive install.

Granted, WHDLoad is *VERY* convenient for play-right-now platforms like the Mini and Retroarch (with built-in direct support for archives and a convenient UI that can handle tons of titles), so my love for it is real...but I think at this point I'm still doggedly pursuing it because a) it should work, and b) it may well indicate a bug in the emulator that would be useful to find and fix if possible.

[jotd]

I would say game bug that triggers with your memory layout or conflict with a workbench component.

Guru 8003 means word read on unaligned address for instance.

Did you try without startup sequence ? That's what whdload basically does. It boots up a fresh virtual system without anything, so memory is unfragmented, free as much as possible, and no other process is running.

[rhester72]

Originally Posted by jotd

I would say game bug that triggers with your memory layout or conflict with a workbench component.

Ordinarily I'd agree - but I've tried five major different major memory layouts (.5C/.5S, 1C, .5C/.5S/2F/1C2F, 2C) without success.

Did you try without startup sequence ? That's what whdload basically does. It boots up a fresh virtual system without anything, so memory is unfragmented, free as much as possible, and no other process is running.

Yes, both with and without SetPatch. The only commands run are the necessary Assigns and the game itself.

[Toni Wilen]

Try increasing stack before starting the game. (stack 10000 or similar, default 4096 is quite small)

[rhester72]

Originally Posted by Toni Wilen

Try increasing stack before starting the game. (stack 10000 or similar, default 4096 is quite small)

Went all the way up to 64K...no joy. (The Workbench icon doesn't override default stack and the Startup-Sequence on the first floppy doesn't change it either, so I'm guessing this is an app that sets its own stack to whatever is needed.)

[QuikSanz]

Works great on my A2000. Using the real thing helps.
Chris

[m1ckey]

That´s the copy protection kicking in.

Try alternative adf´s, ipf´s, what-the-fuck-not´s.

[rhester72]

Originally Posted by m1ckey

That´s the copy protection kicking in.

Try alternative adf´s, ipf´s, what-the-fuck-not´s.

The only copy protection is a simple pass/fail check soon after game start, but the crash happens in exactly the same way even on unaltered binaries (both the release and bugfixed versions).

[rhester72]

Originally Posted by QuikSanz

Works great on my A2000. Using the real thing helps.

That tallies with my experience from back in the day...so it is an emulation issue of some kind. Good to know, thank you!

[Toni Wilen]

I still don't this means anything (similar situations have happened many times already..)

But at least seeing what gets logged when it crashes could help. Guru 00000003 should include some log message which usually does not help much but it can give some hints.

Run with -log command line, enable also "Log illegal memory accesses" misc option before starting.

Only if 68000? What if 68020?

[rhester72]

Originally Posted by Toni Wilen

But at least seeing what gets logged when it crashes could help. Guru 00000003 should include some log message which usually does not help much but it can give some hints.

Run with -log command line, enable also "Log illegal memory accesses" misc option before starting.

Will test.

Only if 68000? What if 68020?

Both crash.

[rhester72]

Originally Posted by Toni Wilen

Run with -log command line, enable also "Log illegal memory accesses" misc option before starting.

civ_crash.zip in The Zone

[Karlos]

For a moment I thought from the title that this thread was a bit of political commentary