Alpha Centauri 2

Sid Meier's Alpha Centauri & Alien Crossfire => Modding => Bug/Patch Discussion => Topic started by: Vox Imperatoris on November 06, 2016, 05:13:37 PM

Title: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Vox Imperatoris on November 06, 2016, 05:13:37 PM

I mentioned a crash in this thread (http://alphacentauri2.info/index.php?topic=18347.0), and I was told it would be fixed in 3.5b.

However, 3.5b has failed to fix the problem. The crash can be reproduced in one or two turns from the attached save file.

3.4b continues to work.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Buster's Uncle on November 06, 2016, 05:32:29 PM

You'll want to PM Yitzi, so he'll see this in a timely fashion.  -Be polite, please.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Vox Imperatoris on November 06, 2016, 05:54:08 PM

Okay, I'll send a PM.

And of course I don't mean to demand that anyone work for free for my benefit! I just wanted to point it out.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Yitzi on November 07, 2016, 03:32:12 AM

Thanks, I'll check it out.  I thought I fixed it...

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: MercantileInterest on November 07, 2016, 05:07:36 AM

 :mad: Yitzi is a valuable resource. Don't treat him lightly.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Yitzi on November 07, 2016, 02:26:45 PM

I'm actually fairly tolerant of such things.

This is strange, as it is working properly for me.

Can you check the version number in-game, just to make sure that you are in fact using 3.5b?  If you are, the only thing I can think of is if you want some instructions on how to use a debugger to gather information from your own computer's behavior.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: gwillybj on November 07, 2016, 05:32:34 PM

I thought I'd mention I'm running 3.5b with no issues.
I installed the game, then the WinXP patch, then Yitzi's patch.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Vox Imperatoris on November 07, 2016, 09:10:49 PM

Yitzi, thanks for your response!

I can confirm that I just tested it again, made sure the version number said 3.5b, and saw it crash again in exactly the same place (in the inter-turn period between 2268 and 2269).

If you can provide me with the instructions to gather information from my system for you, I'd be happy to help. By the way, I'm running Windows 10, 64-bit, using the GOG version of SMACX with your patch on top.

Edit: just to make sure, I tried doing a clean reinstall of the game. Got the same crash.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Yitzi on November 08, 2016, 12:11:47 PM

Quote from: Vox Imperatoris on November 07, 2016, 09:10:49 PM

Yitzi, thanks for your response!

I can confirm that I just tested it again, made sure the version number said 3.5b, and saw it crash again in exactly the same place (in the inter-turn period between 2268 and 2269).

If you can provide me with the instructions to gather information from my system for you, I'd be happy to help. By the way, I'm running Windows 10, 64-bit, using the GOG version of SMACX with your patch on top.

Edit: just to make sure, I tried doing a clean reinstall of the game. Got the same crash.

Ok.  So download ollydbg (http://www.ollydbg.de/ (http://www.ollydbg.de/)), open it up, load the SMACX executable, and hit "run".  The game should start.  Reproduce the bug again, and instead of crashing it should switch to ollydbg, which will have some sort of error message.  Take a screenshot of the resulting screen, and PM it to me, and that should at least give me enough information to give you further instructions.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Vox Imperatoris on November 08, 2016, 04:13:48 PM

Thanks for the advice!

(http://i.imgur.com/15Wd3Gy.png)

This is the error message I get: "Access violation when reading [045CB024]". When I do "shift+run" as the program prompts, it tells me "application was unable to process exception".

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Yitzi on November 08, 2016, 10:07:22 PM

Yeah, shift-run is for certain problems that don't cause actual crashes.

Unfortunately, the problem seems to be in an often-used function, which makes it harder to find the specific point at which the problem happens.  It may be best to modify the program slightly (in a way that won't actually affect anything in how it works) to create a suitable breakpoint:
At 4B428C, it says JMP 005B435B.

We want to replace that with the following instructions (one line after the next; you can double-click in the third column to edit the instructions; highlight multiple NOP's so that there's enough room for it):
CMP DWORD PTR SS:[EBP-8], 1000000 (that's 6 0's)
JL 005B435B
JMP 005B435B

Then set a breakpoint (easiest way is to double-click in the second column, it should turn red) on the JMP instruction (it should be 5B4299), and run it.  When it hits the breakpoint, post a screenshot (preferably with the window maximized so that I can see more), and then run it again and post a screenshot of the next time it hits the breakpoint (and maybe the third as well).  If it hits the crash (i.e. the "access violation when reading" message) before hitting multiple breakpoints, let me know that as well (though I don't think it will).

(When you're done, if you close ollydbg without saving to disk, the changes will not be maintained, though as I said they don't really have any effect other than making a suitable breakpoint.)

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Vox Imperatoris on November 09, 2016, 04:11:58 PM

I'm sorry; I've never used an editor like this, and I'm not very familiar with how they work.

I can't find an address "4B428C". I assume I'm supposed to be looking in the top-left box, in leftmost row?

The addresses go straight from 4B428A to 4B4290.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Yitzi on November 18, 2016, 07:05:22 PM

Sorry, I wrote 4B428C, but that was a typo; I meant 5B428C (so it should be just below the information with the screenshot you showed me).

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Vox Imperatoris on November 22, 2016, 02:20:18 AM

Thanks for the correction!

Sorry for the delay on my part, but I ran it again and unfortunately got the crash before hitting the breakpoint.

You can see the result here:

(http://i.imgur.com/VeNsX56.png)

I assume this is harmless, but in order to put the instructions in where you said, I had to overwrite one of the lines reading "DD 90909090". I've included one of them for reference here:

(http://i.imgur.com/uhLSekQ.png)

It seems like a filler line, and, again, it was necessary in order to put the instructions in where you requested—but I wanted to point that out to you.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Yitzi on November 22, 2016, 12:09:22 PM

Yeah, 90 stands for "no operation" and is used for filler, and DD 90909090 just means that for some reason (probably because it's designed for analyzing code that was compiled, rather than compiled and then heavily modified in assembly) Ollydbg thought that line was data, but it wasn't.

So next, try the same thing with a breakpoint at 5B428C (if you didn't save the changes and they're gone, there's no need to redo them).  It's very possible that it will again hit the bug before the breakpoint, but whether it does or not will give me more information.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Vox Imperatoris on November 22, 2016, 08:05:49 PM

Alright, I put the breakpoint at 5B428C. When it's there, it hits the breakpoint at every end turn sequence (regardless if it's before the crash or not). Here's what it looks like:

(http://i.imgur.com/EYhPuEm.png)

When I disable the breakpoint in order to allow the turn to advance, I get the crash. Interestingly, the "access violation" seems to be located in a different area this time:

(http://i.imgur.com/ynBLK3i.png)

C894B024

I'm also not sure why 5B4299 is highlighted in yellow. I copied over the 3.5b .exe again straight from the download, in order to make sure it wasn't altered by any previous changes on my part.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Yitzi on November 22, 2016, 11:17:01 PM

The yellow means that there's a disabled breakpoint there; it doesn't do anything (but is a good way to remember where you had a breakpoint), but to delete the breakpoint entirely just double-click on the line just like you did to create the breakpoint.  You'll probably want to remove (not just disable) it if you change anything in that area, though.

That said: I still don't know why mine is working while yours is breaking, but you did provide enough to (I think) figure out what's going on.  When I adjusted the social engineering calculation code to accommodate some new faction bonus stuff, I assumed that it wouldn't be calculating social engineering for the native "player", and that must have resulted in something going wrong when it did so.  I should be able to analyze the rest on my own and post a fix.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Yitzi on November 22, 2016, 11:57:13 PM

3.5c should be up shortly; let me know if that fixes it.

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Buster's Uncle on November 23, 2016, 01:44:26 AM

;nod
http://alphacentauri2.info/index.php?action=downloads;sa=view;down=349 (http://alphacentauri2.info/index.php?action=downloads;sa=view;down=349)

Title: Re: 3.5b did not fix the gamebreaking crash bug I pointed out earlier
Post by: Vox Imperatoris on November 23, 2016, 01:29:20 PM

I believe that did fix it!

Thanks so much for your help!