Skip to main content

Homebrew game - the worst bug ever


Exhibit A. (6502 assembly code)

Have you seen it yet?  This took me two sessions to find, God knows how many hours. 

The frustrating problem with this bug is that it caused memory to be corrupted elsewhere, memory that happened to be a part of the program. So the symptoms made no sense. 

The game would crash after losing the first life, during setup for the next life. Once I'd found the place where the crash was happening, it was a routine that had already worked properly once. What was going wrong when it ran the second time?

After pausing the emulator just before the crash and opening the monitor to examine the memory, I could see that my code at that particular spot was now corrupt. But that corruption could have happened at any time after initialising the game. No clue as to the culprit. 

I eventually started returning from the game at various points and using the monitor to check whether the corruption had happened yet. Eventually a pattern in the corruption itself gave a clue as to where it might be coming from. Even then, that loop looked fine. Often you see what you know should be there, rather than what actually is there.

Even if you've not spotted the problem, you may be asking why I needed to decrement X until beyond zero. Decrementing X or Y followed by a bne or beq (branch if equal to zero or not zero) is a very neat way to write a loop.  This loop needs X to count down to and including zero. I guess it's possible to rearrange things a couple of different ways so that you can still use X=0 and not need a compare. But this is the dirty way that I wrote it at the time because I tend to just bash it out willy nilly and tidy up later* 

I've now altered that loop so that it doesn't do that compare at all. 

In case you haven't spotted it, cmp is a 'compare A' not a 'compare X'.

The loop copied a block of data into a table using X as an index. So with the bug in place, the loop wasn't finishing and after 0 it was writing spurious data into memory 255 bytes away, which happened to be in the middle of real game code. 

Lessons learned: As well as taking the time to write more elegant code, I'll also be rearranging things so that all variables and tables are at the very end of my code. If a similar corruption or overflow happens, then it's likely to go into empty memory or into game data or tables. That will leave the game running and may yield more clues as to what's going wrong.

The game is going great now that it's back on track, thanks for asking.





* probably never tidy it up. 

Comments

Popular posts from this blog

RC2024/10 - my entry

A while ago I made this MIDI module for RC2014: It works but a better design would have its own serial chip and port decoding.  As it is, it provides the MIDI interface and a clock signal for the second SIO2 serial port. This means that it requires a little setting up and will only work for RC2014s with an SIO2 (and port B not already used). I think people might reasonably expect it to be plug-and-play and self-contained, ie do all the serial itself. My challenge to myself is to:  learn how to connect a serial chip (probably 68B50 ACIA) to receive the incoming MIDI and to serialise outgoing MIDI design the module, including the port decoding write a library so that it can easily be used on any RC2014. Potential applications include a MIDI sequencer and using incoming MIDI to trigger notes on the AY or SID sound chips. Entering the Retro Challenge 2024 (aka RC2024/10)  has given me an incentive to get on with this! I'm happy to see several more entries in the RC2014 catego...

IM53 8080 birthday cake

 Each year I've been trying to get more creative with ideas for Spencer's birthday cake. The plan this year was to incorporate LEDs in place of candles. I eventually settled on an Altair / IMSAI / PDP -style computer since those are the type of computers that inspired his RC2014. The IMSAI 8080 has the most colourful switches as well as a name that I could twist. The thought that it could show randomly flashing lights (as if the computer were running) and that it could also play a game of 'kill the bit' was very appealing. A plan formed to use a capacitive touch pad on the cake itself. The first job is to bake the fruitcake. I often use a 7" square tin and one of those cut in half and rearranged makes a cake of suitable proportions.  Even after taking a slice off the faces to make them nice and square, there are still some rounded corners, so after putting on the marzipan, I used more marzipan as a filler to flatten the whole thing. Even though I wanted to end up w...

Using the Speccy BEEP routine and a beeper speaker on my Orton 3C

W hen I first learned about the Orton 3C and its very clever serial routine which toggles A15 (meaning that the max 32k of ram appears twice in the address space) I've wanted to repurpose this for driving a beeper speaker, just as you might have done with the early PETs*  I'm writing this post after successfully playing a tune. The video is embedded near the end of this blog post if you'd like to skip straight to that. Yes, it certainly would be possible to toggle a pin on the digital i/o by 'out'ing to port zero. This would potentially allow for 'bit-banging' multiple voices, with a lot more work than I've put in so far. However, what I love about the Orton 3C is the 'bare bones'ness of it. The whole point is that it can work using only three chips and to that end I've unplugged my digital i/o for this. I didn't go to the trouble of removing the port zero protection shim (3C-PO) but its one logic chip isn't being used here since I...