yego.me
💡 Stop wasting time. Read Youtube instead of watch. Download Chrome Extension

Axe Ghost devlog - The bug that ruined the run


3m read
·Nov 8, 2024

My name is Thomas K. I'm working on a game called Axe Ghost. It's currently in development. There's a demo on Steam, and my good friend Rishad Buser, he plays the beta version of Axe Ghost, and when he's able, he streams his daily runs. Richard, by the way, is the brains behind Iban OB, the co-op to The Puzzler game. He made Cello Cello together with me, and right now he's working on Dubo, which is a minimal elegant ball-based puzzle game that he's making together with AJ Cink. Dubo also has a demo that you can play on Steam at the moment.

So Richard was playing the Axe Ghost daily challenge, and he was using two new special weapons for the first time. He was doing really well; he had the highest score that I think anyone has achieved in the game so far, and he had almost beaten Gamar, the end boss of the game. Upon doing that, he would have won the run and gotten the SC multiplayer. As all that was happening, it was going really well. He clicked end the turn, and the game froze.

"Uhoh, no, did I? Uhoh, are you still there, Gamar? Was I clicking too fast? Did I crash the game?" He handled it very diplomatically, but of course, this was no fun for anyone involved. So I went to work trying to figure out what was going on. First of all, I recreated the state of the game board just as the freeze happened to try and see if I could recreate it that way, but I couldn't. The next thing I did was to watch through his stream, and then in my development environment, I would follow through. I would replicate each of his steps that he took, and that way I could get it to freeze, but I didn't immediately see very clear what the problem was. So there was no error message or anything obvious like that.

I decided this was a good moment to create a more robust system, which recorded whole game runs, so that if anyone experiences a problem like this again, I can quickly get their run from them and then be able to play it back locally and more quickly find the problem. Because it took me a long time to recreate all of the moves from Richard's run, and I didn't fancy doing it very often. So I built this whole system, and that works well. The system works once more. I played through each of Rad's moves so I could save them all into that recording system, and thereafter I could play that back repeatedly and iterate through more potential solutions to the problem or more things to explore what was going wrong.

Eventually, I figured it out. I have this function that spawns robots. Now, robots are an enemy that appears late on in a run, and they're different from the other enemies, and they can't be destroyed. You know when the robots arrive; they make things a good deal harder. The code that I had in there checked the spawn for each of the potential spawn locations, 'cause those are up at the top of the grid just out of sight. It picked a random tile, one of those spawn tiles. It checked if the tile was empty, like if there was not something already on it. If it was empty, it would put a robot there.

But the way I'd done it was that I had this little loop. It would check a tile, see if it's empty, and if it wasn't empty, it would randomly pick another of the spawn tiles and check if that was empty and so on. In the event that all of those spawn tiles were occupied, it would just keep checking forever. I don't know what I was thinking at the time I'd written this. I like to think that I was kind of distracted and quickly wanted to get something working just to check that this one particular system worked, and that I would then go back and fix that properly. For whatever reason, that didn't happen, and so it ended up in what was the current build at the time.

This is how the code should have been. From now on, I'm not going to use any loops that say while true. Instead, I'm going to say for i in range 0 to whatever, and that can be some crazy high number if I need it to be. But at least this way, when things go wrong, it's a little bit easier to figure out why.

Anyway, Axe Ghost is on Steam. You can wish list it there. There's a demo; you can try out the demo. The demo has a leaderboard. You can see if you get to the top of it. I'll see you around.

More Articles

View All
Divergence formula, part 2
Hello again. In the last video, we were looking at vector fields that only have an X component, basically meaning all of the vectors point just purely to the left or to the right, with nothing up and down going on. We landed at the idea that the divergenc…
Lecture 18 - Legal and Accounting Basics for Startups (Kirsty Nathoo, Carolynn Levy)
General C. Chrisy: “And Carollyn and I are going to talk about uh finance and legal mechanics for startups. Um this is certainly not the most exciting of the classes, sorry, but if you get this right, uh this is probably the class most… So thank you very …
Kat Manalac's Whale AMA
We usually let the startups in each batch decide when they want to launch. Um, so most of the startups in the winter ‘17 batch haven’t announced yet. But, um, there is one female founder who has announced her company. Um, it’s called Simple Habit. It is a…
Forget Scarecrows—Falcons Protect This Farm | National Geographic
We’re kind of like security guards. We arrived before the sugar content of the fruit starts going up. As the foods ripen, the birds are more and more attracted to it, so we stand guard ten hours a day in that field until basically the fruit is harvested. …
How to sell private jets to billionaires
This is a day in the life of a private jet broker. My day starts in a meeting with No. She’s an interior designer. We discussed renovating my fuselage. Nor presented intriguing designs for new lighting, new carpet, and even suggests adding a bed—a mini be…
Sampling distribution of sample proportion part 2 | AP Statistics | Khan Academy
This right over here is a scratch pad on Khan Academy created by Khan Academy user Charlotte Allen. What you see here is a simulation that allows us to keep sampling from our gumball machine and start approximating the sampling distribution of the sample …