[tstrimple]

None of your points are unique to game development. Moving goal posts and requirements changing constantly is a challenge at literally every single development job I've had. There is an eternal back and forth between what the sales team says the software can do, and what the programmers desperately attempt to account for late in the development cycle. Plenty of teams have idiosyncrasies around their tool chain, refuse to reflect on it and maintain that their project is a special case to justify all the mess.

[Guvante]

Let me first say that game developers should do a better job about testing, especially when it comes to developing isolated systems to support unit tests. But it certainly seems like you misunderstand how much change there is in the vast majority of game development.

> Moving goal posts and requirements changing constantly is a challenge at literally every single development job I've had.

It is hard to describe because it sounds the same. But game development is really different because there are no fundamentals of things you care to test.

Even doing a combat sequence requires a herculean effort. You need basically the whole game running because otherwise what is the point. You fake out most of the data so your test doesn't fail when the designers decide to make combat harder. But now it ends up that pressing A defends instead of attacks because reasons and all of your combat tests now fail.

This is all fixable but it makes the cost per test of anything but the tiniest things hard enough that broad test coverage can sometimes be a determent as you end up testing what the game is now which means it will all be thrown away if your assumptions change.

Sure that can always be the case but "the sum of the lines equals the total" kind of tests are much less likely to backfire in this way.

> Plenty of teams have idiosyncrasies around their tool chain

C++ is chosen because the tooling doesn't exist outside of C++. Full stop. You have to build all the tooling in language X which when talking about 3D graphics is a huge amount of tooling.

Rust is starting to have some cool stuff but if you compare you will see there is a world of difference.

Thus if you choose not C++ you get to write C wrappers around your API and deal with all that nonsense since C++ interop is the worst in most languages.

At some point Rust will get proper C++ interop and then the gap will be smaller but for now you are giving up a ton for a slightly safer language by not choosing C++.

Also note that nearly everybody writes a huge amount of non-C++ code, they just call it a scripting language instead.

[megameter]

A recurring thing that happens in indie dev in the past decade is that some programmer writes up a blog about their "fully test-driven" game.

It always turns out that the game has a trivial state-of-the-art-circa-1980 design with a very small featureset. Nobody doing this is also writing a large RPG or even a Mario style platformer.

So, you can do it, but you spend a massive amount of "scope points" doing it.

The language tooling is much the same way. To do games - big or small - you need lots of I/O handling, and this immediately leads you towards talking to the OS directly, which leads you towards either C or C++ because that's where the tools and resources are. You can get a binding of SDL or whatever for your language, but that's effectively limiting the scope of your engagement with I/O - if the framework doesn't work, you have to debug it across a binding layer which is always iffy. And it can really hurt when talking about console dev.

[kaba0]

Offtopic but may I ask you for some game/OpenGL-related questions?

I’m attending a computer graphics course and have to write a minecraft clone - and I don’t yet “feel” the performance of the CPU vs the GPU and sometimes I have trouble deciding which one to push.

Like, am I allowed to call glDraw* multiple times per different objects when the data is already on the gpu, or should I try to push somewhat dissimilar objects into the same buffer and write more complicated shaders differentiating them (without ifs if possible?) or are gpus so performant nowadays that unless I do something stupid I should not worry about it?

[DavidVoid]

> Thus if you choose not C++ you get to write C wrappers around your API and deal with all that nonsense since C++ interop is the worst in most languages.

An example of this is the Swedish game dev company Embark Studios who, after a significant amount of effort, managed to get NVIDIA PhysX (which is widely used in AAA games) working with Rust [1].

[1] https://www.youtube.com/watch?v=RxtXGeDHu0w