#beckett engine

Actually, pissing me off isn't the right term. I'm not sure what is. Regardless!

Here's my train scene in the Beckett Engine test thing.

I made it so that I can toggle (load/unload) Bob with the press of a key. And sure enough, loading him works.

But then when I unload him, most of the train which shares a shader with him disappears too.

But it's still being rendered?

Maybe afterwards I'll ask ny'all for suggestions. I dunno...

To be honest I'd rather have concrete coding help for my projects, not just the Beckett engine or Project Special K, than suggestions on stuff I probably can't implement anyway.

I suppose the best course of action at this point would be to give up and find another hobby to replace both with a system like @vaartis suggested.

Say I have the common functions dialogue and music, and a reference to the player. Buncha stuff like that. I also have NPC-specific scripts with event functions like onTrigger or onSpawn. I want each of these NPCs to have their own onTrigger functions but share the one set of common stuff.

Edit: I mean say I have an entity defined to use "alpha.lua" as its script source. On spawn, that entity loads "alpha.iua" into its own sol::state, calling the onSpawn function if it exists in that state. That function may set a bunch of "global" variables that should stay in that state, not clobbering any other entities.

But at the same time, all these individual entities with their own states should have access to a bunch of actually global stuff like common functions... that I preferably only set up once.

/edit

To delete one, you set its "dead" flag, so they're removed after iteration. To add one you push it into a separate "add these for the next loop" vector…

…Because doing either of these things directly would invalidate the iterator.

Model loading in Beckett Engine is a bit wonky, especially when trying to unload a model. I'm thinking of rewriting it a bit.

Current state:

Say you have a file monkeyhead.fbx. Call the Model constructor, passing that file name, to get a Model. You can now mess with the textures and armature, disable specific meshes, and Draw it in a given position.

Calling the Model constructor again with the same file name attempts to save on VBOs and such, it's kind of a mess.

Possibility A:

Forgo the cache entirely. Loading two Suzannes means loading two Suzannes, and each of them will have their own data on the OpenGL side of things instead of sharing a VBO and all that. Simple but inefficient. If the two models have to be animated, textured, and/or positioned differently, you can still do that.

Possibility B:

Move the cache out of the Model class' responsibility. Have a specific LoadModel function that either constructs the first copy from scratch, or returns the first Model (smart pointer?). I feel like destroying the OpenGL data when both monkeys are deleted would be the hardest part.

So last month I'd been working on a set for the Project Special K introduction sequence.

And I figured, y'know what I should do? I should export the thing to FBX and see what we get.

So I make sure the scale is right and all that, and edit my development testing thing to use this set instead of the weirdly blocky test scene I used to work out directional lighting.

And I get nothing. Just a blue screen with my cursor and a custom button. Twisting the camera around reveals nothing either.

So I capture it all in RenderDoc, right?

Okay so the first item is the bench frames, second is the cabin itself, third is... I'm not sure but I think it's the door in the back, fourth item is the bench seats, fifth is the posters, sixth is the windows, eight is the overhead racks, and ninth is the window, door, and poster frames.

And nothing shows up.

The vertex shader output is all zeroes.