Package system is being integrated into the core of Redintegrate in a separate branch in case everything catches fire. So far, the core has all file IO migrated and am working on ironing out bugs with resource names being wrong.

<edit>nvidia and SDL2 are up there as being top 2 things eating RAM in the client
Code:


I can try to remove some of the SDL surfaces as not everything needs them.

<edit> I am seeing that massif is reporting quite a bit of heap allocated for boost::filesystem::path::operator==, such as:
Code:


Anyone using boost know what's up with these? (The usage continues to pile up over continuous calling of operator==)

<edit> Just passed 100 commits, will work on merging the package system branch into trunk and get everything working again.

Work on character modelling has begun, will see how well I can model and animate as well as benchmark the engine rendering and animating high-poly models.

<edit>

Finally went ahead and recorded the progress on armature animations:

That looks great, and seems to be almost entirely free of render errors and artifacts to boot. Keep up the dedicated work.

Progress has started again with Blender now that I have a mouse that has all 3 primary buttons working!

Things are happening...

After seeing how to do multitexturing with COLLADA meshes, I have wiped my code for COLLADA models and am starting from scratch. The original code was messy, but couldn't be adapted to multitexturing (loads actually changed). Will be a while until rendering starts working again (I have to fight with the controller again).

Multitexturing is implemented and should work fine with animations.



The cube has 3 sides using one material and the other 3 sides another. Because of this, the cube is drawn in 2 parts, one per material used. Code re-write was needed to load multiple submeshes.

Gotta love that feeling, haha.

Yeah, lately I have been going back and doing major rewrites. A pain to carry out as everything breaks, but much better in the long-run instead of layering hack on top of hack.

This is literally the last 6 years of my life with FreeBuild. I feel your pain. Just keep telling yourself this is definitely the last thing you'll really need to fix and you'll make it through ok

Yeah, I can still see how nice things will be once things are fixed, however there is a growing list of tasks to work on before I step out of core engine components.

Having a very odd problem: I am able to run Redintegrate's win32 build on one machine with wine and Win7, a laptop with win 7, but not another identical laptopn with win 7. It will fail to draw anything :S

<edit>

AMD Catalyst driver was causing the problem. No way to "fix" this when there are 0 errors reported by GL and I am doing things right

A standard question for anyone having graphics problems on something you've had working before is "have you double checked that your graphics drivers are up to date?"

Well, having drivers breaks things, not having them breaks things, so :S I don't think they were the latest, but don't want to go back and check just yet.

After having the multitexture branch basically done, I have also put work into Lua and model making. I haven't been productive lately while sick, but I have seen the first map with a model controlled through Lua. The Lua API is moving ahead, allowing maps to dynamically controls various map elements, and with my test map, setting the position of a cube with the y position oscillating in a sine wave.

This isn't cross-script communication, just interfacing with the model on the server side. The script communication will be re-worked later, but there is still much to be done. There is also a lack of interpolated data, so object movement runs at the server's speed, which for right now is limited to.... some ticks per second, not sure how it is being limited any more, that also needs to be re-made.

Progress!

Now comes the time of Lua sync code. Previously, I would register Lua tables in scripts and they will be synced server -> client. I also had a way to have clients send events to the server, triggering function calls.

Trying to find out other ways of syncing data. Event sending will be done, not sure for data sync. Maybe just remove it and use event triggering to call functions with parameters? Or have a globally synced table with arbitrary data in it? :/

2 month thread bump! Progress has happened!

Changeset 172: Fixed long-standing bug where the client's camera movement sent them to NaN,NaN,NaN (div by 0 error in vector normalization).

I can now move around in the world on the Windows build! Apparently there was no special case for handling normalization of zero vectors in Bullet.

Where was your data coming from that you needed to normalize a zero vector? This is undefined behavior for a pretty sane reason.

This is in the calculation for spectator camera flying mode. My best bet is that it fails to get a direction for the velocity vector (rotation matrix) which sets it to 0 and is then normalized before multiplied by speed.

But yes, something is wrong getting the direction for the velocity vector.

I have been migrating all of my resources into Mercurial as just having an unorganized folder of some useful files and a lot of garbage isn't the best way to manage resources

A week or so ago, I got Jenkins running on a server that can build all dependencies as well as a Redintegrate zip and installer for win32. I also recently improved the build speed from 16 minutes to 5 minutes by optimizing a DLL dependency resolver/packager CMake script to cache DLL searches.

I am now focusing on modelling with a priority on a basic character model, but also working on maps and generic resources.

I need to figure out how to get interfaces implemented. I currently have a material for a model that has a UV map (not 1:1) for the gooey. I probably need to mod the COLLADA exporter script to let me add in custom export information (which will remove the guessing on interfaces and other such things).