v2.0.0-alpha-rc8

Drop a bogus error message

v2.0.0-alpha-rc7

Turn off occlusion queryies

v2.0.0-alpha-rc4

• 2.0.0-alpha

This is a major version change that encloses major rewrites of many core
systems. Being an alpha release comes with the caveat that several core
engine systems are broken and in the process of being ported or having bugs
ferretted out. See Future Work at the end of this Changelog entry

Notable feature improvements are as follows.

• Terrain Generation now happens on the GPU
  This was a pretty big change. I was initially motivated by performance,
  however the real value of the feature is in having terrain gen in regular
  glsl shaders. More info and instructions for operating the terrain
  generation pipeline is available in docs/

• Editing now happens on the GPU
  Following terrain generation, editing also had to be ported to work on the
  GPU. This is not necessarily for performance or usability, but for
  maintainability. I did not want to maintain a GPU terrain generation
  pipeline, and a CPU editing pipeline.

  The editor capabilities have also been significantly upgraded.
  Comprehensive documentation for the editor is WIP, but what's been started
  can be found in docs/

• World storage is octree-based
  As a more robust answer to LoD, the world is now stored hierarchically as an
  octree. This unlocks huge view distances for a stable, linear cost in
  performance.

• Many smaller improvements, optimizations and fixes
  This update is way too large to address every change; it took many months,
  hundreds of commits and and tens of thousands of lines changed.

• Future work
  Several systems remain broken or unstable in the wake of this large change.
  Notable systems include:

  • Asset Loaders
  • Collision Detector
  • Shadow Mapping
  • Transparent and Emissive materials

  These systems are slated to be finished in priority order as listed here.

v1.6.2

• Change color from a 16-bit palette index to 16-bit HSV encoded color
  Handling color in the same way MagicaVoxel does (by using a fixed 255 element palette)
  turned out to be a bad idea for several reasons. Firstly, if you import
  multiple models from different sources, and those sources have more than 255
  combined colors, what do you do? I opted to go for a 16-bit palette, but at
  that point you might as well just do 16-bit color and be done with it. So,
  that's what I did here.

  This change ended up being pretty sweeping, affecting world gen, asset/level
  loading, particle systems, rendering.

• Implement HSV color picker

• Implement a Modal widget in the UI

• Change "SeedBrushWithSelection" to work with destructive brush previews

• Fix a race condition sometimes causing the main thread to unlock before the render thread had completed

• Fix some edge cases with negatively offset brushes and small selection regions

• Fix v3_u8 normal data being incorrectly encoded on negative edge

• Fix a frame of lag when changing the selection region with a brush active

v1.6.2-rc1

Changelog

v1.6.1

• Change untextured_3d_geometry_buffer to have a data_type Type
  This is a pretty major internal change. untextured_3d_geometry_buffers can
  now switch between 32-bit float and 8-bit int. When loading a model from
  disk, the size of the model is taken into account and the appropriate type
  is chosen dynamically.

  This required a fair rewrite of the low-level vertex buffering routines to
  be able to handle the permutations. This is still somewhat of a WIP; I
  didn't finish all of them and the ones not done fall-back to the most
  general routine, which is the slowest. I expect I'll rewrite/delete this
  code again before performance in that code is the biggest problem I have.

• Change world updates to happen in serial on a dedicated thread
  Previous to this, multiple world updates could be in-flight at any given
  time, which if overlapping would stomp on each other. Serializing updates
  onto a single thread ensures that:
  a) Updates are correct
  b) Updates happen in the order they were issued
  c) Chunks affected by multiple updates on a single frame only have their
  mesh rebuilt once

  (c) is accomplished by completing all queued update jobs and then issuing
  the chunk mesh rebuild jobs, one per unique chunk updated.

• Add ability to create new assets from a world selection and serialize .chunk assets to disk
  The selection region is saved out as a world_chunk

• Add ability to load asset .chunks from disk

• Add abiltiy to render AssetType_WorldChunk

• Fix layered brush previews crashing the program

• Fix crash in AreEqual if null pointers were passed

• Fixed failing tests, added test runs to CI/CD

v1.6.1-rc5

Mark chunk_flag as a @bitfield

v1.6.1-rc4

Fix a couple failed assertions

v1.6.1-rc3

Remove an annoying assertion

V1.6.1-rc2

Merge pull request #74 from scallyw4g/develop

Develop