This replaces test.yml with ci.yml and renames the test to draco-ci.
A Github PR with this all put together exists at the following link:
https://github.com/google/draco/pull/806
An example draco-ci test summary exists at this link:
https://github.com/google/draco/actions/runs/1789061363
- Add the following test configurations in a job named
draco-tests.
test-macos-make-release-shared
test-macos-make-release-static
test-macos-xcode-release-shared
test-macos-xcode-release-static
test-ubuntu-make-release-shared
test-ubuntu-make-release-static
test-windows-msvc-release-shared
test-windows-msvc-release-static
test-windows-make-release-shared
test-windows-make-release-static
Each configuration name is intended to be self documenting.
All configurations build Draco with tests enabled, and then
run the tests.
- Add the following test configurations in a job named
draco-install-tests:
install-test-ubuntu-make
install-test-macos-make
install-test-macos-xcode
install-test-windows-make
install-test-windows-msvc
Each configuration runs test.py in verbose mode to allow for
failure diagnostics.
Some additional changes to make all this work (and improve usability):
- draco-config.cmake now defines DRACO_LIBRARY_DLL and DRACO_FOUND.
- DRACO_LIBRARY_DLL will contain the full path to the Draco DLL on systems that
produce DLL files (aka Windows).
- test.py now streams output in verbose mode
Super minor, but I was looking at the repo and noticed the logo was blurry, poked around a bit and found the svg version already in the repo, so why not try and use it for the README?
Updates require_var() and set_variable_if_unset() to use the
environment values for named variables.
CMake processes toolchain files multiple times, and on some
processing runs cache variables are not available. Environment
variables are always available.
* Improved kD-tree based point cloud encoding
* Now applicable to point clouds with any number of attributes
* Support for all integer attribute types and quantized floating point types
* Improved mesh compression up to 10% (on average ~2%)
* For meshes, the 1.3.0 bitstream is fully compatible with 1.2.x decoders
* Improved Javascript API
* Added support for all signed and unsigned integer types
* Added support for point clouds to our Javascript encoder API
* Added support for integer properties to the PLY decoder
* Bug fixes
* Up to 20% faster decoding
* Added support for integer attributes to our Javascript Encoder
* Fixed issues with THREE.DracoLoader not releasing memory associated
with the Draco module
* OBJ decoder can now be used to parse pure point clouds
* Added Unity plugins to support runtime loading and design-time
importing of encoded Draco files
* Javascript and WebAssembly decoder size reduced by 35%
* Added specialized Javascript and Webassembly decoders for GLTF (size
reduction about 50% compared to the previous version)
The latest version of Draco brings a number of new compression
enhancements for even smaller models:
- Improved compression
- Up to 40% better compression of normals
- Up to 5% better compression for models with multiple attributes
- Faster decode speeds
- 30% faster decoding for models with multiple attributes for lower compression levels 4 and below
- Note: Decreases compression by 10%.
- Encoding of metadata to .obj (e.g. Draco can preserve material or sub-object names)
- Security fixes
-Increase version to 1.0.1
-Update Javascript decode timing example
-Fix async creation of DracoDecoderModule
-fixes wasm parallel decoding
-Introduce base class for geometric normal predictor
-Added integer attribute interface to JavaScript decoder
-Move loading Draco decoder into three.js Draco loader
-Add compiler launcher support to CMake build
-Add sanitizer support via CMake build
-Check we are not reading past the end of the attribute buffer
