Skip to content

Latest commit

 

History

History
31 lines (26 loc) · 2.62 KB

ci_testing.md

File metadata and controls

31 lines (26 loc) · 2.62 KB
Raw

CI and testing

BigWheels uses GitHub Actions for its CI and testing solution.

Current testing

The following tests are run as part of every PR submission:

  • On Windows:
    • Full build of BigWheels, all shader targets, and DX12 samples with Visual Studio 2022.
    • Unit tests.
    • Some DX12 samples are run in headless mode using DX12's software renderer.
  • On Linux:
    • Full build of BigWheels, SPIR-V shaders, and Vulkan samples.
    • Unit tests.
    • Some Vulkan samples are run with a virtual framebuffer (using xvfb) and using Mesa's Lavapipe as the software renderer.
  • On Linux, build APKs for mobile.
  • Code formatting check that ensures code is formatted using clang-format, using clang-format version 16.

The samples run as part of the runtime tests produce screenshots that are then uploaded as GitHub artifacts. You can retrieve them by navigating to a successful CI run's page and then downloading the screenshots artifact.

Maintenance

Testing workflows are found under the .github/workflows folder in the repository.

We use the freely-available GitHub-hosted runners. There may be the need of updating dependencies or other settings used by the workflows as GitHub updates their runners, or as BigWheels' needs change.

Common maintenance tasks:

  • Updating the runner's images: we are currently using ubuntu-20.04 and windows-2022.
  • Updating the Vulkan SDK and its components that are downloaded and installed as part of the workflows. This should be a simple change of the version string in the workflows' files:
    • On Windows, both the Vulkan SDK and the RT components are used.
    • On Linux, only the Vulkan SDK is used.
  • Updating the Windows dependencies, mainly the Windows SDK.
  • Maintaining the CMake build flags used to build BigWheels (Linux, Windows).
  • Curating the list of samples that are run (Linux, Windows). Note that we strive to maintain a healthy balance between test coverage and runtime/resource usage minimization of the CI runners.