Description
Expand the qrhi autotest with exploring new (auto)testing approaches.
Something we could never do before is to have the actual graphics implementation "stubbed out", while leaving the API sitting on top functional. This is now possible with the Null backend of QRhi. Any QRhi test application or Qt Quick application can be run with this backend, and while they won't output anything to their window, all QRhi APIs will function in a sane way without the layers on top noticing anything.
This allows testing the QRhi API and perhaps some things in the Qt Quick integration in autotests that can run in CI environments without requiring any actual graphics rendering support.
Note however that meaningful testing needs more thinking. What exactly should the null backend do with data like buffers, textures, draw calls? What exactly do we want to verify here? The real deal is always the backend that renders something for real, let's not get blindsided and focus too much on testing an internal interface instead of the real life behavior.
Attachments
Issue Links
- relates to
-
QTBUG-78971 Extend qrhi autotest exercising APIs with offscreen targets
- Closed
-
QTBUG-78625 Revamp scenegraph (and quick) autotests
- Closed
-
QTBUG-78972 Improve QRhi regression testing - further autotests? Lancelot? etc.
- Closed