Faked HWC for SurfaceFlinger testing
Infrastructure and initial port of transaction tests. Faking the HWC
allows exercising the real path through the SurfaceFlinger, not relying
on screen captures. Faked HWC also opens up the possibility of faking
interactions like display hotplugs.
The tests are verifying the composition rectangles instead of a set of
select pixels. GLES rendering differences won't affect the
results. Also, the test expectations become clearer.
The ported transaction tests ran roughly twice as fast when compared
with the original transaction test. This is mostly due to the thighter
control over the vsyncs.
Test: Running the test on Marlin
Change-Id: I1c876cda78db94c1965498af957e64fdd23459ce
diff --git a/services/surfaceflinger/SurfaceFlinger.h b/services/surfaceflinger/SurfaceFlinger.h
index acfad46..5123b58 100644
--- a/services/surfaceflinger/SurfaceFlinger.h
+++ b/services/surfaceflinger/SurfaceFlinger.h
@@ -300,6 +300,7 @@
HdrCapabilities* outCapabilities) const;
virtual status_t enableVSyncInjections(bool enable);
virtual status_t injectVSync(nsecs_t when);
+ virtual status_t getLayerDebugInfo(std::vector<LayerDebugInfo>* outLayers) const;
/* ------------------------------------------------------------------------
@@ -628,6 +629,7 @@
#ifdef USE_HWC2
HWComposer* mRealHwc;
HWComposer* mVrHwc;
+ const std::string mHwcServiceName; // "default" for real use, something else for testing.
#endif
// constant members (no synchronization needed for access)
RenderEngine* mRenderEngine;