[YevheniiKotyrlo]

Stacked on #550; the diff of its own is the last commit.

Problem

context.unconfigure() calls Surface::Unconfigure() whenever a surface is attached, and Dawn rejects that on a surface that is not configured (Surface::Unconfigure: "is not configured"). GPUCanvasContext.unconfigure() is valid in each of these cases, and each one reaches Dawn as a validation error:

• a second unconfigure();
• an unconfigure() on a context that was never configured while its view's surface is attached;
• on Android, an unconfigure() after device.destroy(), which recreates the canvas's surface unconfigured.

Solution

SurfaceInfo tracks whether its surface is configured: set when it configures the surface, cleared when it unconfigures it, adopts or recreates a surface, or detaches one. unconfigure() calls Surface::Unconfigure() only while it is.

Tests

Three cases in tests/cpp/SurfaceInfoTest.cpp: unconfigure twice, unconfigure a context that was never configured, and configure again after unconfigure and present. Without the change the first two fail with the fake's record of Dawn's rejection (Surface.Unconfigure on a surface that is not configured); with it all 29 C++ tests pass.

Verification

yarn test:cpp on Ubuntu 24.04 (g++ 13.3): 29 passed. The Android library's native build (:react-native-webgpu:externalNativeBuildDebug, x86_64) compiles the change, and so does the example's iOS build for the simulator SDK. clang-format and the package's cpplint command leave the two files unchanged. CI's Android e2e step (debug build, Metro with CI=true) on an Android 15 emulator: 219 of 221, as on main; the two failures, the YUV cases of ExternalTextureColorSpace and NativeBufferWrapper, fail on main there too. On devices the whole suite passes, 221 of 221: on a Galaxy S23 (Android 14), and on an iPad Pro 12.9-inch (4th generation, iOS 27.0.1) built and run as CI's iOS job does on a simulator (Debug build, Metro with CI=true). main passes 221 of 221 on both.

[YevheniiKotyrlo]

The three new tests against #550's SurfaceRegistry.h (ctest -R Unconfigure):

1/3 Test #20: SurfaceInfoTest.UnconfiguresTwice .........................................***Failed    0.00 sec
    Which is: { "Surface.Unconfigure on a surface that is not configured" }
2/3 Test #21: SurfaceInfoTest.UnconfiguresAContextThatWasNeverConfigured ................***Failed    0.00 sec
    Which is: { "Surface.Unconfigure on a surface that is not configured" }
3/3 Test #22: SurfaceInfoTest.PresentsOnTheSurfaceAgainOnceConfiguredAfterUnconfigure ...   Passed    0.00 sec
33% tests passed, 2 tests failed out of 3

With the change, the whole C++ suite:

100% tests passed, 0 tests failed out of 29