Added switch statement tests for default and literal values

[samsinsane]

Related to #3058

[CLAassistant]

All committers have signed the CLA.

[kdashg]

fshaderDeclarationInsideSwitchLiteralDefault fails on Firefox Windows with ANGLE, but passes without ANGLE (native gl amd).

[kdashg]

This might be a D3D bug, given the failure is with this seemingly-reasonable hlsl:

// FRAGMENT SHADER BEGIN
// GLSL BEGIN
#version 300 es
precision highp float;
out vec4 my_FragColor;
void main()
{
    my_FragColor = vec4(1, 0, 0, 1);
    switch (0)
    {
        default:
            ivec2 i;
            i = ivec2(1, 0);
            my_FragColor = vec4(0, i[0], 0, 1);
    }
}
// GLSL END
// INITIAL HLSL BEGIN
float4 vec4_ctor(int x0, int x1, int x2, int x3)
{
    return float4(x0, x1, x2, x3);
}
#ifdef ANGLE_ENABLE_LOOP_FLATTEN
#define LOOP [loop]
#define FLATTEN [flatten]
#else
#define LOOP
#define FLATTEN
#endif
#define ATOMIC_COUNTER_ARRAY_STRIDE 4
// Varyings
static float4 out_my_FragColor = {0, 0, 0, 0};
cbuffer DriverConstants : register(b1)
{
};
@@ PIXEL OUTPUT @@
PS_OUTPUT main(@@ PIXEL MAIN PARAMETERS @@){
@@ MAIN PROLOGUE @@
(out_my_FragColor = float4(1.0, 0.0, 0.0, 1.0));
{
int2 _i2324 = {0, 0};
switch (0) {
default:
(_i2324 = int2(1, 0));
(out_my_FragColor = vec4_ctor(0, _i2324[0], 0, 1));
break;
}
}
return generateOutput();
}
// INITIAL HLSL END
// FRAGMENT SHADER END

[kenrussell]

Thanks for the test case and sorry for the long delay reviewing it. Can confirm that the tests pass on macOS, so this looks like a problem specific to ANGLE's D3D11 backend, and yes, potentially a bug in Microsoft's HLSL compiler fxc.exe. Have filed https://bugs.chromium.org/p/angleproject/issues/detail?id=4807 about this and will try to engage the D3D team at Microsoft to help confirm. CC'ing @RafaelCintron .