Do loop fix (#209)

* Bug fix: emit logic for `do` loops

This case was never tested, and I was outputting some garbage characters. This comit fixes the codegen and adds a test case.

* Bug fix: make sure to pass through `[allow_uav_condition]`

This also fixes the standard library definition of `IncrementCounter()` so that it returns a `uint` instead of `void`.

1 files changed, 35 insertions, 0 deletions

diff --git a/tests/bugs/do-loop.hlsl b/tests/bugs/do-loop.hlsl
new file mode 100644
index 000000000..eea62e92a
--- /dev/null
+++ b/tests/bugs/do-loop.hlsl
@@ -0,0 +1,35 @@
+//TEST:COMPARE_HLSL: -profile vs_5_0 -target dxbc-assembly
+
+// Check output for `do` loops
+
+cbuffer C : register(b0)
+{
+	int n;
+};
+
+float4 main() : SV_Position
+{
+	float4 x = 0;
+	int i = n;
+	{
+		do
+		{
+			x = (x + (float)1) * x;
+
+			// Note(tfoley): The "right" thing here would be
+			// `i--`, but that leads to a subtle difference
+			// in the final code between just invokeing `fxc`
+			// and invoking it on the Slang-generated output
+			// (despite the generated HLSL for this line being
+			// identical, modulo some `#line` directives).
+			//
+			// I'm using a binary operator that will yield
+			// the same code with its operands swapped, just
+			// to work around it. A better long-term fix
+			// is to have this test be an end-to-end test
+			// that we execute.
+			i = i*i;
+		} while((bool) i);
+	}
+    return x;
+}