[clang] [CIR] Add the ability to detect if SwitchOp covers all the cases (PR #171246)

[Andy Kaylor via cfe-commits]

================
@@ -36,3 +36,33 @@ cir.func @s0() {
 // CHECK-NEXT: }
 // CHECK-NEXT: cir.yield
 // CHECK-NEXT: }
+
+
+// Pretends that this is lowered from a C file and was tagged with allEnumCasesCovered = true
+cir.func @s1(%1 : !s32i) {
+  cir.switch (%1 : !s32i) {
+    cir.case (default, []) {
+      cir.return
+    }
+    cir.case (equal, [#cir.int<1> : !s32i]) {
+      cir.yield
+    }
+    cir.case (equal, [#cir.int<2> : !s32i]) {
+      cir.yield
+    }
+    cir.yield
+  } { allEnumCasesCovered = true}
+  cir.return
+} 
+// CHECK: cir.switch (%arg0 : !s32i) {
+// CHECK-NEXT: cir.case(default, []) {
+// CHECK-NEXT:   cir.return
+// CHECK-NEXT: }
+// CHECK-NEXT: cir.case(equal, [#cir.int<1> : !s32i]) {
+// CHECK-NEXT:   cir.yield
+// CHECK-NEXT: }
+// CHECK-NEXT: cir.case(equal, [#cir.int<2> : !s32i]) {
+// CHECK-NEXT:   cir.yield
+// CHECK-NEXT: }
+// CHECK-NEXT: cir.yield
+// CHECK-NEXT: } {allEnumCasesCovered = true}
----------------
andykaylor wrote:

I expected the attribute to be printed before the region, but I'm honestly not sure if there's an MLIR standard for that. I guess since you didn't add any code to print it this must be the default behavior.

@xlauko Does this look right to you?

https://github.com/llvm/llvm-project/pull/171246