[Mlir-commits] [mlir] [mlir] [irdl] Add support for regions in irdl-to-cpp (PR #158540)

[Théo Degioanni]

================
@@ -1,15 +1,29 @@
-// RUN: mlir-opt %s --pass-pipeline="builtin.module(test-irdl-conversion-check)" | FileCheck %s
+// RUN: mlir-opt %s --allow-unregistered-dialect --pass-pipeline="builtin.module(test-irdl-conversion-check)" | FileCheck %s
 // CHECK-LABEL: module {
 module {
-    // CHECK: func.func @test() {
+    // CHECK: func.func @test(%[[test_arg:[^ ]*]]: i1) {
     // CHECK: %[[v0:[^ ]*]] = "test_irdl_to_cpp.bar"() : () -> i32
     // CHECK: %[[v1:[^ ]*]] = "test_irdl_to_cpp.bar"() : () -> i32
     // CHECK: %[[v2:[^ ]*]] = "test_irdl_to_cpp.hash"(%[[v0]], %[[v0]]) : (i32, i32) -> i32
+    // CHECK: scf.if %[[test_arg]]
     // CHECK: return
     // CHECK: }
-    func.func @test() {
+    func.func @test(%test_arg: i1) {
         %0 = "test_irdl_to_cpp.bar"() : () -> i32
         %1 = "test_irdl_to_cpp.beef"(%0, %0) : (i32, i32) -> i32
+        "test_irdl_to_cpp.conditional"(%test_arg) ({
+        ^cond(%test: i1):
+          %3 = "test_irdl_to_cpp.bar"() : () -> i32
+          "fake.terminator"() : () -> ()
+        }, {
+        ^then(%what: i1, %ever: i32):
+          %4 = "test_irdl_to_cpp.bar"() : () -> i32
+          "fake.terminator"() : () -> ()
----------------
Moxinilian wrote:

Do you think it would be possible to use a dummy terminator like one from the TableGen test dialect instead while we don't have trait support? The reason I ask is because one of the points of this test file is to make sure the dialect is correctly registered, which is no longer validated if we allow unregistered dialects.

https://github.com/llvm/llvm-project/pull/158540