[mlir][python] Propagate error diagnostics when an op couldn't be created. (#169499)

mlir/test/python/ir/operation.py

@@ -5,7 +5,7 @@ import io
 from tempfile import NamedTemporaryFile
 from mlir.ir import *
 from mlir.dialects.builtin import ModuleOp
-from mlir.dialects import arith, func, scf
+from mlir.dialects import arith, func, scf, shape
 from mlir.dialects._ods_common import _cext
 from mlir.extras import types as T
 
@@ -774,6 +774,21 @@ def testKnownOpView():
         print(repr(constant))
 
 
+# CHECK-LABEL: TEST: testFailedGenericOperationCreationReportsError
+@run
+def testFailedGenericOperationCreationReportsError():
+    with Context(), Location.unknown():
+        c0 = shape.const_shape([])
+        c1 = shape.const_shape([1, 2, 3])
+        try:
+            shape.MeetOp.build_generic(operands=[c0, c1])
+        except MLIRError as e:
+            # CHECK: unequal shape cardinality
+            print(e)
+        else:
+            assert False, "Expected exception"
+
+
 # CHECK-LABEL: TEST: testSingleResultProperty
 @run
 def testSingleResultProperty():