[PATCH] D23833: [Polly] Introduce unittests.

[Tobias Grosser via llvm-commits]

grosser accepted this revision.
grosser added a comment.
This revision is now accepted and ready to land.

Very cool. I have been looking forward to this since a while!


================
Comment at: CMakeLists.txt:178
@@ -174,3 +177,3 @@
 # Add target to check formatting of polly files
-file( GLOB_RECURSE files *.h lib/*.cpp lib/*.c tools/*.cpp tools/*.c tools/*.h)
+file( GLOB_RECURSE files *.h lib/*.cpp lib/*.c tools/*.cpp tools/*.c tools/*.h unittests/*.cpp)
 file( GLOB_RECURSE jsonfiles lib/JSON/*.h lib/JSON/*.cpp)
----------------
Good idea to check formatting here as well.

================
Comment at: unittests/Isl/IslTest.cpp:1
@@ +1,2 @@
+//===- DeLICMTest.cpp ----------------------------------------------------===//
+//
----------------
DeLICMTest.cpp - This seems to be a leftover?

================
Comment at: unittests/Isl/IslTest.cpp:27
@@ +26,3 @@
+    isl_val_free(IslNOne);
+  }
+
----------------
	​
22		​  {
23		​    APInt APNOne(32, -1, true);
24		​    auto *IslNOne = isl_valFromAPInt(IslCtx, APNOne, false);

with 'false' in valFromAPInt might also be an interesting test case.

================
Comment at: unittests/Isl/IslTest.cpp:61
@@ +60,3 @@
+    // FIXME: The canonical representation of a negative APInt is two's
+    // complement.
+    EXPECT_EQ(APNOne, 1);
----------------
Very interesting observation. Do you have an idea why this is working today? Or does this indeed result in a bug that can be triggered externally?


https://reviews.llvm.org/D23833