[mlir] Fix ambiguity when building with Clang 14.0.6

[aganea]

Summary

When building with clang-cl 14.0.6, the previous "hack" to select the overload, does not seem to be enough:

C:\git\llvm-project>ninja check-mlir -C stage1
ninja: Entering directory `stage1'
[1/3] Building CXX object tools\mlir\unittests\Analysis\Pr...iles\MLIRPresburgerTests.dir\IntegerPolyhedronTest.cpp.obj
FAILED: tools/mlir/unittests/Analysis/Presburger/CMakeFiles/MLIRPresburgerTests.dir/IntegerPolyhedronTest.cpp.obj
C:\PROGRA~1\LLVM\bin\clang-cl.EXE  /nologo -TP -DGTEST_HAS_RTTI=0 -DMLIR_CUDA_CONVERSIONS_ENABLED=1 -DMLIR_INCLUDE_TESTS -DMLIR_ROCM_CONVERSIONS_ENABLED=1 -DUNICODE -D_CRT_NONSTDC_NO_DEPRECATE -D_CRT_NONSTDC_NO_WARNINGS -D_CRT_SECURE_NO_DEPRECATE -D_CRT_SECURE_NO_WARNINGS -D_HAS_EXCEPTIONS=0 -D_SCL_SECURE_NO_DEPRECATE -D_SCL_SECURE_NO_WARNINGS -D_UNICODE -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS -IC:\git\llvm-project\stage1\tools\mlir\unittests\Analysis\Presburger -IC:\git\llvm-project\mlir\unittests\Analysis\Presburger -IC:\git\llvm-project\stage1\include -IC:\git\llvm-project\llvm\include -IC:\git\llvm-project\mlir\include -IC:\git\llvm-project\stage1\tools\mlir\include -IC:\git\llvm-project\llvm\utils\unittest\googletest\include -IC:\git\llvm-project\llvm\utils\unittest\googlemock\include /GS- /D_ITERATOR_DEBUG_LEVEL=0 /arch:AVX /Zc:inline /Zc:__cplusplus /Zi -gcodeview-ghash /Oi /Brepro /bigobj /permissive- /W4  -Wextra -Wno-unused-parameter -Wwrite-strings -Wcast-qual -Wmissing-field-initializers -Wimplicit-fallthrough -Wcovered-switch-default -Wno-noexcept-type -Wnon-virtual-dtor -Wdelete-non-virtual-dtor -Wsuggest-override -Wstring-conversion -Wmisleading-indentation -Wctad-maybe-unsupported /Gw -Werror=mismatched-tags /MT /O2 /Ob2  -Wno-variadic-macros -Wno-gnu-zero-variadic-macro-arguments /EHs-c- /GR- -UNDEBUG -Wno-suggest-override -std:c++17 /showIncludes /Fotools\mlir\unittests\Analysis\Presburger\CMakeFiles\MLIRPresburgerTests.dir\IntegerPolyhedronTest.cpp.obj /Fdtools\mlir\unittests\Analysis\Presburger\CMakeFiles\MLIRPresburgerTests.dir\ -c -- C:\git\llvm-project\mlir\unittests\Analysis\Presburger\IntegerPolyhedronTest.cpp
C:\git\llvm-project\mlir\unittests\Analysis\Presburger\IntegerPolyhedronTest.cpp(1484,21): error: call to member function 'containsPointNoLocal' is ambiguous
  EXPECT_TRUE(poly3.containsPointNoLocal({-0, 0}));
              ~~~~~~^~~~~~~~~~~~~~~~~~~~
C:\git\llvm-project\llvm\utils\unittest\googletest\include\gtest/gtest.h(1968,23): note: expanded from macro 'EXPECT_TRUE'
  GTEST_TEST_BOOLEAN_(condition, #condition, false, true, \
                      ^~~~~~~~~
C:\git\llvm-project\llvm\utils\unittest\googletest\include\gtest/internal/gtest-internal.h(1329,34): note: expanded from macro 'GTEST_TEST_BOOLEAN_'
      ::testing::AssertionResult(expression)) \
                                 ^~~~~~~~~~
C:\git\llvm-project\mlir\include\mlir/Analysis/Presburger/IntegerRelation.h(380,3): note: candidate function
  containsPointNoLocal(ArrayRef<MPInt> point) const;
  ^
C:\git\llvm-project\mlir\include\mlir/Analysis/Presburger/IntegerRelation.h(382,3): note: candidate function
  containsPointNoLocal(ArrayRef<int64_t> point) const {
  ^
1 error generated.
ninja: build stopped: subcommand failed.

[ftynse]

Explicit conversion is almost always better than relying on hacks, especially the clever ones, that may not be obvious to anyone or even the author in retrospect. With the conversion, the comment isn't really necessary while adding the two-line comment to explain the hack defies the purpose of using the hack for brevity.

[arjunp]

In D134219#3801949, @ftynse wrote:

Explicit conversion is almost always better than relying on hacks, especially the clever ones, that may not be obvious to anyone or even the author in retrospect. With the conversion, the comment isn't really necessary while adding the two-line comment to explain the hack defies the purpose of using the hack for brevity.

Yes, this was a mistake and shouldn't have made it in. We'll keep it in mind for next time.

mlir/unittests/Analysis/Presburger/IntegerPolyhedronTest.cpp

TEST(IntegerPolyhedronTest, containsPointNoLocal) {
   IntegerPolyhedron poly2 = parseIntegerPolyhedron(
       "(x) : (x - 2*(x floordiv 2) == 0, x - 4*(x floordiv 4) - 2 == 0)");
   EXPECT_TRUE(containsPointNoLocal(poly2, {6}));
   EXPECT_FALSE(containsPointNoLocal(poly2, {4}));
 
   IntegerPolyhedron poly3 =
       parseIntegerPolyhedron("(x, y) : (2*x - y >= 0, y - 3*x >= 0)");
 
-  // -0 instead of 0 to prevent unwanted conversion to pointer types,
-  // which would lead to ambiguity in overload resolution.
-  EXPECT_TRUE(poly3.containsPointNoLocal({-0, 0}));
+  EXPECT_TRUE(poly3.containsPointNoLocal(ArrayRef<int64_t>({0, 0})));
   EXPECT_FALSE(poly3.containsPointNoLocal({1, 0}));
 }
 
 TEST(IntegerPolyhedronTest, truncateEqualityRegressionTest) {
   // IntegerRelation::truncate was truncating inequalities to the number of
   // equalities.
   IntegerRelation set(PresburgerSpace::getSetSpace(1));
   IntegerRelation::CountsSnapshot snapshot = set.getCounts();
   set.addEquality({1, 0});
   set.truncate(snapshot);
   EXPECT_EQ(set.getNumEqualities(), 0u);
 }