Unit tests to improve code coverage with IR test too, also for the test in request D118234

[arindamsharma8]

Summary

Adds a batch of C tests that have been found to cover several hundred
lines of Clang/LLVM that are not covered by the unit and regression
tests of the main LLVM project, nor by the test suite when run with the
-O3 configuration.

The tests were originally generated using our fuzzer, and were then reduced
using C-Reduce. These test have been checked for undefined behaviour-freedom
using Frama-C and code sanitisers, manually checked to eliminate
implementation-defined behaviour.

Most of the new coverage achieved by these 7 C code tests is in:
llvm-project/clang/lib/AST (2 testcases)
llvm-project/clang/lib/Sema (2 testcase)
llvm-project/llvm/lib/Analysis (1 testcase)
llvm-project/llvm/lib/CodeGen (4 testcases)
llvm-project/llvm/lib/Transforms(1 testcase)

For the 13 IR tests (of the .c tests here and in request D118234):
llvm-project/llvm/lib/Transforms/ (4 testcases)
llvm-project/llvm/lib/IR/ (1 testcase)
llvm-project/llvm/lib/Support/ (2 testcases)
llvm-project/clang/lib/AST (5 testcases)
llvm-project/clang/lib/Sema (2 testcase)
llvm-project/llvm/lib/CodeGen (4 testcases)

[afd]

I reviewed the C programs, and they look good except for one request for change and one query.

For the cases where there are C programs it makes sense to me that IR tests are also contributed.

What I'm unsure about is why IR tests are contributed without C programs. Are these the IR tests corresponding to previously-contributed C test cases?

SingleSource/UnitTests/testcase-Float2Int.c

3

Could the blank space be changed to a numeric value? To understand the code below, one needs to know what a blank space char value looks like when cast to double.

SingleSource/UnitTests/testcase-SemaStmt-1.c

5

Is it definitely OK for __builtin_expect to use a side-effecting operation as its second argument?

If so then this looks fine.

But if it would be OK for an implementation to #define the __builtin_expect expression to nothing, then this test would yield a different result, so I'm a bit unsure.

[karineek]

@lenary: I assume the .ll files shall go to the tests in the llvm project (e.g. llvm-project/tree/main/llvm/test/CodeGen/X86), right?
Also, how shall we build these? (we did not use emit-llvm flag, but maybe should have?)

SingleSource/UnitTests/testcase-Float2Int.c

3

Yes, it is not changing the coverage. I will edit it to 32.

SingleSource/UnitTests/testcase-SemaStmt-1.c

5

When moving the ++d outside (and just using d), we don't get the coverage. I think I will remove this test and maybe later we will find a test that covers that function.

[lenary]

In D142638#4202443, @karineek wrote:

@lenary: I assume the .ll files shall go to the tests in the llvm project (e.g. llvm-project/tree/main/llvm/test/CodeGen/X86), right?
Also, how shall we build these? (we did not use emit-llvm flag, but maybe should have?)

If you're going to put the llvm ir tests in-tree, then they should not rely on being linked/executed as they currently are. But it does sound like a better place for them given they are x86-specific, rather than target-independent (while LLVM IR semantics are portable, the IR generated by clang for different targets ends up being different for a variety of reasons).

I don't work on the x86 backend, so you'd have to work out a better person to review the test additions there.

[arindamsharma8]

Addresses @afd comments and update testcases accordingly.
Thanks for the feedback @afd.
@lenary we will move the IR tetscases into the appropriate location and make them either target independent or have a separate for each target x-86, ARM etc.

[arindamsharma8]

We are closing this review request due to the added IR tests being mixed with the original C single source tests. We will create 2 separate requests for these.

Arindam

SingleSource/UnitTests/testcase-CodeGen-1.c

 // COVERAGE: llvm-project/llvm/lib/Analysis/InlineCost.cpp (1 new functions)
-// COVERAGE: llvm-project/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp (2 new functions)
+// COVERAGE: llvm-project/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp (2 new functions) 
 // COVERAGE: llvm-project/llvm/lib/CodeGen/TargetLoweringBase.cpp (2 new functions)
 extern void abort();
 #ifndef __SIZEOF_INT128__
 #define __int128 long long // Appears in test case of pr61375.c
 #endif
 unsigned __int128 a;
 void b() { a = (double)a + 0; }
 int main() {
   b();
   if (a)
     abort();
   return 0;
 }

SingleSource/UnitTests/testcase-CodeGen-2.c

-// COVERAGE: llvm-project/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp (1 new functions)
+// COVERAGE: llvm-project/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp (1 new functions) 
 // COVERAGE: llvm-project/llvm/lib/CodeGen/TargetLoweringBase.cpp (1 new functions)
 extern void abort();
 #ifndef __SIZEOF_INT128__
 #define __int128 long long // Appears in test case of pr61375.c
 #endif
 float a;
 int main() {
   __int128 b = a;
   if (b + (double)10)
     ++b;
   while (b != 1)
     abort();
   return 0;
 }

SingleSource/UnitTests/testcase-CodeGen-3.c

 // COVERAGE: llvm-project/clang/lib/AST/ExprConstant.cpp (1 new functions)
-// COVERAGE: llvm-project/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp (1 new functions)
+// COVERAGE: llvm-project/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp (1 new functions) 
 // COVERAGE: llvm-project/llvm/lib/CodeGen/TargetLoweringBase.cpp (1 new functions)
 #include "stdio.h"
 #ifdef __SIZEOF_INT128__
-typedef __int128 L;  // Appears in pr49218.c test case
+typedef __int128 L; // Appears in pr49218.c test case
 #else
 typedef long long L;
 #endif
 float f;
 int main() {
   L i = f;
   if (i <= 10 +
-                41 * +((long)(((int)10) >> ((int)10))) *
-                    ((long)(((int)(i)) % ((int)10))) -
-                ((long)(((double)(i)) + ((double)10))) -
-                ((long)(((int)10) ^ ((int)10))))
     do {
       ++i;
-      printf("%lld", (long long) i);
+      printf("%lld", (long long)i);
     } while (i != 11);
   return 0;
 }

SingleSource/UnitTests/testcase-Float2Int.c

 // COVERAGE: llvm-project/llvm/lib/Transforms/Scalar/Float2Int.cpp (1 new functions)
 #include "stdio.h"
-char b = ' ';
+char b = 32;
 int main() {
   for (b = ((long)(((double)(b)) * ((double)(b)))); b >= -29; b--)
     printf("%d", (int)b);
   printf("\n");
   return 0;
 }

SingleSource/UnitTests/testcase-SelectionDAG.c

+// COVERAGE: llvm-project/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp (1 new functions)
+int a;
+int main() {
+  int b = 0;
+  for (; a; a = 2 - (0 | a)) {
+    b++;
+    if (b <= 4)
+      continue;
+    break;
+  }
+  return b;
+}

SingleSource/UnitTests/testcase-SemaStmt-1-IR.ll

-; ModuleID = 'testcase-SemaStmt-1.c'
-source_filename = "testcase-SemaStmt-1.c"
-target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
-target triple = "x86_64-pc-linux-gnu"
-
-@d = dso_local local_unnamed_addr global i32 10, align 4
-@b = dso_local local_unnamed_addr global i32 0, align 4
-@c = dso_local local_unnamed_addr global i32 0, align 4
-
-; Function Attrs: nounwind uwtable
-define dso_local i32 @main() local_unnamed_addr #0 {
-  store i32 0, ptr @b, align 4, !tbaa !5
-  %1 = load i32, ptr @d, align 4, !tbaa !5
-  %2 = add nsw i32 %1, 1
-  store i32 %2, ptr @d, align 4, !tbaa !5
-  store i32 %2, ptr @c, align 4, !tbaa !5
-  %3 = icmp eq i32 %2, 11
-  br i1 %3, label %5, label %4
-
-4:                                                ; preds = %0
-  tail call void @abort() #2
-  unreachable
-
-5:                                                ; preds = %0
-  ret i32 0
-}
-
-; Function Attrs: noreturn nounwind
-declare void @abort() local_unnamed_addr #1
-
-attributes #0 = { nounwind uwtable "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
-attributes #1 = { noreturn nounwind "frame-pointer"="none" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
-attributes #2 = { noreturn nounwind }
-
-!llvm.module.flags = !{!0, !1, !2, !3}
-!llvm.ident = !{!4}
-
-!0 = !{i32 1, !"wchar_size", i32 4}
-!1 = !{i32 7, !"PIC Level", i32 2}
-!2 = !{i32 7, !"PIE Level", i32 2}
-!3 = !{i32 7, !"uwtable", i32 2}
-!4 = !{!"Ubuntu clang version 15.0.0-++20220502063808+e547a333a433-1~exp1~20220502183820.463"}
-!5 = !{!6, !6, i64 0}
-!6 = !{!"int", !7, i64 0}
-!7 = !{!"omnipotent char", !8, i64 0}
-!8 = !{!"Simple C/C++ TBAA"}

SingleSource/UnitTests/testcase-SemaStmt-1.c

-// COVERAGE: llvm-project/clang/lib/Sema/SemaStmt.cpp (1 new functions)
-// Compilation of this code should give unused value warning on return value of function declared with const attribute
-int b, c, d = 10;
-int main(void) {
-  __builtin_expect(1 ? b = 0 : 0, d++); // __builtin_expect used in other PRs test cases: like pr43385.c and pr85156.c
-  
-  c = d;
-  if (d != 11)
-    __builtin_abort();
-  return 0;
-}

SingleSource/UnitTests/testcase-SimplifyIndVar.c

+// COVERAGE: llvm-project/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp (1 new functions)
+short a;
+void foo() {
+  if (a < 0)
+    __builtin_abort();
+}
+int main() {
+  int b = 8;
+  while (a++ <= 0) {
+    b++;
+    if (b <= 0 + 1 * (b % 40) * (double)0 - 0 - 0 -
+                 ((040) * ((b))) * ((((int)(b)) << ((int)(b)))) +
+                 ((long)(((double)(b)) - ((double)0))) +
+                 ((long)(((double)0) - ((double)(b)))) -
+                 ((short)(((double)(b)) - ((double)0))) +
+                 ((long)(((int)0) ^ ((int)0))) -
+                 ((long)(((double)(b)) + ((double)0))))
+      break;
+  }
+  foo();
+  if (b != 9)
+    __builtin_abort();
+  return 0;
+}

SingleSource/UnitTests/testcase-StmtPrinter.c

+// COVERAGE: llvm-project/clang/lib/AST/StmtPrinter.cpp (1 new functions)
+char a, b, g, h, i;
+short c, d, k;
+int e, f, j;
+void l() {
+  k = i > 0 || d > 1 >> d << i;
+  for (; j; j++)
+    h = c < 0 || h || c << (8 + 2 - (long)0);
+  g = f + 1 + 2 * 0 * 0;
+  for (; f < 1; f = 1 + 2 - 0)
+    b = a + 1 + 2 * 0;
+  if (b == 0)
+    __builtin_abort();
+}
+int main() {
+  for (; e < 1; e++)
+    l();
+  if (h != 0)
+    __builtin_abort();
+  return 0;
+}

SingleSource/UnitTests/testcase-StmtPrinter.reference_output

+exit 0

SingleSource/UnitTests/testcase-clang-lib-1.c

-// COVERAGE: llvm-project/clang/include/clang/AST/EvaluatedExprVisitor.h (1 new functions)
-// COVERAGE: llvm-project/clang/lib/AST/ComputeDependence.cpp (1 new functions)
+// COVERAGE: llvm-project/clang/include/clang/AST/EvaluatedExprVisitor.h (1 new functions) 
+// COVERAGE: llvm-project/clang/lib/AST/ComputeDependence.cpp (1 new functions) 
 // COVERAGE: llvm-project/clang/lib/AST/Expr.cpp (2 new functions)
 // COVERAGE: llvm-project/clang/lib/CodeGen/CGExprScalar.cpp (1 new functions)
 // COVERAGE: llvm-project/clang/lib/Parse/ParseExpr.cpp (1 new functions)
 // COVERAGE: llvm-project/clang/lib/Sema/SemaExpr.cpp (2 new functions)
 #include <stdio.h>
 typedef int (*fptr)(int);
 int foo(int i) { return i; }
 int main() {
   int b = _Generic(foo, fptr : 3, int : 4);
   printf("%d\n", b);
   return 0;
 }