[clang] [ubsan][test] Add allow-ubsan-check-divergence.c (PR #164673)

via cfe-commits cfe-commits at lists.llvm.org
Wed Oct 22 10:45:01 PDT 2025


llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-clang

Author: Thurston Dang (thurstond)

<details>
<summary>Changes</summary>

This testcase shows that adding a ubsan check and then removing it during the LowerAllowCheck pass does not entirely undo the effects of adding the check.

---
Full diff: https://github.com/llvm/llvm-project/pull/164673.diff


1 Files Affected:

- (added) clang/test/CodeGen/allow-ubsan-check-divergence.c (+30) 


``````````diff
diff --git a/clang/test/CodeGen/allow-ubsan-check-divergence.c b/clang/test/CodeGen/allow-ubsan-check-divergence.c
new file mode 100644
index 0000000000000..a21d4f693269b
--- /dev/null
+++ b/clang/test/CodeGen/allow-ubsan-check-divergence.c
@@ -0,0 +1,30 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 6
+
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -emit-llvm -O1 -o - %s \
+// RUN:     | FileCheck %s --check-prefixes=CLEAN-O1
+
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -emit-llvm -O1 -o - %s \
+// RUN:     -fsanitize=signed-integer-overflow \
+// RUN:     -fsanitize-skip-hot-cutoff=signed-integer-overflow=1.0 \
+// RUN:     -fallow-runtime-check-skip-hot-cutoff=1.0 \
+// RUN:     | FileCheck %s --check-prefixes=UBSAN-O1
+
+// This test shows that -fsanitize-skip-hot-cutoff=...=1.0 plus
+// -fallow-runtime-check-skip-hot-cutoff=1.0 does not perfectly undo the
+// effects of -fsanitize.
+
+// CLEAN-O1-LABEL: define dso_local i32 @overflow(
+// CLEAN-O1-SAME: i32 noundef [[X:%.*]], i32 noundef [[Y:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// CLEAN-O1-NEXT:  [[ENTRY:.*:]]
+// CLEAN-O1-NEXT:    [[ADD:%.*]] = add nsw i32 [[Y]], [[X]]
+// CLEAN-O1-NEXT:    ret i32 [[ADD]]
+//
+// UBSAN-O1-LABEL: define dso_local noundef i32 @overflow(
+// UBSAN-O1-SAME: i32 noundef [[X:%.*]], i32 noundef [[Y:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// UBSAN-O1-NEXT:  [[ENTRY:.*:]]
+// UBSAN-O1-NEXT:    [[TMP0:%.*]] = add i32 [[X]], [[Y]]
+// UBSAN-O1-NEXT:    ret i32 [[TMP0]]
+//
+int overflow(int x, int y) {
+  return x + y;
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/164673


More information about the cfe-commits mailing list