[ubsan] Change "Type mismatch in operation" trap reason to "Alignment, null, or object-size error"

[thurstond]

I originally proposed this rewording when trap reasons were introduced in #145967 (comment). This was not adopted because there was a counter-proposal to split the enum; however, that work appears to have stalled (#151243). In the meantime, there has been an additional datapoint that the current wording is confusing to users. Thus, let's reword it now to prevent further confusion.

[llvmbot]

@llvm/pr-subscribers-clang

Author: Thurston Dang (thurstond)

Changes

I originally proposed this rewording when trap reasons were introduced in #145967 (comment). At the time, there was a counter-proposal to split the enum; however, that work appears to have stalled (#151243). In the meantime, there has been an additional datapoint that the current wording is confusing to users. Thus, let's reword it now to prevent further confusion.

---

Full diff: https://github.com/llvm/llvm-project/pull/169752.diff

2 Files Affected:

• (modified) clang/lib/CodeGen/SanitizerHandler.h (+1-1)
• (modified) clang/test/DebugInfo/Generic/ubsan-trap-reason-type-mismatch.c (+1-1)

diff --git a/clang/lib/CodeGen/SanitizerHandler.h b/clang/lib/CodeGen/SanitizerHandler.h
index a66e7ab354eb2..ff932a9c452e1 100644
--- a/clang/lib/CodeGen/SanitizerHandler.h
+++ b/clang/lib/CodeGen/SanitizerHandler.h
@@ -64,7 +64,7 @@
   SANITIZER_CHECK(SubOverflow, sub_overflow, 0,                                \
                   "Integer subtraction overflowed")                            \
   SANITIZER_CHECK(TypeMismatch, type_mismatch, 1,                              \
-                  "Type mismatch in operation")                                \
+                  "Alignment, null and/or object-size error")                  \
   SANITIZER_CHECK(AlignmentAssumption, alignment_assumption, 0,                \
                   "Alignment assumption violated")                             \
   SANITIZER_CHECK(                                                             \
diff --git a/clang/test/DebugInfo/Generic/ubsan-trap-reason-type-mismatch.c b/clang/test/DebugInfo/Generic/ubsan-trap-reason-type-mismatch.c
index 802ec91b53a0d..a4c4c8cd8db2f 100644
--- a/clang/test/DebugInfo/Generic/ubsan-trap-reason-type-mismatch.c
+++ b/clang/test/DebugInfo/Generic/ubsan-trap-reason-type-mismatch.c
@@ -6,4 +6,4 @@ int type_mismatch(int *p) { return *p; }
 // CHECK-LABEL: @type_mismatch
 // CHECK: call void @llvm.ubsantrap(i8 22) {{.*}}!dbg [[LOC:![0-9]+]]
 // CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
-// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Type mismatch in operation"
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Alignment, null and/or object-size error"

[llvmbot]

@llvm/pr-subscribers-clang-codegen

Author: Thurston Dang (thurstond)

Changes

I originally proposed this rewording when trap reasons were introduced in #145967 (comment). At the time, there was a counter-proposal to split the enum; however, that work appears to have stalled (#151243). In the meantime, there has been an additional datapoint that the current wording is confusing to users. Thus, let's reword it now to prevent further confusion.

---

Full diff: https://github.com/llvm/llvm-project/pull/169752.diff

2 Files Affected:

• (modified) clang/lib/CodeGen/SanitizerHandler.h (+1-1)
• (modified) clang/test/DebugInfo/Generic/ubsan-trap-reason-type-mismatch.c (+1-1)

diff --git a/clang/lib/CodeGen/SanitizerHandler.h b/clang/lib/CodeGen/SanitizerHandler.h
index a66e7ab354eb2..ff932a9c452e1 100644
--- a/clang/lib/CodeGen/SanitizerHandler.h
+++ b/clang/lib/CodeGen/SanitizerHandler.h
@@ -64,7 +64,7 @@
   SANITIZER_CHECK(SubOverflow, sub_overflow, 0,                                \
                   "Integer subtraction overflowed")                            \
   SANITIZER_CHECK(TypeMismatch, type_mismatch, 1,                              \
-                  "Type mismatch in operation")                                \
+                  "Alignment, null and/or object-size error")                  \
   SANITIZER_CHECK(AlignmentAssumption, alignment_assumption, 0,                \
                   "Alignment assumption violated")                             \
   SANITIZER_CHECK(                                                             \
diff --git a/clang/test/DebugInfo/Generic/ubsan-trap-reason-type-mismatch.c b/clang/test/DebugInfo/Generic/ubsan-trap-reason-type-mismatch.c
index 802ec91b53a0d..a4c4c8cd8db2f 100644
--- a/clang/test/DebugInfo/Generic/ubsan-trap-reason-type-mismatch.c
+++ b/clang/test/DebugInfo/Generic/ubsan-trap-reason-type-mismatch.c
@@ -6,4 +6,4 @@ int type_mismatch(int *p) { return *p; }
 // CHECK-LABEL: @type_mismatch
 // CHECK: call void @llvm.ubsantrap(i8 22) {{.*}}!dbg [[LOC:![0-9]+]]
 // CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
-// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Type mismatch in operation"
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Alignment, null and/or object-size error"

[thurstond]

@anthonyhatran @alazarev FYI (GitHub won't let me add you as reviewers)

[EmployedRussian]

Alignment, null and/or object-size error

May I suggest "Alignment, null, or object-size error" to make it clearer that this could be an error related to any of the three.

[vitalybuka]

Enum is not split because we realized that it increases the binary size. These checks are most expensive, and it will make the impact even worse.

Maybe it's OK to split in "no-merge" mode.

Please wait for @delcypher

[thurstond]

Alignment, null and/or object-size error

May I suggest "Alignment, null, or object-size error" to make it clearer that this could be an error related to any of the three.

Thanks for the suggestion, changed.

[thurstond]

Please wait for @delcypher

Ack

[delcypher]

LGTM. Thanks.