[Clang][LLDB] Refactor trap reason demangling out of LLDB and into Clang (llvm#165996)

This patch refactors the trap reason demangling logic in
`lldb_private::VerboseTrapFrameRecognizer::RecognizeFrame` into a new
public function `clang::CodeGen::DemangleTrapReasonInDebugInfo`.

There are two reasons for doing this:

1. In a future patch the logic for demangling needs to be used somewhere
else in LLDB and thus the logic needs refactoring to avoid duplicating
code.
2. The logic for demangling shouldn't really be in LLDB anyway because
it's a Clang implementation detail and thus the logic really belongs
inside Clang, not LLDB.

Unit tests have been added for the new function that demonstrate how to
use the new API.

The function names recognized by VerboseTrapFrameRecognizer are
identical to before. However, this patch isn't NFC because:

* The `lldbTarget` library now links against `clangCodeGen` which it
didn't previously.
* The LLDB logging output is a little different now. The previous code
tried to log failures for an invalid regex pattern and for the
`Regex::match` API not returning the correct number of matches. These
failure conditions are unreachable via unit testing so they have been
made assertions failures inside the `DemangleTrapReasonInDebugInfo`
implementation instead of trying to log them in LLDB.

rdar://163230807

Author: delcypher

clang/include/clang/CodeGen/ModuleBuilder.h

@@ -120,6 +120,23 @@ CodeGenerator *CreateLLVMCodeGen(DiagnosticsEngine &Diags,
                                  llvm::LLVMContext &C,
                                  CoverageSourceInfo *CoverageInfo = nullptr);
 
+namespace CodeGen {
+/// Demangle the artificial function name (\param FuncName) used to encode trap
+/// reasons used in debug info for traps (e.g. __builtin_verbose_trap). See
+/// `CGDebugInfo::CreateTrapFailureMessageFor`.
+///
+/// \param FuncName - The function name to demangle.
+///
+/// \return A std::optional. If demangling succeeds the optional will contain
+/// a pair of StringRefs where the first field is the trap category and the
+/// second is the trap message. These can both be empty. If demangling fails the
+/// optional will not contain a value. Note the returned StringRefs if non-empty
+/// point into the underlying storage for \param FuncName and thus have the same
+/// lifetime.
+std::optional<std::pair<StringRef, StringRef>>
+DemangleTrapReasonInDebugInfo(StringRef FuncName);
+} // namespace CodeGen
+
 } // end namespace clang
 
 #endif

clang/lib/CodeGen/ModuleBuilder.cpp

@@ -23,6 +23,7 @@
 #include "llvm/IR/DataLayout.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/Module.h"
+#include "llvm/Support/FormatVariadic.h"
 #include "llvm/Support/VirtualFileSystem.h"
 #include <memory>
 
@@ -378,3 +379,31 @@ clang::CreateLLVMCodeGen(DiagnosticsEngine &Diags, llvm::StringRef ModuleName,
                                HeaderSearchOpts, PreprocessorOpts, CGO, C,
                                CoverageInfo);
 }
+
+namespace clang {
+namespace CodeGen {
+std::optional<std::pair<StringRef, StringRef>>
+DemangleTrapReasonInDebugInfo(StringRef FuncName) {
+  static auto TrapRegex =
+      llvm::Regex(llvm::formatv("^{0}\\$(.*)\\$(.*)$", ClangTrapPrefix).str());
+  llvm::SmallVector<llvm::StringRef, 3> Matches;
+  std::string *ErrorPtr = nullptr;
+#ifndef NDEBUG
+  std::string Error;
+  ErrorPtr = &Error;
+#endif
+  if (!TrapRegex.match(FuncName, &Matches, ErrorPtr)) {
+    assert(ErrorPtr && ErrorPtr->empty() && "Invalid regex pattern");
+    return {};
+  }
+
+  if (Matches.size() != 3) {
+    assert(0 && "Expected 3 matches from Regex::match");
+    return {};
+  }
+
+  // Returns { Trap Category, Trap Message }
+  return std::make_pair(Matches[1], Matches[2]);
+}
+} // namespace CodeGen
+} // namespace clang

clang/unittests/CodeGen/CMakeLists.txt

@@ -1,6 +1,7 @@
 add_clang_unittest(ClangCodeGenTests
   BufferSourceTest.cpp
   CodeGenExternalTest.cpp
+  DemangleTrapReasonInDebugInfo.cpp
   TBAAMetadataTest.cpp
   CheckTargetFeaturesTest.cpp
   CLANG_LIBS

clang/unittests/CodeGen/DemangleTrapReasonInDebugInfo.cpp

@@ -0,0 +1,67 @@
+//=== unittests/CodeGen/DemangleTrapReasonInDebugInfo.cpp -----------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/CodeGen/ModuleBuilder.h"
+#include "llvm/ADT/StringRef.h"
+#include "gtest/gtest.h"
+
+using namespace clang::CodeGen;
+
+void CheckValidCommon(llvm::StringRef FuncName, const char *ExpectedCategory,
+                      const char *ExpectedMessage) {
+  auto MaybeTrapReason = DemangleTrapReasonInDebugInfo(FuncName);
+  ASSERT_TRUE(MaybeTrapReason.has_value());
+  auto [Category, Message] = MaybeTrapReason.value();
+  ASSERT_STREQ(Category.str().c_str(), ExpectedCategory);
+  ASSERT_STREQ(Message.str().c_str(), ExpectedMessage);
+}
+
+void CheckInvalidCommon(llvm::StringRef FuncName) {
+  auto MaybeTrapReason = DemangleTrapReasonInDebugInfo(FuncName);
+  ASSERT_TRUE(!MaybeTrapReason.has_value());
+}
+
+TEST(DemangleTrapReasonInDebugInfo, Valid) {
+  std::string FuncName(ClangTrapPrefix);
+  FuncName += "$trap category$trap message";
+  CheckValidCommon(FuncName, "trap category", "trap message");
+}
+
+TEST(DemangleTrapReasonInDebugInfo, ValidEmptyCategory) {
+  std::string FuncName(ClangTrapPrefix);
+  FuncName += "$$trap message";
+  CheckValidCommon(FuncName, "", "trap message");
+}
+
+TEST(DemangleTrapReasonInDebugInfo, ValidEmptyMessage) {
+  std::string FuncName(ClangTrapPrefix);
+  FuncName += "$trap category$";
+  CheckValidCommon(FuncName, "trap category", "");
+}
+
+TEST(DemangleTrapReasonInDebugInfo, ValidAllEmpty) {
+  //  `__builtin_verbose_trap` actually allows this
+  // currently. However, we should probably disallow this in Sema because having
+  // an empty category and message completely defeats the point of using the
+  // builtin (#165981).
+  std::string FuncName(ClangTrapPrefix);
+  FuncName += "$$";
+  CheckValidCommon(FuncName, "", "");
+}
+
+TEST(DemangleTrapReasonInDebugInfo, InvalidOnlyPrefix) {
+  std::string FuncName(ClangTrapPrefix);
+  CheckInvalidCommon(FuncName);
+}
+
+TEST(DemangleTrapReasonInDebugInfo, Invalid) {
+  std::string FuncName("foo");
+  CheckInvalidCommon(FuncName);
+}
+
+TEST(DemangleTrapReasonInDebugInfo, InvalidEmpty) { CheckInvalidCommon(""); }

lldb/source/Plugins/LanguageRuntime/CPlusPlus/CMakeLists.txt

@@ -6,6 +6,8 @@ add_lldb_library(lldbPluginCPPRuntime
     lldbCore
     lldbSymbol
     lldbTarget
+  CLANG_LIBS
+    clangCodeGen
 )
 
 add_subdirectory(ItaniumABI)

lldb/source/Plugins/LanguageRuntime/CPlusPlus/VerboseTrapFrameRecognizer.cpp

@@ -95,33 +95,14 @@ VerboseTrapFrameRecognizer::RecognizeFrame(lldb::StackFrameSP frame_sp) {
   if (func_name.empty())
     return {};
 
-  static auto trap_regex =
-      llvm::Regex(llvm::formatv("^{0}\\$(.*)\\$(.*)$", ClangTrapPrefix).str());
-  SmallVector<llvm::StringRef, 3> matches;
-  std::string regex_err_msg;
-  if (!trap_regex.match(func_name, &matches, &regex_err_msg)) {
-    LLDB_LOGF(GetLog(LLDBLog::Unwind),
-              "Failed to parse match trap regex for '%s': %s", func_name.data(),
-              regex_err_msg.c_str());
-
-    return {};
-  }
-
-  // For `__clang_trap_msg$category$message$` we expect 3 matches:
-  // 1. entire string
-  // 2. category
-  // 3. message
-  if (matches.size() != 3) {
-    LLDB_LOGF(GetLog(LLDBLog::Unwind),
-              "Unexpected function name format. Expected '<trap prefix>$<trap "
-              "category>$<trap message>'$ but got: '%s'.",
-              func_name.data());
-
+  auto maybe_trap_reason =
+      clang::CodeGen::DemangleTrapReasonInDebugInfo(func_name);
+  if (!maybe_trap_reason.has_value()) {
+    LLDB_LOGF(GetLog(LLDBLog::Unwind), "Failed to demangle '%s' as trap reason",
+              func_name.str().c_str());
     return {};
   }
-
-  auto category = matches[1];
-  auto message = matches[2];
+  auto [category, message] = maybe_trap_reason.value();
 
   std::string stop_reason =
       category.empty() ? "<empty category>" : category.str();