Merge branch 'main' into main

Author: epistax

.github/workflows/containers/github-action-ci/Dockerfile

@@ -32,7 +32,7 @@ RUN cmake -B ./build -G Ninja ./llvm \
   -DLLVM_ENABLE_RUNTIMES="compiler-rt" \
   -DCMAKE_INSTALL_PREFIX="$LLVM_SYSROOT" \
   -DLLVM_ENABLE_PROJECTS="bolt;clang;lld;clang-tools-extra" \
-  -DLLVM_DISTRIBUTION_COMPONENTS="lld;compiler-rt;clang-format;scan-build" \
+  -DLLVM_DISTRIBUTION_COMPONENTS="lld;compiler-rt;clang-format;scan-build;llvm-symbolizer" \
   -DCLANG_DEFAULT_LINKER="lld"
 
 RUN ninja -C ./build stage2-clang-bolt stage2-install-distribution && ninja -C ./build install-distribution

.mailmap

@@ -29,7 +29,8 @@
 <[email protected]> <[email protected]>
 <[email protected]> <[email protected]>
 <[email protected]> <[email protected]>
-<[email protected]> <[email protected]>
+Min Hsu <[email protected]> <[email protected]>
+Min Hsu <[email protected]> <[email protected]>
 <[email protected]> <[email protected]> <[email protected]>
 <[email protected]> <[email protected]>
 <[email protected]> <[email protected]>

bolt/lib/Core/BinaryContext.cpp

@@ -663,6 +663,7 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
     const BinaryFunction *TargetBF = getBinaryFunctionContainingAddress(Value);
     if (!TargetBF || !areRelatedFragments(TargetBF, &BF)) {
       LLVM_DEBUG(printEntryDiagnostics(dbgs(), TargetBF));
+      (void)printEntryDiagnostics;
       break;
     }
 
@@ -843,6 +844,7 @@ BinaryContext::getOrCreateJumpTable(BinaryFunction &Function, uint64_t Address,
                                &Function, std::placeholders::_1);
     assert(llvm::all_of(JT->Parents, isSibling) &&
            "cannot re-use jump table of a different function");
+    (void)isSibling;
     if (opts::Verbosity > 2) {
       this->outs() << "BOLT-INFO: multiple fragments access the same jump table"
                    << ": " << *JT->Parents[0] << "; " << Function << '\n';

bolt/lib/Profile/DataAggregator.cpp

@@ -871,18 +871,27 @@ DataAggregator::getFallthroughsInTrace(BinaryFunction &BF,
 
   BinaryContext &BC = BF.getBinaryContext();
 
-  if (!BF.isSimple())
-    return std::nullopt;
-
-  assert(BF.hasCFG() && "can only record traces in CFG state");
-
   // Offsets of the trace within this function.
   const uint64_t From = FirstLBR.To - BF.getAddress();
   const uint64_t To = SecondLBR.From - BF.getAddress();
 
   if (From > To)
     return std::nullopt;
 
+  // Accept fall-throughs inside pseudo functions (PLT/thunks).
+  // This check has to be above BF.empty as pseudo functions would pass it:
+  // pseudo => ignored => CFG not built => empty.
+  // If we return nullopt, trace would be reported as mismatching disassembled
+  // function contents which it is not. To avoid this, return an empty
+  // fall-through list instead.
+  if (BF.isPseudo())
+    return Branches;
+
+  if (!BF.isSimple())
+    return std::nullopt;
+
+  assert(BF.hasCFG() && "can only record traces in CFG state");
+
   const BinaryBasicBlock *FromBB = BF.getBasicBlockContainingOffset(From);
   const BinaryBasicBlock *ToBB = BF.getBasicBlockContainingOffset(To);
 

bolt/test/X86/callcont-fallthru.s

@@ -9,6 +9,7 @@
 # RUN: link_fdata %s %t %t.pa3 PREAGG3
 # RUN: link_fdata %s %t %t.pat PREAGGT1
 # RUN: link_fdata %s %t %t.pat2 PREAGGT2
+# RUN: link_fdata %s %t %t.patplt PREAGGPLT
 
 ## Check normal case: fallthrough is not LP or secondary entry.
 # RUN: llvm-strip --strip-unneeded %t -o %t.strip
@@ -42,6 +43,12 @@
 # RUN: llvm-bolt %t.strip --pa -p %t.pat2 -o %t.out \
 # RUN:   --print-cfg --print-only=main | FileCheck %s --check-prefix=CHECK3
 
+## Check pre-aggregated traces don't report zero-sized PLT fall-through as
+## invalid trace
+# RUN: llvm-bolt %t.strip --pa -p %t.patplt -o %t.out | FileCheck %s \
+# RUN:   --check-prefix=CHECK-PLT
+# CHECK-PLT: traces mismatching disassembled function contents: 0
+
   .globl foo
   .type foo, %function
 foo:
@@ -65,7 +72,10 @@ main:
 	movl	$0x0, -0x4(%rbp)
 	movl	%edi, -0x8(%rbp)
 	movq	%rsi, -0x10(%rbp)
+Ltmp0_br:
 	callq	puts@PLT
+## Check PLT traces are accepted
+# PREAGGPLT: T #Ltmp0_br# #puts@plt# #puts@plt# 3
 ## Target is an external-origin call continuation
 # PREAGG1: B X:0 #Ltmp1# 2 0
 # PREAGGT1: T X:0 #Ltmp1# #Ltmp4_br# 2

bolt/test/link_fdata.py

@@ -8,6 +8,7 @@
 """
 
 import argparse
+import os
 import subprocess
 import sys
 import re
@@ -84,8 +85,16 @@
             exit("ERROR: unexpected input:\n%s" % line)
 
 # Read nm output: <symbol value> <symbol type> <symbol name>
+is_llvm_nm = os.path.basename(args.nmtool) == "llvm-nm"
 nm_output = subprocess.run(
-    [args.nmtool, "--defined-only", args.objfile], text=True, capture_output=True
+    [
+        args.nmtool,
+        "--defined-only",
+        "--special-syms" if is_llvm_nm else "--synthetic",
+        args.objfile,
+    ],
+    text=True,
+    capture_output=True,
 ).stdout
 # Populate symbol map
 symbols = {}

clang-tools-extra/clang-tidy/bugprone/UnintendedCharOstreamOutputCheck.cpp

@@ -7,6 +7,8 @@
 //===----------------------------------------------------------------------===//
 
 #include "UnintendedCharOstreamOutputCheck.h"
+#include "../utils/Matchers.h"
+#include "../utils/OptionsUtils.h"
 #include "clang/AST/Type.h"
 #include "clang/ASTMatchers/ASTMatchFinder.h"
 #include "clang/ASTMatchers/ASTMatchers.h"
@@ -35,10 +37,14 @@ AST_MATCHER(Type, isChar) {
 
 UnintendedCharOstreamOutputCheck::UnintendedCharOstreamOutputCheck(
     StringRef Name, ClangTidyContext *Context)
-    : ClangTidyCheck(Name, Context), CastTypeName(Options.get("CastTypeName")) {
-}
+    : ClangTidyCheck(Name, Context),
+      AllowedTypes(utils::options::parseStringList(
+          Options.get("AllowedTypes", "unsigned char;signed char"))),
+      CastTypeName(Options.get("CastTypeName")) {}
 void UnintendedCharOstreamOutputCheck::storeOptions(
     ClangTidyOptions::OptionMap &Opts) {
+  Options.store(Opts, "AllowedTypes",
+                utils::options::serializeStringList(AllowedTypes));
   if (CastTypeName.has_value())
     Options.store(Opts, "CastTypeName", CastTypeName.value());
 }
@@ -50,13 +56,20 @@ void UnintendedCharOstreamOutputCheck::registerMatchers(MatchFinder *Finder) {
                     // with char / unsigned char / signed char
                     classTemplateSpecializationDecl(
                         hasTemplateArgument(0, refersToType(isChar()))));
+  auto IsDeclRefExprFromAllowedTypes = declRefExpr(to(varDecl(
+      hasType(matchers::matchesAnyListedTypeName(AllowedTypes, false)))));
+  auto IsExplicitCastExprFromAllowedTypes = explicitCastExpr(hasDestinationType(
+      matchers::matchesAnyListedTypeName(AllowedTypes, false)));
   Finder->addMatcher(
       cxxOperatorCallExpr(
           hasOverloadedOperatorName("<<"),
           hasLHS(hasType(hasUnqualifiedDesugaredType(
               recordType(hasDeclaration(cxxRecordDecl(
                   anyOf(BasicOstream, isDerivedFrom(BasicOstream)))))))),
-          hasRHS(hasType(hasUnqualifiedDesugaredType(isNumericChar()))))
+          hasRHS(expr(hasType(hasUnqualifiedDesugaredType(isNumericChar())),
+                      unless(ignoringParenImpCasts(
+                          anyOf(IsDeclRefExprFromAllowedTypes,
+                                IsExplicitCastExprFromAllowedTypes))))))
           .bind("x"),
       this);
 }

clang-tools-extra/clang-tidy/bugprone/UnintendedCharOstreamOutputCheck.h

@@ -30,6 +30,7 @@ class UnintendedCharOstreamOutputCheck : public ClangTidyCheck {
   }
 
 private:
+  const std::vector<StringRef> AllowedTypes;
   const std::optional<StringRef> CastTypeName;
 };
 

clang-tools-extra/clang-tidy/utils/Matchers.cpp

@@ -18,8 +18,9 @@ bool NotIdenticalStatementsPredicate::operator()(
 }
 
 MatchesAnyListedTypeNameMatcher::MatchesAnyListedTypeNameMatcher(
-    llvm::ArrayRef<StringRef> NameList)
-    : NameMatchers(NameList.begin(), NameList.end()) {}
+    llvm::ArrayRef<StringRef> NameList, bool CanonicalTypes)
+    : NameMatchers(NameList.begin(), NameList.end()),
+      CanonicalTypes(CanonicalTypes) {}
 
 MatchesAnyListedTypeNameMatcher::~MatchesAnyListedTypeNameMatcher() = default;
 
@@ -32,7 +33,7 @@ bool MatchesAnyListedTypeNameMatcher::matches(
 
   PrintingPolicy PrintingPolicyWithSuppressedTag(
       Finder->getASTContext().getLangOpts());
-  PrintingPolicyWithSuppressedTag.PrintCanonicalTypes = true;
+  PrintingPolicyWithSuppressedTag.PrintCanonicalTypes = CanonicalTypes;
   PrintingPolicyWithSuppressedTag.SuppressElaboration = true;
   PrintingPolicyWithSuppressedTag.SuppressScope = false;
   PrintingPolicyWithSuppressedTag.SuppressTagKeyword = true;

clang-tools-extra/clang-tidy/utils/Matchers.h

@@ -172,21 +172,28 @@ AST_MATCHER_P(Stmt, isStatementIdenticalToBoundNode, std::string, ID) {
 class MatchesAnyListedTypeNameMatcher
     : public ast_matchers::internal::MatcherInterface<QualType> {
 public:
-  explicit MatchesAnyListedTypeNameMatcher(llvm::ArrayRef<StringRef> NameList);
+  explicit MatchesAnyListedTypeNameMatcher(llvm::ArrayRef<StringRef> NameList,
+                                           bool CanonicalTypes);
   ~MatchesAnyListedTypeNameMatcher() override;
   bool matches(
       const QualType &Node, ast_matchers::internal::ASTMatchFinder *Finder,
       ast_matchers::internal::BoundNodesTreeBuilder *Builder) const override;
 
 private:
   std::vector<llvm::Regex> NameMatchers;
+  bool CanonicalTypes;
 };
 
 // Returns a matcher that matches QualType against a list of provided regular.
 inline ::clang::ast_matchers::internal::Matcher<QualType>
-matchesAnyListedTypeName(llvm::ArrayRef<StringRef> NameList) {
+matchesAnyListedTypeName(llvm::ArrayRef<StringRef> NameList,
+                         bool CanonicalTypes) {
   return ::clang::ast_matchers::internal::makeMatcher(
-      new MatchesAnyListedTypeNameMatcher(NameList));
+      new MatchesAnyListedTypeNameMatcher(NameList, CanonicalTypes));
+}
+inline ::clang::ast_matchers::internal::Matcher<QualType>
+matchesAnyListedTypeName(llvm::ArrayRef<StringRef> NameList) {
+  return matchesAnyListedTypeName(NameList, true);
 }
 
 } // namespace clang::tidy::matchers