[analyzer] Syntactic matcher for leaks associated with run loop and autoreleasepool

[george.karpenkov]

Summary

A checker for detecting leaks resulting from allocating temporary
autoreleasing objects before starting the main run loop.

Checks for two antipatterns:
1. ObjCMessageExpr followed by [[NARunLoop mainRunLoop] run] in the same
autorelease pool.
2. ObjCMessageExpr followed by [[NARunLoop mainRunLoop] run] in no
autorelease pool.

Happens-before relationship is modeled purely syntactically.

rdar://39299145

[NoQ]

clang/lib/StaticAnalyzer/Checkers/RunLoopAutoreleaseLeakChecker.cpp

61

*makes a joke about QuadBoolTy and runs away*

89

I wonder if we could make a reusable matcher out of this: hasDescendantsInOrder(A, ..., B).

90

This memoization map is short-lived. I don't quite understand how would we ever encounter the same statement twice during our traversal, could you give an example? Like, you're traversing the AST in a manner in which RecursiveASTVisitors usually do that, and i'm pretty sure that they encounter every statement just once, and you throw away the map immediately after that. What am i missing?

92

So is the optional here only to add an assertion? It's essentially a non-optional in release mode(?) Maybe just assert in the callee that we never return None?

189–190

That'd match non-functions and C++ methods and stuff like that.

In the non-matcher world i'd have checked FunctionDecl::isMain() here, but i don't think we have a matcher for that, but i guess we may add it.

clang/test/Analysis/html_diagnostics/relevant_lines/macros_same_file.c

16 ↗

(On Diff #156198)

I guess this is accidental.

[NoQ]

clang/lib/StaticAnalyzer/Checkers/RunLoopAutoreleaseLeakChecker.cpp

89

Aha, i guess you're already discussing it on the mailing list, i didn't intend to block the patch on that.

[george.karpenkov]

clang/lib/StaticAnalyzer/Checkers/RunLoopAutoreleaseLeakChecker.cpp

90

i think we've discussed that some statements can still be DAGs (e.g. elvis operator)

[NoQ]

Looks good!

clang/lib/StaticAnalyzer/Checkers/RunLoopAutoreleaseLeakChecker.cpp

90

Yeah, apparently it has as many as 4 (!!) sub-expressions, 3 of which partially overlap. Still not sure it's worth the memoization, but let's add a comment explaining why we did that.

[dcoughlin]

It is great to see a check for this!

clang/include/clang/StaticAnalyzer/Checkers/Checkers.td

572

Mentioning "loops" three times here doesn't seem right.

"Check for leaked memory in autorelease pools that will never be popped"

clang/lib/StaticAnalyzer/Checkers/RunLoopAutoreleaseLeakChecker.cpp

12

"autoreleasing" --> "autoreleased"

19

I think it would be good to explain why this pattern may lead to leaks/abandoned memory.

Something like: "Any temporary objects autoreleased in code called in the expression will not be deallocated until the program exits and so are effectively leaks."

133

Should "mainRunLoop" really be camelCase? Does "mainRunLoop" make sense when the call is to xpc_main?

Why not just "the main run loop"?

134

The comma here should be a semi-colon to match our other diagnostics: "may never get released; consider moving them to a".

[dcoughlin]

clang/lib/StaticAnalyzer/Checkers/RunLoopAutoreleaseLeakChecker.cpp

15

Also note the typo: "NARunLoop" --> "NSRunLoop"

clang/include/clang/StaticAnalyzer/Checkers/Checkers.td

 def ObjCPropertyChecker : Checker<"ObjCProperty">,
   HelpText<"Check for proper uses of Objective-C properties">,
   DescFile<"ObjCPropertyChecker.cpp">;
 
 } // end "osx"
 
 let ParentPackage = Cocoa in {
 
+def RunLoopAutoreleaseLeakChecker : Checker<"RunLoopAutoreleaseLeak">,
+  HelpText<"Check for detecting memory loops from autorelease loops wrapping run loops">,
+  DescFile<"RunLoopAutoreleaseLeakChecker.cpp">;
+
 def ObjCAtSyncChecker : Checker<"AtSync">,
   HelpText<"Check for nil pointers used as mutexes for @synchronized">,
   DescFile<"ObjCAtSyncChecker.cpp">;
 
 def NilArgChecker : Checker<"NilArg">,
   HelpText<"Check for prohibited nil arguments to ObjC method calls">,
   DescFile<"BasicObjCFoundationChecks.cpp">;
 

clang/lib/StaticAnalyzer/Checkers/CMakeLists.txt

add_clang_library(clangStaticAnalyzerCheckers
   ObjCUnusedIVarsChecker.cpp
   PaddingChecker.cpp
   PointerArithChecker.cpp
   PointerSubChecker.cpp
   PthreadLockChecker.cpp
   RetainCountChecker.cpp
   ReturnPointerRangeChecker.cpp
   ReturnUndefChecker.cpp
+  RunLoopAutoreleaseLeakChecker.cpp
   SimpleStreamChecker.cpp
   StackAddrEscapeChecker.cpp
   StdLibraryFunctionsChecker.cpp
   StreamChecker.cpp
   TaintTesterChecker.cpp
   TestAfterDivZeroChecker.cpp
   TraversalChecker.cpp
   TrustNonnullChecker.cpp

clang/lib/StaticAnalyzer/Checkers/RunLoopAutoreleaseLeakChecker.cpp

+//=- RunLoopAutoreleaseLeakChecker.cpp --------------------------*- C++ -*-==//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//
+//===----------------------------------------------------------------------===//
+//
+// A checker for detecting leaks resulting from allocating temporary
+// autoreleasing objects before starting the main run loop.
+//
+// Checks for two antipatterns:
+// 1. ObjCMessageExpr followed by [[NARunLoop mainRunLoop] run] in the same
+// autorelease pool.
+// 2. ObjCMessageExpr followed by [[NARunLoop mainRunLoop] run] in no
+// autorelease pool.
+//
+//===----------------------------------------------------------------------===//
+//
+
+#include "ClangSACheckers.h"
+#include "clang/AST/Decl.h"
+#include "clang/AST/DeclObjC.h"
+#include "clang/ASTMatchers/ASTMatchFinder.h"
+#include "clang/StaticAnalyzer/Core/BugReporter/BugReporter.h"
+#include "clang/StaticAnalyzer/Core/BugReporter/BugType.h"
+#include "clang/StaticAnalyzer/Core/Checker.h"
+#include "clang/StaticAnalyzer/Core/CheckerManager.h"
+#include "clang/StaticAnalyzer/Core/PathSensitive/CallEvent.h"
+#include "clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h"
+#include "clang/StaticAnalyzer/Core/PathSensitive/ExprEngine.h"
+
+using namespace clang;
+using namespace ento;
+using namespace ast_matchers;
+
+namespace {
+
+const char * RunLoopBind = "NSRunLoopM";
+const char * RunLoopRunBind = "RunLoopRunM";
+const char * OtherMsgBind = "OtherMessageSentM";
+const char * AutoreleasePoolBind = "AutoreleasePoolM";
+
+class RunLoopAutoreleaseLeakChecker : public Checker<
+                                      check::ASTCodeBody> {
+
+public:
+  void checkASTCodeBody(const Decl *D,
+                        AnalysisManager &AM,
+                        BugReporter &BR) const;
+
+};
+
+} // end anonymous namespace
+
+
+using TriBoolTy = Optional<bool>;
+using MemoizationMapTy = llvm::DenseMap<const Stmt *, Optional<TriBoolTy>>;
+
+static TriBoolTy
+seenBeforeRec(const Stmt *Parent, const Stmt *A, const Stmt *B,
+              MemoizationMapTy &Memoization) {
+  for (const Stmt *C : Parent->children()) {
+    if (C == A)
+      return true;
+
+    if (C == B)
+      return false;
+
+    Optional<TriBoolTy> &Cached = Memoization[C];
+    if (!Cached)
+      Cached = seenBeforeRec(C, A, B, Memoization);
+
+    if (Cached->hasValue())
+      return Cached->getValue();
+  }
+
+  return None;
+}
+
+/// \return Whether {@code A} occurs before {@code B} in traversal of
+/// {@code Parent}.
+/// Conceptually a very incomplete/unsound approximation of happens-before
+/// relationship (A is likely to be evaluated before B),
+/// but useful enough in this case.
+static bool seenBefore(const Stmt *Parent, const Stmt *A, const Stmt *B) {
+  MemoizationMapTy Memoization;
+  TriBoolTy Val = seenBeforeRec(Parent, A, B, Memoization);
+  return Val.getValue();
+}
+
+static void emitDiagnostics(BoundNodes &Match,
+                            const Decl *D,
+                            BugReporter &BR,
+                            AnalysisManager &AM,
+                            const RunLoopAutoreleaseLeakChecker *Checker) {
+
+  assert(D->hasBody());
+  const Stmt *DeclBody = D->getBody();
+
+  AnalysisDeclContext *ADC = AM.getAnalysisDeclContext(D);
+
+  const auto *ME = Match.getNodeAs<ObjCMessageExpr>(OtherMsgBind);
+  assert(ME);
+
+  const auto *AP =
+      Match.getNodeAs<ObjCAutoreleasePoolStmt>(AutoreleasePoolBind);
+  bool HasAutoreleasePool = (AP != nullptr);
+
+  const auto *RLR = Match.getNodeAs<Stmt>(RunLoopRunBind);
+  assert(RLR && "Run loop launch not found");
+
+  assert(ME != RLR);
+  if (HasAutoreleasePool && seenBefore(AP, RLR, ME))
+    return;
+
+  if (!HasAutoreleasePool && seenBefore(DeclBody, RLR, ME))
+    return;
+
+  PathDiagnosticLocation Location = PathDiagnosticLocation::createBegin(
+    ME, BR.getSourceManager(), ADC);
+  SourceRange Range = ME->getSourceRange();
+
+  BR.EmitBasicReport(ADC->getDecl(), Checker,
+                     /*Name=*/"Memory leak inside autorelease pool",
+                     /*Category=*/"Memory",
+                     /*Name=*/
+                     (Twine("Temporary objects allocated in the") +
+                      " autorelease pool " +
+                      (HasAutoreleasePool ? "" : "of last resort ") +
+                      "followed by the launch of mainRunLoop "
+                      "may never get released, consider moving them to a "
+                      "separate autorelease pool")
+                         .str(),
+                     Location, Range);
+}
+
+static StatementMatcher getRunLoopRunM(StatementMatcher Extra = anything()) {
+  StatementMatcher MainRunLoopM =
+      objcMessageExpr(hasSelector("mainRunLoop"),
+                      hasReceiverType(asString("NSRunLoop")),
+                      Extra)
+          .bind(RunLoopBind);
+
+  StatementMatcher MainRunLoopRunM = objcMessageExpr(hasSelector("run"),
+                         hasReceiver(MainRunLoopM),
+                         Extra).bind(RunLoopRunBind);
+
+  StatementMatcher XPCRunM =
+      callExpr(callee(functionDecl(hasName("xpc_main")))).bind(RunLoopRunBind);
+  return anyOf(MainRunLoopRunM, XPCRunM);
+}
+
+static StatementMatcher getOtherMessageSentM(StatementMatcher Extra = anything()) {
+  return objcMessageExpr(unless(anyOf(equalsBoundNode(RunLoopBind),
+                                      equalsBoundNode(RunLoopRunBind))),
+                         Extra)
+      .bind(OtherMsgBind);
+}
+
+static void
+checkTempObjectsInSamePool(const Decl *D, AnalysisManager &AM, BugReporter &BR,
+                           const RunLoopAutoreleaseLeakChecker *Chkr) {
+  StatementMatcher RunLoopRunM = getRunLoopRunM();
+  StatementMatcher OtherMessageSentM = getOtherMessageSentM();
+
+  StatementMatcher RunLoopInAutorelease =
+      autoreleasePoolStmt(
+        hasDescendant(RunLoopRunM),
+        hasDescendant(OtherMessageSentM)).bind(AutoreleasePoolBind);
+
+  DeclarationMatcher GroupM = decl(hasDescendant(RunLoopInAutorelease));
+
+  auto Matches = match(GroupM, *D, AM.getASTContext());
+  for (BoundNodes Match : Matches)
+    emitDiagnostics(Match, D, BR, AM, Chkr);
+}
+
+static void
+checkTempObjectsInNoPool(const Decl *D, AnalysisManager &AM, BugReporter &BR,
+                         const RunLoopAutoreleaseLeakChecker *Chkr) {
+
+  auto NoPoolM = unless(hasAncestor(autoreleasePoolStmt()));
+
+  StatementMatcher RunLoopRunM = getRunLoopRunM(NoPoolM);
+  StatementMatcher OtherMessageSentM = getOtherMessageSentM(NoPoolM);
+
+  DeclarationMatcher GroupM = functionDecl(
+    isMain(),
+    hasDescendant(RunLoopRunM),
+    hasDescendant(OtherMessageSentM)
+  );
+
+  auto Matches = match(GroupM, *D, AM.getASTContext());
+
+  for (BoundNodes Match : Matches)
+    emitDiagnostics(Match, D, BR, AM, Chkr);
+
+}
+
+void RunLoopAutoreleaseLeakChecker::checkASTCodeBody(const Decl *D,
+                        AnalysisManager &AM,
+                        BugReporter &BR) const {
+  checkTempObjectsInSamePool(D, AM, BR, this);
+  checkTempObjectsInNoPool(D, AM, BR, this);
+}
+
+void ento::registerRunLoopAutoreleaseLeakChecker(CheckerManager &mgr) {
+  mgr.registerChecker<RunLoopAutoreleaseLeakChecker>();
+}

clang/test/Analysis/Checkers/RunLoopAutoreleaseLeakChecker.m

+// UNSUPPORTED: system-windows
+// RUN: %clang_analyze_cc1 -fobjc-arc -analyzer-checker=core,osx.cocoa.RunLoopAutoreleaseLeak %s -triple x86_64-darwin -verify
+// RUN: %clang_analyze_cc1 -DEXTRA=1 -DAP1=1 -fobjc-arc -analyzer-checker=core,osx.cocoa.RunLoopAutoreleaseLeak %s -triple x86_64-darwin -verify
+// RUN: %clang_analyze_cc1 -DEXTRA=1 -DAP2=1 -fobjc-arc -analyzer-checker=core,osx.cocoa.RunLoopAutoreleaseLeak %s -triple x86_64-darwin -verify
+// RUN: %clang_analyze_cc1 -DEXTRA=1 -DAP3=1 -fobjc-arc -analyzer-checker=core,osx.cocoa.RunLoopAutoreleaseLeak %s -triple x86_64-darwin -verify
+// RUN: %clang_analyze_cc1 -DEXTRA=1 -DAP4=1 -fobjc-arc -analyzer-checker=core,osx.cocoa.RunLoopAutoreleaseLeak %s -triple x86_64-darwin -verify
+
+#include "../Inputs/system-header-simulator-for-objc-dealloc.h"
+
+#ifndef EXTRA
+
+void just_runloop() { // No warning: no statements in between
+  @autoreleasepool {
+    [[NSRunLoop mainRunLoop] run]; // no-warning
+  }
+}
+
+void just_xpcmain() { // No warning: no statements in between
+  @autoreleasepool {
+    xpc_main(); // no-warning
+  }
+}
+
+void runloop_init_before() { // Warning: object created before the loop.
+  @autoreleasepool {
+    NSObject *object = [[NSObject alloc] init]; // expected-warning{{Temporary objects allocated in the autorelease pool followed by the launch of mainRunLoop may never get released, consider moving them to a separate autorelease pool}}
+    (void) object;
+    [[NSRunLoop mainRunLoop] run]; 
+  }
+}
+
+void xpcmain_init_before() { // Warning: object created before the loop.
+  @autoreleasepool {
+    NSObject *object = [[NSObject alloc] init]; // expected-warning{{Temporary objects allocated in the autorelease pool followed by the launch of mainRunLoop may never get released, consider moving them to a separate autorelease pool}}
+    (void) object;
+    xpc_main(); 
+  }
+}
+
+void runloop_init_before_two_objects() { // Warning: object created before the loop.
+  @autoreleasepool {
+    NSObject *object = [[NSObject alloc] init]; // expected-warning{{Temporary objects allocated in the autorelease pool followed by the launch of mainRunLoop may never get released, consider moving them to a separate autorelease pool}}
+    NSObject *object2 = [[NSObject alloc] init]; // no-warning, warning on the first one is enough.
+    (void) object;
+    (void) object2;
+    [[NSRunLoop mainRunLoop] run]; 
+  }
+}
+
+void runloop_no_autoreleasepool() {
+  NSObject *object = [[NSObject alloc] init]; // no-warning
+  (void)object;
+  [[NSRunLoop mainRunLoop] run];
+}
+
+void runloop_init_after() { // No warning: objects created after the loop
+  @autoreleasepool {
+    [[NSRunLoop mainRunLoop] run]; 
+    NSObject *object = [[NSObject alloc] init]; // no-warning
+    (void) object;
+  }
+}
+
+#endif
+
+#ifdef AP1
+int main() {
+    NSObject *object = [[NSObject alloc] init]; // expected-warning{{Temporary objects allocated in the autorelease pool of last resort followed by the launch of mainRunLoop may never get released, consider moving them to a separate autorelease pool}}
+    (void) object;
+    [[NSRunLoop mainRunLoop] run]; 
+    return 0;
+}
+#endif
+
+#ifdef AP2
+// expected-no-diagnostics
+int main() {
+  NSObject *object = [[NSObject alloc] init]; // no-warning
+  (void) object;
+  @autoreleasepool {
+    [[NSRunLoop mainRunLoop] run]; 
+  }
+  return 0;
+}
+#endif
+
+#ifdef AP3
+// expected-no-diagnostics
+int main() {
+    [[NSRunLoop mainRunLoop] run];
+    NSObject *object = [[NSObject alloc] init]; // no-warning
+    (void) object;
+    return 0;
+}
+#endif
+
+#ifdef AP4
+int main() {
+    NSObject *object = [[NSObject alloc] init]; // expected-warning{{Temporary objects allocated in the autorelease pool of last resort followed by the launch of mainRunLoop may never get released, consider moving them to a separate autorelease pool}}
+    (void) object;
+    xpc_main();
+    return 0;
+}
+#endif

clang/test/Analysis/Inputs/system-header-simulator-for-objc-dealloc.h

 - (void)dealloc;
 - (id)init;
 - (id)retain;
 - (oneway void)release;
 @end
 
 @interface NSRunLoop : NSObject
 + (NSRunLoop *)currentRunLoop;
++ (NSRunLoop *)mainRunLoop;
+- (void) run;
 - (void)cancelPerformSelectorsWithTarget:(id)target;
 @end
 
 @interface NSNotificationCenter : NSObject
 + (NSNotificationCenter *)defaultCenter;
 - (void)removeObserver:(id)observer;
 @end
 
 typedef struct objc_selector *SEL;
 
 void _Block_release(const void *aBlock);
 #define Block_release(...) _Block_release((const void *)(__VA_ARGS__))
 
 @interface CIFilter : NSObject
 @end
+
+extern void xpc_main(void);