llvm-project/clang/lib/StaticAnalyzer/Checkers/CheckObjCDealloc.cpp

263 lines
8.6 KiB
C++
Raw Normal View History

//==- CheckObjCDealloc.cpp - Check ObjC -dealloc implementation --*- C++ -*-==//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
2008-07-12 04:53:14 +08:00
// This file defines a CheckObjCDealloc, a checker that
// analyzes an Objective-C class's implementation to determine if it
// correctly implements -dealloc.
//
//===----------------------------------------------------------------------===//
#include "clang/StaticAnalyzer/Checkers/LocalCheckers.h"
#include "clang/StaticAnalyzer/Core/BugReporter/PathDiagnostic.h"
#include "clang/StaticAnalyzer/Core/BugReporter/BugReporter.h"
#include "clang/AST/ExprObjC.h"
#include "clang/AST/Expr.h"
#include "clang/AST/DeclObjC.h"
#include "clang/Basic/LangOptions.h"
#include "llvm/Support/raw_ostream.h"
using namespace clang;
using namespace ento;
static bool scan_dealloc(Stmt* S, Selector Dealloc) {
if (ObjCMessageExpr* ME = dyn_cast<ObjCMessageExpr>(S))
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. llvm-svn: 101972
2010-04-21 08:45:42 +08:00
if (ME->getSelector() == Dealloc) {
switch (ME->getReceiverKind()) {
case ObjCMessageExpr::Instance: return false;
case ObjCMessageExpr::SuperInstance: return true;
case ObjCMessageExpr::Class: break;
case ObjCMessageExpr::SuperClass: break;
}
}
// Recurse to children.
for (Stmt::child_iterator I = S->child_begin(), E= S->child_end(); I!=E; ++I)
if (*I && scan_dealloc(*I, Dealloc))
return true;
return false;
}
static bool scan_ivar_release(Stmt* S, ObjCIvarDecl* ID,
const ObjCPropertyDecl* PD,
Selector Release,
IdentifierInfo* SelfII,
ASTContext& Ctx) {
// [mMyIvar release]
if (ObjCMessageExpr* ME = dyn_cast<ObjCMessageExpr>(S))
if (ME->getSelector() == Release)
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. llvm-svn: 101972
2010-04-21 08:45:42 +08:00
if (ME->getInstanceReceiver())
if (Expr* Receiver = ME->getInstanceReceiver()->IgnoreParenCasts())
if (ObjCIvarRefExpr* E = dyn_cast<ObjCIvarRefExpr>(Receiver))
if (E->getDecl() == ID)
return true;
// [self setMyIvar:nil];
if (ObjCMessageExpr* ME = dyn_cast<ObjCMessageExpr>(S))
Overhaul the AST representation of Objective-C message send expressions, to improve source-location information, clarify the actual receiver of the message, and pave the way for proper C++ support. The ObjCMessageExpr node represents four different kinds of message sends in a single AST node: 1) Send to a object instance described by an expression (e.g., [x method:5]) 2) Send to a class described by the class name (e.g., [NSString method:5]) 3) Send to a superclass class (e.g, [super method:5] in class method) 4) Send to a superclass instance (e.g., [super method:5] in instance method) Previously these four cases where tangled together. Now, they have more distinct representations. Specific changes: 1) Unchanged; the object instance is represented by an Expr*. 2) Previously stored the ObjCInterfaceDecl* referring to the class receiving the message. Now stores a TypeSourceInfo* so that we know how the class was spelled. This both maintains typedef information and opens the door for more complicated C++ types (e.g., dependent types). There was an alternative, unused representation of these sends by naming the class via an IdentifierInfo *. In practice, we either had an ObjCInterfaceDecl *, from which we would get the IdentifierInfo *, or we fell into the case below... 3) Previously represented by a class message whose IdentifierInfo * referred to "super". Sema and CodeGen would use isStr("super") to determine if they had a send to super. Now represented as a "class super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). 4) Previously represented by an instance message whose receiver is a an ObjCSuperExpr, which Sema and CodeGen would check for via isa<ObjCSuperExpr>(). Now represented as an "instance super" send, where we have both the location of the "super" keyword and the ObjCInterfaceDecl* of the superclass we're targetting (statically). Note that ObjCSuperExpr only has one remaining use in the AST, which is for "super.prop" references. The new representation of ObjCMessageExpr is 2 pointers smaller than the old one, since it combines more storage. It also eliminates a leak when we loaded message-send expressions from a precompiled header. The representation also feels much cleaner to me; comments welcome! This patch attempts to maintain the same semantics we previously had with Objective-C message sends. In several places, there are massive changes that boil down to simply replacing a nested-if structure such as: if (message has a receiver expression) { // instance message if (isa<ObjCSuperExpr>(...)) { // send to super } else { // send to an object } } else { // class message if (name->isStr("super")) { // class send to super } else { // send to class } } with a switch switch (E->getReceiverKind()) { case ObjCMessageExpr::SuperInstance: ... case ObjCMessageExpr::Instance: ... case ObjCMessageExpr::SuperClass: ... case ObjCMessageExpr::Class:... } There are quite a few places (particularly in the checkers) where send-to-super is effectively ignored. I've placed FIXMEs in most of them, and attempted to address send-to-super in a reasonable way. This could use some review. llvm-svn: 101972
2010-04-21 08:45:42 +08:00
if (ME->getInstanceReceiver())
if (Expr* Receiver = ME->getInstanceReceiver()->IgnoreParenCasts())
if (DeclRefExpr* E = dyn_cast<DeclRefExpr>(Receiver))
if (E->getDecl()->getIdentifier() == SelfII)
if (ME->getMethodDecl() == PD->getSetterMethodDecl() &&
ME->getNumArgs() == 1 &&
ME->getArg(0)->isNullPointerConstant(Ctx,
Expr::NPC_ValueDependentIsNull))
return true;
// self.myIvar = nil;
if (BinaryOperator* BO = dyn_cast<BinaryOperator>(S))
if (BO->isAssignmentOp())
if (ObjCPropertyRefExpr* PRE =
dyn_cast<ObjCPropertyRefExpr>(BO->getLHS()->IgnoreParenCasts()))
if (PRE->isExplicitProperty() && PRE->getExplicitProperty() == PD)
if (BO->getRHS()->isNullPointerConstant(Ctx,
Expr::NPC_ValueDependentIsNull)) {
// This is only a 'release' if the property kind is not
// 'assign'.
return PD->getSetterKind() != ObjCPropertyDecl::Assign;;
}
// Recurse to children.
for (Stmt::child_iterator I = S->child_begin(), E= S->child_end(); I!=E; ++I)
if (*I && scan_ivar_release(*I, ID, PD, Release, SelfII, Ctx))
return true;
return false;
}
void ento::CheckObjCDealloc(const ObjCImplementationDecl* D,
const LangOptions& LOpts, BugReporter& BR) {
assert (LOpts.getGCMode() != LangOptions::GCOnly);
ASTContext& Ctx = BR.getContext();
const ObjCInterfaceDecl* ID = D->getClassInterface();
// Does the class contain any ivars that are pointers (or id<...>)?
// If not, skip the check entirely.
// NOTE: This is motivated by PR 2517:
// http://llvm.org/bugs/show_bug.cgi?id=2517
bool containsPointerIvar = false;
for (ObjCInterfaceDecl::ivar_iterator I=ID->ivar_begin(), E=ID->ivar_end();
I!=E; ++I) {
ObjCIvarDecl* ID = *I;
QualType T = ID->getType();
if (!T->isObjCObjectPointerType() ||
ID->getAttr<IBOutletAttr>() || // Skip IBOutlets.
ID->getAttr<IBOutletCollectionAttr>()) // Skip IBOutletCollections.
continue;
containsPointerIvar = true;
break;
}
if (!containsPointerIvar)
return;
// Determine if the class subclasses NSObject.
IdentifierInfo* NSObjectII = &Ctx.Idents.get("NSObject");
IdentifierInfo* SenTestCaseII = &Ctx.Idents.get("SenTestCase");
for ( ; ID ; ID = ID->getSuperClass()) {
IdentifierInfo *II = ID->getIdentifier();
if (II == NSObjectII)
break;
// FIXME: For now, ignore classes that subclass SenTestCase, as these don't
// need to implement -dealloc. They implement tear down in another way,
// which we should try and catch later.
// http://llvm.org/bugs/show_bug.cgi?id=3187
if (II == SenTestCaseII)
return;
}
if (!ID)
return;
// Get the "dealloc" selector.
IdentifierInfo* II = &Ctx.Idents.get("dealloc");
Selector S = Ctx.Selectors.getSelector(0, &II);
ObjCMethodDecl* MD = 0;
// Scan the instance methods for "dealloc".
for (ObjCImplementationDecl::instmeth_iterator I = D->instmeth_begin(),
E = D->instmeth_end(); I!=E; ++I) {
if ((*I)->getSelector() == S) {
MD = *I;
break;
}
}
if (!MD) { // No dealloc found.
const char* name = LOpts.getGCMode() == LangOptions::NonGC
? "missing -dealloc"
: "missing -dealloc (Hybrid MM, non-GC)";
std::string buf;
llvm::raw_string_ostream os(buf);
os << "Objective-C class '" << D << "' lacks a 'dealloc' instance method";
BR.EmitBasicReport(name, os.str(), D->getLocStart());
return;
}
// dealloc found. Scan for missing [super dealloc].
if (MD->getBody() && !scan_dealloc(MD->getBody(), S)) {
const char* name = LOpts.getGCMode() == LangOptions::NonGC
? "missing [super dealloc]"
: "missing [super dealloc] (Hybrid MM, non-GC)";
std::string buf;
llvm::raw_string_ostream os(buf);
os << "The 'dealloc' instance method in Objective-C class '" << D
<< "' does not send a 'dealloc' message to its super class"
" (missing [super dealloc])";
BR.EmitBasicReport(name, os.str(), D->getLocStart());
return;
}
// Get the "release" selector.
IdentifierInfo* RII = &Ctx.Idents.get("release");
Selector RS = Ctx.Selectors.getSelector(0, &RII);
// Get the "self" identifier
IdentifierInfo* SelfII = &Ctx.Idents.get("self");
// Scan for missing and extra releases of ivars used by implementations
// of synthesized properties
for (ObjCImplementationDecl::propimpl_iterator I = D->propimpl_begin(),
E = D->propimpl_end(); I!=E; ++I) {
// We can only check the synthesized properties
if ((*I)->getPropertyImplementation() != ObjCPropertyImplDecl::Synthesize)
continue;
ObjCIvarDecl* ID = (*I)->getPropertyIvarDecl();
if (!ID)
continue;
QualType T = ID->getType();
if (!T->isObjCObjectPointerType()) // Skip non-pointer ivars
continue;
const ObjCPropertyDecl* PD = (*I)->getPropertyDecl();
if (!PD)
continue;
// ivars cannot be set via read-only properties, so we'll skip them
if (PD->isReadOnly())
continue;
// ivar must be released if and only if the kind of setter was not 'assign'
bool requiresRelease = PD->getSetterKind() != ObjCPropertyDecl::Assign;
if (scan_ivar_release(MD->getBody(), ID, PD, RS, SelfII, Ctx)
!= requiresRelease) {
const char *name;
const char* category = "Memory (Core Foundation/Objective-C)";
std::string buf;
llvm::raw_string_ostream os(buf);
if (requiresRelease) {
name = LOpts.getGCMode() == LangOptions::NonGC
? "missing ivar release (leak)"
: "missing ivar release (Hybrid MM, non-GC)";
os << "The '" << ID
<< "' instance variable was retained by a synthesized property but "
"wasn't released in 'dealloc'";
} else {
name = LOpts.getGCMode() == LangOptions::NonGC
? "extra ivar release (use-after-release)"
: "extra ivar release (Hybrid MM, non-GC)";
os << "The '" << ID
<< "' instance variable was not retained by a synthesized property "
"but was released in 'dealloc'";
}
BR.EmitBasicReport(name, category, os.str(), (*I)->getLocation());
}
}
}