llvm-project/clang/test/Analysis/misc-ps.m

511 lines
14 KiB
Mathematica
Raw Normal View History

// RUN: clang-cc -analyze -checker-cfref --analyzer-store=basic -analyzer-constraints=basic --verify -fblocks %s &&
// RUN: clang-cc -analyze -checker-cfref --analyzer-store=basic -analyzer-constraints=range --verify -fblocks %s &&
// RUN: clang-cc -analyze -checker-cfref --analyzer-store=region -analyzer-constraints=basic --verify -fblocks %s &&
// RUN: clang-cc -analyze -checker-cfref --analyzer-store=region -analyzer-constraints=range --verify -fblocks %s
This is a fairly large patch, which resulted from a cascade of changes made to RegionStore (and related classes) in order to handle some analyzer failures involving casts and manipulation of symbolic memory. The root of the change is in StoreManager::CastRegion(). Instead of using ad hoc heuristics to decide when to layer an ElementRegion on a casted MemRegion, we now always layer an ElementRegion when the cast type is different than the original type of the region. This carries the current cast information associated with a region around without resorting to the error prone recording of "casted types" in GRState. Along with this new policy of layering ElementRegions, I added a new algorithm to strip away existing ElementRegions when they simply represented casts of a base memory object. This algorithm computes the raw "byte offset" that an ElementRegion represents from the base region, and allows the new ElementRegion to be based off that offset. The added benefit is that this naturally handles a series of casts of a MemRegion without building up a set of redundant ElementRegions (thus canonicalizing the region view). Other related changes that cascaded from this one (as tests were failing in RegionStore): - Revamped RegionStoreManager::InvalidateRegion() to completely remove all bindings and default values from a region and all subregions. Now invalidated fields are not bound directly to new symbolic values; instead the base region has a "default" symbol value from which "derived symbols" can be created. The main advantage of this approach is that it allows us to invalidate a region hierarchy and then lazily instantiate new values no matter how deep the hierarchy went (i.e., regardless of the number of field accesses, e.g. x->f->y->z->...). The previous approach did not do this. - Slightly reworked RegionStoreManager::RemoveDeadBindings() to also incorporate live symbols and live regions that do not have direct bindings but also have "default values" used for lazy instantiation. The changes to 'InvalidateRegion' revealed that these were necessary in order to achieve lazy instantiation of values in the region store with those bindings being removed too early. - The changes to InvalidateRegion() and RemoveDeadBindings() revealed a serious bug in 'getSubRegionMap()' where not all region -> subregion relationships involved in actually bindings (explicit and implicit) were being recorded. This has been fixed by using a worklist algorithm to iteratively fill in the region map. - Added special support to RegionStoreManager::Bind()/Retrieve() to handle OSAtomicCompareAndSwap in light of the new 'CastRegion' changes and the layering of ElementRegions. - Fixed a bug in SymbolReaper::isLive() where derived symbols were not being marked live if the symbol they were derived from was also live. This fix was critical for getting lazy instantiation in RegionStore to work. - Tidied up the implementation of ValueManager::getXXXSymbolVal() methods to use SymbolManager::canSymbolicate() to decide whether or not a symbol should be symbolicated. - 'test/Analysis/misc-ps-xfail.m' now passes; that test case has been moved to 'test/Analysis/misc-ps.m'. - Tweaked some pretty-printing of MemRegions, and implemented 'ElementRegion::getRawOffset()' for use with the CastRegion changes. llvm-svn: 77782
2009-08-01 14:17:29 +08:00
typedef struct objc_ivar *Ivar;
typedef struct objc_selector *SEL;
typedef signed char BOOL;
typedef int NSInteger;
typedef unsigned int NSUInteger;
typedef struct _NSZone NSZone;
@class NSInvocation, NSArray, NSMethodSignature, NSCoder, NSString, NSEnumerator;
This is a fairly large patch, which resulted from a cascade of changes made to RegionStore (and related classes) in order to handle some analyzer failures involving casts and manipulation of symbolic memory. The root of the change is in StoreManager::CastRegion(). Instead of using ad hoc heuristics to decide when to layer an ElementRegion on a casted MemRegion, we now always layer an ElementRegion when the cast type is different than the original type of the region. This carries the current cast information associated with a region around without resorting to the error prone recording of "casted types" in GRState. Along with this new policy of layering ElementRegions, I added a new algorithm to strip away existing ElementRegions when they simply represented casts of a base memory object. This algorithm computes the raw "byte offset" that an ElementRegion represents from the base region, and allows the new ElementRegion to be based off that offset. The added benefit is that this naturally handles a series of casts of a MemRegion without building up a set of redundant ElementRegions (thus canonicalizing the region view). Other related changes that cascaded from this one (as tests were failing in RegionStore): - Revamped RegionStoreManager::InvalidateRegion() to completely remove all bindings and default values from a region and all subregions. Now invalidated fields are not bound directly to new symbolic values; instead the base region has a "default" symbol value from which "derived symbols" can be created. The main advantage of this approach is that it allows us to invalidate a region hierarchy and then lazily instantiate new values no matter how deep the hierarchy went (i.e., regardless of the number of field accesses, e.g. x->f->y->z->...). The previous approach did not do this. - Slightly reworked RegionStoreManager::RemoveDeadBindings() to also incorporate live symbols and live regions that do not have direct bindings but also have "default values" used for lazy instantiation. The changes to 'InvalidateRegion' revealed that these were necessary in order to achieve lazy instantiation of values in the region store with those bindings being removed too early. - The changes to InvalidateRegion() and RemoveDeadBindings() revealed a serious bug in 'getSubRegionMap()' where not all region -> subregion relationships involved in actually bindings (explicit and implicit) were being recorded. This has been fixed by using a worklist algorithm to iteratively fill in the region map. - Added special support to RegionStoreManager::Bind()/Retrieve() to handle OSAtomicCompareAndSwap in light of the new 'CastRegion' changes and the layering of ElementRegions. - Fixed a bug in SymbolReaper::isLive() where derived symbols were not being marked live if the symbol they were derived from was also live. This fix was critical for getting lazy instantiation in RegionStore to work. - Tidied up the implementation of ValueManager::getXXXSymbolVal() methods to use SymbolManager::canSymbolicate() to decide whether or not a symbol should be symbolicated. - 'test/Analysis/misc-ps-xfail.m' now passes; that test case has been moved to 'test/Analysis/misc-ps.m'. - Tweaked some pretty-printing of MemRegions, and implemented 'ElementRegion::getRawOffset()' for use with the CastRegion changes. llvm-svn: 77782
2009-08-01 14:17:29 +08:00
@protocol NSObject
- (BOOL)isEqual:(id)object;
- (id)autorelease;
@end
@protocol NSCopying
- (id)copyWithZone:(NSZone *)zone;
@end
@protocol NSMutableCopying - (id)mutableCopyWithZone:(NSZone *)zone; @end
This is a fairly large patch, which resulted from a cascade of changes made to RegionStore (and related classes) in order to handle some analyzer failures involving casts and manipulation of symbolic memory. The root of the change is in StoreManager::CastRegion(). Instead of using ad hoc heuristics to decide when to layer an ElementRegion on a casted MemRegion, we now always layer an ElementRegion when the cast type is different than the original type of the region. This carries the current cast information associated with a region around without resorting to the error prone recording of "casted types" in GRState. Along with this new policy of layering ElementRegions, I added a new algorithm to strip away existing ElementRegions when they simply represented casts of a base memory object. This algorithm computes the raw "byte offset" that an ElementRegion represents from the base region, and allows the new ElementRegion to be based off that offset. The added benefit is that this naturally handles a series of casts of a MemRegion without building up a set of redundant ElementRegions (thus canonicalizing the region view). Other related changes that cascaded from this one (as tests were failing in RegionStore): - Revamped RegionStoreManager::InvalidateRegion() to completely remove all bindings and default values from a region and all subregions. Now invalidated fields are not bound directly to new symbolic values; instead the base region has a "default" symbol value from which "derived symbols" can be created. The main advantage of this approach is that it allows us to invalidate a region hierarchy and then lazily instantiate new values no matter how deep the hierarchy went (i.e., regardless of the number of field accesses, e.g. x->f->y->z->...). The previous approach did not do this. - Slightly reworked RegionStoreManager::RemoveDeadBindings() to also incorporate live symbols and live regions that do not have direct bindings but also have "default values" used for lazy instantiation. The changes to 'InvalidateRegion' revealed that these were necessary in order to achieve lazy instantiation of values in the region store with those bindings being removed too early. - The changes to InvalidateRegion() and RemoveDeadBindings() revealed a serious bug in 'getSubRegionMap()' where not all region -> subregion relationships involved in actually bindings (explicit and implicit) were being recorded. This has been fixed by using a worklist algorithm to iteratively fill in the region map. - Added special support to RegionStoreManager::Bind()/Retrieve() to handle OSAtomicCompareAndSwap in light of the new 'CastRegion' changes and the layering of ElementRegions. - Fixed a bug in SymbolReaper::isLive() where derived symbols were not being marked live if the symbol they were derived from was also live. This fix was critical for getting lazy instantiation in RegionStore to work. - Tidied up the implementation of ValueManager::getXXXSymbolVal() methods to use SymbolManager::canSymbolicate() to decide whether or not a symbol should be symbolicated. - 'test/Analysis/misc-ps-xfail.m' now passes; that test case has been moved to 'test/Analysis/misc-ps.m'. - Tweaked some pretty-printing of MemRegions, and implemented 'ElementRegion::getRawOffset()' for use with the CastRegion changes. llvm-svn: 77782
2009-08-01 14:17:29 +08:00
@protocol NSCoding
- (void)encodeWithCoder:(NSCoder *)aCoder;
@end
@interface NSObject <NSObject> {}
- (id)init;
+ (id)allocWithZone:(NSZone *)zone;
@end
extern id NSAllocateObject(Class aClass, NSUInteger extraBytes, NSZone *zone);
@interface NSString : NSObject <NSCopying, NSMutableCopying, NSCoding>
- (NSUInteger)length;
+ (id)stringWithUTF8String:(const char *)nullTerminatedCString;
@end extern NSString * const NSBundleDidLoadNotification;
This is a fairly large patch, which resulted from a cascade of changes made to RegionStore (and related classes) in order to handle some analyzer failures involving casts and manipulation of symbolic memory. The root of the change is in StoreManager::CastRegion(). Instead of using ad hoc heuristics to decide when to layer an ElementRegion on a casted MemRegion, we now always layer an ElementRegion when the cast type is different than the original type of the region. This carries the current cast information associated with a region around without resorting to the error prone recording of "casted types" in GRState. Along with this new policy of layering ElementRegions, I added a new algorithm to strip away existing ElementRegions when they simply represented casts of a base memory object. This algorithm computes the raw "byte offset" that an ElementRegion represents from the base region, and allows the new ElementRegion to be based off that offset. The added benefit is that this naturally handles a series of casts of a MemRegion without building up a set of redundant ElementRegions (thus canonicalizing the region view). Other related changes that cascaded from this one (as tests were failing in RegionStore): - Revamped RegionStoreManager::InvalidateRegion() to completely remove all bindings and default values from a region and all subregions. Now invalidated fields are not bound directly to new symbolic values; instead the base region has a "default" symbol value from which "derived symbols" can be created. The main advantage of this approach is that it allows us to invalidate a region hierarchy and then lazily instantiate new values no matter how deep the hierarchy went (i.e., regardless of the number of field accesses, e.g. x->f->y->z->...). The previous approach did not do this. - Slightly reworked RegionStoreManager::RemoveDeadBindings() to also incorporate live symbols and live regions that do not have direct bindings but also have "default values" used for lazy instantiation. The changes to 'InvalidateRegion' revealed that these were necessary in order to achieve lazy instantiation of values in the region store with those bindings being removed too early. - The changes to InvalidateRegion() and RemoveDeadBindings() revealed a serious bug in 'getSubRegionMap()' where not all region -> subregion relationships involved in actually bindings (explicit and implicit) were being recorded. This has been fixed by using a worklist algorithm to iteratively fill in the region map. - Added special support to RegionStoreManager::Bind()/Retrieve() to handle OSAtomicCompareAndSwap in light of the new 'CastRegion' changes and the layering of ElementRegions. - Fixed a bug in SymbolReaper::isLive() where derived symbols were not being marked live if the symbol they were derived from was also live. This fix was critical for getting lazy instantiation in RegionStore to work. - Tidied up the implementation of ValueManager::getXXXSymbolVal() methods to use SymbolManager::canSymbolicate() to decide whether or not a symbol should be symbolicated. - 'test/Analysis/misc-ps-xfail.m' now passes; that test case has been moved to 'test/Analysis/misc-ps.m'. - Tweaked some pretty-printing of MemRegions, and implemented 'ElementRegion::getRawOffset()' for use with the CastRegion changes. llvm-svn: 77782
2009-08-01 14:17:29 +08:00
@interface NSValue : NSObject <NSCopying, NSCoding>
- (void)getValue:(void *)value;
@end
@interface NSNumber : NSValue
- (char)charValue;
- (id)initWithBool:(BOOL)value;
@end
@interface NSAssertionHandler : NSObject {}
+ (NSAssertionHandler *)currentHandler;
- (void)handleFailureInMethod:(SEL)selector object:(id)object file:(NSString *)fileName lineNumber:(NSInteger)line description:(NSString *)format,...;
@end
extern NSString * const NSConnectionReplyMode;
typedef float CGFloat;
typedef struct _NSPoint {
CGFloat x;
CGFloat y;
} NSPoint;
typedef struct _NSSize {
CGFloat width;
CGFloat height;
} NSSize;
typedef struct _NSRect {
NSPoint origin;
NSSize size;
} NSRect;
2008-10-01 13:05:46 +08:00
// Reduced test case from crash in <rdar://problem/6253157>
@interface A @end
@implementation A
- (void)foo:(void (^)(NSObject *x))block {
if (!((block != ((void *)0)))) {}
}
@end
// Reduced test case from crash in PR 2796;
// http://llvm.org/bugs/show_bug.cgi?id=2796
unsigned foo(unsigned x) { return __alignof__((x)) + sizeof(x); }
// Improvement to path-sensitivity involving compound assignments.
// Addresses false positive in <rdar://problem/6268365>
//
unsigned r6268365Aux();
void r6268365() {
unsigned x = 0;
x &= r6268365Aux();
unsigned j = 0;
if (x == 0) ++j;
if (x == 0) x = x / j; // no-warning
}
void divzeroassume(unsigned x, unsigned j) {
x /= j;
if (j == 0) x /= 0; // no-warning
if (j == 0) x /= j; // no-warning
if (j == 0) x = x / 0; // no-warning
}
void divzeroassumeB(unsigned x, unsigned j) {
x = x / j;
if (j == 0) x /= 0; // no-warning
if (j == 0) x /= j; // no-warning
if (j == 0) x = x / 0; // no-warning
}
// InitListExpr processing
typedef float __m128 __attribute__((__vector_size__(16), __may_alias__));
__m128 return128() {
// This compound literal has a Vector type. We currently just
// return UnknownVal.
return __extension__(__m128) { 0.0f, 0.0f, 0.0f, 0.0f };
}
typedef long long __v2di __attribute__ ((__vector_size__ (16)));
typedef long long __m128i __attribute__ ((__vector_size__ (16), __may_alias__));
__m128i vec128i(long long __q1, long long __q0) {
// This compound literal returns true for both isVectorType() and
// isIntegerType().
return __extension__ (__m128i)(__v2di){ __q0, __q1 };
}
// Zero-sized VLAs.
void check_zero_sized_VLA(int x) {
if (x)
return;
int vla[x]; // expected-warning{{Variable-length array 'vla' has zero elements (undefined behavior)}}
}
void check_uninit_sized_VLA() {
int x;
int vla[x]; // expected-warning{{Variable-length array 'vla' garbage value for array size}}
}
// sizeof(void)
// - Tests a regression reported in PR 3211: http://llvm.org/bugs/show_bug.cgi?id=3211
void handle_sizeof_void(unsigned flag) {
int* p = 0;
if (flag) {
if (sizeof(void) == 1)
return;
// Infeasible.
*p = 1; // no-warning
}
void* q;
if (!flag) {
if (sizeof(*q) == 1)
return;
// Infeasibe.
*p = 1; // no-warning
}
// Infeasible.
*p = 1; // no-warning
}
// PR 3422
void pr3422_helper(char *p);
void pr3422() {
char buf[100];
char *q = &buf[10];
pr3422_helper(&q[1]);
}
// PR 3543 (handle empty statement expressions)
2009-07-22 02:59:16 +08:00
void pr_3543(void) {
({});
}
// <rdar://problem/6611677>
// This test case test the use of a vector type within an array subscript
// expression.
typedef long long __a64vector __attribute__((__vector_size__(8)));
typedef long long __a128vector __attribute__((__vector_size__(16)));
static inline __a64vector __attribute__((__always_inline__, __nodebug__))
my_test_mm_movepi64_pi64(__a128vector a) {
return (__a64vector)a[0];
}
// Test basic tracking of ivars associated with 'self'.
@interface SelfIvarTest : NSObject {
int flag;
}
- (void)test_self_tracking;
@end
@implementation SelfIvarTest
- (void)test_self_tracking {
char *p = 0;
char c;
if (flag)
p = "hello";
if (flag)
c = *p; // no-warning
}
@end
// PR 3770
char pr3770(int x) {
int y = x & 0x2;
char *p = 0;
if (y == 1)
p = "hello";
if (y == 1)
return p[0]; // no-warning
return 'a';
}
// PR 3772
// - We just want to test that this doesn't crash the analyzer.
typedef struct st ST;
struct st { char *name; };
extern ST *Cur_Pu;
void pr3772(void)
{
static ST *last_Cur_Pu;
if (last_Cur_Pu == Cur_Pu) {
return;
}
}
// PR 3780 - This tests that StmtIterator isn't broken for VLAs in DeclGroups.
void pr3780(int sz) { typedef double MAT[sz][sz]; }
// <rdar://problem/6695527> - Test that we don't symbolicate doubles before
// we are ready to do something with them.
int rdar6695527(double x) {
if (!x) { return 0; }
return 1;
}
// <rdar://problem/6708148> - Test that we properly invalidate structs
// passed-by-reference to a function.
void pr6708148_invalidate(NSRect *x);
void pr6708148_use(NSRect x);
void pr6708148_test(void) {
NSRect x;
pr6708148_invalidate(&x);
pr6708148_use(x); // no-warning
}
// Handle both kinds of noreturn attributes for pruning paths.
void rdar_6777003_noret() __attribute__((noreturn));
void rdar_6777003_analyzer_noret() __attribute__((analyzer_noreturn));
void rdar_6777003(int x) {
int *p = 0;
if (x == 1) {
rdar_6777003_noret();
*p = 1; // no-warning;
}
if (x == 2) {
rdar_6777003_analyzer_noret();
*p = 1; // no-warning;
}
*p = 1; // expected-warning{{Dereference of null pointer}}
}
// For pointer arithmetic, --/++ should be treated as preserving non-nullness,
// regardless of how well the underlying StoreManager reasons about pointer
// arithmetic.
// <rdar://problem/6777209>
void rdar_6777209(char *p) {
if (p == 0)
return;
++p;
// This branch should always be infeasible.
if (p == 0)
*p = 'c'; // no-warning
}
// PR 4033. A symbolic 'void *' pointer can be used as the address for a
// computed goto.
typedef void *Opcode;
Opcode pr_4033_getOpcode();
void pr_4033(void) {
next_opcode:
{
Opcode op = pr_4033_getOpcode();
if (op) goto *op;
}
}
// Test invalidating pointers-to-pointers with slightly different types. This
// example came from a recent false positive due to a regression where the
// branch condition was falsely reported as being uninitialized.
void invalidate_by_ref(char **x);
int test_invalidate_by_ref() {
unsigned short y;
invalidate_by_ref((char**) &y);
if (y) // no-warning
return 1;
return 0;
}
// Test for <rdar://problem/7027684>. This just tests that the CFG is
// constructed correctly. Previously, the successor block of the entrance
// was the block containing the merge for '?', which would trigger an
// assertion failure.
int rdar_7027684_aux();
int rdar_7027684_aux_2() __attribute__((noreturn));
void rdar_7027684(int x, int y) {
{}; // this empty compound statement is critical.
(rdar_7027684_aux() ? rdar_7027684_aux_2() : (void) 0);
}
// Test that we handle casts of string literals to arbitrary types.
unsigned const char *string_literal_test1() {
return (const unsigned char*) "hello";
}
const float *string_literal_test2() {
return (const float*) "hello";
}
// Test that we handle casts *from* incomplete struct types.
extern const struct _FooAssertStruct _cmd;
void test_cast_from_incomplete_struct_aux(volatile const void *x);
void test_cast_from_incomplete_struct() {
test_cast_from_incomplete_struct_aux(&_cmd);
}
// Test for <rdar://problem/7034511>
// "ValueManager::makeIntVal(uint64_t X, QualType T) should return a 'Loc'
// when 'T' is a pointer"
//
// Previously this case would crash.
void test_rdar_7034511(NSArray *y) {
NSObject *x;
for (x in y) {}
if (x == ((void*) 0)) {}
}
// Handle casts of function pointers (CodeTextRegions) to arbitrary pointer
// types. This was previously causing a crash in CastRegion.
void handle_funcptr_voidptr_casts() {
void **ptr;
typedef void *PVOID;
typedef void *PCHAR;
typedef long INT_PTR, *PINT_PTR;
typedef INT_PTR (*FARPROC)();
FARPROC handle_funcptr_voidptr_casts_aux();
PVOID handle_funcptr_voidptr_casts_aux_2(PVOID volatile *x);
PVOID handle_funcptr_voidptr_casts_aux_3(PCHAR volatile *x);
ptr = (void**) handle_funcptr_voidptr_casts_aux();
handle_funcptr_voidptr_casts_aux_2(ptr);
handle_funcptr_voidptr_casts_aux_3(ptr);
}
// RegionStore::Retrieve previously crashed on this example. This example
// was previously in the test file 'xfail_regionstore_wine_crash.c'.
void testA() {
long x = 0;
char *y = (char *) &x;
if (!*y)
return;
}
// RegionStoreManager previously crashed on this example. The problem is that
// the value bound to the field of b->grue after the call to testB_aux is
// a symbolic region. The second '*__gruep__' involves performing a load
// from a 'int*' that really is a 'void**'. The loaded location must be
// implicitly converted to an integer that wraps a location. Previosly we would
// get a crash here due to an assertion failure.
typedef struct _BStruct { void *grue; } BStruct;
void testB_aux(void *ptr);
void testB(BStruct *b) {
{
int *__gruep__ = ((int *)&((b)->grue));
int __gruev__ = *__gruep__;
testB_aux(__gruep__);
}
{
int *__gruep__ = ((int *)&((b)->grue));
int __gruev__ = *__gruep__;
if (~0 != __gruev__) {}
}
}
void test_trivial_symbolic_comparison(int *x) {
int test_trivial_symbolic_comparison_aux();
int a = test_trivial_symbolic_comparison_aux();
int b = a;
if (a != b) {
int *p = 0;
*p = 0xDEADBEEF; // no-warning
}
a = a == 1;
b = b == 1;
if (a != b) {
int *p = 0;
*p = 0xDEADBEEF; // no-warning
}
}
// Test for:
// <rdar://problem/7062158> false positive null dereference due to
// BasicStoreManager not tracking *static* globals
//
// This just tests the proper tracking of symbolic values for globals (both
// static and non-static).
//
static int* x_rdar_7062158;
void rdar_7062158() {
int *current = x_rdar_7062158;
if (current == x_rdar_7062158)
return;
int *p = 0;
*p = 0xDEADBEEF; // no-warning
}
int* x_rdar_7062158_2;
void rdar_7062158_2() {
int *current = x_rdar_7062158_2;
if (current == x_rdar_7062158_2)
return;
int *p = 0;
*p = 0xDEADBEEF; // no-warning
}
// This test reproduces a case for a crash when analyzing ClamAV using
// RegionStoreManager (the crash doesn't exhibit in BasicStoreManager because
// it isn't doing anything smart about arrays). The problem is that on the
// second line, 'p = &p[i]', p is assigned an ElementRegion whose index
// is a 16-bit integer. On the third line, a new ElementRegion is created
// based on the previous region, but there the region uses a 32-bit integer,
// resulting in a clash of values (an assertion failure at best). We resolve
// this problem by implicitly converting index values to 'int' when the
// ElementRegion is created.
unsigned char test_array_index_bitwidth(const unsigned char *p) {
unsigned short i = 0;
for (i = 0; i < 2; i++) p = &p[i];
return p[i+1];
}
// This case tests that CastRegion handles casts involving BlockPointerTypes.
// It should not crash.
void test_block_cast() {
id test_block_cast_aux();
(void (^)(void *))test_block_cast_aux(); // expected-warning{{expression result unused}}
}
2009-07-21 05:00:55 +08:00
// Test comparison of 'id' instance variable to a null void* constant after
// performing an OSAtomicCompareAndSwap32Barrier.
// This previously was a crash in RegionStoreManager.
@interface TestIdNull {
id x;
}
-(int)foo;
@end
@implementation TestIdNull
-(int)foo {
OSAtomicCompareAndSwap32Barrier(0, (signed)2, (signed*)&x);
if (x == (void*) 0) { return 0; }
return 1;
}
@end
// PR 4594 - This was a crash when handling casts in SimpleSValuator.
void PR4594() {
char *buf[1];
char **foo = buf;
*foo = "test";
}
// Test invalidation logic where an integer is casted to an array with a
// different sign and then invalidated.
void test_invalidate_cast_int() {
void test_invalidate_cast_int_aux(unsigned *i);
signed i;
test_invalidate_cast_int_aux((unsigned*) &i);
if (i < 0)
return;
}
This is a fairly large patch, which resulted from a cascade of changes made to RegionStore (and related classes) in order to handle some analyzer failures involving casts and manipulation of symbolic memory. The root of the change is in StoreManager::CastRegion(). Instead of using ad hoc heuristics to decide when to layer an ElementRegion on a casted MemRegion, we now always layer an ElementRegion when the cast type is different than the original type of the region. This carries the current cast information associated with a region around without resorting to the error prone recording of "casted types" in GRState. Along with this new policy of layering ElementRegions, I added a new algorithm to strip away existing ElementRegions when they simply represented casts of a base memory object. This algorithm computes the raw "byte offset" that an ElementRegion represents from the base region, and allows the new ElementRegion to be based off that offset. The added benefit is that this naturally handles a series of casts of a MemRegion without building up a set of redundant ElementRegions (thus canonicalizing the region view). Other related changes that cascaded from this one (as tests were failing in RegionStore): - Revamped RegionStoreManager::InvalidateRegion() to completely remove all bindings and default values from a region and all subregions. Now invalidated fields are not bound directly to new symbolic values; instead the base region has a "default" symbol value from which "derived symbols" can be created. The main advantage of this approach is that it allows us to invalidate a region hierarchy and then lazily instantiate new values no matter how deep the hierarchy went (i.e., regardless of the number of field accesses, e.g. x->f->y->z->...). The previous approach did not do this. - Slightly reworked RegionStoreManager::RemoveDeadBindings() to also incorporate live symbols and live regions that do not have direct bindings but also have "default values" used for lazy instantiation. The changes to 'InvalidateRegion' revealed that these were necessary in order to achieve lazy instantiation of values in the region store with those bindings being removed too early. - The changes to InvalidateRegion() and RemoveDeadBindings() revealed a serious bug in 'getSubRegionMap()' where not all region -> subregion relationships involved in actually bindings (explicit and implicit) were being recorded. This has been fixed by using a worklist algorithm to iteratively fill in the region map. - Added special support to RegionStoreManager::Bind()/Retrieve() to handle OSAtomicCompareAndSwap in light of the new 'CastRegion' changes and the layering of ElementRegions. - Fixed a bug in SymbolReaper::isLive() where derived symbols were not being marked live if the symbol they were derived from was also live. This fix was critical for getting lazy instantiation in RegionStore to work. - Tidied up the implementation of ValueManager::getXXXSymbolVal() methods to use SymbolManager::canSymbolicate() to decide whether or not a symbol should be symbolicated. - 'test/Analysis/misc-ps-xfail.m' now passes; that test case has been moved to 'test/Analysis/misc-ps.m'. - Tweaked some pretty-printing of MemRegions, and implemented 'ElementRegion::getRawOffset()' for use with the CastRegion changes. llvm-svn: 77782
2009-08-01 14:17:29 +08:00
// Reduced from a crash involving the cast of an Objective-C symbolic region to
// 'char *'
static NSNumber *test_ivar_offset(id self, SEL _cmd, Ivar inIvar) {
return [[[NSNumber allocWithZone:((void*)0)] initWithBool:*(_Bool *)((char *)self + ivar_getOffset(inIvar))] autorelease];
}
// Reduced from a crash in StoreManager::CastRegion involving a divide-by-zero.
// This resulted from not properly handling region casts to 'const void*'.
void test_cast_const_voidptr() {
char x[10];
char *p = &x[1];
const void* q = p;
}
// Reduced from a crash when analyzing Wine. This test handles loads from
// function addresses.
typedef long (*FARPROC)();
FARPROC test_load_func(FARPROC origfun) {
if (!*(unsigned char*) origfun)
return origfun;
return 0;
}