llvm-project/clang/test/CodeGenCXX/microsoft-abi-eh-cleanups.cpp

340 lines
12 KiB
C++
Raw Normal View History

// RUN: %clang_cc1 -std=c++11 -emit-llvm %s -o - -triple=i386-pc-win32 -mconstructor-aliases -fexceptions -fcxx-exceptions -fno-rtti | FileCheck -check-prefix WIN32 -check-prefix WIN32-O0 %s
// RUN: %clang_cc1 -std=c++11 -emit-llvm -O3 -disable-llvm-passes %s -o - -triple=i386-pc-win32 -mconstructor-aliases -fexceptions -fcxx-exceptions -fno-rtti | FileCheck -check-prefix WIN32 -check-prefix WIN32-O3 -check-prefix WIN32-LIFETIME %s
struct A {
A();
~A();
int a;
};
A getA();
int TakesTwo(A a, A b);
void HasEHCleanup() {
TakesTwo(getA(), getA());
}
// With exceptions, we need to clean up at least one of these temporaries.
// WIN32-LABEL: define dso_local void @"?HasEHCleanup@@YAXXZ"() {{.*}} {
// WIN32: %[[base:.*]] = call i8* @llvm.stacksave()
// If this call throws, we have to restore the stack.
// WIN32: call void @"?getA@@YA?AUA@@XZ"(%struct.A* sret %{{.*}})
// If this call throws, we have to cleanup the first temporary.
// WIN32: invoke void @"?getA@@YA?AUA@@XZ"(%struct.A* sret %{{.*}})
// If this call throws, we have to cleanup the stacksave.
// WIN32: call i32 @"?TakesTwo@@YAHUA@@0@Z"
// WIN32: call void @llvm.stackrestore
// WIN32: ret void
//
// There should be one dtor call for unwinding from the second getA.
// WIN32: cleanuppad
// WIN32: call x86_thiscallcc void @"??1A@@QAE@XZ"({{.*}})
// WIN32-NOT: @"??1A@@QAE@XZ"
// WIN32: }
void TakeRef(const A &a);
int HasDeactivatedCleanups() {
return TakesTwo((TakeRef(A()), A()), (TakeRef(A()), A()));
}
// WIN32-LABEL: define dso_local i32 @"?HasDeactivatedCleanups@@YAHXZ"() {{.*}} {
// WIN32: %[[isactive:.*]] = alloca i1
// WIN32: call i8* @llvm.stacksave()
// WIN32: %[[argmem:.*]] = alloca inalloca [[argmem_ty:<{ %struct.A, %struct.A }>]]
// WIN32: %[[arg1:.*]] = getelementptr inbounds [[argmem_ty]], [[argmem_ty]]* %[[argmem]], i32 0, i32 1
// WIN32: call x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32: invoke void @"?TakeRef@@YAXABUA@@@Z"
//
// WIN32: invoke x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"(%struct.A* %[[arg1]])
// WIN32: store i1 true, i1* %[[isactive]]
//
// WIN32: %[[arg0:.*]] = getelementptr inbounds [[argmem_ty]], [[argmem_ty]]* %[[argmem]], i32 0, i32 0
// WIN32: invoke x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32: invoke void @"?TakeRef@@YAXABUA@@@Z"
// WIN32: invoke x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32: store i1 false, i1* %[[isactive]]
//
// WIN32: invoke i32 @"?TakesTwo@@YAHUA@@0@Z"([[argmem_ty]]* inalloca %[[argmem]])
// Destroy the two const ref temporaries.
// WIN32: call x86_thiscallcc void @"??1A@@QAE@XZ"({{.*}})
// WIN32: call x86_thiscallcc void @"??1A@@QAE@XZ"({{.*}})
// WIN32: ret i32
//
// Conditionally destroy arg1.
// WIN32: %[[cond:.*]] = load i1, i1* %[[isactive]]
// WIN32: br i1 %[[cond]]
// WIN32: call x86_thiscallcc void @"??1A@@QAE@XZ"(%struct.A* %[[arg1]])
// WIN32: }
// Test putting the cleanups inside a conditional.
int CouldThrow();
int HasConditionalCleanup(bool cond) {
return (cond ? TakesTwo(A(), A()) : CouldThrow());
}
// WIN32-LABEL: define dso_local i32 @"?HasConditionalCleanup@@YAH_N@Z"(i1 zeroext %{{.*}}) {{.*}} {
// WIN32: store i1 false
// WIN32: br i1
// WIN32: call i8* @llvm.stacksave()
// WIN32: call x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"(%struct.A* %{{.*}})
// WIN32: store i1 true
// WIN32: invoke x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"(%struct.A* %{{.*}})
// WIN32: call i32 @"?TakesTwo@@YAHUA@@0@Z"
//
// WIN32: call void @llvm.stackrestore
//
// WIN32: call i32 @"?CouldThrow@@YAHXZ"()
//
// Only one dtor in the invoke for arg1
// WIN32: call x86_thiscallcc void @"??1A@@QAE@XZ"({{.*}})
// WIN32-NOT: invoke x86_thiscallcc void @"??1A@@QAE@XZ"
// WIN32: }
// Now test both.
int HasConditionalDeactivatedCleanups(bool cond) {
return (cond ? TakesTwo((TakeRef(A()), A()), (TakeRef(A()), A())) : CouldThrow());
}
// WIN32-O0-LABEL: define dso_local i32 @"?HasConditionalDeactivatedCleanups@@YAH_N@Z"{{.*}} {
// WIN32-O0: alloca i1
// WIN32-O0: %[[arg1_cond:.*]] = alloca i1
// Start all four cleanups as deactivated.
// WIN32-O0: store i1 false
// WIN32-O0: store i1 false
// WIN32-O0: store i1 false
// WIN32-O0: store i1 false
// WIN32-O0: br i1
// True condition.
// WIN32-O0: call x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32-O0: store i1 true
// WIN32-O0: invoke void @"?TakeRef@@YAXABUA@@@Z"
// WIN32-O0: invoke x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32-O0: store i1 true, i1* %[[arg1_cond]]
// WIN32-O0: invoke x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32-O0: store i1 true
// WIN32-O0: invoke void @"?TakeRef@@YAXABUA@@@Z"
// WIN32-O0: invoke x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32-O0: store i1 true
// WIN32-O0: store i1 false, i1* %[[arg1_cond]]
// WIN32-O0: invoke i32 @"?TakesTwo@@YAHUA@@0@Z"
// False condition.
// WIN32-O0: invoke i32 @"?CouldThrow@@YAHXZ"()
// Two normal cleanups for TakeRef args.
// WIN32-O0: call x86_thiscallcc void @"??1A@@QAE@XZ"({{.*}})
// WIN32-O0-NOT: invoke x86_thiscallcc void @"??1A@@QAE@XZ"
// WIN32-O0: ret i32
//
// Somewhere in the landing pad soup, we conditionally destroy arg1.
// WIN32-O0: %[[isactive:.*]] = load i1, i1* %[[arg1_cond]]
// WIN32-O0: br i1 %[[isactive]]
// WIN32-O0: call x86_thiscallcc void @"??1A@@QAE@XZ"({{.*}})
// WIN32-O0: }
// WIN32-O3-LABEL: define dso_local i32 @"?HasConditionalDeactivatedCleanups@@YAH_N@Z"{{.*}} {
// WIN32-O3: alloca i1
// WIN32-O3: alloca i1
// WIN32-O3: %[[arg1_cond:.*]] = alloca i1
// Start all four cleanups as deactivated.
// WIN32-O3: store i1 false
// WIN32-O3: store i1 false
// WIN32-O3: store i1 false
// WIN32-O3: store i1 false
// WIN32-O3: store i1 false
// WIN32-O3: store i1 false
// WIN32-O3: br i1
// True condition.
// WIN32-O3: call x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32-O3: store i1 true
// WIN32-O3: invoke void @"?TakeRef@@YAXABUA@@@Z"
// WIN32-O3: invoke x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32-O3: store i1 true, i1* %[[arg1_cond]]
// WIN32-O3: invoke x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32-O3: store i1 true
// WIN32-O3: invoke void @"?TakeRef@@YAXABUA@@@Z"
// WIN32-O3: invoke x86_thiscallcc %struct.A* @"??0A@@QAE@XZ"
// WIN32-O3: store i1 true
// WIN32-O3: store i1 false, i1* %[[arg1_cond]]
// WIN32-O3: invoke i32 @"?TakesTwo@@YAHUA@@0@Z"
// False condition.
// WIN32-O3: invoke i32 @"?CouldThrow@@YAHXZ"()
// Two normal cleanups for TakeRef args.
// WIN32-O3: call x86_thiscallcc void @"??1A@@QAE@XZ"({{.*}})
// WIN32-O3-NOT: invoke x86_thiscallcc void @"??1A@@QAE@XZ"
// WIN32-O3: ret i32
//
// Somewhere in the landing pad soup, we conditionally destroy arg1.
// WIN32-O3: %[[isactive:.*]] = load i1, i1* %[[arg1_cond]]
// WIN32-O3: br i1 %[[isactive]]
// WIN32-O3: call x86_thiscallcc void @"??1A@@QAE@XZ"({{.*}})
// WIN32-O3: }
namespace crash_on_partial_destroy {
struct A {
virtual ~A();
};
struct B : virtual A {
// Has an implicit destructor.
};
struct C : B {
C();
};
void foo();
// We used to crash when emitting this.
C::C() { foo(); }
// Verify that we don't bother with a vbtable lookup when adjusting the this
// pointer to call a base destructor from a constructor while unwinding.
// WIN32-LABEL: define dso_local {{.*}} @"??0C@crash_on_partial_destroy@@QAE@XZ"{{.*}} {
// WIN32: cleanuppad
//
// We shouldn't do any vbptr loads, just constant GEPs.
// WIN32-NOT: load
// WIN32: getelementptr i8, i8* %{{.*}}, i32 4
// WIN32-NOT: load
// WIN32: bitcast i8* %{{.*}} to %"struct.crash_on_partial_destroy::B"*
// WIN32: call x86_thiscallcc void @"??1B@crash_on_partial_destroy@@UAE@XZ"
//
// WIN32-NOT: load
// WIN32: bitcast %"struct.crash_on_partial_destroy::C"* %{{.*}} to i8*
// WIN32-NOT: load
Compute and preserve alignment more faithfully in IR-generation. Introduce an Address type to bundle a pointer value with an alignment. Introduce APIs on CGBuilderTy to work with Address values. Change core APIs on CGF/CGM to traffic in Address where appropriate. Require alignments to be non-zero. Update a ton of code to compute and propagate alignment information. As part of this, I've promoted CGBuiltin's EmitPointerWithAlignment helper function to CGF and made use of it in a number of places in the expression emitter. The end result is that we should now be significantly more correct when performing operations on objects that are locally known to be under-aligned. Since alignment is not reliably tracked in the type system, there are inherent limits to this, but at least we are no longer confused by standard operations like derived-to-base conversions and array-to-pointer decay. I've also fixed a large number of bugs where we were applying the complete-object alignment to a pointer instead of the non-virtual alignment, although most of these were hidden by the very conservative approach we took with member alignment. Also, because IRGen now reliably asserts on zero alignments, we should no longer be subject to an absurd but frustrating recurring bug where an incomplete type would report a zero alignment and then we'd naively do a alignmentAtOffset on it and emit code using an alignment equal to the largest power-of-two factor of the offset. We should also now be emitting much more aggressive alignment attributes in the presence of over-alignment. In particular, field access now uses alignmentAtOffset instead of min. Several times in this patch, I had to change the existing code-generation pattern in order to more effectively use the Address APIs. For the most part, this seems to be a strict improvement, like doing pointer arithmetic with GEPs instead of ptrtoint. That said, I've tried very hard to not change semantics, but it is likely that I've failed in a few places, for which I apologize. ABIArgInfo now always carries the assumed alignment of indirect and indirect byval arguments. In order to cut down on what was already a dauntingly large patch, I changed the code to never set align attributes in the IR on non-byval indirect arguments. That is, we still generate code which assumes that indirect arguments have the given alignment, but we don't express this information to the backend except where it's semantically required (i.e. on byvals). This is likely a minor regression for those targets that did provide this information, but it'll be trivial to add it back in a later patch. I partially punted on applying this work to CGBuiltin. Please do not add more uses of the CreateDefaultAligned{Load,Store} APIs; they will be going away eventually. llvm-svn: 246985
2015-09-08 16:05:57 +08:00
// WIN32: getelementptr inbounds i8, i8* %{{.*}}, i32 4
// WIN32-NOT: load
// WIN32: bitcast i8* %{{.*}} to %"struct.crash_on_partial_destroy::A"*
// WIN32: call x86_thiscallcc void @"??1A@crash_on_partial_destroy@@UAE@XZ"({{.*}})
// WIN32: }
}
namespace dont_call_terminate {
struct C {
~C();
};
void g();
void f() {
C c;
g();
}
// WIN32-LABEL: define dso_local void @"?f@dont_call_terminate@@YAXXZ"()
// WIN32: invoke void @"?g@dont_call_terminate@@YAXXZ"()
// WIN32-NEXT: to label %[[cont:[^ ]*]] unwind label %[[lpad:[^ ]*]]
//
// WIN32: [[cont]]
// WIN32: call x86_thiscallcc void @"??1C@dont_call_terminate@@QAE@XZ"({{.*}})
//
// WIN32: [[lpad]]
// WIN32-NEXT: cleanuppad
// WIN32: call x86_thiscallcc void @"??1C@dont_call_terminate@@QAE@XZ"({{.*}})
}
namespace noexcept_false_dtor {
struct D {
~D() noexcept(false);
};
void f() {
D d;
CouldThrow();
}
}
// WIN32-LABEL: define dso_local void @"?f@noexcept_false_dtor@@YAXXZ"()
// WIN32: invoke i32 @"?CouldThrow@@YAHXZ"()
// WIN32: call x86_thiscallcc void @"??1D@noexcept_false_dtor@@QAE@XZ"(%"struct.noexcept_false_dtor::D"* %{{.*}})
// WIN32: cleanuppad
// WIN32: call x86_thiscallcc void @"??1D@noexcept_false_dtor@@QAE@XZ"(%"struct.noexcept_false_dtor::D"* %{{.*}})
// WIN32: cleanupret
namespace lifetime_marker {
struct C {
~C();
};
void g();
void f() {
C c;
g();
}
// WIN32-LIFETIME-LABEL: define dso_local void @"?f@lifetime_marker@@YAXXZ"()
// WIN32-LIFETIME: %[[c:.*]] = alloca %"struct.lifetime_marker::C"
// WIN32-LIFETIME: %[[bc0:.*]] = bitcast %"struct.lifetime_marker::C"* %c to i8*
// WIN32-LIFETIME: call void @llvm.lifetime.start.p0i8(i64 1, i8* %[[bc0]])
// WIN32-LIFETIME: invoke void @"?g@lifetime_marker@@YAXXZ"()
// WIN32-LIFETIME-NEXT: to label %[[cont:[^ ]*]] unwind label %[[lpad0:[^ ]*]]
//
// WIN32-LIFETIME: [[cont]]
// WIN32-LIFETIME: call x86_thiscallcc void @"??1C@lifetime_marker@@QAE@XZ"({{.*}})
// WIN32-LIFETIME: %[[bc1:.*]] = bitcast %"struct.lifetime_marker::C"* %[[c]] to i8*
// WIN32-LIFETIME: call void @llvm.lifetime.end.p0i8(i64 1, i8* %[[bc1]])
//
// WIN32-LIFETIME: [[lpad0]]
// WIN32-LIFETIME-NEXT: cleanuppad
// WIN32-LIFETIME: call x86_thiscallcc void @"??1C@lifetime_marker@@QAE@XZ"({{.*}})
// WIN32-LIFETIME: cleanupret {{.*}} unwind label %[[lpad1:[^ ]*]]
//
// WIN32-LIFETIME: [[lpad1]]
// WIN32-LIFETIME-NEXT: cleanuppad
// WIN32-LIFETIME: %[[bc2:.*]] = bitcast %"struct.lifetime_marker::C"* %[[c]] to i8*
// WIN32-LIFETIME: call void @llvm.lifetime.end.p0i8(i64 1, i8* %[[bc2]])
}
[MS-ABI]V-base dtor called more than needed when throw happens in v-base ctor in window. Need add "complete object flag" check in eh cleanup code. The problem only happen on window ( A MS-ABI issuer ) The nature of the problem is virtual base dtor called more than it is needed after exception throw in inheriting base class(with virtual bases) ctor. The root problem is when throw happen, not all virtual base classes have been contructed, so not all virtual base dtors are need to call for ehcleanup. clang has code to handle vbase initialization: basically add check for "complete object flag" before call to v-base ctor. But that part is missing for cleanup code. To fix this add similar code as v-base init to cleanup code, same algorithm. 1> Add new routine: EmitDtorCompleteObjectHandler With corresponding to EmitCtorCompleteObjectHandler 2> In the EmitDestructorCal Call EmitDtorCompleteObjectHandler when generate ehcleanup inside ctor. Just add check for "complete object flag" before call to v-base dtor. Without my change: ehcleanup: ; preds = %ctor.skip_vbases %13 = cleanuppad within none [], !dbg !66 %14 = bitcast %struct.class_0* %this1 to i8*, !dbg !66 %15 = getelementptr inbounds i8, i8* %14, i64 8, !dbg !66 %16 = bitcast i8* %15 to %struct.class_2*, !dbg !66 call void @"\01??1class_2@@UEAA@XZ"(%struct.class_2* %16) #6 [ "funclet"(token %13) ], !dbg !66 cleanupret from %13 unwind to caller, !dbg !66 with my change: ehcleanup: ; preds = %ctor.skip_vbases %13 = cleanuppad within none [], !dbg !66 %14 = bitcast %struct.class_0* %this1 to i8*, !dbg !66 %15 = getelementptr inbounds i8, i8* %14, i64 8, !dbg !66 %16 = bitcast i8* %15 to %struct.class_2*, !dbg !66 %is_complete_object4 = icmp ne i32 %is_most_derived2, 0, !dbg !66 br i1 %is_complete_object4, label %Dtor.dtor_vbase, label %Dtor.skip_vbase, !d bg !66 Dtor.dtor_vbase: ; preds = %ehcleanup call void @"\01??1class_2@@UEAA@XZ"(%struct.class_2* %16) #6 [ "funclet"(token %13) ], !dbg !66 br label %Dtor.skip_vbase, !dbg !66 Dtor.skip_vbase: ; preds = %Dtor.dtor_vbase, %ehcleanup cleanupret from %13 unwind to caller, !dbg !66 Please let me know you need more info. Patch by Jennifer Yu. Differential Revision: https://reviews.llvm.org/D27358 llvm-svn: 288869
2016-12-07 08:21:45 +08:00
struct class_2 {
class_2();
virtual ~class_2();
};
struct class_1 : virtual class_2 {
class_1(){throw "Unhandled exception";}
virtual ~class_1() {}
};
struct class_0 : class_1 {
class_0() ;
virtual ~class_0() {}
};
class_0::class_0() {
// WIN32: define dso_local x86_thiscallcc %struct.class_0* @"??0class_0@@QAE@XZ"(%struct.class_0* returned %this, i32 %is_most_derived)
[MS-ABI]V-base dtor called more than needed when throw happens in v-base ctor in window. Need add "complete object flag" check in eh cleanup code. The problem only happen on window ( A MS-ABI issuer ) The nature of the problem is virtual base dtor called more than it is needed after exception throw in inheriting base class(with virtual bases) ctor. The root problem is when throw happen, not all virtual base classes have been contructed, so not all virtual base dtors are need to call for ehcleanup. clang has code to handle vbase initialization: basically add check for "complete object flag" before call to v-base ctor. But that part is missing for cleanup code. To fix this add similar code as v-base init to cleanup code, same algorithm. 1> Add new routine: EmitDtorCompleteObjectHandler With corresponding to EmitCtorCompleteObjectHandler 2> In the EmitDestructorCal Call EmitDtorCompleteObjectHandler when generate ehcleanup inside ctor. Just add check for "complete object flag" before call to v-base dtor. Without my change: ehcleanup: ; preds = %ctor.skip_vbases %13 = cleanuppad within none [], !dbg !66 %14 = bitcast %struct.class_0* %this1 to i8*, !dbg !66 %15 = getelementptr inbounds i8, i8* %14, i64 8, !dbg !66 %16 = bitcast i8* %15 to %struct.class_2*, !dbg !66 call void @"\01??1class_2@@UEAA@XZ"(%struct.class_2* %16) #6 [ "funclet"(token %13) ], !dbg !66 cleanupret from %13 unwind to caller, !dbg !66 with my change: ehcleanup: ; preds = %ctor.skip_vbases %13 = cleanuppad within none [], !dbg !66 %14 = bitcast %struct.class_0* %this1 to i8*, !dbg !66 %15 = getelementptr inbounds i8, i8* %14, i64 8, !dbg !66 %16 = bitcast i8* %15 to %struct.class_2*, !dbg !66 %is_complete_object4 = icmp ne i32 %is_most_derived2, 0, !dbg !66 br i1 %is_complete_object4, label %Dtor.dtor_vbase, label %Dtor.skip_vbase, !d bg !66 Dtor.dtor_vbase: ; preds = %ehcleanup call void @"\01??1class_2@@UEAA@XZ"(%struct.class_2* %16) #6 [ "funclet"(token %13) ], !dbg !66 br label %Dtor.skip_vbase, !dbg !66 Dtor.skip_vbase: ; preds = %Dtor.dtor_vbase, %ehcleanup cleanupret from %13 unwind to caller, !dbg !66 Please let me know you need more info. Patch by Jennifer Yu. Differential Revision: https://reviews.llvm.org/D27358 llvm-svn: 288869
2016-12-07 08:21:45 +08:00
// WIN32: store i32 %is_most_derived, i32* %[[IS_MOST_DERIVED_VAR:.*]], align 4
// WIN32: %[[IS_MOST_DERIVED_VAL:.*]] = load i32, i32* %[[IS_MOST_DERIVED_VAR]]
// WIN32: %[[SHOULD_CALL_VBASE_CTORS:.*]] = icmp ne i32 %[[IS_MOST_DERIVED_VAL]], 0
// WIN32: br i1 %[[SHOULD_CALL_VBASE_CTORS]], label %[[INIT_VBASES:.*]], label %[[SKIP_VBASES:.*]]
// WIN32: [[INIT_VBASES]]
// WIN32: br label %[[SKIP_VBASES]]
// WIN32: [[SKIP_VBASES]]
// ehcleanup:
// WIN32: %[[CLEANUPPAD:.*]] = cleanuppad within none []
// WIN32-NEXT: bitcast %{{.*}}* %{{.*}} to i8*
// WIN32-NEXT: getelementptr inbounds i8, i8* %{{.*}}, i{{.*}} {{.}}
[MS-ABI]V-base dtor called more than needed when throw happens in v-base ctor in window. Need add "complete object flag" check in eh cleanup code. The problem only happen on window ( A MS-ABI issuer ) The nature of the problem is virtual base dtor called more than it is needed after exception throw in inheriting base class(with virtual bases) ctor. The root problem is when throw happen, not all virtual base classes have been contructed, so not all virtual base dtors are need to call for ehcleanup. clang has code to handle vbase initialization: basically add check for "complete object flag" before call to v-base ctor. But that part is missing for cleanup code. To fix this add similar code as v-base init to cleanup code, same algorithm. 1> Add new routine: EmitDtorCompleteObjectHandler With corresponding to EmitCtorCompleteObjectHandler 2> In the EmitDestructorCal Call EmitDtorCompleteObjectHandler when generate ehcleanup inside ctor. Just add check for "complete object flag" before call to v-base dtor. Without my change: ehcleanup: ; preds = %ctor.skip_vbases %13 = cleanuppad within none [], !dbg !66 %14 = bitcast %struct.class_0* %this1 to i8*, !dbg !66 %15 = getelementptr inbounds i8, i8* %14, i64 8, !dbg !66 %16 = bitcast i8* %15 to %struct.class_2*, !dbg !66 call void @"\01??1class_2@@UEAA@XZ"(%struct.class_2* %16) #6 [ "funclet"(token %13) ], !dbg !66 cleanupret from %13 unwind to caller, !dbg !66 with my change: ehcleanup: ; preds = %ctor.skip_vbases %13 = cleanuppad within none [], !dbg !66 %14 = bitcast %struct.class_0* %this1 to i8*, !dbg !66 %15 = getelementptr inbounds i8, i8* %14, i64 8, !dbg !66 %16 = bitcast i8* %15 to %struct.class_2*, !dbg !66 %is_complete_object4 = icmp ne i32 %is_most_derived2, 0, !dbg !66 br i1 %is_complete_object4, label %Dtor.dtor_vbase, label %Dtor.skip_vbase, !d bg !66 Dtor.dtor_vbase: ; preds = %ehcleanup call void @"\01??1class_2@@UEAA@XZ"(%struct.class_2* %16) #6 [ "funclet"(token %13) ], !dbg !66 br label %Dtor.skip_vbase, !dbg !66 Dtor.skip_vbase: ; preds = %Dtor.dtor_vbase, %ehcleanup cleanupret from %13 unwind to caller, !dbg !66 Please let me know you need more info. Patch by Jennifer Yu. Differential Revision: https://reviews.llvm.org/D27358 llvm-svn: 288869
2016-12-07 08:21:45 +08:00
// WIN32-NEXT: bitcast i8* %{{.*}} to %{{.*}}*
// WIN32-NEXT: %[[SHOULD_CALL_VBASE_DTOR:.*]] = icmp ne i32 %[[IS_MOST_DERIVED_VAL]], 0
// WIN32-NEXT: br i1 %[[SHOULD_CALL_VBASE_DTOR]], label %[[DTOR_VBASE:.*]], label %[[SKIP_VBASE:.*]]
// WIN32: [[DTOR_VBASE]]
// WIN32-NEXT: call x86_thiscallcc void @"??1class_2@@UAE@XZ"
[MS-ABI]V-base dtor called more than needed when throw happens in v-base ctor in window. Need add "complete object flag" check in eh cleanup code. The problem only happen on window ( A MS-ABI issuer ) The nature of the problem is virtual base dtor called more than it is needed after exception throw in inheriting base class(with virtual bases) ctor. The root problem is when throw happen, not all virtual base classes have been contructed, so not all virtual base dtors are need to call for ehcleanup. clang has code to handle vbase initialization: basically add check for "complete object flag" before call to v-base ctor. But that part is missing for cleanup code. To fix this add similar code as v-base init to cleanup code, same algorithm. 1> Add new routine: EmitDtorCompleteObjectHandler With corresponding to EmitCtorCompleteObjectHandler 2> In the EmitDestructorCal Call EmitDtorCompleteObjectHandler when generate ehcleanup inside ctor. Just add check for "complete object flag" before call to v-base dtor. Without my change: ehcleanup: ; preds = %ctor.skip_vbases %13 = cleanuppad within none [], !dbg !66 %14 = bitcast %struct.class_0* %this1 to i8*, !dbg !66 %15 = getelementptr inbounds i8, i8* %14, i64 8, !dbg !66 %16 = bitcast i8* %15 to %struct.class_2*, !dbg !66 call void @"\01??1class_2@@UEAA@XZ"(%struct.class_2* %16) #6 [ "funclet"(token %13) ], !dbg !66 cleanupret from %13 unwind to caller, !dbg !66 with my change: ehcleanup: ; preds = %ctor.skip_vbases %13 = cleanuppad within none [], !dbg !66 %14 = bitcast %struct.class_0* %this1 to i8*, !dbg !66 %15 = getelementptr inbounds i8, i8* %14, i64 8, !dbg !66 %16 = bitcast i8* %15 to %struct.class_2*, !dbg !66 %is_complete_object4 = icmp ne i32 %is_most_derived2, 0, !dbg !66 br i1 %is_complete_object4, label %Dtor.dtor_vbase, label %Dtor.skip_vbase, !d bg !66 Dtor.dtor_vbase: ; preds = %ehcleanup call void @"\01??1class_2@@UEAA@XZ"(%struct.class_2* %16) #6 [ "funclet"(token %13) ], !dbg !66 br label %Dtor.skip_vbase, !dbg !66 Dtor.skip_vbase: ; preds = %Dtor.dtor_vbase, %ehcleanup cleanupret from %13 unwind to caller, !dbg !66 Please let me know you need more info. Patch by Jennifer Yu. Differential Revision: https://reviews.llvm.org/D27358 llvm-svn: 288869
2016-12-07 08:21:45 +08:00
// WIN32: br label %[[SKIP_VBASE]]
// WIN32: [[SKIP_VBASE]]
}
namespace PR37146 {
// Check that IRGen doesn't emit calls to synthesized destructors for
// non-trival C structs.
// WIN32: define dso_local void @"?test@PR37146@@YAXXZ"()
// WIN32: call void @llvm.memset.p0i8.i32(
// WIN32: call i32 @"?getS@PR37146@@YA?AUS@1@XZ"(
// WIN32: call void @"?func@PR37146@@YAXUS@1@0@Z"(
// WIN32-NEXT: ret void
// WIN32-NEXT: {{^}$}}
struct S {
int f;
};
void func(S, S);
S getS();
void test() {
func(getS(), S());
}
}