On Jan 4, 2013, at 4:34 PM, Fariborz Jahanian <[email protected]> wrote: > On Jan 4, 2013, at 3:56 PM, John McCall wrote: >> On Jan 4, 2013, at 3:32 PM, Fariborz Jahanian <[email protected]> wrote: >>> Author: fjahanian >>> Date: Fri Jan 4 17:32:24 2013 >>> New Revision: 171555 >>> >>> URL: http://llvm.org/viewvc/llvm-project?rev=171555&view=rev >>> Log: >>> objective-C arc: in copy helper function for >>> __strong __block variables, perform objc_storeStrong on >>> source and destination instead of direct move. This >>> is done with -O0 and to improve some analysis. >>> // rdar://12530881 >>> >>> Added: >>> cfe/trunk/test/CodeGenObjC/arc-unoptimized-byref-var.m >>> Modified: >>> cfe/trunk/lib/CodeGen/CGBlocks.cpp >>> >>> Modified: cfe/trunk/lib/CodeGen/CGBlocks.cpp >>> URL: >>> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGBlocks.cpp?rev=171555&r1=171554&r2=171555&view=diff >>> ============================================================================== >>> --- cfe/trunk/lib/CodeGen/CGBlocks.cpp (original) >>> +++ cfe/trunk/lib/CodeGen/CGBlocks.cpp Fri Jan 4 17:32:24 2013 >>> @@ -1565,6 +1565,11 @@ >>> llvm::Value *null = >>> >>> llvm::ConstantPointerNull::get(cast<llvm::PointerType>(value->getType())); >>> >>> + if (CGF.CGM.getCodeGenOpts().OptimizationLevel == 0) { >>> + CGF.EmitARCStoreStrongCall(destField, value, /*ignored*/ true); >>> + CGF.EmitARCStoreStrongCall(srcField, null, /*ignored*/ true); >>> + return; >>> + } >>> llvm::StoreInst *store = CGF.Builder.CreateStore(value, destField); >>> store->setAlignment(Alignment.getQuantity()); >>> >>> >>> Added: cfe/trunk/test/CodeGenObjC/arc-unoptimized-byref-var.m >>> URL: >>> http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGenObjC/arc-unoptimized-byref-var.m?rev=171555&view=auto >>> ============================================================================== >>> --- cfe/trunk/test/CodeGenObjC/arc-unoptimized-byref-var.m (added) >>> +++ cfe/trunk/test/CodeGenObjC/arc-unoptimized-byref-var.m Fri Jan 4 >>> 17:32:24 2013 >>> @@ -0,0 +1,14 @@ >>> +// RUN: %clang_cc1 -triple x86_64-apple-darwin10 -emit-llvm -fblocks >>> -fobjc-arc -fobjc-runtime-has-weak -o - %s | FileCheck >>> -check-prefix=CHECK-UNOPT %s >>> +// rdar://12530881 >>> + >>> +void test19() { >>> + __block id x; >>> +// CHECK-UNOPT: define internal void @__Block_byref_object_copy >>> +// CHECK-UNOPT: [[X:%.*]] = getelementptr inbounds [[BYREF_T:%.*]]* >>> [[VAR:%.*]], i32 0, i32 6 >>> +// CHECK-UNOPT: [[X2:%.*]] = getelementptr inbounds [[BYREF_T:%.*]]* >>> [[VAR1:%.*]], i32 0, i32 6 >>> +// CHECK-UNOPT-NEXT: [[SIX:%.*]] = load i8** [[X2]], align 8 >>> +// CHECK-UNOPT-NEXT: call void @objc_storeStrong(i8** [[X]], i8* [[SIX]]) >>> nounwind >>> +// CHECK-UNOPT-NEXT: call void @objc_storeStrong(i8** [[X2]], i8* null) >>> nounwind >>> +// CHECK-UNOPT-NEXT: ret void >> >> You need to initialize the destination to null (with a simple store) before >> doing a storeStrong to it. > In r171572. This is I assume to prevent storeStrong from releasing the old > value. I guess this should not happen > in this context,
Right. objc_storeStrong is an assignment, not an initialization; it reads the old value out of the destination and releases it. When moving a __block variable to the heap, libclosure basically allocates some memory, sets up the byref header, and then checks whether there's a copy helper; if there is, it invokes it immediately. Therefore, when we're invoked, the destination address is freshly-allocated memory, and we cannot rely on it having been zero-initialized. John. _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
