Add a codegen test that checks destructors for temporaries inside asm statements are called.

I think the assertion is assuming StmtExpr is created only for GNU statement expressions, but it's created for asm statements too (see https://github.com/llvm/llvm-project/commit/3050d9bdb84e322e122219d54aedfe2d8ef7c51c).

Does debug info need information on deduction guides? I think it's safe to skip deduction guides as they are used only for template argument deduction.

The assertion assert(FromType->isIntegralOrUnscopedEnumerationType()) in StandardConversionSequence::getNarrowingKind fails when the invalid initialization is performed.

Without this fix, CXXNameMangler::mangleUnresolvedName crashes when it tries to mangle the name of a CXXDeductionGuideDecl.

Is it not possible to handle this similarly to volatile unsigned? If I replace _Atomic unsigned with volatile unsigned, I see LookupOverloadedBinOp succeed without having to strip volatile because addAssignmentArithmeticOverloads adds candidates with volatile types.

LGTM with one minor comment.

It looks like this commit is causing https://green.lab.llvm.org/green/job/clang-stage1-cmake-RA-expensive/22234/testReport/junit/ThreadSanitizer-Unit/rtl___TsanRtlTest-x86_64-Test_ThreadSanitizer/RaceWithDeadThread/ to fail. Can you take a look?

LGTM

It looks like this is causing a test to fail:
https://green.lab.llvm.org/green/job/clang-san-iossim/5229/consoleFull#711521816a1ca8a51-895e-46c6-af87-ce24fa4cd561

I removed -O1 from the command line:
https://github.com/llvm/llvm-project/commit/6ec79a15cbe9539faf121b5ad39f195dc611fc09

Sorry, I didn't notice you left a comment until after committing the patch. I agree that -O1 can be removed.

Yes, this test was susceptible to changes made to llvm passes and the pass pipeline.

Yes, check-clang passes.

LGTM

The ObjC ARC changes look good.

Update comments.

Build the map (Captures) after building the vector (SortedCaptures).

This commit seems to have caused a test to fail: https://green.lab.llvm.org/green/job/clang-stage1-cmake-RA-incremental/26118/testReport/

Since the strings are used in other places (e.g., PreISelIntrinsicLowering.cpp) too, you could place them in a header in llvm/IR or llvm/Analysis. But I think it's okay to do so in a separate patch.

In D115230#3180259, @fhahn wrote:

Updated to only use bundles during expansion for Darwin platforms. This keeps it in sync with TargetMachine (for now the bundle expansion pass is only scheduled on Darwin platforms).

@ille, are you still working on this patch?

LGTM

LGTM

The lifetime of an object won't be extended in the following case since the variable isn't captured by a block:

LGTM

LGTM

LGTM

Improve comment.

No, only arm64 and x86-64 can correctly handle the operand bundle in the backend. I should mention that this only changes the middle-end passes and we still need to make changes to the backend to handle arbitrary targets.

id is a typedef and @class is an ObjCInterface now, so it looks like it's safe to remove ObjCObjectPointer case then.

I'm not sure why ObjCObjectPointerType was originally a specifier, but this patch looks safe to me. ObjCObjectPointerType is never passed to TypePrinter::printBefore and QualType GetBaseType, which are the two functions that use isSpecifierType, AFAICT.

I was just asking for test cases as I thought it was already a driver option, but it turns out it wasn't. I agree that this shouldn't be a driver option.

Can you add a driver test, which tests both -fobjc-avoid-heapify-local-blocks and -fno-objc-avoid-heapify-local-blocks?

Do you need arc-blocks-avoid-heapify.m? It seems like the other tests already cover all the cases we care about.

Rebase and ping.