C++ code can return values that depend on the storage that backs the
references that were passed in as argument. Thus, swift should not
introdue temporary copies of that storage before invoking those
functions as they could result in lifetime issues.
In GenCall, fix the IR gen for C++ method calls as under MSVC as the
calling conventions for free functions and C++ methods can be
different. This also fixes the missing inreg (on sret arguments)
issues on Windows ARM64. Also refactor to use CGFunctionInfo
returnInfo isSretAfterThis to detect when to reorder the sret and the
this arguments under MSVC.
In ClagImporter, don't drop the return type for the compound
assignment operators such as operator+= when the return value is a
reference so that the CGFunctionInfo will be correctly indicate an
indirect return for the compound assignment operators.
On Windows/AArch64, a different register is used between when an
arugment is both inreg and sret (X0 or X1) and when it is just sret
(X8) as the following comment indicates:
46fe36a429/llvm/lib/Target/AArch64/AArch64CallingConvention.td (L42)
```
// In AAPCS, an SRet is passed in X8, not X0 like a normal pointer parameter.
// However, on windows, in some circumstances, the SRet is passed in X0 or X1
// instead. The presence of the inreg attribute indicates that SRet is
// passed in the alternative register (X0 or X1), not X8:
// - X0 for non-instance methods.
// - X1 for instance methods.
// The "sret" attribute identifies indirect returns.
// The "inreg" attribute identifies non-aggregate types.
// The position of the "sret" attribute identifies instance/non-instance
// methods.
// "sret" on argument 0 means non-instance methods.
// "sret" on argument 1 means instance methods.
CCIfInReg<CCIfType<[i64],
CCIfSRet<CCIfType<[i64], CCAssignToReg<[X0, X1]>>>>>,
CCIfSRet<CCIfType<[i64], CCAssignToReg<[X8]>>>,
```
So missing/dropping inreg can cause a codegen bug.
This is a partial fix for #74866
This fixes `Interop/Cxx/class/method/methods-this-and-indirect-return-irgen-itanium.swift` by adjusting the expected function signature.
On rebranch, the actual signature is:
```
define linkonce_odr void @_ZN10HasMethods28nonConstPassThroughAsWrapperEi(ptr dead_on_unwind noalias writable sret(%struct.NonTrivialInWrapper) align 4 %agg.result, ptr noundef nonnull align 1 dereferenceable(1) %this, i32 noundef %a)
```
rdar://127263407
When we have both const and non-const version of a function, we import
the non-cont version with the "Mutating" suffix. This logic, however, is
redundant for static member functions as those can never be marked as
"const" since they don't have a "self" or "this" to mutate.
rdar://120858502
Clang is using C++17 standard version by default since Clang 16.
Swift’s ClangImporter should do the same, to make sure that clients who run clang and then swiftc without explicit std version see consistent behavior.
rdar://125777068
Fix a bug in expandExternalSignatureTypes where it wasn't annotating a function call parameter type with sret when the result was being returned indirectly.
The bug was causing calls to ObjC methods that return their results indirectly to crash.
Additionally, fix the return type for C++ constructors computed in expandExternalSignatureTypes. Previously, the return type was always void even on targets that require constructors to return this (e.g., Apple arm64), which was causing C++ constructor thunks to be emitted needlessly.
Resolves rdar://121618707
Handle a slight LLVM IR difference in the second parameter (inreg) of
a function between x86_64 and aarch64. These are how Clang on windows
emits the code.
The test msvc-abi-return-indirect-trivial-record.swift fails on
windows aarch64 because the type of the third argument of
@"??YLoadableIntWrapper@@qeaa?AU0@U0@@z"
(@LoadableIntWrapper::operator+=(LoadableIntWrapper)) is i64 which
differs from i32 for x86_64. Accommodate this as this is how Clang
generates code. See https://godbolt.org/z/G76aMr5M9.
Fixes https://github.com/apple/swift/issues/66326
This allows us to reneable Windows method tests. Note that Windows still has
a broken convention for non-trivial record with non-trivial destructor but
trivial copy-constructor, so classes in the methods.swift test need an explicit
copy constructor.
Fixes rdar://88391102
When importing a C++ struct that contains two methods that only differ in const-ness, we append `Mutating` to the name of the non-const method to make it possible to call from Swift unambiguously.
Unfortunately that logic was dependent on the order in which we import methods of a class: the `Mutating` suffix was added when another method with the same name was already imported.
This caused lookup failures, and the behavior was incorrect when the pair of methods return instances of an unsafe type: the const overload was renamed as `Unsafe` properly, but the non-const overload was not renamed.
