* [IRGen] Handle complex single payload enum cases
rdar://110138498
Handles single payload enum cases with more complex bit patterns (e.g. >64 bits or scattered) by storing a relative pointer to a function that reads the tag.
* Use proper symbol for enum tag helper
rdar://110088270
When extra tag bytes are used in single payload enums, the generated layout string does not include the tag bytes in its offset after the enum payload. This causes subsequent ref count operations to use the wrong memory location.
When performing an implicit module build, the frontend was prepending
`-target-min-inlining-target target` to the command line. This was overriding
the implicit `-target-min-inlining-target min` argument that is implied when
`-library-level api` is specified. As a result, the wrong overload could be
picked when compiling the body of an inlinable function to SIL for emission
into the client, potentially resulting in crashes when the client of the module
is back deployed to an older OS.
Resolves rdar://109336472
Some notes:
1. I put in both a swiftpm like test case and a library evolution test case. I
also updated the moveonly_deinit serialization swift test to show that we
actually serialize the deinit.
2. I changed when we emit the deinit table to only be when we have a type with
an actual value type destructor. Notably this doesn't include classes today so
as a side-effect, we no longer attempt to devirtualize moveonly class deinits.
This doesn't affect anything we are trying to actually do since we do not
support noncopyable classes today. With that in mind, I changed one test that
was showing that deinit devirtualization worked to use a struct with deinit
instead of a class.
rdar://109679168
In LLVM unified builds `%swift_obj_root` points to `<LLVM build dir>/tools/swift`,
and folders like `bin`, `lib` and `share` are not under `swift_obj_root`, which
makes some tests fail.
For the cases in which `%swift_obj_root/lib` was used, replace it by
using `%swift-lib-dir` instead. Replicate `%swift-lib-dir` to create
`%swift-bin-dir` and `%swift-share-dir`, and use those instead of
`%swift_obj_root/bin` and `%swift_obj_root/share`.
This alternates work both in Swift build-script builds and also in LLVM
unified builds.
* [IRGen] Support additional single payload enum cases in layout strings
rdar://105837101
Adds layout string support for single payload enums with simple (non-scattered) extra inhabitant patterns
* Add more test cases
When `-unavailable-decl-optimization=complete` is specified, exclude
unavailable enum cases from the runtime layout of enums with payloads. Without
this, the type metadata for unavailable types may be referenced by enum cases
with unavailable payloads and cause linker failures.
Resolves rdar://107483852
You can overload a function based on its `async`-ness, and
resolution is carried out based on async-ness of calling
context.
But during protocol conformance checking, for an `async`
requirement we were accidentally choosing the non-`async
overload instead of the `async` one. The `async` one is
the choice people would expect, since the `async` requirement
is in essence the "context" that forwards to the underlying
witness. This intended behavior is also inferred from:
https://github.com/apple/swift/pull/40088
The problem boiled down to a bad check when categorizing the
witness matches prior to ranking them.
Resolves rdar://109135488 / https://github.com/apple/swift/issues/60318
SE-390 concluded with choosing the keyword discard rather than forget for
the statement that disables the deinit of a noncopyable type. This commit
adds parsing support for `discard self` and adds a deprecation warning for
`_forget self`.
rdar://108859077
* [IRGen] Respect optionality of unowned(unsafe) reference properties
rdar://108705703
When generating the type info for fields marked unowned(unsafe) of an optional reference (e.g. AnyObject?), we dropped the fact that it was optional along the way. This caused incorrect tags to be returned when on object of the type was stored in an optional itself and the unowned property contained `nil`. In those cases the outer optional appeared to be `nil` as well, even if it in fact contained a value.
* Fix additional case
When `-unavailable-decl-optimization=stub` is specified, insert a call to
`_diagnoseUnavailableCodeReached()` at the beginning of the function to cause
it to trap if executed at run time.
Part of rdar://107388493
These are the same semantically, just the mangling is slightly different. The
benefit of doing this is that we are actually testing what we expect our users
to do.
rdar://108511703
The reason to do this is that:
1. Otherwise, we do not emit markers when someone attempts to consume the let.
We need the no_consume_or_assign to be there.
2. We need to insert assign_but_not_consuming so that DI can properly check lets
that are conditionally initialized and convert them to
initable_but_not_consuming.
I included a full definite_init SIL test that validates that we get the correct
codegen after DI in this case and emit the appropriate error as well.
rdar://108511534
The reason why I am doing this is that otherwise if one has a function that
takes both a guaranteed and an owned parameter, we will break OSSA invariants
since the load [take] will invalidate the load_borrow. So instead, we put in a
load_borrow knowing that the move checker will convert it to a load_borrow
assuming that the two pass exclusivity checking.
NOTE: Because of some missing functionality in subsequent tests, I had to
disable one test (moveonly_escaping_definite_initialization.swift) and also add
some checks for copy of noncopyable object errors. They will go away in the next
2 commits.
rdar://108510987
* [IRGen] Fix layout string generation for pre-specialized metadata
rdar://108012057
Pre-specialized metadata has to be specifically handled by using the bound generic type instead of the unbound one. All the necessary information is already being passed down as BoundGenericTypeCharacteristics, we just need to apply them when present.
* Add tests and a few fixes
* Fixes after rebase
* Attempt to fix Windows linker issue in test
Normally, if we project from a mutable class ivar or global variable, we'll
load a copy in a tight access scope and then project from the copy, in order to
minimize the potential for exclusivity violations while working with classes and
copyable values. However, this is undesirable when a value is move-only, since
copying is not allowed; borrowing the value in place is the expected and only possible
behavior. rdar://105794506
* [IRGen] Reject enums with inaccessible tpye metadata in layout string generation
rdar://107679697
Because we are currently handling most enums in layout strings by going through the metadata, we have to ensure that the metadata is accessible from the current module and reject the enum otherwise.
* Use proper mechanism to create and reference dylib in test
* Fix linking
* Add rpath to test dylib
