Just because the type of the initializer expression is an opaque return type,
does not mean it is the opaque return type *for the variable being initialized*.
It looks like there is a bit of duplicated logic and layering violations going
on so I only fixed one caller of openOpaqueType(). This addresses the test case
in the issue. For the remaining calls I added FIXMEs to investigate what is
going on.
Fixes https://github.com/swiftlang/swift/issues/73245.
Fixes rdar://127180656.
Instead, each scan's 'ModuleDependenciesCache' will hold all of the data corresponding to discovered module dependencies.
The initial design presumed the possibility of sharing a global scanning cache amongs different scanner invocations, possibly even different concurrent scanner invocations.
This change also deprecates two libSwiftScan entry-points: 'swiftscan_scanner_cache_load' and 'swiftscan_scanner_cache_serialize'. They never ended up getting used, and since this code has been largely stale, we are confident they have not otherwise had users, and they do not fit with this design.
A follow-up change will re-introduce moduele dependency cache serialization on a per-query basis and bring the binary format up-to-date.
When serializing the module interface path of an interface that
is part of the SDK, we serialize relative to the SDK path. During
deserialization we need to know if a path was serialized relative
to the SDK or not. The existing logic assumes any relative path
has been serialized relative to the SDK, which makes it impossible
to compile modules from relative swiftinterface paths that are not
part of the SDK.
Update the swiftmodule file to include an attribute to show if the
path was serialized relative to the SDK or not, which is used
during deserialization to correctly reconstruct the interface path.
In rare scenarios, Swift was emitting diagnostics that looked like this:
```
warning: 'import_owned' swift attribute ignored on type 'basic_string': type is not copyable or destructible
```
This change makes sure the compiler does not emit these (incorrect) warnings. See the inline comment for more details.
Now that IUOs are supported for compound function
references, we can properly set the compound bit
here.
This is a source breaking change since this used
to be legal:
```swift
struct S {
static func foo(x: Int) -> Self { .init() }
}
let _: S = .foo(x:)(x: 0)
```
However I somewhat doubt anyone is intentionally
writing code like that.
Many APIs using nonescapable types would like to vend interior pointers to their
parameter bindings, but this isn't normally always possible because of representation
changes the caller may do around the call, such as moving the value in or out of memory,
bridging or reabstracting it, etc. `@_addressable` forces the corresponding parameter
to be passed indirectly in memory, in its maximally-abstracted representation.
[TODO] If return values have a lifetime dependency on this parameter, the caller must
keep this in-memory representation alive for the duration of the dependent value's
lifetime.
The symbol graph output from a module can contain an arbitrary number of
files, depending on what extensions it contains, so cache a list of
symbol graph files with their base name and contents so that they can be
replayed.
rdar://140286819
If the output loading failed after cache key lookup, treat that as a
warning and resume as if that is a cache miss. This is not a valid
configuration for builtin CAS but can happen for a remote CAS service
that failed to serve the output. Instead of failing, we should continue
to compile to avoid disruptive failures.
rdar://140822432
This change addresses the following issue: when an error is being wrapped in a warning, the diagnostic message will use the wrapper's `DiagGroupID` as the warning's name. However, we want to retain the original error's group for use. For example, in Swift 5, async_unavailable_decl is wrapped in error_in_future_swift_version. When we print a diagnostic of this kind, we want to keep the `DiagGroupID` of `async_unavailable_decl`, not that of `error_in_future_swift_version`.
To achieve this, we add `DiagGroupID` to the `Diagnostic` class. When an active diagnostic is wrapped in DiagnosticEngine, we retain the original `DiagGroupID`.
For illustration purposes, this change also introduces a new group: `DeclarationUnavailableFromAsynchronousContext`.
With this change, we produce errors and warnings of this kind with messages like the following:
```
global function 'fNoAsync' is unavailable from asynchronous contexts [DeclarationUnavailableFromAsynchronousContext]
global function 'fNoAsync' is unavailable from asynchronous contexts; this is an error in the Swift 6 language mode [DeclarationUnavailableFromAsynchronousContext]
```
Also remove the underlying `SemanticUnavailableAttrRequest`, which used memory
very inefficiently in order to cache a detailed answer to what was usually a
much simpler question.
The only remaining use of `Decl::getSemanticUnavailableAttr()` that actually
needed to locate the semantic attribute making a declaration unavailable was in
`TypeCheckAttr.cpp`. The implementation of the request could just be used
directly in that one location. The other remaining callers only needed to know
if the decl was unavailable or not, which there are simpler queries for.
# Please enter the commit message for your changes. Lines starting
Previously we would not propagate those into the generated distributed
actor, making a lot of generic distributed actor protocols impossible to
express.
We indeed cannot handle protocols WITHOUT primary associated types, but
we certainly can handle them with!
This resolves rdar://139332556
Many existing C APIs for retaining references, including Apple's own, return
the reference. Support this pattern, along with the existing void return
signature, with when importing reference types from C++.
As the utility runs, new gens may become local: as access scopes are
determined to contain deinit barriers, their `end_access` instructions
become kills; if such an `end_access` occurs in the same block above an
initially-non-local gen, that gen is now local.
Previously, it was asserted that initially-non-local gens would not
encounter when visiting the block backwards from that gen. Iteration
would also _stop_ at the discovered kill, if any. As described above,
the assertion was incorrect.
Stopping at the discovered kill was also incorrect. It's necessary to
continue walking the block after finding such a new kill because the
book-keeping the utility does for which access scopes contain barriers.
Concretely, there are two cases:
(1) It may contain another `end_access` and above it a deinit barrier
which must result in that second scope becoming a deinit barrier.
(2) Some of its predecessors may be in the region, all the access scopes
which are open at the begin of this block must be unioned into the set
of scopes open at each predecessors' end, and more such access scopes
may be discovered above the just-visited `end_access`.
Here, both the assertion failure and the early bailout are fixed by
walking from the indicated initially-non-local gen backwards over the
entire block, regardless of whether a kill was encountered. If a kill
is encountered, it is asserted that the kill is an `end_access` to
account for the case described above.
rdar://139840307
In terms of the test suite the only difference is that we allow for non-Sendable
types to be returned from nonisolated functions. This is safe due to the rules
of rbi. We do still error when we return non-Sendable functions across isolation
boundaries though.
The reason that I am doing this now is that I am implementing a prototype that
allows for nonisolated functions to inherit isolation from their caller. This
would have required me to implement support both in Sema for results and
arguments in SIL. Rather than implement results in Sema, I just finished the
work of transitioning the result checking out of Sema and into SIL. The actual
prototype will land in a subsequent change.
rdar://127477211
