Previously, when evaluating a `#if canImport(Module, _version: 42)` directive the compiler could diagnose and ignore the directive under the following conditions:
- The associated binary module is corrupt/bogus.
- The .tbd for an underlying Clang module is missing a current-version field.
This behavior is surprising when there is a valid `.swiftinterface` available and it only becomes apparent when building against an SDK with an old enough version of the module that the version in the `.swiftinterface` is too low, making this failure easy to miss. Some modules have different versioning systems for their Swift and Clang modules and it can also be intentional for a distributed binary `.swiftmodule` to contain bogus data (to force the compiler to recompile the `.swiftinterface`) so we need to handle both of these cases gracefully and predictably.
Now the compiler will enumerate all module loaders, ask each of them to attempt to parse the module version and then consistently use the parsed version from a single source. The `.swiftinterface` is preferred if present, then the binary module if present, and then finally the `.tbd`. The `.tbd` is still always used exclusively for the `_underlyingVersion` variant of `canImport()`.
Resolves rdar://88723492
When we get rid of `LeaveClosureBodiesUnchecked` we no longer save closure types to the AST and thus also don’t save their actor isolation to the AST. Hence, we need to extract types and actor isolations of parent closures from the constraint system solution instead of the AST. This prepares `ActorIsolationChecker` to take custom functions to determine the type of an expression or the actor isolation of a closure.
function body, map the result builder type into context.
This was already done for inferred result builder attributes; now,
the constraint system will map the builder type into context for all
result builder attributes applied to computed properties/functions.
We intended to introduce AST conversions that strip concurrency
attributes off of types associated with `@preconcurrency` decls.
But for VarDecl references, we stripped it too early, leading to
things like a MemberVarDecl that doesn't have `@Sendable` in its
result type, but the VarDecl it refers to does have it.
That caused crashes in SIL where types didn't match up. This patch
fixes things by delaying the stripping until the right point.
resolves rdar://98018067
This change tweaks the 'GlobalModuleDependenciesCache', which persists across scanner invocations with the same 'DependencyScanningTool' to no longer cache discovered Clang modules.
Doing so felt like a premature optimization, and we should instead attempt to share as much state as possible by keeping around the actual Clang scanner's state, which performs its own caching. Caching discovered dependencies both in the Clang scanner instance, and in our own cache is much more error-prone - the Clang scanner has a richer context for what is okay and not okay to cache/re-use.
Instead, we still cache discovered Clang dependencies *within* a given scan, since those are discovered using a common Clang scanner instance and should be safe to keep for the duration of the scan.
This change should make it simpler to pin down the core functionality and correctness of the scanner.
Once we turn our attention to the scanner's performance, we can revisit this strategy and optimize the caching behaviour.
If `buildBlock` is also unavailable, or the
builder itself is unavailable, continue to solve
using `buildPartialBlock` to get better
diagnostics.
This behavior technically differs from what is
specified in SE-0348, but only affects the invalid
case where no builder methods are available to use.
In particular, this improves diagnostics for
RegexComponentBuilder when the deployment target
is too low. Previously we would try to solve using
`buildBlock` (as `buildPartialBlock` is unavailable),
but RegexComponentBuilder only defines `buildBlock`
for the empty body case, leading to unhelpful
diagnostics that ultimately preferred not to use
the result builder at all.
rdar://97533700
Previously we would cache the result of the first
query, with any further query of
`ResultBuilder::supports` ignoring the
`checkAvailability` parameter. Separate out the
availability checking such that we compute the
information up front, and adjust the result
depending on `checkAvailability`.
An extraneous argument doesn't have a corresponding parameter so the
information object for such an argument is not safe to produce.
Resolves: https://github.com/apple/swift/issues/60436
Resolves: rdar://98304482
In some circumstances opened type might not have a fixed binding.
A good example could be dependent sub-component produced for
result builder transformed code (connected via one-way constraints),
in such a case `OpenedTypes` would have outer generic parameters
but they might not be bound yet, so they have to be printed as
type variables.
Since result builder is just an AST transformation, the result
of a successful transform could be cache and reused with a different
`$__builderSelf` type.
The transform changes closure body into a multi-statement closure
with all of the implicit result builder calls and type-checks it
like a regular closure.
There are a couple of result builder specific changes mentioned below,
otherwise the logic to generate constraints and apply solutions is
unchanged:
- Placeholder variable: A variable declaration that doesn't have a
type deduced and infers it from its first use. If such a variable
has an initializer, it would be type-checked during solution application.
- TypeJoinExpr - an implicit expression that refers to a "join" variable
and a set of expressions that should all produce the same type that
becomes a type of a "join" variable.
Closures and functions that have result builder transform applied have to be
handled specially by the constraint generator because transformed body is
associated with the context (function or closure) only during solution application.
Move off `Type` based requests and onto `Decl`
based requests, utilizing name lookup's
`extractDirectlyReferencedNominalTypes` utility.
This allows us to better cache the results, and
avoids the need to guard against type variable
inputs when deciding whether or not to cache.
In ambiguity scenarios solutions are not applied back to the constraint
system, so it might not always have contextual type information when it
was recorded e.g. for a multi-statement closure.
Resolves: rdar://97396399
This was already enabled as part of `-enable-implicit-dynamic` but this
new flag allows turning on opaque type erasure all by itself whether or
not `dynamic` is added explicitly.
rdar://97375478
