This broke when we split `@execution(...)` into `@concurrent` and
`nonisolated(nonsending)` because the latter became its own `TypeRepr`,
whereas the condition for whether to attempt migration diagnostics
inside `resolveASTFunctionType` is still based on the function type's
attributes alone.
Use `resetBenignCodeGenOptions()` from clang dependency scanner to clear
the swift explicit module build cc1 arguments. This fixes the problem
that CurrentWorkingDirectory is leaking through
`-fcoverage-compilation-dir` that can cause extra module variants when
caching is enabled. This also avoid the duplicating the logics for
clearing CodeGen options inside Swift.
rdar://151395300
rdar://151176697
While generic types generally have layout strings (when enabled), non-copyable types don't, so
we have to make sure the flag does not get set.
We deallocate an instruction's packs at points where no further
control flow path uses the value. In the case of an alloc_stack,
this will be right after the dealloc_stack. Thus, if alloc_stack
allocates some packs to build type metadata for a tuple type
that contains a pack, and then proceeds to allocate a value
large enough to hold the tuple, we will free the second allocation
first, before we free the pack, as expected.
However, after stack allocating the value, alloc_stack does
some further work to emit debug info. This could result in
emission of additional metadata packs.
Split up the debug info emission into two parts; the first we do
before we perform the stack allocation, the rest we do after.
- Fixes https://github.com/swiftlang/swift/issues/67702.
- Fixes rdar://problem/141363236.
We need the C++ interop support library to make the build work, as
the Runtime module uses C++ interop.
(Also, we should build it, because some user programs might want to
use C++ interop.)
rdar://147201087
There were two problems that have been there for years:
- SubstitutionMap::lookupConformance() assumes that every concrete conformance
in the path has a root normal conformance. But calling getRootNormalConformance()
on a self conformance would assert.
- SelfProtocolConformance::getAssociatedConformance() was not implemented. While
self-conforming protocols do not have associated types, they can inherit from
other protocols, and we model this with an associated conformance requirement
having a subject type of `Self`.
Both problems were hidden by the fact that ProtocolConformanceRef::subst()
directly handled self-conforming existentials without calling into the
substitution map. But that is the wrong place for other reasons. The refactoring
in a209ff8869 exposed both of the above issues.
Fixes rdar://problem/151162470.
`getLoweredLocalCaptures` seems the best place to do it because
during Sema there is a chicken and an egg situation where isolation
depends on captures and in this case captures depend on isolation.
By default (currently) the closure passed to a parameter with `@_inheritActorContext`
would only inherit isolation from `nonisolated`, global actor isolated or actor
context when "self" is captured by the closure. `always` changes this behavior to
always inherit actor isolation from context regardless of whether it's captured
or not.
Start printing `#include` for headers referenced from `@cdecl` function
signatures. This adds on top of the existing tiered imports. We already
print each module referenced from decls printed in the compatibility
header. Previously we printed mostly `@import` with an option to
fallback on a `#import`. This change adds a third fallback to `#include`
when the module is referenced from a `@cdecl` function signature. The
bridging header can also be imported in a similar way.
Close all the pipe file descriptors in the child process after duping
them to the standard I/O. This is not necessary but it's a good thing to
do anyway.
Previously, the compiler waited the plugin process with
`llvm::sys::Wait(pid, /*SecondsToWait=*/1)`. But in the `Wait`
implementation, it sets the `alarm(SecondsToWait)`, then
`wait4(pid, ..)` so if the alarm fires before the `wait4` call,
it may miss the timeout and can wait indefinitely. To mitigate that
risk,use `10` for the timeout value.
rdar://150474701
There are four attributes which serialize out a DeclNameRef, sometimes by dropping some of its components. Standardize them with a representation that can handle module selectors.
Currently we delete dead drop_deinit instructions in InstructionDeleter.
For address results, we may end up with ownership errors after being promoted to value forms.
For value results, fixLifetimes mode of InstructionDeleter will insert an illegal destroy_value
rdar://151104993
