Replaces generic `expression is 'async' but is not marked with 'await`
diagnostic with a tailed one for cases where there is an access to an
actor-isolated value outside of its actor without `await` keyword.
This makes the diagnostics for async and sync contexts consistent
and actually identifies a problem instead of simply pointing out
the solution.
Resolves: rdar://151720646
We started building iOS sanitizers when switching to
`LLVM_ENABLE_RUNTIMES` to build compiler-rt, and turns out they
currently provide only the arm64 slice.
Addresses rdar://151782340
This matches send non sendable but importantly also makes it clear that we are
talking about something that doesn't conform to the Sendable protocol which is
capitalized.
rdar://151802975
To determine whether an instruction may require pack metadata, the types
of its operands are examined.
Previously, the top level type was checked for having a pack in its
signature, and the whole type was checked for having a type anywhere in
its layout (via TypeLowering). This didn't account for the case where
the metadata was required for a resilient type which uses a pack in its
signature.
Here, during type lowering, a type having a pack in its signature is
counted towards the type having a pack.
Fixes a compiler crash.
rdar://147207926
Apply the MoveOnlyAddressChecker change from
https://github.com/swiftlang/swift/pull/73358 to the
MoveOnly[Value]Checker.
After 7713eef817, before running value
checking, all lifetimes in the function are completed. That doesn't
quite work because lifetime completion expects not to encounter
reborrows or their adjacent phis.
rdar://151325025
Use the higher level APIs on SourceManager that handle locations in
parent vs child buffers. This then allows us to fix `walkToDeclPre`
such that we don't set the found DeclContext unless the location is
actually within that decl (here the location may well be in a
separate buffer as we may have a replaced function body).
Both the syntax and relative order of the LLVM `nocapture` parameter
attribute changed upstream in 29441e4f5fa5f5c7709f7cf180815ba97f611297.
To reduce conflicts with rebranch, adjust FileCheck patterns to expect
both syntaxes and orders anywhere the presence of the attribute is not
critical to the test. These changes are temporary and will be cleaned
up once rebranch is merged into main.
The parameter attribute was introduced in
5ece35df8586d0cb8c104a9f44eaae771de025f5 (llvm-project) and is expected
here according to the lowered function body.
Match it using a wildcard regex, since it is not relevant to this test.
This is intended to reduce future conflicts with rebranch.
This attribute was introduced in
7eca38ce76d5d1915f4ab7e665964062c0b37697 (llvm-project).
Match it using a wildcard regex, since it is not relevant to these
tests.
This is intended to reduce future conflicts with rebranch.
rdar://149985633
Using at least Int8 here allows LLVM to apply more optimizations, reducing code size, avoiding stack allocations and as a result often eliminating complete stack frames.
This fixes a timeout failure in CI. It looks like that some CI machines are slower than expected.
I could not reproduce this locally. There doesn't seem to be a compile time regression.
Enable walking into `TypeOffsetSizePair`s from an existential into an
archetype. And set the access kind on `open_existential_addr`
instructions which are the sources of rewritten `copy_addr`s to mutable.
rdar://141279635
Kicking the interface type request of the base decl here is wrong
if the decl is e.g a `self` capture in a closure, since we'll be in
the middle of type-checking the closure. I'm planning on properly
fixing this by folding the lookup into the constraint system, but for
now let's at least avoid kicking the request if we don't have an enum
case or enum var. That at least prevents it from affecting cases where
e.g you're pattern matching against a property in a class.
We could potentially tighten up the checking here even further, but
that could potentially impact source compatibility for ambiguous
cases. I'd like to keep this patch low risk, and then deal with any
fallout as part of the pattern type-checking work.
rdar://146952007
This takes about twice as long to build in CI, but runs significantly
quicker (~22x). The exact time taken varies quite a bit between runs,
but this brings the worst-case overall run-time down from ~500s to
~200s. While here, make sure we pass `--skip-build` to the second
invocation to ensure we avoid building twice. Also disable using the
dependency cache since we want to test a clean build, and we just
want swiftpm to use the local deps.
LLVM-21 plans to remove the legacy method for building compiler-rt
in the same invocation as LLVM using `LLVM_ENABLED_PROJECTS` and
`LLVM_BUILD_EXTERNAL_COMPILER_RT`.
Support the new way of building compiler-rt with a new build-script
opt-in flag `--llvm-build-compiler-rt-with-use-runtimes` --
this will allow a staged introduction, and will ensure we can revert
back to the old behaviour temporarily in case of unforeseen regression.
Since this flag is meant to be short lived, in an attempt to keep the
logic simple we are gating on it only the
CMake cache entries that strictly control the compilation mode, all the
other entries used for configuring are added in both modes.
Take this chance to remove some stale code from `build-script-impl`, and
move some code in the generic CMake product to the LLVM one.
Addresses rdar://147505298
Fixes a crash when existing logic cannot properly determine whether
argument is resolved or not.
Instead of ad-hoc code that checks `ResolvedOverloads` directly, let's
use a method that would reach of a callee locator to determine whether
argument is a reference to a generic function or not.
Resolves: rdar://137825558
`typeCheck{Expression, Target}` has a pre-check phase which would
replace some invalid AST nodes (i.e. name references that are not
available in the given declaration context) with `ErrorExpr`s and
emit a diagnostic. Such diagnostics were then dropped by `abort()`
call to a diagnostic transaction. This results in invalid code being
accepted by Sema and forwarded to SILGen.
Resolves: https://github.com/swiftlang/swift/issues/73986
Resolves: rdar://131732245
A big test was disabled due to one module failing to verify with
sil-opt. This PR reenables the test for all other modules to catch
errors while this particular problem is being looked into.
rdar://143050566
The test was using the default value for SWIFT_BUILD_ROOT, which tried
to write into `.build_script_log`. Use an environment variable so the
tests use their own directory and write files that do not append to the
actual `.build_script_log`.
Follow up to #80102
