Diagnostics are suppressed when parsing swiftinterface files, since the
warnings emitted from compiling the swiftinterface of a dependency would just
be a nuisance. It follows that warnings generated when parsing the arguments in
a swiftinterface file should also be suppressed, but that wasn't happening
because the diagnostic engine of the main compile was used for parsing. Pass
the diagnostic engine of the compiler subinstance instead, and proactively
suppress warnings before parsing begins.
Resolves rdar://142814164.
The issue here is that the demangler (since we have a postfix mangling) parses
parameters/results/etc and then uses earlier postfix type arguments to attach
the relevant types to the parameters/results/etc. Since the flag for a sending
result was placed in between the parameters and results, we get an off by one
error.
Rather than fix that specific issue by introducing an offset for the off by one
error, I used the fact that the impl-function part of the mangling is not ABI
and can be modified to move the bit used to signify a sending result to before
the parameters so the whole problem is avoided.
I also while I was doing this looked through the sending result mangling for any
further issues and fixed them as I found them.
rdar://141962865
I discovered this while trying to figure out why we were dropping a sending bit
when computing a reabstraction thunk in the tests for the following commit. This
turned out to be the reason why.
Now that AvailableAttr has storage for its cached AvailabilityDomain, it's no
longer necessary to store an AvailabilityDomain inline in
SemanticAvailableAttr.
NFC.
`x declared here` is not helpful and clear enough, especially when there
are other notes attached. Swap it for a new note that says
`requirement x declared here`.
Represent an AvailabilityDomain as a pointer union where one member of the
union is inline storage for built-in domains with known requirements and the
other member of the union is a pointer to an externally allocated domain
definition.
NFC.
Types annotated as `@_addressableForDependencies` are still usable by older
compilers that don't know about nonescapable types or lifetime dependencies,
since it only affects the behavior of the type when it's the source of a
dependency.
This diagnostic was staged in
https://github.com/swiftlang/swift/pull/38991 as a warning to not break
source. Give it a reasonable margin and, in turn, convey that it is in
fact a compatibility warning and will become an error.
It was difficult to preserve the existing, buggy behavior of availability
attribute inference with respect to attributes specifying availability for
non-platform-specific domains. Instead, this change improves attribute merging
by tracking every domain independently, and only merging attributes from the
same domain.
AvailableAttr::Kind and AvailabilityDomain are designed to replace
PlatformAgnosticAvailabilityKind, allowing AvailableAttr to more flexibly model
availability for arbitrary domains. For now, the new constructor just
translates its inputs into inputs for the existing constructor. Once all of the
callers of the existing AvailableAttr constructor have been updated to use the
new constructor, the representation of AvailableAttr will be updated to store
the new properties.
Fixes the immediate problem, but the presence of demangling code in the
runtime means that we'll need a follow-up to fix the compiler so that it
doesn't try to use the demangler to materialize metadata for function types
that have both isolation and a sending result.
rdar://142443925
Introduce an `unsafe` expression akin to `try` and `await` that notes
that there are unsafe constructs in the expression to the right-hand
side. Extend the effects checker to also check for unsafety along with
throwing and async operations. This will result in diagnostics like
the following:
10 | func sum() -> Int {
11 | withUnsafeBufferPointer { buffer in
12 | let value = buffer[0]
| | `- note: reference to unsafe subscript 'subscript(_:)'
| |- warning: expression uses unsafe constructs but is not marked with 'unsafe'
| `- note: reference to parameter 'buffer' involves unsafe type 'UnsafeBufferPointer<Int>'
13 | tryWithP(X())
14 | return fastAdd(buffer.baseAddress, buffer.count)
These will come with a Fix-It that inserts `unsafe` into the proper
place. There's also a warning that appears when `unsafe` doesn't cover
any unsafe code, making it easier to clean up extraneous `unsafe`.
This approach requires that `@unsafe` be present on any declaration
that involves unsafe constructs within its signature. Outside of the
signature, the `unsafe` expression is used to identify unsafe code.
As an example, use this for the "`@preconcurrency` on import has no
effect" warning, which is not yet working correctly. This disables it
by default but leaves it in place for our testing.
We have some problems on Linux where Glibc pulls in `<elf.h>` and then
we end up with conflicting definitions. Fix by using C++ interop and
putting our definitions into a namespace.
rdar://137201928
