Specifically:
1. I made it so that thunks from caller -> concurrent properly ignore the
isolated parameter of the thunk when calling the concurrent function.
rdar://148112362
2. I made it so that thunks from concurrent -> caller properly create a
Optional<any Actor>.none and pass that into the caller function.
rdar://148112384
3. I made it so that in cases where we are assigning an @Sendable caller to a
non-sendable caller variable, we allow for the conversion as long as the
parameters/results are sendable as well.
rdar://148112532
4. I made it so that when we generate a thunk from @execution(caller) ->
@GlobalActor, we mangle in @GlobalActor into the thunk.
rdar://148112569
5. I discovered that due to the way we handle function conversion expr/decl ref
expr, we were emitted two thunks when we assigned a global @caller function to a
local @caller variable. The result is that we would first cast from @caller ->
@concurrent and then back to @caller. The result of this would be that the
@caller function would always be called on the global queue.
rdar://148112646
I also added a bunch of basic tests as well that showed that this behavior was
broken.
Print diagnostic groups as part of the LLVM printer in the same manner as the
Swift one does, always. Make `-print-diagnostic-groups` an inert option, since we
always print diagnostic group names with the `[#GroupName]` syntax.
As part of this, we no longer render the diagnostic group name as part
of the diagnostic *text*, instead leaving it up to the diagnostic
renderer to handle the category appropriately. Update all of the tests
that were depending on `-print-diagnostic-groups` putting it into the
text to instead use the `{{documentation-file=<file name>}}`
diagnostic verification syntax.
We've been converging the implementations of educational notes and
diagnostic groups, where both provide category information in
diagnostics (e.g., `[#StrictMemorySafety]`) and corresponding
short-form documentation files. The diagnostic group model is more
useful in a few ways:
* It provides warnings-as-errors control for warnings in the group
* It is easier to associate a diagnostic with a group with
GROUPED_ERROR/GROUPED_WARNING than it is to have a separate diagnostic
ID -> mapping.
* It is easier to see our progress on diagnostic-group coverage
* It provides an easy name to use for diagnostic purposes.
Collapse the educational-notes infrastructure into diagnostic groups,
migrating all of the existing educational notes into new groups.
Simplify the code paths that dealt with multiple educational notes to
have a single, possibly-missing "category documentation URL", which is
how we're treating this.
This feature only exists as a mechanism to suppress the warning introduced in
https://github.com/swiftlang/swift/pull/75378. The RegexParser module, which is
effectively part of the standard library, declares a Swift runtime symbol and
as a result every build of the compiler and stdlib produces warnings which
there are no plans to address. Warnings that are not going to be addressed need
some way of being suppressed, and an experimental features seems like a
reasonable mechanism for this one.
If an `@objc implementation extension` had a public stored property with an observer, Swift would print `@_hasStorage` on the extension. This is Not Good because in a module interface, an objcImpl extension appears to be an ordinary extension, and properties in ordinary extensions are not supposed to have storage.
Suppress printing this attribute in objcImpl extensions to avoid this problem.
Partially fixes rdar://144811653 by suppressing emission of bad attributes.
The last step in building a generic signature is to sort the requirements.
Requirements are sorted by comparing their subject types. If two
requirements have the same subject type, which can only happen with
conformance requirements, we break the tie by comparing protocol
declarations.
We compare protocol declarations using TypeDecl::compare(), which is a
shortlex order on the components of the fully qualified name of a
protocol (eg, Swift.Sequence, etc.)
While this order is part of the ABI, it has not been updated over the
years for several important changes:
- It did not handle module aliases; if we import a module via an
alias, we should use the real module name to compare protocols, and
not the aliased name. This produced inconsistent results if the
same module was imported under different names, which can happen
with module interface files that use module aliases.
- It did not handle the -module-abi-name flag. Changing the ABI name
of a module changes how we mangle protocol names, and the order
should match the mangling.
This change fixes the first case only. The second requires more
careful staging, because of _Concurrency and CompilerSwiftSyntax.
Fixes rdar://147441890.
Update the logic to correctly handle replacement ranges, and re-enable
the assertion. For now, carve out an exception for attributes on
extensions, I'll try and tackle those in a follow-up.
`visitParsedAccessors` currently iterates over any non-implicit
accessor that happens to be present. This means we end up with
accessors added by macro expansions if the expansion happens before
the ASTScope expansion. Make sure we avoid adding such accessors to
the ASTScope tree, they should instead use their own tree. Ideally
we'd better formalize `visitParsedAccessors` such that it only ever
visits non-expansion accessors, but I'm leaving that for another day.
Rather than looking at their DeclContext, look at the location of
their storage. This allows us to differentiate accessors added by
macro expansions vs accessors that are attached to storage that's
in a macro expansion.
This PR adds a set of DeclAttr.def flags for specifying how a given attribute interacts with `@abi`, and declares a behavior for each existing attribute. Future attributes will be *required* to declare an `@abi` behavior lest they fail a static assert.
Note that the behavior is not actually enforced in this commit—it is merely specified here.
Adds a new method to `DeclAttribute` which can compare two attributes and see if they would be “equivalent” in light of the given decl. “Equivalent” here means that they would have the same effect on the declaration; for instance, two attrs with different source locations can be equivalent, and two attributes with the same arguments in a different order are equivalent if the ordering of those argumetns is not semantically equivalent.
This capability is not yet used in this commit, but in a future commit `@abi` will check if certain attributes are equivalent or not.
ABI-only declarations now inherit `@available`, `@backDeployed`, etc. from their ABI counterpart. This will make it unnecessary to specify these attributes in `@abi`. Also some changes to make sure we suggest inserting `@available` in the correct place.
No tests because the enforcement is not yet in.
ABI-only declarations now inherit access control modifiers like `public` or `private(set)`, as well as `@usableFromInline` and `@_spi`, from their API counterpart. This means these attributes and modifiers don’t need to be specified in an `@abi` attribute.
Very few tests because we aren’t yet enforcing the absence of these attributes.
ABI-only declarations now query their API counterpart for things like `isObjC()`, their ObjC name, dynamic status, etc. This means that `@objc` and friends can simply be omitted from an `@abi` attribute.
No tests in this commit since attribute checking hasn’t landed yet.
Correct an issue with `ImplementsAttr` that would come up if you ever tried to clone an attribute that had been deserialized.
No tests because there’s nothing in the compiler yet that might actually do so.
I am doing this in preparation for adding the ability to represent in the SIL
type system that a function is global actor isolated. Since we have isolated
parameters in SIL, we do not need to represent parameter, nonisolated, or
nonisolated caller in the type system. So this should be sufficient for our
purposes.
I am adding this since I need to ensure that we mangle into thunks that convert
execution(caller) functions to `global actor` functions what the global actor
is. Otherwise, we cannot tell the difference in between such a thunk and a thunk
that converts execution(caller) to execution(concurrent).
