Lookup into C++ namespaces uses a different path from C++ record declarations.
Augment the C++ namespace lookup path to also account for the auxiliary
declarations introduced by peer macro expansions.
It is possible for a module interface (e.g., ModuleA) to be generated
with C++ interop disabled, and then rebuilt with C++ interop enabled
(e.g., because ModuleB, which imports ModuleA, has C++ interop enabled).
This circumstance can lead to various issues when name lookup behaves
differently depending on whether C++ interop is enabled, e.g., when
a module name is shadowed by a namespace of the same name---this only
happens in C++ because namespaces do not exist in C. Unfortunately,
naming namespaces the same as a module is a common C++ convention,
leading to many textual interfaces whose fully-qualified identifiers
(e.g., c_module.c_member) cannot be correctly resolved when C++ interop
is enabled (because c_module is shadowed by a namespace of the same
name).
This patch does two things. First, it introduces a new frontend flag,
-formal-cxx-interoperability-mode, which records the C++ interop mode
a module interface was originally compiled with. Doing so allows
subsequent consumers of that interface to interpret it according to the
formal C++ interop mode. Note that the actual "versioning" used by this
flag is very crude: "off" means disabled, and "swift-6" means enabled.
This is done to be compatible with C++ interop compat versioning scheme,
which seems to produce some invalid (but unused) version numbers. The
versioning scheme for both the formal and actual C++ interop modes
should be clarified and fixed in a subsequent patch.
The second thing this patch does is fix the module/namespace collision
issue in module interface files. It uses the formal C++ interop mode to
determine whether it should resolve C++-only decls during name lookup.
For now, the fix is very minimal and conservative: it only filters out
C++ namespaces during unqualified name lookup in an interface that was
originally generated without C++ interop. Doing so should fix the issue
while minimizing the chance for collateral breakge. More cases other
than C++ namespaces should be added in subsequent patches, with
sufficient testing and careful consideration.
rdar://144566922
Switch over to split caching for the conformance isolation request,
which optimizes for the common case where the conformance is
nonisolated. Also put the explicit global actor TypeExpr* in an
ASTContext side table, so we don't take a pointer's worth of storage
in every conformance.
For that side table, introduce a new "ASTContext::GlobalCache" that's
there only for side tables, so we don't have to go add get/set
operations to ASTContext and recompile the world every time we want to
add a side table like this.
Thanks, Slava!
Added an `-executor-factory` argument to the compiler to let you safely
specify the executors you wish to use (by naming a type that returns
them).
Also added some tests of the new functionality.
rdar://141348916
When code in the current module defaults to main actor (under SE-0466),
also infer main-actor isolation for protocol conformances of main-actor
isolated types.
The NormalProtocolConformance APIs for checking for an explicitly-written
isolation on a conformance were easy to get to, and the real semantic
API was buried in the type checker, leading to some unprincipled
checking. Instead, create a central ProtocolConformance::getIsolation()
to get the (semantic) actor isolation, and let that be the only place
that will access the explicitly-written global actor isolation for a
conformance. Update all call sites appropriately.
Instead of using the `isolated P` syntax, switch to specifying the
global actor type directly, e.g.,
class MyClass: @MainActor MyProto { ... }
No functionality change at this point
Store `CustomAvailabilityDomain` instances in a folding set on `ASTContext`.
This instances of custom domains to be created without needing to cache them in
disparate locations.
ASTDumper now has the ability to dump attributes in the usual S-expression format, but `DeclAttribute::dump()` and `DeclAttributes::dump()` are still using the printing infrastructure. Use ASTDumper for these functions instead to provide a more “raw” view of the attributes.
We’re running out of bits in DeclAttrOptions, so split it in two: DeclAttrRequirements contains all the `On*` options that describe the declarations allowed to have the attribute, while the other options are now DeclAttrBehaviors.
This commit also sorts the entries in DeclAttr.def by serialization code and improves the formatting of the file.
[Dependency Scanning] Setting Default Clang Optimize Args Option to All
https://github.com/llvm/llvm-project/pull/129809 turned off current working directory (CWD) optimization for clang dependency scanning, because clang needs to coordinate with the build system to safely optimize the CWD. Swift can already handle the CWDs. This PR turns the CWD optimization on for Swift.
Imported C++ template specializations receive identifiers that contain
their type signature; e.g., `X<Y, Z>`. Since this means the identifier
contains non-identifier characters, the new behavior was trying to
escape them with backticks in ASTPrinter, ASTMangler, and the runtime
metadata. This pulls that back to preserve the current behavior for
specifically those types.
`#fileID` never accounted for the possibility that someone one have
a module alias _itself_, so it always generated the module's real
(physical) name. This _technically_ changes the behavior of `#fileID`
for self-aliased modules, but since nobody would have ever had a reason
to do that before raw identifiers, it's unlikely that this change would
affect anyone in practice.
For build systems that already generate these files, it makes sense to include the aliases so that the map file serves as a comprehensive index of how the module inputs are referenced.
Raw identifiers are backtick-delimited identifiers that can contain any
non-identifier character other than the backtick itself, CR, LF, or other
non-printable ASCII code units, and which are also not composed entirely
of operator characters.
We use experimental features to let people know that the construct is
subject to change and users should not rely on this unless they are
willing to rewrite the uses of this feature later. However, in compiler
generated code everything should be fair game, we will update the
compiler when these features change. This is a requirement to be able to
turn safe wrapper generation on by default.
We introduce a new macro called #SwiftSettings that can be used in conjunction
with a new stdlib type called SwiftSetting to control the default isolation at
the file level. It overrides the current default isolation whether it is the
current nonisolated state or main actor (when -enable-experimental-feature
UnspecifiedMeansMainActorIsolated is set).
Normal conformances, self conformances, and availability contexts cannot
contain types with type variables, so there is no reason to duplicate
the uniquing maps between the permanent arena and solver arena.
When printing diagnostics, category names are printed as [#<category-name>]
at the end of a diagnostic. For all of the category names that are mentioned
in this manner, print "footnotes" at the end of compilation providing
documentation references to each category, e.g.,
[#deprecated]: <http://example.com/deprecated>
[#StrictMemorySafety]: <http://example.com/memory-safety>
Right now, these point into the markdown files in the installed toolchain,
same as the URLs behind references. That is subject to change in the future.
