Sema now type-checks the alternate ABI-providing decls inside of @abi attributes.
Making this work—particularly, making redeclaration checking work—required making name lookup aware of ABI decls. Name lookup now evaluates both API-providing and ABI-providing declarations. In most cases, it will filter ABI-only decls out unless a specific flag is passed, in which case it will filter API-only decls out instead. Calls that simply retrieve a list of declarations, like `IterableDeclContext::getMembers()` and friends, typically only return API-providing decls; you have to access the ABI-providing ones through those.
As part of that work, I have also added some basic compiler interfaces for working with the API-providing and ABI-providing variants. `ABIRole` encodes whether a declaration provides only API, only ABI, or both, and `ABIRoleInfo` combines that with a pointer to the counterpart providing the other role (for a declaration that provides both, that’ll just be a pointer to `this`).
Decl checking of behavior specific to @abi will come in a future commit.
Note that this probably doesn’t properly exercise some of the new code (ASTScope::lookupEnclosingABIAttributeScope(), for instance); I expect that to happen only once we can rename types using an @abi attribute, since that will create distinguishable behavior differences when resolving TypeReprs in other @abi attributes.
Introduce a number of fixes to allow us to fully use declarations that
are produced by applying a peer macro to an imported declarations.
These changes include:
* Ensuring that we have the right set of imports in the source file
containing the macro expansion, because it depends only on the module
it comes from
* Ensuring that name lookup looks in that file even when the
DeclContext hierarchy doesn't contain the source file (because it's
based on the Clang module structure)
Expand testing to be sure that we're getting the right calls,
diagnostics, and generated IR symbols.
Both `try?` and `try!` are catch nodes, because they catch an error
thrown in their subexpression and handle it. Introduce an ASTScope for
all `try/try?/try1` expressions so we can find them, and model them as
catch nodes.
Fixes rdar://119216455.
The implementation of "lookup the catch node" is scattered throughout
various `ASTScopeImpl` subclasses. Remove that, and instead centralize
the code to find the innermost catch scope.
The subclasses of `GenericTypeOrExtensionScope` are customized by a `Portion`
mix-in class that effectively makes each into several different scope
nodes: one for the "whole decl", one for the where clause, one for the
body. Re-introduce the customized `getClassName()` operation so we get
the nicer debugging dumps.
Replace a number of virtual query functions like `getDeclIfAny`, which had
to be overridden at various points in the hierarchy, with simpler
implementations based on the scope kind and `dyn_cast`.
Macro-metaprogram whether each scope node is associated with a
particular declaration/statement/expression/attribute so these
implementations can be efficiently macro-metaprogrammed.
The ASTScope class hierarchy rooted at ASTScopeImpl is relying entirely
on overridden virtual functions and, in some cases, string comparisons
on the class name, to provide customization for the various classes.
Introduce proper LLVM dynamic casting support to this class hierarchy,
eliminating the string comparisons and the manually-specified
implementations of `getClassName` (which is now used only for printing).
We've been meaning to do this for years, but today I got angry enough.
ASTScope assertions that a new child node inserted into its parent list
is after all other child nodes in the source, which included some bespoke
code for dealing with macro expansions. Replace this bespoke code with
uses of the newer SourceManager operations that check ordering of
source locations in a manner that respects macro expansions.
Introduce a new API to find the AST node that catches or rethrows an
error thrown from the given source location. Use it to determine the
thrown error type to use for type checking a `throw` statement, which
begins as `any Error` within a `do..catch` and is later refined.
Addresses recently discovered performance regression.
`findChildContaining` has to handle macros and for that
it needs a parent module, looking it up every time is
expensive because it involves walking up back to a source
file scope.
Resolves: rdar://114667408
Resolves: rdar://114667496
Reformatting everything now that we have `llvm` namespaces. I've
separated this from the main commit to help manage merge-conflicts and
for making it a bit easier to read the mega-patch.
This is phase-1 of switching from llvm::Optional to std::optional in the
next rebranch. llvm::Optional was removed from upstream LLVM, so we need
to migrate off rather soon. On Darwin, std::optional, and llvm::Optional
have the same layout, so we don't need to be as concerned about ABI
beyond the name mangling. `llvm::Optional` is only returned from one
function in
```
getStandardTypeSubst(StringRef TypeName,
bool allowConcurrencyManglings);
```
It's the return value, so it should not impact the mangling of the
function, and the layout is the same as `std::optional`, so it should be
mostly okay. This function doesn't appear to have users, and the ABI was
already broken 2 years ago for concurrency and no one seemed to notice
so this should be "okay".
I'm doing the migration incrementally so that folks working on main can
cherry-pick back to the release/5.9 branch. Once 5.9 is done and locked
away, then we can go through and finish the replacement. Since `None`
and `Optional` show up in contexts where they are not `llvm::None` and
`llvm::Optional`, I'm preparing the work now by going through and
removing the namespace unwrapping and making the `llvm` namespace
explicit. This should make it fairly mechanical to go through and
replace llvm::Optional with std::optional, and llvm::None with
std::nullopt. It's also a change that can be brought onto the
release/5.9 with minimal impact. This should be an NFC change.
ConditionalClauseInitializerScope often create redundant scopes, however, they
are needed to correctly represent the scopes in an expression such as:
if case .something(let foo), let foo = foo.
This patch changes SILGen to lower them 1:1 from ASTScopes.
rdar://110841130
namespace.
This moves the `isInMacroArgument` predicate and `lookupMacros` into `namelookup`.
ASTScope still encapsulates the scope tree and contains the operation to lookup
the enclosing macro scope, which then invokes a callback to determine whether a
potential macro scope is indeed a macro, because answering this question requires
name lookup.
so that it can represent the scope for any custom attribute and its arguments.
This commit is NFC, but CustomAttributeScope is now applicable to attached macros.
Before this patch the parents of SILDebugScopes representing macro expansions
were missing the inlinedAt field, which resulted in incorrent LLVM IR being
produced. This is fixed by first computing the inlined call site for a macro
expansion and then computing the nested SILDebugScope for the ASTScope of the
expanded nodes; adding the inlinedAt field to all of levels of parent scopes.
rdar://108323748
This patch replaces the stateful generation of SILScope information in
SILGenFunction with data derived from the ASTScope hierarchy, which should be
100% in sync with the scopes needed for local variables. The goal is to
eliminate the surprising effects that the stack of cleanup operations can have
on the current state of SILBuilder leading to a fully deterministic (in the
sense of: predictible by a human) association of SILDebugScopes with
SILInstructions. The patch also eliminates the need to many workarounds. There
are still some accomodations for several Sema transformation passes such as
ResultBuilders, which don't correctly update the source locations when moving
around nodes. If these were implemented as macros, this problem would disappear.
This necessary rewrite of the macro scope handling included in this patch also
adds proper support nested macro expansions.
This fixes
rdar://88274783
and either fixes or at least partially addresses the following:
rdar://89252827
rdar://105186946
rdar://105757810
rdar://105997826
rdar://105102288
Teach ASTScope to create child nodes for each of the arguments of a
macro expansion declaration, so that we can perform name lookup into
them. Otherwise, we cannot have nontrivial closures in arguments to a
macro expansion declaration.
Fixes rdar://107388151.
Introduce SingleValueStmtExpr, which allows the
embedding of a statement in an expression context.
This then allows us to parse and type-check `if`
and `switch` statements as expressions, gated
behind the `IfSwitchExpression` experimental
feature for now. In the future,
SingleValueStmtExpr could also be used for e.g
`do` expressions.
For now, only single expression branches are
supported for producing a value from an
`if`/`switch` expression, and each branch is
type-checked independently. A multi-statement
branch may only appear if it ends with a `throw`,
and it may not `break`, `continue`, or `return`.
The placement of `if`/`switch` expressions is also
currently limited by a syntactic use diagnostic.
Currently they're only allowed in bindings,
assignments, throws, and returns. But this could
be lifted in the future if desired.
Teach ASTScope how to reason about macro expansions. When we create an
ASTScope for a source file that represents a macro expansion, its
parent scope node is the macro expansion itself. When performing
unqualified lookup (of any form), find the starting source file based
on the location, not on the source file provided---this ensures that
we start lookups within the macro expansion (for example).
The effect of this is to enable macro expansions to work in nested
contexts, where they refer to names in the scope in which the macro is
expanded.
This is decidedly unhiegenic, but it fits with our syntactic model.
Semantically, the capture list binding behavior doesn't require the scope
to be an explicit closure, and forming new ClosureExprs during type-checking
is difficult because they have to have preassigned discriminators, unlike
AutoClosureExprs which get discriminators introduced by Sema itself. Allowing
CaptureListExpr to hold an AutoClosureExpr makes it easier to synthesize
CaptureListExprs during type checking, to represent expressions with closure
semantics that evaluate some parts of the expression eagerly in the surrounding
context.
Many, many, many types in the Swift compiler are intended to only be allocated in the ASTContext. We have previously implemented this by writing several `operator new` and `operator delete` implementations into these types. Factor those out into a new base class instead.
When LLDB wraps a user-defined expression in the REPL, it takes something like this
```
<expr>
```
and turns it into (very very abstractly)
```
var result
do {
result = <expr>
}
print(result)
```
In the process, it creates an implicit pattern binding and an implicit do block. Of these, only the implicit do is considered by ASTScope lookup to be relevant. This presents a problem when <expr> is or contains a closure, as the parameters of that closure are defined within a scope that will never be expanded. Thus,
```
> [42].map { x in x } // <- cannot find 'x' in scope
```
This patch provides the Swift half of the fix wherein we privilege pattern bindings created by the debugger and look through them to the underlying user expression when performing ASTScope expansion.
rdar://78256873
Combine the wasExpanded flag with the parent pointer, and remove
the haveAddedCleanup flag since it's no longer necessary now that
"re-expansion" is gone.
This doesn't really fit the request evaluator model since the
result of evaluating the request is the new insertion point,
but we don't have a way to get the insertion point of an
already-expanded scope.
Instead, let's change the callers of the now-removed
expandAndBeCurrentDetectingRecursion() to instead call
expandAndBeCurrent(), while checking first if the scope was
already expanded.
Also, set the expanded flag before calling expandSpecifically()
from inside expandAndBeCurrent(), to ensure that re-entrant
calls to expandAndBeCurrent() are flagged by the existing
assertion there.
Finally, replace a couple of existing counters, and the
now-gone request counter with a single ASTScopeExpansions
counter to track expansions.
Scopes need to register cleanups for non-trivial ivars in the ASTContext.
Instead of doing that here, let's just change the SmallVectors into
ArrayRefs.
Fixes <rdar://problem/69908937>.
The top-level scope for a conditional clause with a pattern is now
ConditionalClausePatternUseScope, which introduces the pattern's
bindings.
Since the patterns are not visible in their own initializer, a new
ConditionalClauseInitializerScope is used for the initializer.
While it is nested inside of the ConditionalClausePatternUseScope,
it's lookup parent skips one level, giving us the desired behavior.
Today, the reported source range of a GuardStmtScope is just that of
the statement itself, ending after the 'else' block. Similarly, a
PatternEntryDeclScope ends immediately after the initializer
expression.
Since these scopes introduce names until the end of the parent
BraceStmt (meaning they change the insertion point), we were
relying on the addition of child scopes to extend the source
range.
While we still need the hack for extending source ranges to deal
with string interpolation, at least in the other cases we can
get rid of it.
Note that this fixes a bug where a GuardStmtScope would end
before an inactive '#if' block if the inactive '#if' block was
the last element of a BraceStmt.
