Add a platform kind and availability attributes for macCatalyst. macCatalyst
uses iOS version numbers and inherits availability from iOS attributes unless
a macCatalyst attribute is explicitly provided.
For protocol requirements and class members with `@differentiable` attribute,
conforming types and subclasses must have the same `@differentiable` attribute
(or one with a superset of differentiability parameters) on implementing/
overriding declarations.
For implementing/overriding declarations that are missing a `@differentiable`
attribute, emit a fix-it that adds the missing attribute.
Resolves TF-1118.
When wrapping a function which is supposed to capture the caller’s location, there’s always a risk that the wrapper won’t capture the information the wrapped function wants; for instance, you might pass `(…, line, column)` where the callee expected `(…, column, line)`.
This commit emits a warning when a call passes an explicit argument to something that has a default argument, and that explicit argument is itself a parameter with a default argument, and both parameters use magic identifiers, but they use *different* magic identifiers. This is partially in support of concise #file, but applies to all magic identifiers.
Fixes rdar://problem/58588633.
* If there is a disjunction associated with closure type e.g.
coercion to some other type `_ = { $0 } as (Int32) -> Void`
* If there is a disjunction associated with a collection which
could provide more context to the constraint solver.
Collect all references to parameters whose types involve type variables,
including in closures that aren’t single-expression. This fixes a type
checker assertion that occurs when the constraint graph can get disconnected
with the combination of delayed constraint generation for single-expression
closures and the use of function builders.
Fixes rdar://problem/58695803.
ASTScopes completely skip implicit pattern bindings, so don't mark
generated ones as implicit. Instead, give them suitable source
location information.
Fixes rdar://problem/58710568.
The `@differentiable` attribute marks a function as differentiable.
Example:
```
@differentiable(wrt: x, jvp: derivativeFoo where T: Differentiable)
func id<T>(_ x: T) -> T { x }
```
The `@differentiable` attribute has an optional `wrt:` clause specifying the
parameters that are differentiated "with respect to", i.e. the differentiability
parameters. The differentiability parameters must conform to the
`Differentiable` protocol.
If the `wrt:` clause is unspecified, the differentiability parameters are
currently inferred to be all parameters that conform to `Differentiable`.
The `@differentiable` attribute also has optional `jvp:` and `vjp:` labels
for registering derivative functions. These labels are deprecated in favor of
the `@derivative` attribute and will be removed soon.
The `@differentiable` attribute also has an optional `where` clause, specifying
extra differentiability requirements for generic functions.
The `@differentiable` attribute is gated by the
`-enable-experimental-differentiable-programming` flag.
Code changes:
- Add `DifferentiableAttributeTypeCheckRequest`.
- Currently, the request returns differentiability parameter indices, while
also resolving `JVPFunction`, `VJPFunction`, and
`DerivativeGenericSignature` and mutating them in-place in
`DifferentiableAttr`. This was the simplest approach that worked without
introducing request cycles.
- Add "is type-checked" bit to `DifferentiableAttr`.
- Alternatively, I tried changing `DifferentiableAttributeTypeCheckRequest` to
use `CacheKind::Cache` instead of `CacheKind::SeparatelyCached`, but it did
not seem to work: `@differentiable` attributes in non-primary-files were
left unchecked.
Type-checking rules (summary):
- `@differentiable` attribute must be declared on a function-like "original"
declaration: `func`, `init`, `subscript`, `var` (computed properties only).
- Parsed differentiability parameters must be valid (if they exist).
- Parsed `where` clause must be valid (if it exists).
- Differentiability parameters must all conform to `Differentiable`.
- Original result must all conform to `Differentiable`.
- If JVP/VJP functions are specified, they must match the expected type.
- `@differentiable(jvp:vjp:)` for derivative registration is deprecated in
favor of `@derivative` attribute, and will be removed soon.
- Duplicate `@differentiable` attributes with the same differentiability
parameters are invalid.
- For protocol requirements and class members with `@differentiable` attribute,
conforming types and subclasses must have the same `@differentiable` attribute
(or one with a superset of differentiability parameter indices) on
implementing/overriding declarations.
The right solution is to extend the notion of the "anchor" of a locator
to also cover statements (and TypeReprs, and Patterns, and more), so
this is a stop-gap.
We used to get this contextualization "for free" because closures that
had function builders applied to them would get translated into
single-expression closures. Now, we need to check for this explicitly.
When generating constraints in the context of the "old" diagnostic
path that type-checks subexpressions independently, we might not have
types for closure parameters. Cope with the case with a narrower form
of the original hack.
A couple of trivial fixes as part of the builder transform refactoring:
* When recording captured expressions, record the source entity appropriately
* Make sure to check (and skip) #if declarations in application
* Check and diagnostic #warning/#error as part of application (not generation)
* Simplify the body result type for return coercion during application
* When checking for applicability of a function builder, don't do any
constraint generation.
Delayed constraint generation for the body of the single-statement
closures regressed variadic parameter handling. Fix it by using
`FunctionType::Param::getParameterType` for "internal" version of
the parameter type which translates variadic/inout correctly.
Resolved: rdar://problem/58647769