User code should not be diagnosed as "unreachable" by the SIL optimizer when
the no-return function that made the code unreachable is a compiler inserted
call to `_diagnoseUnavailableCodeReached()`.
Part of rdar://107388493
When `-unavailable-decl-optimization=stub` is specified, insert a call to
`_diagnoseUnavailableCodeReached()` at the beginning of the function to cause
it to trap if executed at run time.
Part of rdar://107388493
There is a strong desire for "no-implicit-copy" semantics to be
applied to Copyable values marked with the newer `borrowing` and
`consuming` ownership specifiers. If we lock ourselves into what
SE-390 specifies now for Copyable types in the implementation, then
we'll have to introduce a source break when we want such values to
have the desired semantics.
Before there's wider adoption, I'm making it an error to use those
newer names. If for some reason this is enabling a critical
optimization for your use case, please use the old names instead:
- `__owned` for `consuming` parameters
- `__consuming` for `consuming` methods
- `__shared` for `borrowing` parameters
NOTE: the older names have their ownership attribute mangled into
the name of the function itself, so there's a possibility of ABI
breaks in the future if you move from the old name to the new one.
There is some consideration being made to allow for a migration to
the new names in the future without breaking ABI, so do reach out if
this is an issue for you.
resolves rdar://108538971
Does not fix the fix-it. The current fix it will be left as a stop-gap solution and we can hopefully update this fix it in the near future to actually plop in a for loop (too much work for this PR though).
Optimizations can rely on alias analysis to know that an in-argument (or parts of it) is not actually read.
We have to do the same in the verifier: if alias analysis says that an in-argument is not read, there is no need that the memory location is initialized.
Fixes a false verifier error.
rdar://106806899
This just creates a new SILUndef with the same type as the SILValue. This just
prevents one from having to write
SILUndef::get(value->getType(), *value->getModule()).
Previously, when constructing a container within a StackState, a
SILFunction that was passed to the StackState's constructor was passed
along. Here, arbitrary arguments are forwarded along.
Diagnose situation when a single argument to tuple type is passed to
a value pack expansion parameter that expects distinct N elements:
```swift
struct S<each T> {
func test(x: Int, _: repeat each T) {}
}
S<Int, String>().test(x: 42, (2, "b"))
```
The constraint takes two pack types and makes sure that their
reduced shapes are equal. This helps with diagnostics because
constraint has access to the original pack expansion pattern
types.
If there are explicit generic arguments that fully resolves the
pack expansion, let's bind opened pack expansion to its contextual
type early (while resolving pack expansion variable), doing so
helps with performance and diagnostics.
`openType` didn't need a locator before it was simply replacing generic
parameters with corresponding type variables but now, with opening of
pack expansions types, a locator is needed for pack expansion variables.
A pack expansion type variable gets bound to its original pattern
type when solver can infer a contextual type for it. This makes
sure that pack expansion types are always matched via `matchTypes`
without `simplifyType` involvement which can expand them prematurely.
Models `PackExpansionType` as a type variable that can only
bind to `PackExpansionType` and `expansion of` constraint that
connects expansion variable to its pattern, shape types.
This could cause us to double up on the same note
for nested macro expansions, as we'd generate the
note, then when recursing back through the function
we'd compute the same note again. Also remove the
seemingly unused `lastBufferID` param.
We need to teach code completion how to invoke the type checker for attached macro attributes. After that, everything started working.
rdar://105232015
We parse `~Copyable` in an inheritance clause of enum and
struct decls as a synonym for the `@_moveOnly` attribute
being added to that decl. This completely side-steps the
additional infrastructure for generalized suppressed
conformances in favor of a minimal solution. One benefit of
this minimal solution is that it doesn't risk introducing
any back-compat issues with older compilers or stdlibs.
The trade-off is that we're more committed to supporting
`@_moveOnly` in compiled modules in the future. In fact,
this change does not deprecate `@_moveOnly` in any way.
resolves rdar://106775103
We were not type checking the default argument initializer because `MacroDecl` is not an `AbstractFucntionDecl`. Add `MacroDecl` to the list of nodes we know how to get parameters from.
rdar://108163564
When completing after `names:`, completion should offer the different ways you can specify the names, i.e. `arbitrary`, `named`, etc.
```
@freestanding(declaration, names: #^COMPLETE^#)
```
rdar://108535077
