With support for redundant conformance declarations via macros,
the `Observable` protocol can be a non-marker protocol, which
provides more flexibility for evolution in the future.
rdar://111463883
This change also switches to the new ExtensionMacro protocol,
the requirement for which includes information about whether the
conformance to the Observable protocol has already been added, either
in the declaration or in a superclass to the macro-attributed type.
This allows the @Observable macro to be applied to subclasses of
observable types without redundant-conformance errors.
* Make ObservationRegistrar Codable/Hashable
These conformances enable automatic Codable synthesis for Observable
types, and smooth the runway for structs being supported by the
Observable macro in the future.
* Limit Observable macro to classes
This removes the ability for the Observable macro to apply to structs,
and adds diagnostic tests for the three disallowed declaration kinds.
The review of SE-0395 is down to small details at this point that won't
affect the overall shape of the API much. Rename the model in
anticipation of that.
* Fix sign of fractional units in Duration initializers from Double.
When separating a double duration into integral and fractional parts, we got the sign wrong. This fixes that bug.
* Fixup variable name.
Now that we've made accessor macro expansion more lazy, ensure that
when querying for init accessors (e.g., to build a memberwise
initializer), we also expand any accessor macros that might produce an
init accessor.
This is a partial step toward the real goal, which is that
`AbstractStorageDecl::getAccessor()` should lazily expand macros if
needed.
Update the Observable macro to document that it produces an `init`
accessor.
* [Observation] Transition to peer macros instead of arbitrary members
* [Observation] Lift the initializer requirement by utilizing init accessors for fully formed definite initialization
* [Observation] Gate enabling of peer macros by flag
* [Observation] Enable feature for InitAccessors in the observation tests
* [Observation] Add tests to validate memberwise and definite initialization
Previously we would wait until CSApply, which
would trigger their type-checking in
`coercePatternToType`. This caused a number of
bugs, and hampered solver-based completion, which
does not run CSApply. Instead, form a conjunction
of all the ExprPatterns present, which preserves
some of the previous isolation behavior (though
does not provide complete isolation).
We can then modify `coercePatternToType` to accept
a closure, which allows the solver to take over
rewriting the ExprPatterns it has already solved.
This then sets the stage for the complete removal
of `coercePatternToType`, and doing all pattern
type-checking in the solver.
Balance the call to allocateBufferIn with a call to deallocateBufferIn. When an error value is small, the missing deallocateBufferIn doesn't do anything. But when the error value is a larger struct that doesn't fit inline, we need deallocateBufferIn to avoid leaking the allocation.
rdar://109933822
When constructing a Duration from Double, we should do it in such a way that exact integer inputs are preserved exactly, so long as they are represented as a Duration. This was not previously the case. Now it is.
Now that Array is eagerMove, a different type needs to be used to test
the behavior of `consume` regarding uniqueness. Used a class and
otherwise restored the test case to its previous usage of consume,
replacing `._buffer.isUniquelyReferenced()` with `_isUnique_native`.
rdar://107642615
Run DestroyAddrHoisting in the pipeline where DestroyHoisting was
previously running. Avoid extra ARC traffic that having no form of
destroy hoisting in the mandatory pipeline results in.
rdar://90495704
The macro tests were all using "REQUIRES: OS=macosx" as a proxy for
"have the Swift Swift parser". There was an existing feature for this,
but it was just checking whether the path was passed through. Fix that
to use the same variable as in CMake.
Also remove all extraneous `-I` and `-L` to the host libs in the target
invocations.
Types that have "value semantics" should not have lexical lifetimes.
Value types are not expected to have custom deinits. Are not expected to
expose unsafe interior pointers. And cannot have weak references because
they are structs. Therefore, deinitialization barriers are irrelevant.
rdar://107076869
We ran into an issue on minimal stdlib build configurations, because they
pull slightly older toolchains to build and test the stdlib. Adding this flag
doesn't hurt anything and just ensures things will work with older compilers.
But it is not needed and should eventually be safe to remove.
part of resolving rdar://106849189
* [Observation] Change visibility of observation runtime functions to be hidden
* [Observation] Update API requirements for Observable AsyncSequence names and alter the behavior of value emissions to be based upon transactionality
* [Observation] Slight naming alteration of isolation -> isolatedTo
Note: This re-enables the previously disabled tests since the implementation should be considerably more robust to hangs.
