The `_Backtracing` module has a number of private implementation only
imports that aren't used outside of the module and that don't require
any additional libraries (hence they aren't relevant to the outside
world). `verify_all_overlays.py` needs to know about these when it
does its test, because it loadas the module as the main module, which
results in implementation only imports being required instead of
ignored.
rdar://110261712
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.
Deallocate dynamic allocas done for metadata/wtable packs. These
stackrestore calls are inserted on the dominance frontier and then stack
nesting is fixed up. That was achieved as follows:
Added a new IRGen pass PackMetadataMarkerInserter; it
- determines if there are any instructions which might allocate on-stack
pack metadata
- if there aren't, no changes are made
- if there are, alloc_pack_metadata just before instructions that could
allocate pack metadata on the stack and dealloc_pack_metadata on the
dominance frontier of those instructions
- fixup stack nesting
During IRGen, the allocations done for metadata/wtable packs are
recorded and IRGenSILFunction associates them with the instruction that
lowered. It must be the instruction after some alloc_pack_metadata
instruction. Then, when visiting the dealloc_pack_metadata instructions
corresponding to that alloc_pack_metadata, deallocate those packs.
watchOS devices can have different allocation characteristics
from other devices. This modifies the string capacity growth
tests to be more flexible about measuring the growth in
capacity, specifically when more is allocated than requested.
The main change here is in IRGen which needs to be able to emit constant enum values.
Use `emitValueInjection` to create the enum constant.
Usually this method creates code in the current function.
But if all arguments to the enum are constant, the builder never has to emit an instruction.
Instead it can constant fold everything and just returns the final constant.
Also, create statically initialized let-globals as constant global (`constant` instead of `global`).
This is needed short term to allow the test to pass when running on
Python 3.11 and later and avoid errors like
```
<stdin>:44:1: note: non-matching line after previous match is here
<string>:1: DeprecationWarning: 'pipes' is deprecated and slated for removal in Python 3.13
^
```
Long term we want to replace uses of the obsolete `pipes.quotes`
function with `shlex.quotes`.
Addresses rdar://109664710
Fix the type of the `alloca` created by `GenPack`'s for type metadata
and witness tables by fixing its callee, `emitDynamicAlloca` to always
return a `StackAddress` whose `Address`' type is the one specified by
the caller.
rdar://109540863
When called on a string that is not uniquely referenced,
`String.reserveCapacity(_:)` ignores the current capacity, using
the passed-in capacity for the size of its new storage. This can
result in an underallocation and write past the end of the new
buffer.
This fix changes the new size calculation to use the current UTF-8
count as the minimum. Non-native or non-unique strings
now allocate the requested capacity (or space enough for the
current contents, if that's larger than what's requested).
rdar://109275875
Fixes#53483
When completing, we call `getExistentialType` on the contextual type to get a nice and concise description of the contextual parameter’s type that doesn’t contain archetypes and which we can also serialize into a USR so we are able to calculate type relations for code completion results from the code completion cache.
When completing in a position that has a contextual type which only constrains one of two primary associated protocol types, this fails because `getExistentialType` (which calls `getDependentUpperBounds`) tries to form a `ParameterizedProtocolType`, which fails since not all primary associated types have been constrained.
AFAICT the fix here is to just fall back to the default behavior of returning the plain protocol type instead of `abort`ing.
rdar://108835466
We shouldn't be allocating placeholders for type
variables in the permanent arena, and we should be
caching them such that equality works.
To achieve this, we need to introduce a new
"solver allocated" type property. This is required
because we don't want to mark placeholder types
with type variable originators as themselves having
type variables, as it's not part of their structural
type. Also update ErrorType to use this bit, though
I don't believe we currently create ErrorTypes
with type variable originators.
In some Linux setups, one might need to provide extra arguments to setup
the environment for invocations done during the tests. The Swift
frontend and driver already had a mechanism to provide those arguments
in the shape of `SWIFT_FRONTEND_TEST_OPTIONS` and
`SWIFT_DRIVER_TEST_OPTIONS`. `swift-ide-test` was sadly missing the same
feature, which meant that many tests might fail or tests against the
incorrect environment.
The changes implement the `SWIFT_IDE_TEST_TEST_OPTIONS` mechanism to
cover that gap.
We can't assume that we will always find a viable witness for a protocol
requirement when typechecking a `.swiftinterface` file. The assert that assumed
there would be a match could fail when building a broken interface or building
a valid interface with a broken SDK, causing crashes instead of emitting
diagnostics.
Resolves rdar://108688535
Rather than eagerly binding them to holes if the
sequence element type ends up being Any, let's
record the CollectionElementContextualMismatch fix,
and then if the patterns end up becoming holes,
skip penalizing them if we know the fix was
recorded. This avoids prematurely turning type
variables for ExprPatterns into holes, which
should be able to get better bindings from the
expression provided. Also this means we'll apply
the logic to non-Any sequence types, which
previously we would give a confusing diagnostic
to.
As an optimization, there is a fast-cast to non-final classes that just
compares isa pointers. If the source of the cast is an Optional<class>,
though, it's not allowed to "directly compare the isa-pointer"; because
the source might be Optional.none, i.e. null.
Add a check for nil when emitting the fast class cast.
rdar://108614878
We started using clang to emit the _OBJC_PROTOCOL_ definition.
But we would use a different name for the proto_list definition than clang.
"OBJC_LABEL_PROTOCOL$" (objc)
|--> OBJC_PROTOCOL
"\01l_OBJC_LABEL_PROTOCOL$_" (swift)
|--> OBJC_PROTOCOL
If an Objective C object also emitted the same protocol definition you could
end up in a situation where both clang's and swift's proto_list definitions
point to the same protocol definition.
Older linkers don't like that.
rdar://108505376
Only fully resolved substitutions are eligible to be considered
as conflicting, if holes are involved in any position that automatically
disqualifies a generic parameter.
Resolves: rdar://108534555
Resolves: https://github.com/apple/swift/issues/63450
Code like that is usually indicative of programmer error, and does not
round-trip through module interface files since there is no source
syntax to refer to an outer generic parameter.
For source compatibility this is a warning, but becomes an error with
-swift-version 6.
Fixes rdar://problem/108385980 and https://github.com/apple/swift/issues/62767.
