This is going to let me just pass through the error struct to the diagnostic
rather than having the CRTP and then constructing an info object per CRTP.
Currently, to make it easier to refactor, I changed the code in
TransferNonSendable to just take in the new error and call the current CRTP
routines. In the next commit, I am going to refactor TransferNonSendable.cpp
itself. This just makes it easier to test that I did not break anything.
Extends checkForSameInstMultipleUseErrors to ignore irrelevant unowned uses,
including the mark_dependence base operand, which is currently considered a
pointer escape.
While the FIXME to derive whether an address begins initialized requires
auditing all sites where the instruction is emitted to begin with,
making this a predicate that depends only on the instruction can be done
now.
Such destroys mark the lifetime end of their operands along their
availability boundary. They are currently inserted in this test case
by the ClosureLifetimeFixup pass, but in the fullness of time they will
be present for every value which is not explicitly destroyed (that's
what complete OSSA lifetimes is mostly about).
Currently, such destroys are diagnosed by DiagnoseUnreachable. Fix the
diagnostic pass not to diagnose these valid instructions.
rdar://137960229
PredictableMemOpt only runs with OSSA. The recent refactoring did not
preserve non-OSSA conditions because it adds complexity. This commit
just makes it official.
In the near future, this pass will be replaced by an OSSA-only pass
anyway, so this brings us one small step closer to that.
The reason that I am modifying this error is that in situations like the
following one can have a Sendable type that triggers this error since the box
containing the value is non-Sendable.
```
func methodConsuming(x: consuming SendableKlass) async throws {
try await withThrowingTaskGroup(of: Void.self) { group in
group.addTask { // expected-tns-warning {{passing closure as a 'sending' parameter risks causing data races between code in the current task and concurrent execution of the closure}}
useValue(x) // expected-tns-note {{closure captures reference to mutable var 'x' which is accessible to code in the current task}}
}
try await group.waitForAll()
}
}
```
rdar://133813644
(cherry picked from commit 36c2b3cc1330c07dcf9715f8ae88d31f9dba58c4)
In Embedded Swift, witness method lookup is done from specialized witness tables.
For this to work, the type of witness_method must be specialized as well.
Otherwise the method call would be done with wrong parameter conventions (indirect instead of direct).
This requires two major changes.
The first is that we need to teach SILGen that the isolation of an initializer
is essentially dynamic (as far as SILGen is concerned) --- that it needs to emit
code in order to get the isolation reference. To make this work, I needed to
refactor how we store the expected executor of a function so that it's not
always a constant value; instead, we'll need to emit code that DI will lower
properly. Fortunately, I can largely build on top of the work that Doug previously
did to support #isolation in these functions. The SIL we emit here around delegating
initializer calls is not ideal --- the breadcrumb hop ends up jumping to the
generic executor, and then DI actually emits the hop to the actor. This is a little
silly, but it's hard to eliminate without special-casing the self-rebinding, which
honestly we should consider rather than the weirdly global handling of that in
SILGen today. The optimizer should eliminate this hop pretty reliably, at least.
The second is that we need to teach DI to handle the pattern of code we get in
delegating initializers, where the builtin actually has to be passed the self var
rather than a class reference. This is because we don't *have* a class reference
that's consistently correct in these cases. This ended up being a fairly
straightforward generalization.
I also taught the hop_to_executor optimizer to skip over the initialization of
the default-actor header; there are a lot of simple cases where we still do emit
the prologue generic-executor hop, but at least the most trivial case is handled.
To do this better, we'd need to teach this bit of the optimizer that the properties
of self can be stored to in an initializer prior to the object having escaped, and
we don't have that information easily at hand, I think.
Fixes rdar://87485045.
This fixes a logic error that will only be exposed by the additional
functionality that motivated the cleanup. When the additional
functionality is checked in, it will contain a unit test for this fix.
This bug was introduced in:
commit 755a146730 (public/predmem_rewrite)
Author: Andrew Trick <atrick@apple.com>
Date: Fri Sep 27 22:18:29 2024
[NFC] rewrite PredictableMemOpts dead allocation elimination
Generalize the code that promotes the remaining uses of an allocation to make it
readable and extensible. We need to be able to promote allocations with more
interesting uses, namely mark_dependence.
MandatoryPerformanceOptimizations already did most of the vtable specialization work.
So it makes sense to remove the VTableSpecializerPass completely and do everything in MandatoryPerformanceOptimizations.
Motivated by need for protocol-based dynamic dispatch, which hasn't been possible in Embedded Swift due to a full ban on existentials. This lifts that restriction but only for class-bound existentials: Class-bound existentials are already (even in desktop Swift) much more lightweight than full existentials, as they don't need type metadata, their containers are typically 2 words only (reference + wtable pointer), don't incur copies (only retains+releases).
Included in this PR:
[x] Non-generic class-bound existentials, executable tests for those.
[x] Extension methods on protocols and using those from a class-bound existential.
[x] RuntimeEffects now differentiate between Existential and ExistentialClassBound.
[x] PerformanceDiagnostics don't flag ExistentialClassBound in Embedded Swift.
[x] WTables are generated in IRGen when needed.
Left for follow-up PRs:
[ ] Generic classes support
The "buffer ID" in a SourceFile, which is used to find the source file's
contents in the SourceManager, has always been optional. However, the
effectively every SourceFile actually does have a buffer ID, and the
vast majority of accesses to this information dereference the optional
without checking.
Update the handful of call sites that provided `nullopt` as the buffer
ID to provide a proper buffer instead. These were mostly unit tests
and testing programs, with a few places that passed a never-empty
optional through to the SourceFile constructor.
Then, remove optionality from the representation and accessors. It is
now the case that every SourceFile has a buffer ID, simplying a bunch
of code.
No update is needed for the values they produce. This pass should
really be refactored not to crash on instructions that aren't explicitly
listed or at least not to compile if not every instruction is listed.
rdar://133779160
In order for availability checks in iOS apps to be evaluated correctly when
running on macOS, the application binary must call a copy of
`_stdlib_isOSVersionAtLeast_AEIC()` that was emitted into the app, instead of
calling the `_stdlib_isOSVersionAtLeast()` function provided by the standard
library. This is because the call to the underlying compiler-rt function
`__isPlatformVersionAtLeast()` must be given the correct platform identifier
argument; if the call is not emitted into the client, then the macOS platform
identifier is used and the iOS version number will be mistakenly interpreted as
a macOS version number at runtime.
The `_stdlib_isOSVersionAtLeast()` function in the standard library is marked
`@_transparent` on iOS so that its call to `_stdlib_isOSVersionAtLeast_AEIC()`
is always inlined into the client. This works for the code generated by normal
`if #available` checks, but for the `@backDeployed` function thunks, the calls
to `_stdlib_isOSVersionAtLeast()` were not being inlined and that was causing
calls to `@backDeployed` functions to crash in iOS apps running on macOS since
their availability checks were being misevaluated.
The SIL optimizer has a heuristic which inhibits mandatory inlining in
functions that are classified as thunks, in order to save code size. This
heuristic needs to be relaxed in `@backDeployed` thunks, so that mandatory
inlining of `_stdlib_isOSVersionAtLeast()` can behave as expected. The change
should be safe since the only `@_transparent` function a `@backDeployed` thunk
is ever expected to call is `_stdlib_isOSVersionAtLeast()`.
Resolves rdar://134793410.
The reason why I am making this change is because I want to put a merge
operation on DiagnosticBehavior. This merge operation allows for
DiagnosticBehavior to work like a lattice. When one merges, one moves
potentially from fatal, error to things like note, ignore.
