clang is miscompiling some swiftcall functions on armv7s.
Stop using swiftcall in some places until it is fixed.
Reverts c5bf2ec (#13299).
rdar://35973477
This makes them consistent no matter what shenanigans are pulled by
the importer, particularly NS_ENUM vs. NS_OPTIONS and NS_SWIFT_NAME.
The 'NSErrorDomain' API note /nearly/ works with this, but the
synthesized error struct is still mangled as a Swift declaration,
which means it's not rename-stable. See follow-up commits.
The main place where this still falls down is NS_STRING_ENUM: when
this is applied, a typedef is imported as a unique struct, but without
it it's just a typealias for the underlying type. There's also still a
problem with synthesized conformances, which have a module mangled
into the witness table symbol even though that symbol is linkonce_odr.
rdar://problem/31616162
* [runtime] Clean up symbols in error machinery.
* [runtime] Clean up symbols in Foundation overlay.
* [runtime] Clean up symbols in collections and hashing.
* [runtime] Remove symbol controls from the Linux definition of swift_allocError.
* [tests] Add more stub functions for tests that link directly to the runtime.
Previously it was part of swiftBasic.
The demangler library does not depend on llvm (except some header-only utilities like StringRef). Putting it into its own library makes sure that no llvm stuff will be linked into clients which use the demangler library.
This change also contains other refactoring, like moving demangler code into different files. This makes it easier to remove the old demangler from the runtime library when we switch to the new symbol mangling.
Also in this commit: remove some unused API functions from the demangler Context.
fixes rdar://problem/30503344
The ABI mismatch here would cause a crash in cases when the Foundation overlay wasn't available, or its implementation of swift_Foundation_getErrorDefaultUserInfo wasn't dynamically resolvable, such as in a stripped statically linked binary. Fixes rdar://problem/29173132.
The lazy population of the NSError fields is ordered such that the domain is written into the object last with acq/rel ordering, so another thread only needs to check the domain to see whether the initialization has already happened. (The initialization itself is idempotent, so we can optimistically perform the initialization and discard the results if we race.) Checking the user info is redundant, and is also wrong for real NSError objects, since [NSError errorWithDomain:d code:c userInfo:nil] will in fact plant nil in the userInfo field of the object, leading us to attempt to bridge an already-native NSError.
Use the generic type lowering algorithm described in
"docs/CallingConvention.rst#physical-lowering" to map from IRGen's explosion
type to the type expected by the ABI.
Change IRGen to use the swift calling convention (swiftcc) for native swift
functions.
Use the 'swiftself' attribute on self parameters and for closures contexts.
Use the 'swifterror' parameter for swift error parameters.
Change functions in the runtime that are called as native swift functions to use
the swift calling convention.
rdar://19978563
The runtime and stubs are built for ALL targets, not specific ones. This allows
us to configure when cross-compiling to Windows again. Collapse the dual
addition of the swiftRuntime into a single build. This unifies the runtime
build for the apple and non-Apple SDKs. The difference here was the ObjC
interop sources. In order to deal with that unification add a CPP macro to
indicate whether the interop sources should be included or not.
Provide default implementations for all of the CustomNSError requirements:
* errorDomain: default to the name of the enum type
* errorCode: default to the same thing "_code" gets, e.g., the enum tag or
raw value
* errorUserInfo: default to empty
This makes it significantly easier to customize just one aspect of the
NSError view of an error type, e.g., just the user-info dictionary,
without having to write boilerplate for the others. This was actually
part of SE-0112, but I missed it in the original implementation and we
thought it was an amendment.
Fixes rdar://problem/23511842.
If the Swift error wrapped in a _SwiftNativeNSError box conforms to
Hashable, the box now uses the Swift's conformance to Hashable.
Part of rdar://problem/27574348.
This partially reverts commit
47f01ee98d. In other words, I'm bring
back the changes to the Swift runtime that should eliminate the data
race in swift_bridgeErrorToNSError, but leaving the actual test
disabled while we do more investigation.
Huge thanks to John for noting that 'consume' didn't provide the
guarantees we wanted, and to Michael G. for getting a TSan build up
and running to identify/verify this race.
It's possible that this is overlay strict, and that we only need to
look at the domain to ensure that the code and userInfo are
visible. However, TSan seems to prefix the form in this patch, so
we'll be more conservative for now.
Fixes rdar://problem/27541751.
If there's no better mapping for a Swift value into an Objective-C object for bridging purposes, we can fall back to boxing the value in a class. This class doesn't have any public interface beyond being `NSObject`-conforming in Objective-C, but is recognized by the Swift runtime so that it can be dynamically cast back to the boxed type.
An error type can conform to one or more of these new protocols to
customize its behavior and representation. From an implementation
standpoint, the protocol conformances are used to fill in the
user-info dictionary in NSError to interoperate with the Cocoa
error-handling system.
There are a few outstanding problems with this implementation,
although it is fully functional:
* Population of the userInfo dictionary is currently eager; we
should use user info providers on platforms where they are
available.
* At present, the Swift dynamic casting machinery is unable to unbox a
_SwiftNativeNSError when trying to cast from it to (e.g.) an
existential, which makes it impossible to retrieve the
RecoverableError from the NSError. Instead, just capture the original
error---hey, they're supposed to be value types anyway!---and use that
to implement the entry points for the informal
NSErrorRecoveryAttempting protocol.
This is part (1) of the proposal solution.
_SwiftNativeNSError is a runtime-private subclass, and NSError's inherited NSCoding implementation doesn't handle the Swift payload of bridged errors. We can't really archive arbitrary Swift values anyway yet, so just archive bridged NSError subclasses as regular NSErrors. Fixes rdar://problem/23051728.
The default implementation from NSObject doesn't know how to copy the Swift error payload, and the object's immutable anyway, so we just need to retain and return the object we already have.
The general rule here is that something needs to be SWIFT_CC(swift)
if it's just declared in Swift code using _silgen_name, as opposed to
importing something via a header.
Of course, SWIFT_CC(swift) expands to nothing by default for now, and
I haven't made an effort yet to add the indirect-result / context
parameter ABI attributes. This is just a best-effort first pass.
I also took the opportunity to shift a few files to just implement
their shims header and to demote a few things to be private stdlib
interfaces.
...and explicitly mark symbols we export, either for use by executables or for runtime-stdlib interaction. Until the stdlib supports resilience we have to allow programs to link to these SPI symbols.
Reuses the enum metadata layout and builder because most of the logic is
also required for Optional (generic arg and payload). We may want to
optimize this at some point (Optional doesn't have a Parent), but I
don't see much opportunity.
Note that with this approach there will be no change in metadata layout.
Changing the kind still breaks the ABI of course.
Also leaves the MirrorData summary string as "(Enum Value)". We should
consider changing it.
