If we're using the macro-specific local discriminator, we need to
make sure we avoid mangling the regular local discriminator in
`appendDeclName`, since that could prematurely kick local discriminator
assignment before type-checking has finished.
rdar://143834482
What’s implemented now is actually *far* more thorough than what the surface syntax can currently express, mainly because I can’t apply @abi to nominal types yet.
The mangling of macro expansions relies on having a type-checked AST
for its enclosing context. When that enclosing context is within a
local context (say, a local type), mangling would trigger type
checking of that local type, which could then involve assigning local
discriminators. However, if this happens before type checking of the
enclosing function body, we would end up failing to assign closure
discriminators to (e.g.) autoclosures within the body.
The fundamental problem here is the interaction between discriminator
assignment (which can only happen after type checking) and mangling of
macro expansion buffers (which can happen during that type checking).
Break this cycle by providing a different approach to mangling macro
expansions within local contexts as the innermost non-local context +
a name-based discriminator within that local context. These manglings
are not ABI and are not stable, so we can adjust them later if we come
up with a scheme we like better. However, by breaking this cycle, we
eliminate assertions and miscompiles that come from missing
discriminators in this case.
Fixes rdar://139734958.
The main change here is to associate a witness table with a `ProtocolConformance` instead of a `RootProtocolConformance`.
A `ProtocolConformance` is the base class and can be a `RootProtocolConformance` or a `SpecializedProtocolConformance`.
Mangling this information for future directions like component lifetimes
becomes complex and the current mangling scheme isn't scalable anyway.
Deleting this support for now.
We want to ensure that functions/methods themselves do not have sending mangled
into their names, but we do want sending mangled in non-top level positions. For
example: we do not want to mangle sending into a function like the following:
```swift
// We don't want to mangle this.
func test(_ x: sending NonSendableKlass) -> ()
```
But when it comes to actually storing functions into memory, we do want to
distinguish in between function values that use sending vs those that do not
since we do not want to allow for them to alias. Thus we want to mangle sending
into things like the following:
```swift
// We want to distinguish in between Array<(sending T) -> ()> and
// Array((T) -> ()>
let a = Array<(sending T) -> ()>
// We want to distinguish in between a global contianing (sending T) -> () and a
// global containing (T) -> ().
var global: (sending T) -> ()
```
This commit achieves that by making changes to the ASTMangler in getDeclType
which causes getDeclType to set a flag that says that we have not yet recursed
through the system and thus should suppress the printing of sendable. Once we
get further into the system and recurse, that flag is by default set to true, so
we get the old sending parameter without having to update large amounts of code.
rdar://127383107
The names of the private witness table accessor thunks we generate for
an opaque return type mangle the concrete conformance of the underlying
type.
If a conformance requirement of the opaque return type was witnessed by
a conditional conformance of a variadic generic type, we would crash
because of an unimplemented case in the mangler.
Fixes rdar://problem/125668798.
Invertible protocols are currently always mangled with `Ri`, followed by
a single letter for each invertible protocol (e.g., `c` and `e` for
`Copyable` and `Escapable`, respectively), followed by the generic
parameter index. However, this requires that we extend the mangling
for any future invertible protocols, which mean they won't be
backward compatible.
Replace this mangling with one that mangles the bit # for the
invertible protocol, e.g., `Ri_` (followed by the generic parameter
index) is bit 0, which is `Copyable`. `Ri0_` (then generic parameter
index) is bit 1, which is `Escapable`. This allows us to round-trip
through mangled names for any invertible protocol, without any
knowledge of what the invertible protocol is, providing forward
compatibility. The same forward compatibility is present in all
metadata and the runtime, allowing us to add more invertible
protocols in the future without updating any of them, and also
allowing backward compatibility.
Only the demangling to human-readable strings maps the bit numbers
back to their names, and there's a fallback printing with just the bit
number when appropriate.
Also generalize the mangling a bit to allow for mangling of invertible
requirements on associated types, e.g., `S.Sequence: ~Copyable`. This
is currently unsupported by the compiler or runtime, but that may
change, and it was easy enough to finish off the mangling work for it.
For entities that must be part of the primary definition of a type,
mangle without inverses on the generic parameters of the enclosing
type. This ensures that we can adopt noncopyable on the generic
parameters without breaking the mangling of the fundamental entities
that describe the layout of the type.
Do this for enum elements first, so we don't break the mangling of
`Optional`. There will be other cases to consider as well.
Reformatting everything now that we have `llvm` namespaces. I've
separated this from the main commit to help manage merge-conflicts and
for making it a bit easier to read the mega-patch.
This is phase-1 of switching from llvm::Optional to std::optional in the
next rebranch. llvm::Optional was removed from upstream LLVM, so we need
to migrate off rather soon. On Darwin, std::optional, and llvm::Optional
have the same layout, so we don't need to be as concerned about ABI
beyond the name mangling. `llvm::Optional` is only returned from one
function in
```
getStandardTypeSubst(StringRef TypeName,
bool allowConcurrencyManglings);
```
It's the return value, so it should not impact the mangling of the
function, and the layout is the same as `std::optional`, so it should be
mostly okay. This function doesn't appear to have users, and the ABI was
already broken 2 years ago for concurrency and no one seemed to notice
so this should be "okay".
I'm doing the migration incrementally so that folks working on main can
cherry-pick back to the release/5.9 branch. Once 5.9 is done and locked
away, then we can go through and finish the replacement. Since `None`
and `Optional` show up in contexts where they are not `llvm::None` and
`llvm::Optional`, I'm preparing the work now by going through and
removing the namespace unwrapping and making the `llvm` namespace
explicit. This should make it fairly mechanical to go through and
replace llvm::Optional with std::optional, and llvm::None with
std::nullopt. It's also a change that can be brought onto the
release/5.9 with minimal impact. This should be an NFC change.
'MacroExpansionDecl' and 'MacroExpansionExpr' have many common methods.
Introduce a common base class 'FreestandingMacroExpansion' that holds
'MacroExpansionInfo'.
Factor out common expansion logic to 'evaluateFreestandingMacro'
function that resembles 'evaluateAttachedMacro'.
The mangling of attached macro expansions based on the declaration to
which they are attached requires semantic information (specifically,
the interface type of that declaration) that caused cyclic
dependencies during type checking. Replace the mangling with a
less-complete mangling that only requires syntactic information from
the declaration, i.e., the name of the declaration to which the macro
was attached.
This eliminates reference cycles that occur with attached macros that
produce arbitrary names.
CF_OPTIONS is defined differently in the SDK based on
a __cplusplus preprocessor branch. As a result, declarations
referencing CF_OPTIONS are mangled differently depending
on if C++ interop is enabled.
This meant a module compiled with cxx interop on could
not be linked with a module compiled without and vice versa.
This patch modifies the mangler such that the mangled names
are consistent. This is achieved by feeding the mangler a modified
AST node that looks like the Objective-C definition of CF_OPTIONS,
even when we have cxx interop enabled.
Extend the name mangling scheme for macro expansions to cover attached
macros, and use that scheme for the names of macro expansions buffers.
Finishes rdar://104038303, stabilizing file/buffer names for macro
expansion buffers.
