Implement the toolchain side of proposal #7254, removing the `:!`
binding
syntax for generic and template parameters in favor of the keywords
`generic`,
`template`, and `runtime` plus contextual defaults for phase.
For valid programs this is semantics-preserving: each binding resolves
to the
same phase, and produces the same SemIR, as it did under `:!`/`:`. The
parser
derives a binding's phase from its syntactic context plus any explicit
phase
keyword; new diagnostics and error recovery for misused keywords are
described
below.
Implementation details for each component:
- Lexer: remove the `:!` (`ColonExclaim`) token, move its virtual
parse-node
budget onto `:`, and add the `generic` and `runtime` keywords.
- Parser: thread a `BindingContext` (`ExplicitParam`, `DeducedParam`, or
`CompileTimeEntityParam`) from declaration introducers down through
parameter
lists to each binding pattern, using a one-token lookahead to
distinguish a
name-qualifier parameter list from a declaration's own final list.
Parameters
of a compile-time entity (`class`, `interface`, `constraint`, `choice`,
`alias`, `export`, `namespace`) and deduced `[]` parameters default to
checked
generic; explicit function parameters and local bindings default to
runtime.
`HandleBindingPattern` resolves the phase from that context plus the
keyword: a
`generic` keyword needs no node of its own (the phase is carried by the
binding's node kind), while a `runtime` keyword is preserved as a
`RuntimeBindingName` node so `check` can name it in a diagnostic. A
phase
keyword that is merely redundant with the contextual default is
diagnosed
here, without invalidating the parse tree.
- Check: a phase keyword that is invalid for its context (for example
`runtime`
on a checked-generic parameter) is diagnosed here, and recovers by
building an
error binding that still introduces the name so that later uses of it do
not
produce cascading errors.
The removed `:!` syntax is now rejected as an ordinary parse error.
The `form`/`:?`/`->?` ("extended types") portion of proposal #7254 is
left for a
separate change.
Assisted-by: Claude Code
In #7436 we stopped substituting `.Self` when collecting witnesses out
of a facet type. While this was correct, it did not capture all the
cases that need to avoid substituting `.Self`. And it poisoned the
`IdentifiedFacetType` cache by not replacing `.Self` but storing the
result in the cache. This led to incoherent behaviour, where the result
of an impl lookup would change depending on which ones had been done
previously.
Now we use a flag to track for each `.Self` if we're currently
type-checking inside the scope where it was introduced in a facet type.
While inside that scope, identify should not replace the `.Self`. Any
use of it should remain as-is since we don't yet know what value will
replace it. We call this state "frozen" since it should not be modified
by identify. This requires a substitution step when we leave the scope
that introduced the `.Self`, to remove the flag. The flag is set in the
`EntityName` of the `SymbolicBinding`, and is part of the canonical
value, since `.Self` can become part of types, which are constants, and
the flag needs to follow it for correct behaviour.
We also have to ensure the flag is the same when doing comparison with
constants from inside a facet type and constants from outside. For
instance in `(Z where .Z1 = ()) where .Z2 = .Z1`, when we arrive at the
second `.Z1` its `.Self` will be frozen, while the `.Z1 = ()` contains a
non-frozen `.Self`. So we add the frozen flag to the first when storing
it in `where_stack` in order to compare the constant values of the two
`.Z1`.
The `WhereExpr` requirement inst kinds now have an `InstConstantKind` of
`AlwaysUnique` instead of `Never`. This allows us to add them to the
usual InstBlocks, and in an `eval fn` body they have a constant value,
so eval does not fail when trying to call that function. We have to be
careful to not consider `AlwaysUnique` as being actually concrete
though, since their constant value erases `.Self`-dependence. This
allows us to stop special casing them when thawing the requirements
block in a `WhereExpr`, and we can just thaw each `InstId` in the block
in a straightforward manner.
We add the new flag to the instruction's fingerprint and name in
formatted semir.
This adds SemIR structs and implements building `observe` lists, as well
as naming, formatting, and importing `observe` declarations.
---------
Co-authored-by: Dana Jansens <danakj@orodu.net>
This allows C++ to call Carbon functions with generic type parameters,
with some conditions. Example:
```carbon
interface I {
fn Doit(self);
}
class A {
impl as I { fn Doit(unused self) {} }
}
class B {
impl as I { fn Doit(unused self) {} }
}
fn F[T:! I](t: T) {
t.Doit();
}
inline Cpp '''
void G() {
Carbon::A a;
Carbon::B b;
Carbon::F(a);
Carbon::F(b);
}
''';
```
The initial support is limited; only explicit parameters are handled
currently.
`CarbonExternalASTSource::GetOrExportFunctionToCpp` now generates a
`clang::FunctionTemplateDecl` for generic Carbon functions. If C++ code
attempts to call that templated function,
`CarbonExternalASTSource::LoadExternalSpecializations` will be called
with the template argument types of that call site. Then we can generate
a specialized thunk for those argument types for C++ to call.
In some cases the pattern block can depend on the initializer, so it
must be sequenced after it. See #7469 for a more detailed explanation of
why this is necessary.
The bulk of this change is changing most pattern insts to be `Always`
rather than `AlwaysUnique` constants, so that they can be wrapped in
`SpecificConstant`s to perform substitution. That then lets thunking
rely much more on `SpecificConstant` wrappers instead of deep-copying
the inst tree with modified types.
This approach to thunking should scale better, particularly as things
like form generics make function signatures more complex, because we can
leverage the existing support for constant evaluation and substitution.
Unfortunately, applying this approach to binding patterns will require
more work; see the TODO near the top of `thunk.cpp` for details.
---------
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
Adds support for arithmetic and comparison operators on
`Core.CharLiteral`s, as well as conversions between `CharLiteral` and
integer types.
Make some minor tweaks to fix skill issues encountered while making this
change.
Assisted-by: Gemini via Antigravity
Adapters were erroneously satisfying `Core.Destroy` because we were
directly getting the object's representation without consideration for
abstract and adapted types. This change ensures that adapted types'
representations are used instead of the adapter types.
This ensures the clang_decls map is used as a cache - without this,
visiting the same entity twice could cause it to be
re-exported/duplicated. See attached test case.
Fix import logic to make all imported packages be children of the
`NameScopeId::Package` scope. Previously, indirectly-imported packages
would end up as children of their importing package's scope, which
resulted in them not being treated as packages at all, and in particular
not being fingerprinted as packages.
Fixing that caused a failure in the fingerprinting logic as we started
to encounter packages with no correspoding import scopes. Instead of
looking for import scopes, use a simpler mechanism to map packages to
their package names, and clean up.
Unfortunately the latter change churns all the fingerprints again :(
Hopefully this is the last time for a while.
Implement support for floating-point <-> integer type conversions as
described in #820 and #845, extended to support `unsafe as` conversions
for the conversions that can't be expressed as either implicit
conversions or `as` conversions.
One tricky part here is conversions from floating-point literals to
integer types. Such literals may have both a very large mantissa and a
corresponding somewhat large negative exponent, and still produce a
result that is in the range of values that a small integer type can
represent. In order to support that while avoiding building very large
2^N or 10^N constants in general, we first compute a conservative
approximation of the number of bits necessary to represent the integer
result, with an early exit if the number is either definitely too large
or definitely zero. The remaining cases have a reasonable bound on the
size of integer necessary to compute the base^exponent multiplicand.
Assisted-by: Gemini via Antigravity
When we import from another library in the same package, its entities
end up with our library as their parent scope, resulting in cross-file
fingerprint mismatches. Instead, only include the library ID when
fingerprinting either a package-private entity or an `ImportIRId` that
refers to a particular `SemIR::File`.
For now, disable the use of array types as by-var paramters and by-init
return types when exporting Carbon functions to C++, as C++ does not
support raw arrays being passed or returned by value.
Assisted-by: Gemini via Antigravity
Include the library name in the fingerprint of an entity declared
`private` at namespace scope. Include the entity's fingerprint in the
mangling of a library-private entity.
This fixes miscompiles if two libraries in the same package declare
`private` entites with the same name. We can't fix this with internal
linkage because library-private entities can be reachable through
generics defined in the API file of the library.
Assisted-by: Gemini via Antigravity
Carbon-side thunks (for example the `Copy`/`Destroy` witness thunks
generated for imported C++ types) are mangled by Carbon, and their names
incorporate a fingerprint of the involved types. The instruction
fingerprinter identifies a class only by its name and parent scope,
which is sufficient for Carbon classes but not for imported C++ classes:
different specializations of one class template (and other cases such as
types in anonymous namespaces) share a Carbon name and parent scope. As
a result, the thunks for two distinct specializations could mangle to
the same name, producing a single LLVM function with two definitions and
failing `verifyModule` during lowering.
When fingerprinting a class imported from C++, also include the Clang
mangled name of its type.
Test: toolchain/lower/testdata/interop/cpp/thunks.carbon gains a split
with two specializations of one class template, each requiring a thunk;
their thunks now get distinct mangled names instead of colliding.
Assisted-by: Claude Code
---------
Co-authored-by: Christopher Di Bella <cjdb.ns@gmail.com>
Allow a `ref`-tagged expression to be converted to match a reference
parameter. Move the `ref` checks to the start of `Convert`. Remove the
diagnostic for applying `ref` to a non-reference expression so that
non-reference expressions that convert to a reference would be accepted
(although we don't currently have any such conversions).
Assisted-by: Gemini via Antigravity
Fixes link failures when referencing a symbol involving a fingerprint
from a different package.
Previously we included the `Namespace`'s `import_id` as part of its
fingerprint, which caused local and imported namespaces to get different
fingerprints. We now store the `import_id` on the `NameScope` instead of
on the `Namespace` inst to avoid this problem.
Also, when we reach a package-level `NameScopeId`, consistently
fingerprint it as a (package name, library name) pair. Previously the
fingerprinting depended on whether it was imported or not, as an
imported `NameScopeId` had a parent scope (the current package). We need
to include the library name here so that private entities with the same
name in different libraries have different fingerprints.
This also fixes passing a value expression to a forwarding reference,
since we currently deduce a `const T&&` parameter in that case.
We were accidentally looking at the type of the thunk parameter (which
is never an rvalue reference) rather than the type of the callee
parameter.
Previously, we picked a single Carbon parameter pattern for each C++
parameter pattern. This doesn't work well in cases where the Carbon
semantics and the C++ semantics are not perfectly aligned. In
particular, when a parameter is passed by value in C++, that might mean
either pass-by-move (which in Carbon would best be modeled by a `var`
pattern, as no other form of parameter would perform a move) or
pass-by-copy (which in Carbon would best be modeled by a value
parameter, as a `var` parameter would force an extra copy).
After this change, we compute a passing mode for each parameter based on
the implicit conversion sequence from the argument to the parameter as
determined by C++ overload resolution, and use that to determine the
Carbon pattern corresponding to each C++ parameter. This results in
potentially generating multiple different thunks for the same C++
function if it's called in different ways, but we already did that to
handle default arguments and list-initialization. The passing modes are
included in the thunk mangling.
Add a new value store for clang decl signatures, which capture the
information about parameter passing mode as well as the other existing
information about different ways that a C++ function might be imported
to Carbon.
Most of the rules for computing passing modes are the same as before:
const references use pass by value, non-const lvalue references use
pass-by-ref, non-const rvalue references use pass-by-var. But for C++
non-reference parameters, pick between pass-by-value and pass-by-var
based on whether the implicit conversion sequence was effectively
performing a copy. Prefer pass-by-value if either would work and they'd
do the same thing. We still use pass-by-value for const references, even
when the argument is an lvalue and we could pass a reference; we may
want to change this in future.
For virtual functions, we try to pick a worst-case passing mode, as we
can only pick a single signature for what goes in the vtable. Calls to
virtual functions will still use a thunk to C++, allowing variance in
the calling convention at call sites. We don't allow variance in the
overriders as we don't implement support for thunks for virtual
functions yet. We currently use pass-by-value for const reference
parameters here, but that should probably change at some point.
Assisted-by: Gemini via Antigravity
Fixes mangling collisions when two thunks with the same name (eg, `Op`)
are created in the same context, which in turn would lead to LLVM
verifier failures and miscompiles.
To support this, add a new value store to track a little more
information about thunks beyond what's in the `Function`.
This lets us stop eliding it in textual semir tests with dump ranges.
Previously it would always get elided, even though it was part of the
range being dumped, and was referred to by other instructions in the
dump range.
Since each `.Self` is unique (can change its type if not its value) in a
facet type, having each one distinct by location also aids
understanding.
See
[here](https://docs.google.com/document/d/1rWcueFwIfZox6GKVGxiUG4cBzjrZ6djXiIDGyJDtrE4/edit?tab=t.0)
for the design doc.
This also removes the default value of the `result_type_inst_id`
parameter of `HandleAction`, moves it before the action in the parameter
list, and documents it. This solves two problems:
- The default made it easy to forget, leading to unnecessary
`TypeOfInst` instructions.
- When it was present, putting it after the fairly "bulky" action
argument tended to make the callsite harder to read.
When importing a C++ function with an rvalue reference parameter, we
previously produced a Carbon value parameter. This would lead to the
toolchain believing it could pass the address of a non-expiring object
to the function, which would lead to a use-after-move.
Instead, we now map non-const rvalue reference parameters to Carbon
`var` parameters. This forces the object passed into C++ to be unique
and owned by the call. While that's not an exact match for C++ rvalue
reference parameters, given that it provides "always move" not
"conditionally move", it's the closest match we have at the moment.
When creating the C++ thunk, make the parameters references if the
corresponding callee parameters are `ref`s.
When creating the Carbon thunk, tag the call arguments as `ref` if the
corresponding callee parameters are `ref`s.
The FunctionDecl created for calling the Carbon thunk now takes a `self`
parameter for non-static methods, and the C++ thunk now passes an extra
argument for that `self` parameter when needed.
The CXXMethodDecl thunk created for calling methods now sets the storage
class appropriate depending on whether the method is static or not.
To reduce the number of parameters being passed around to thunk-building
functions, added a `FunctionInfo` struct and pass that around instead.
Add import/ and export/ under function/. Move most top-level tests to a
new basics/ with subdirectories for `import` directives and `inline
Cpp`. Add subdirectory for primitive type handling. Move all `reverse/`
tests to somewhere else, typically under an `export/` directory.
I split two test files up: constexpr.carbon got split into var/ and
function/ pieces, and reverse/simple.carbon was inlined into
namespace/export.carbon. The rest are just simple renames.
Instead of allowing lower to pick whatever type layout it desires,
compute the layouts of types as part of completing the type, and make
lower build types that match that representation.
For now we assume that all pointers are 64-bit, since we don't have
access to target information. We allow tail padding reuse for structs
and tuple types (and by extension, for classes, since they use structs
as their object representation), but not for arrays.
In order to build matching LLVM types, we create LLVM packed structs
where necessary, and we insert inter-field padding on the end of the
previous field so that GEP indexes still always match Carbon's
ElementIndexes.
We don't yet use the computed alignment much in LLVM IR generation -- in
particular, `alloca`s, `load`s, and `store`s should probably use the
computed type alignment, but don't.
Assisted-by: Gemini via Antigravity
Instead of exporting a class or namespace each time a new C++ name
lookup discovers it, track that we have exported the entity on its name
scope, and if a new name lookup finds the same entity, produce the same
clang declaration.
Also declare them `inline` since we're putting the `always_inline`
attribute on them. Use the `internal_linkage` attribute rather than
`SC_Static` since it's a more precise mechanism and matches what we do
for static member functions in reverse interop (where `SC_Static` means
something else and would not give the function internal linkage).
This is already allowed as a builtin conversion, but the impl allows the
generics system to know about it, so that conversions like
`Optional(T*)` to `Optional(const T*)` are allowed. This in turn allows
a C++ `T*` to be implicitly converted to a C++ `const T*` in Carbon
code.
Instead of recursing back into Convert, make CppThunkRef conversion just
add an extra step to category conversion, performing a copy conversion
followed by an ephemeral reference binding conversion.
This is only fixing the decision about *whether* to produce a witness.
Implementation of the witness is still a TODO, though where a body is
generated, it should also precisely reflect where one _needs_ to be
generated.
Note the tests:
- toolchain/lower/testdata/function/generic/import_core_witness.carbon
- toolchain/lower/testdata/function/generic/import_unused_def.carbon
These tests can probably be produced _without_ Core.Destroy, but I found
the essence of them while trying to build //examples with Core.Destroy
and a simpler minimization wasn't striking me.
Assisted-by: Google Antigravity with Gemini
---------
Co-authored-by: jonmeow <jperkins@google.com>
This will be used for const-evaling functions. Splitting into a separate
commit since it touches a lot of test files, and a couple fail_todo
tests are no longer failing.
Instead of injecting code to declare an `operator new`, generate AST for
it directly. In order to use this, directly generate a `CXXNewExpr`
rather than asking Clang to build one.
This is less of a hack, and doesn't visibly leak an `operator new`
declaration that inline C++ code or template instantiations might see.
It also avoids generating a warning in C++26 and later that the
`constexpr` declaration of `operator new` is used but not defined.
Assisted-by: Gemini 3.1 Pro via Antigravity
When a `var` is not explicitly given an initializer, initialize it in
one of two ways:
* If its type implements the new interface `Core.Default`, call
`Core.Default.Op` to initialize it.
* Otherwise, if its type implements `UnformedInit`, leave it in an
unformed state. For now, this is always an uninitialized state, but that
will change in the future.
* If neither of those apply, the `var` declaration is ill-formed.
This is a step towards implementing leads decision #6739 and proposals
#257 and #5913.
Assisted-by: Gemini 3.1 Pro via Antigravity
---------
Co-authored-by: Geoff Romer <gromer@google.com>
Chandler Carruth
Dana Jansens
Özgür T. ÖnsoyandDana Jansens
Nicholas Bishop
Geoff Romer
Christopher Di Bella
David Blaikie
josh11bandJosh L
Jon Ross-Perkinsandjonmeow