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.
Rather than defining our own action groups, work to re-use the
`rules_cc` ones, as they are (much) more comprehensive. Also, completely
eliminate the `codegen` action group as it was not well used. For
example, `-march` flags and `-O` flags change the preprocessor macros
defined. There isn't a really great "codegen" heuristic, so just pass
those flags to all compiles which is simpler anyways.
I'm tempted to do the same with preprocessor actions, but maybe it makes
sense to have that one stay separate.
Assisted-by: Antigravity with Gemini
Use constant values in code that can work with a canonical value
(`TypeStructureReferencesSelf`).
Give explicit location ids for the full require decl and the constraint
to `ValidateRequire`.
Go directly from `InstId` to `TypeId` in `ValidateRequire`.
Correctly/explicitly handle constraint instructions which are not types
instead of calling `SemIR::TypeId::ForTypeConstant` and hoping for the
best.
Leave a clear spot where we will subst `.Self` out of the contraint.
This takes the bootstrap support that was added and makes it available
under convenient user-facing flags for while we're doing development.
For example, to build a bootstrap compiler and use it to build and run
the tests under `//common/...` you can now use:
```
bazel test --//:bootstrap_stage=1 --//:bootstrap_exec_config=true //common/...
```
This will use the stage1 bootstrap compiler, and it will build that
compiler in the exec config (so it is optimized and the above even works
when cross-building with Bazel).
Assisted-by: Antigravity with Gemini
The target was renamed in `82fad290285baf9763132a13b1f73de1e7919074`
from `//toolchain/install:carbon_toolchain_tar_gz_rule` to
`//toolchain/install:carbon_toolchain_tar_gz`
This still only works if the hash of the distinct types are identical
(so it still doesn't address the derived pointer v base pointer case -
well, not in the way we would want to address it, we could use this
change to make derived pointer and base pointer not compare equal, but
that's not very ergonomic)
I think in a follow up maybe I can use a `TranslatingKeyContext` to
translate `Derived*` to `Base*` in general.
No test coverage for this change, since it's a no-compile situation and
we don't seem to generally do no-compile tests.
Discovered while working on #6940
---------
Co-authored-by: Geoff Romer <gromer@google.com>
- Rename `ActionIsDependent` to `ActionIsPerformable` (with negated
meaning), because that name is more concrete and, um, actionable.
- Replace `OperandIsDependent` with `OperandDependence`, which returns a
`ConstantDependence` instead of a bool. We need this additional
generality for handling form actions, where we sometimes need to ask
whether something has _any_ dependence, not just whether it has template
dependence.
Instead, use the inst category to select the right block stack. This
simplifies the API for adding insts, and in subsequent changes it will
enable certain inst kinds like `SpliceInst` to seamlessly function as
either procedural insts or pattern insts.
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
Noticed this when testing the Carbon toolchain with a more complex
environment, don't have any way to observe this at the moment in Bazel
though.
Assisted-by: Antigravity with Gemini
This worked correctly in the system Clang toolchain, but was not
configured correctly in the Carbon toolchains. The test is designed to
let us cover all of these.
Assisted-by: Antigravity with Gemini
Hopefully this significantly reduces how often agents try to run `bazel`
directly without repeatedly including that in prompts. Also tried to
generally give useful skills for building, testing, and running things.
Also added a specific admonition to the `AGENTS.md` as there is a chance
that agents don't think they need to look at any skills for "standard"
build system commands like `bazel`, as those are "trivial". It seems
like a small chunk of context to spend to avoid churning with bad build
commands.
Assisted-by: Antigravity with Gemini
These turn up frequently in real-world code, for example when converting
a mutable global `Cpp.std.string_view` to a `Cpp.std.string`. Only
reject a non-constant call if the callee is `consteval`, not if it's
`constexpr`.
For calling non-`()` functions, the Carbon->Carbon thunk now takes an
extra reference parameter and writes the target function's return value
out to that parameter. (At the SemIR level this is how returns already
work, but adding this extra reference parameter is needed so that the
function is lowered correctly.) The C++ thunk now creates a local
variable to be initialized by the Carbon thunk, and then returns that
value to the original C++ caller.
The runtimes and bootstrap Bazel logic was previously built around
defining custom Bazel platforms constrained with `constraint_settings`.
The use of platforms added significant complexity, including the need to
"save" and "restore" the original platform, and other complexity
stemming from changing the platform as a whole.
This PR switches to use the simpler tool of build settings, and
`target_settings` on the toolchain rather than platform compatibility.
This remove the entire need to save and restore the platform, and also
generally simplifies things.
This PR also fixes some bugs in the bootstrap that were hidden by the
use of platforms, such as the need to carefully manage the different
inputs to the runtimes build so that generated inputs pick up the
correct exec configuration -- the exec transition happened to do this
"automatically", but it seems better to handle explicitly. And it cleans
up an extraneous copy of `carbon_runtimes.bzl` that snuck in somehow.
Assisted-by: Antigravity with Gemini
---------
Co-authored-by: Dana Jansens <danakj@orodu.net>
This has been working really well for me, is incredibly faster than the
other approach, and some commits continue to hit bugs in the old system
where files that aren't even going to be run through `clangd-tidy` end
up tripping up the execution. Hopefully all of that is resolved with the
new version.
Allow any type that has a mapping from Carbon to C++ to be exposed to
C++ via name lookup. This also exposes the logic to export Carbon
classes to C++ to apply during type mapping, which gives very slight
support for passing Carbon types to C++ functions from Carbon, but not
really enough to sensibly test yet.
Depends on #7042.
Previously we'd create a *huge* array here as the tagged ID produced a
very large index value, and spend multiple seconds allocating it and
filling it with zeroes the first time `GetCppLocation` was called.
Reduces test runtime from 26s -> 6s wall time, 450s -> 320s total time
on my machine for `-c dbg`.
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.
`DoVarPreWorkImpl` was provided an incorrect pattern type while trying
to match `var` parameters, which caused the toolchain to crash in
`Convert`. This commit changes `DoVarPreWorkImpl`'s API so it derives
the pattern type from the work item's pattern ID, rather than relying on
an external source.
---------
Co-authored-by: Dana Jansens <danakj@orodu.net>
This makes them part of the identified facet type, and we can see the
constraints as part of stringify and format output.
But this does not do enough to make them useful yet: Any `T impls X`
constraint must contain a reference to `.Self` somewhere. And `.Self`
references do not get substituted, so neither `T(.Self) impls X` and `T
impls X(.Self)` will match against an incoming facet value derived from
an `impl T(U) as X` or `impl T as X(U)`, since `U` and `.Self` are never
the same thing until `.Self` can be substituted.
Now that impl lookup runs into facet values containing `.Self` (a
symbolic binding), such as in `C(.Self)`, we were crashing assuming the
type of `.Self` is a FacetType, but it can be `type` in the case of
`type where C(.Self) impls...`. Instead, use an empty facet type for the
type of `.Self` so it is always a facet. This assists with substituting
other facets into it, without having to insert an extra FacetAccessType.
`MakePeriodSelfFacetValue()` now enforces this requirement.
---------
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
Fixes#7031
Also switches the previous symlinks test to be a more full integration
test. While a bit slow, it does seem worthwhile to have something that
tests things end-to-end, both with the prebuilt runtimes and the
on-demand runtimes. This test is already reasonably well separated from
the rest of the toolchain so incremental development shouldn't be
negatively impacted. And since we turned off ASan by default, it isn't
completely infeasibly expensive.
Assisted-by: Antigravity with Gemini
Note that this will require changing the branch protections to use new
names for all of the checks and be somewhat disruptive. There aren't any
really good ways I could find of fixing this. Some options that I
explored:
- Have a single `pre-merge` workflow file that contains all of the other
workflows, splitting as much of the logic as we can into re-usable
files. This would basically merge testing, `pre-commit`, and
`clangd-tidy` checking into a single workflow file. However, it would
also delay the pre-commit suggestions action to only run once _all_ of
these finish, rather than as soon as pre-commit finishes.
- Serialize `pre-commit` and the rest of `pre-merge` to get the effect
of the above option but without the downside. Instead, the downside
would be serializing some of our actions.
- Have a single `pre-merge` workflow that triggers whenever any of the
other workflows completes, and have it check whether all the others have
completed. It will fail until it reaches that point. This requires
passing in GitHub keys to the workflow so that it can check the status
of other checks, and documentation online seems to indicate it is
sometimes flaky, I assume because of racing triggers of events or
check-status not being guaranteed consistent in the queries.
- Have a single `pre-merge` workflow that polls, waiting for all the
other workflows to finish using some Python logic. This requires
building and maintaining code to poll GitHub, keys to authorize that
polling, and handling all of the failure modes of a polling operation --
timeouts, network issues, etc.
Maybe there are others, but not sure what they look like. Suggestions
welcome here.
I'm hesitant to either delay the pre-commit suggestions or serialize
pre-commit execution. And the complexity or flakiness of the other two
options seem worse than having to re-work the branch protections each
time the naming here changes. But interested if folks think a different
direction would be better.
Assisted-by: Antigravity with Gemini
TypeIterator has both SymbolicType and SymbolicBinding and these overlap
in their meaning. Clarify the API by removing SymbolicBinding and just
using SymbolicType for `SymbolicBinding` insts and when they are
converted to `type` to make a `SymbolicBindingType` inst. Add the
EntityNameId to the SymbolicType for when it is available, when the
instruction is just a simple reference to a binding.
---------
Co-authored-by: Chandler Carruth <chandlerc@gmail.com>
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).
The type must be complete to look for a witness for Destroy. Do this
check through type completion rather than just checking to see if the
ClassInfo says the definition is closed, since completing the type has
side effects (resolves the self specific definition).
Then look for whether the class is abstract through the CompleteTypeInfo
instead of just looking at the inheritance type on ClassInfo, like type
completion does.
Last, FacetTypes are trivially destroyed just like TypeType.
We don't yet populate the bases or fields, so the class types show up as
empty classes in C++ for now. But we do allow calls to static member
functions.
This works by generating two thunks, one in C++ and one in Carbon. For
example, given this input:
```c++
// Carbon:
fn Callme(f: f32) {}
// C++:
void F() {
// This will call `Callme__cpp_thunk`
Carbon::Callme(1.0);
}
```
These functions are generated:
```c++
// Carbon:
fn Callme__carbon_thunk(ref f: f32) {
// Call the target function.
Callme(f);
}
// C++:
// C++ declaration for the Carbon thunk.
void Callme__carbon_thunk(float& f);
void Callme__cpp_thunk(float f) {
// Call the Carbon thunk with args passed by reference.
Callme__carbon_thunk(f);
}
```
For now, all arguments are passed by reference, even if they are simple
types like pointers or i32.
Functions with non-void return types are not supported yet.
This splits off the functionality to handle the base facet type,
rewrites, and impls constraints into separate functions.
We use the Context instead of EvalContext throughout, as the goal is to
move this code to EvalConstantInst in time. That means we do not apply
specifics to the functions in the requirements inst block. That is fine
because WhereExpr never evaluates to an WhereExpr, so this instruction
never survives as a constant value long enough to be re-evaluated with a
specific applied to it.
Use the same C++ -> Carbon map for both interop directions, and when
importing an entity from Carbon -> C++, check whether it was originally
a C++ entity and if so return the original.
Assisted-by: Gemini via Google Antigravity
This implements p7016 for tree_sitter. It also updates the build and
source file to allow this to build successfully and documents how to
successfully run these tests with Bazel given that it is fundamentally
not hermetic.
Assisted-by: Antigravity with Gemini
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.
Fix some situations where we'd drop the storage argument when building
an in-place initializing expression. We now guarantee that an expression
with the in-place initializing category always has a storage argument.
When a function call appears in a generic, and calls another generic
that has a concrete type in its call-site signature, that concrete type
will be completed only in the file that contains the call. The generic
containing the call won't require completeness to be checked again when
forming a specific call, because the type was concrete. This means that
when lowering the call instruction, there is no single file that is
guaranteed to contain complete types for all of the callee's parameters
-- the file containing the specific callee won't necessarily have
completed the concrete parts of the signature, and the files containing
the definition and call won't necessarily have completed the symbolic
parts of the signature.
To handle this, look at both versions of the function when building its
lowered signature -- the version that we saw when forming the `call`
instruction and the version corresponding to the concrete, specific
callee, and combine information from both to form the LLVM function
type.
---------
Co-authored-by: Geoff Romer <gromer@google.com>
Richard Smith
Chandler Carruth
Dana Jansens
David BlaikieandGeoff Romer
Nicholas Bishop
rit
Christopher Di BellaandDana Jansens
dependabot[bot]