This fixes a bug where Explorer would not detect when:
- an implementation used a method to implement a class function in an interface
- an implementation used a class function to implement a method in an interface
- an implementation method used `addr self` when the interface method did not
- an interface method used `addr self` when the implementation method did not
at type checking time. This would then cause a crash at runtime.
Closes#2857
---------
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
Add partial support for initializing expressions for variable declaration. This is based on https://github.com/carbon-language/carbon-lang/pull/2006, which introduces expression categories, and how it is possible to convert to/from those different categories.
## Functional changes
* Initializing expressions initialize directly the provided storage when used to initialize a variable.
* Allows initializing expressions to avoid a copy when using `[var|let] name: type = call_expression(...)` by initializing `name` in-place.
* Support `returned var: ...` and `return <expr>`
* Support nested initializing expressions
## Main implementation changes
* Updated PatternMatch logic to handle expression categories
* Updated `VariableDefinition` interpreter statement to allocate and pass a location to initializing expressions
* Update statement actions to allow passing an allocation, used by return expr or returned var
* Modified the RuntimeScope API to be one step closer to the memory model we want to have
* Remove `GetAllocationId` and older `Bind` which don't apply
* New set of tests to highlight those different situations
* Added a new intrinsic to print the allocation stack (and make sure we behave correctly, beyond visible side effects)
## Next work
* Dedicated `Action` to retrieve expression category information in the interpreter (https://github.com/carbon-language/carbon-lang/pull/2927)
* Avoid copies when initializing value expression from reference expression and prevent mutations for the duration of the "pinning" (https://github.com/carbon-language/carbon-lang/pull/2927)
* Avoid unnecessary copies from value expression to value expression, after ensuring that even value expression temporaries are registered for destruction.
* Avoid unnecessary copies when binding function arguments
The crash occurred because of an alias target of a mixin. While
eventually, we should probably have some ability to alias mixins, this
isn't yet setup in the explorer and doesn't seem like a current
priority.
We got here because the mixin type checking made it far enough to not
reject this within the type checker, but the next step wasn't prepared
for this to come out of the type checker. The simplest fix seems to be
to reflect that it *can* escape the type checker, but still isn't (yet)
a valid alias target.
Test case added.
There are a few implicit conversions that are implemented by an `impl` of `ImplicitAs` that delegates to code in explorer:
- Converting between tuple types
- Converting from tuples of types to `type`
- Converting from tuples of values to an array type
- Converting between struct types
- Converting from a struct type to a class type
These conversions can all rely on performing more conversions for elements or subobjects, but previously those inner conversions could only be performed if they were built into explorer. This change instead uses the full implicit conversion machinery in explorer to perform these conversions, including searching for a user-defined `impl` of `ImplicitAs` when necessary.
For example, this permits a conversion from `{.a: T}` to `{.a: U}`, or from `(T, T)` to `(U, U)`, or from `(T, T)` to `[U; 2]` when there is a user-defined conversion from `T` to `U`.
Depends on #2878
When a template has an argument that involves a generic parameter, we're supposed to delay instantiation until we know the concrete value, but explorer is not set up to do that yet, so for now we instead instantiate the template with the symbolic argument. When that happens, bring the constraints on the generic parameter into scope so they can be used inside the template instantiation.
This requires adding a new search over a value for the generic parameters that appear within it; a `VisitNestedValues` visitor is added to visit all the `Value`s nested with a value, and also convert an existing place where we were doing the same thing in a way that was incorrect (but harmlessly incorrect for now) to use it.
This is needed by #2881, which needs implementations of `ImplicitAs` for nested types when instantiating a builtin impl of `ImplicitAs` for an aggregate type.
Co-authored-by: Geoff Romer <gromer@google.com>
Fix misidentification of a potential cycle in the case where the inner match is missing labels from the outer match, and the inner match is strictly more complex when considering only its labels. We previously ignored labels in the outer match that are absent in the inner match, but the existence of any such label should cause us to treat the inner match as not being strictly more complex.
Per #257, we should be treating unformedness as all-or-nothing, rather than being a per-field or per-array-element property. Previously we initialized an array with no explicit initializer as containing a sequence of uninitialized values, but that led to crashes when attempting to access those values, as the checks for reading an uninitialized value only expected values to be uninitialized at the top level.
Also, we had existing tests that attempt to store to an element of an uninitialized array. We now detect that and treat it as UB during evaluation, rather than crashing due to trying to perform field access into an uninitialized value.
Finally, many of these problems can be detected statically, but the resolve_unformed pass wasn't catching them because it missed a few expression and declaration forms. Support for those cases has been added too. This causes the pass to recurse more often, and in particular our existing recursion test started hitting a stack overflow after this, so resolve_unformed now uses `RunWithExtraStack`. In passing, remove the need to explicitly tell `RunWithExtraStack` the return type, and infer it as the return type of the callable instead.
The intent of this change is that instead of paths looking like `explorer/testdata/foo/bar.carbon` (repo-relative), they're now just `bar.carbon` (local). The consequence is that paths should be a bit more durable in various environments, and just paths should be shorter and easier to read.
The explorer's prelude is an exception to this since it comes from data, rather than being the test target. Due to the change in approaches, it needs the regex again.
Uses #2829
Isolating test execution time (no build time included):
- Linux, `lit` test-per-file: Elapsed time: 28.214s, Critical Path: 13.97s
- Linux, `cc_test`-per-file: Elapsed time: 11.534s, Critical Path: 6.05s
- Linux, merged `cc_test` with 50 shards: Elapsed time: 11.677s, Critical Path: 11.17s
- Mac, `lit` test-per-file: Elapsed time: 295.686s, Critical Path: 20.00s
- Mac, `cc_test`-per-file: Elapsed time: 55.788s, Critical Path: 3.81s
- Mac, merged `cc_test` with 50 shards: Elapsed time: 16.269s, Critical Path: 7.54s
In GH actions:
- [Before](https://github.com/carbon-language/carbon-lang/actions/runs/4866602695/jobs/8678306144?pr=2799):
- test / test (ubuntu-22.04, fastbuild) (pull_request_target) Successful in 20m
- test / test (ubuntu-22.04, opt) (pull_request_target) Successful in 15m
- test / test (macos-12, fastbuild) (pull_request_target) Successful in 36m
- test / test (macos-12, opt) (pull_request_target) Successful in 21m
- [After](https://github.com/carbon-language/carbon-lang/actions/runs/4875154751/jobs/8697004066?pr=2811):
- test / test (ubuntu-22.04, fastbuild) (pull_request_target) Successful in 10m
- test / test (ubuntu-22.04, opt) (pull_request_target) Successful in 9m
- test / test (macos-12, fastbuild) (pull_request_target) Successful in 12m
- test / test (macos-12, opt) (pull_request_target) Successful in 9m
I'm still leaving a handful of `lit` tests to test end-to-end binary execution. This is why testdata directories are split (`lit` tests next to the `explorer` binary, the `cc_test`s next to `ParseAndExecute`).
This is really part of #2811, but is extracted out to allow a little review in parallelism because #2811 expects #2813. Getting this in will allow migration of toolchain tests, whereas #2811 is focused on explorer tests. For explorer test timing information, see #2811.
The syntax being used for matching deliberately mirrors the `FileCheck` setup, partly for compatibility if something changes, partly so there's nothing new to learn, partly so that we don't need to build more test updating.
Individual tests look like:
```
[ RUN ] ParseAndExecuteTestFile.explorer/parse_and_execute/testdata/assert/convert.carbon
To test this file alone, run:
bazel test //explorer/parse_and_execute:file_test.subset --test_arg=explorer/parse_and_execute/testdata/assert/convert.carbon
[ OK ] ParseAndExecuteTestFile.explorer/parse_and_execute/testdata/assert/convert.carbon (202 ms)
```
The printed command line is intended to assist developers in debugging a single test, particularly when sharding the main test. The use of a single `.subset` target means the total number of targets is constant even as the number of test files increases, which may be important for some `bazel` execution environments. I plan to make similar changes to the `glob_sh_run` implementation so that we have consistent setups, i.e. that we no longer create target-per-file scaling risks.
This uses `native_test` to share the test binary, avoiding re-linking if files are individually run.
Investigation did reveal a mistake where STDOUT/STDERR wasn't prefixed on empty output lines; this PR fixes that mistake, so that output is fully covered.
Co-authored-by: Chandler Carruth <chandlerc@gmail.com>
I couldn't figure out a way to actually hit a reasonable out-of-memory case once I add the maximum interpreter step count. However, the step count limit seems more important.
I've moved the todo stack limit out of function calls because there are plenty of ways to build up the todo stack without any function calls.
Fixes#2791
adds trailing comma to single element tuple
example:
```carbon
package ExplorerTest api;
fn Main() -> i32
{
var a: (i32,) = (1, 2);
return 0;
}
```
change in error message:
```diff
-type error in initializer of variable: '(i32, i32)' is not implicitly convertible to '(i32)'
+type error in initializer of variable: '(i32, i32)' is not implicitly convertible to '(i32,)'
```
This also clusters the various `var` tests. This tests similar patterns for a few syntaxes, but `fn f(x: i32, i32) {}` and local variables `var (x: i32, i32);` and `var i32;` were the crashers I found.
In `fn`, the `x: i32,` is helpful to convince the parser that this is valid syntax.
Fixes#2778
Removes `__continuation`, `__await`, and `__run`.
In part here, the discussion was that while the feature had been useful for validating the early explorer design, it's no longer needed for that role as the explorer is now quite robust. Continuations have been experimental and, at this point, don't have an owner pushing to a proposal.
The triggering factor is that, as we push to address fuzzer issues, I ran into a crash bug in this code; basically, `fn Main() -> i32 { __await; return 0; }`. When I mentioned this, the reaction seemed to trend towards removal of the feature.
This is adding more validation of intrinsics. It addresses a bug in rand where CHECK-fails would occur for bad range inputs, instead of a runtime error. I'm also addressing what I think was an int32 range issue in the handling by running the generator with int64.
This case previously crashed. Trying to handle allow_implicit_conversion better per discussion; I've split out the for loop for each path because I think it's now clearer this way.
Move concrete type check for Variable Definition (Local variable) from interpreter to type checker using the `ExpectConcreteType` function created in #2748
Rename value categories to expression categories based on [Discord discussion](https://discord.com/channels/655572317891461132/753021843459538996/1092924035517665332) regarding naming and behavior.
>* let expression -> value expression
>* var expression -> reference expression
>* located expression -> initializing expression
>So:
>- "value expressions" produce values (with no associated location). "reference expressions" produce a location of an existing value. "initializing expressions" take a location and initialize it.
>- A let binding is initialized by a value expression, because lets represent values (with category conversions performed as needed, but if a conversion is performed from a different category of expression, the value of the object is pinned for the lifetime of the let).
>- A var binding is initialized by an initializing expression, without performing a copy (with category conversions performed as needed, calling a copy constructor if the initializer is a different expression category).
>- The & operator requires a reference expression, and it's an error to give it other kinds.
>- The left-hand side of . requires a value expression when calling a function with a non-addr receiver, and requires a reference expression when calling a function with an addr receiver (it's an error to give it a value expression, and for an initializing expression, a temporary is materialized).
Changes
* Rename "value category" to "expression category"
* Rename Var and Let value categories to Value, Reference, and Initializing expression
* Rename `lvalue` to `location` (most of the time)
This PR is making two main changes to the Explorer and Toolchain:
- Replace the `is` keyword in `where SomeType is SomeInterface` with `impls`, so it is `where SomeType impls SomeInterface`
- Rewrite uses of the "impls" to something else to avoid, frequently "`impl` declarations" or "implementations", to avoid confusion with the `impls` keyword.
---------
Co-authored-by: Geoff Romer <gromer@google.com>
The strategy that we use for now to support template instantiation is to check the impl declaration as if it were a generic, but to defer all checking of the impl definition until we see a use in which all template parameters have arguments. At that point, we clone the impl definition and type-check the whole thing, with constant values set on the template parameters corresponding to the given arguments.
No caching of template instantiations is performed yet; each time we form a reference to a template instantiation, we instantiate it afresh. We also don't implement the name lookup rule from #949 yet; lookups during template instantiation look only in the actual type and not in the constraint.
Depends on #2699
### Features
* Add virtual `destructor`s support (Closes#2521)
* Check virtual override for virtual destructors
* Error if attempting to `Delete` a class that does not have virtual destructors from a base class pointer
### Implementation
* Update parser to support virtual override introducers for destructors
* Check virtual override for class destructor and add to class vtable if necessary
* Add corresponding tests
### Notes
Contrary to initial implementation, this implementation leverages the `Address` structure and implements a new `Address::DowncastedAddress()` method to get address from child most class from a base class address. This avoids the need to use `GetAllocationId` and its issues when it comes to having multiple values for an `AllocationId`.
### Next work
Following this PR, we need to:
* Check when using `Delete` that the class was allocated with `New` (WIP)
* Drop the old `GetAllocationId(Value*)` in favor of a better system (WIP)
As described in [the generics design](https://github.com/carbon-language/carbon-lang/blob/trunk/docs/design/generics/details.md#type-structure-of-an-impl-declaration), `impl` declarations are prioritized by type structure. Given two `impl` declarations that match a `type as interface` query, the one that describes the longest prefix of the query without using placeholders is preferred.
We implement this by putting all impls in a total order, first by type structure equivalence classes and then by lexical order. When matching an impl, we walk this total order, and stop once we find a match and reach the end of its equivalence class.
Equivalence classes are determined by finding the locations of the "holes" (the positions where deduced parameters appear) within the type structure, viewed as a tree. Two impls are in the same equivalence class if their holes are in the same place, and equivalence classes are ordered based on a reverse lexicographical ordering of their holes.
Explorer doesn't keep the `Bindings` list for a parameterized type in any particular order, but the type structure rule requires that we consider them in lexical order. In order to support this, we now track an index on the declared parameters of each generic. This is a simple numbering of enclosing generic parameters, both on that generic and on all lexically enclosing generics.
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>