Conversation
When working on rpit we needed to change to use a monomorphized clone of the result of function calls. This ended up causing a verify gimple issue with tuples because: fn test<A, B>(a: A, b: B) -> (A, B) When passing for example: let a = test::<i32, i32> (123, 456) -> (A=i32, B=i32) The resulting gimple types became: const struct (A=i32, B=i32) vs struct (i32, i32) We removed the VIEW_CONVERT_EXPR support to auto fix this stuff a good while ago because it hides these kinds of issues because the type hasher included the A=i32, B=i32 vs the i32, i32 name so updating this to use get_name instead keeps the naming the same as well as the fields meaning these types are 100% equivilant and therefore no conversion is required. This only occurs because tuples are not named types we should really add more rust specific info on our gimple TYPES. gcc/rust/ChangeLog: * backend/rust-compile-type.cc (TyTyResolveCompile::visit): use get_name * typecheck/rust-tyty.cc (TupleType::get_name): likewise Signed-off-by: Philip Herron <herron.philip@googlemail.com>
This is the initial patch for RPIT, we can build on this to handle the
more complex cases but there are enough distinct changes going on here
that it should just get merged now.
RPIT is really a sneaky generic so for example:
fn foo() -> impl Bar {
Baz
}
This is represented as: fn () -> OpaqueType Bar. But when we handle the
coercion site for Baz on impl Bar when we type resolve the function we
know that the underlying type is Baz. Note this function is _not_ generic
so its using this special OpaqueType and keeping track of the underlying type
in its ty_ref reference hir-id which will resolve to Baz.
This also means if we have a case where maybe this was in an if statement:
fn foo(a: i32) -> impl Bar {
if a > 10 {
Baz
} else {
Qux
}
}
The rules of impl Bar is that Baz is handled but Baz and Qux are different
underlying types so this is not allowed. The reason is impl traits are not
generic and although from a programmer perspective the callers dont know what
the underlying type is, the compiler _knows_ what it is. So really when
you call a function and get its return position impl trait the compiler knows
what to do and does all whats nessecary to handle calling functions using that
type etc.
gcc/rust/ChangeLog:
* backend/rust-compile-type.cc (TyTyResolveCompile::visit): we need to resolve the
underlying type
* typecheck/rust-substitution-mapper.cc (SubstMapperInternal::visit): just clone
* typecheck/rust-tyty-call.cc (TypeCheckCallExpr::visit):
ensure we monomphize to get the underlying
* typecheck/rust-tyty.cc (BaseType::destructure): handle opaque types
(OpaqueType::resolve): this is much simpler now
(OpaqueType::handle_substitions): no longer needed
* typecheck/rust-tyty.h: update header
* typecheck/rust-unify.cc (UnifyRules::expect_opaque): unify rules for opaque
gcc/testsuite/ChangeLog:
* rust/compile/bad-rpit1.rs: New test.
* rust/execute/torture/impl_rpit1.rs: New test.
* rust/execute/torture/impl_rpit2.rs: New test.
* rust/execute/torture/impl_rpit3.rs: New test.
Signed-off-by: Philip Herron <herron.philip@googlemail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This is the initial patch for RPIT, we can build on this to handle the more complex cases but there are enough distinct changes going on here that it should just get merged now.
RPIT is really a sneaky generic so for example:
This is represented as: fn () -> OpaqueType Bar. But when we handle the coercion site for Baz on impl Bar when we type resolve the function we know that the underlying type is Baz. Note this function is not generic so its using this special OpaqueType and keeping track of the underlying type in its ty_ref reference hir-id which will resolve to Baz.
This also means if we have a case where maybe this was in an if statement:
The rules of impl Bar is that Baz is handled but Baz and Qux are different underlying types so this is not allowed. The reason is impl traits are not generic and although from a programmer perspective the callers dont know what the underlying type is, the compiler knows what it is. So really when you call a function and get its return position impl trait the compiler knows what to do and does all whats nessecary to handle calling functions using that type etc.
gcc/rust/ChangeLog:
gcc/testsuite/ChangeLog: