gccrs: expand: Add stub function for attribute expansion
Add a stub function and utility functions that should be called on some
nodes to expand attribute procedural macros.
gcc/rust/ChangeLog:
* expand/rust-expand-visitor.cc (ExpandVisitor::expand_outer_attribute):
Stub for a single attribute expansion.
(ExpandVisitor::visit_outer_attrs): Visit the attributes to
expand on a given item.
* expand/rust-expand-visitor.h: Add function prototypes.
gccrs: ast: Add outer attribute getter to Expr class
We need to retrieve outer attributes from some Expressions depending on
their context. This means this should be retrieved from their parent
node. But expr did not have a getter for outer attributes since some
expr can't have any outer attribute.
gcc/rust/ChangeLog:
* ast/rust-ast.h: Add getter to Expr class.
* ast/rust-expr.h: Add override attribute to existing getters.
Also implement it for RangeExpr, attempting to retrieve outer
attributes on those types will crash the compiler.
* ast/rust-macro.h: Add override attribute to existing getters.
* ast/rust-path.h: Likewise.
Add function stubs for inner attribute proc macros expansion.
gcc/rust/ChangeLog:
* expand/rust-expand-visitor.cc (ExpandVisitor::visit): Add call
to inner attribute stub.
(ExpandVisitor::expand_inner_attribute): Expand one single
attribute.
(ExpandVisitor::visit_inner_using_attrs): Expand a given item
using a vector of attributes.
(ExpandVisitor::visit_inner_attrs): Visit inner attributes of a
given item.
* expand/rust-expand-visitor.h: Add function prototypes.
* expand/rust-expand-visitor.cc (ExpandVisitor::visit): Add call
to derive expander.
(ExpandVisitor::expand_derive): Expand a single derive.
(ExpandVisitor::visit_attrs_with_derive): Visit an item with
derive attributes.
(ExpandVisitor::is_derive): Identify a derive attribute.
* expand/rust-expand-visitor.h: Add function prototypes.
Attributes were not converted to a correct tokenstream, furthermore meta
items containers and token trees attribute arguments were not even
implemented. This commit fix the literal attribute type and implement
both unimplemented types.
gcc/rust/ChangeLog:
* ast/rust-ast-tokenstream.cc (TokenStream::visit): Fix existing
and implement remaining attribute visitors.
Even though the comma may be optional with blocks of expressions, it is
mandatory for a single expression alone. By changing this behavior we
ensure a working behavior with any kind of expressions.
gcc/rust/ChangeLog:
* ast/rust-ast-tokenstream.cc (TokenStream::visit): Make comma
mandatory.
The visitor created some additional semicolon between statements in
extern blocks, this lead to empty statement.
gcc/rust/ChangeLog:
* ast/rust-ast-tokenstream.cc (TokenStream::visit): Remove
additional semicolon output.
* ast/rust-ast-tokenstream.h: Change block visitor prototype
with a default value for trailing tokens.
Philip Herron [Sun, 2 Apr 2023 20:59:50 +0000 (21:59 +0100)]
gccrs: Fix ICE using constructors for intilizers in statics
We are getting constant expressions for the initilizers for static items
this hits an assertion in the GCC middle-end which is looking for a
constructor so we need to unwrap the constant expression using DECL_INITIAL
as the initilizer to the global static.
Fixes #2080
gcc/rust/ChangeLog:
* backend/rust-compile-item.cc (CompileItem::visit): unwrap the constant expression
gcc/testsuite/ChangeLog:
* rust/execute/torture/issue-2080.rs: New test.
Signed-off-by: Philip Herron <herron.philip@googlemail.com>
In rust cast to char is allowed only from u8 type.
This patch handles fix the case when the type is infered from an integer value,
allowing only the u8 case'
Fixes #2027
gcc/rust/ChangeLog:
* typecheck/rust-casts.cc (TypeCastRules::cast_rules):
case INTEGRAL handles TypeKind::CHAR
Philip Herron [Wed, 29 Mar 2023 12:35:53 +0000 (13:35 +0100)]
gccrs: Fix handling of generic arguments for lifetimes only
We might have generics that are only for lifetimes this is a bad error
check. We can simply rely on the function to marshall the HIR generics
arguements for the type system to do the error handling for us which will
be more acurate anyway.
Fixes #2043 #2039
gcc/rust/ChangeLog:
* resolve/rust-ast-resolve-item.cc (ResolveTraitItems::visit):
add check for reference marker and type's in self params
(ResolveItem::visit): likewise
* typecheck/rust-hir-type-check-path.cc (TypeCheckExpr::resolve_root_path):
remove bad generics check
(TypeCheckExpr::resolve_segments): likewise
* typecheck/rust-hir-type-check-type.cc (TypeCheckType::resolve_root_path): likewise
gcc/testsuite/ChangeLog:
* rust/compile/const_generics_5.rs: update test case
* rust/compile/issue-2043.rs: New test.
* rust/compile/issue-2039.rs: New test.
Signed-off-by: Philip Herron <herron.philip@googlemail.com>
Philip Herron [Wed, 29 Mar 2023 13:24:12 +0000 (14:24 +0100)]
gccrs: add missing name resolution to self params with specified types
In rust it is possible to write method where you explicitly specify the
type of a self param. It does not however allow for you to use reference
destructuring in this senario.
gcc/rust/ChangeLog:
* resolve/rust-ast-resolve-item.cc
(ResolveTraitItems::visit): add name resolution self param
(ResolveItem::visit): likewise
Signed-off-by: Philip Herron <herron.philip@googlemail.com>
Philip Herron [Wed, 29 Mar 2023 14:00:39 +0000 (15:00 +0100)]
gccrs: make distinction between error and empty for substution args
When handling generics of only lifetimes we have an empty
SubstitutionArgumentMappings which was being detected as an error but this
is not the case as we don't currently handle const generics or generic
lifetimes so this makes the distinction between an error occurred or its
simply empty because it was all only lifetime arguments and we don't care
about it.
Addresses #2043 #2039
gcc/rust/ChangeLog:
* typecheck/rust-tyty-bounds.cc (TypeBoundPredicate::TypeBoundPredicate): this is an empty
(TypeBoundPredicate::operator=): likewise
* typecheck/rust-tyty-subst.cc (SubstitutionArgumentMappings::empty): new interface
(SubstitutionArgumentMappings::is_error): this marks it as an error
* typecheck/rust-tyty-subst.h: update prototypes
Signed-off-by: Philip Herron <herron.philip@googlemail.com>
Add an option to the command line interface to dump the ast back to a
tokenstream.
gcc/rust/ChangeLog:
* rust-session-manager.cc (Session::enable_dump): Enable new
dump.
(Session::compile_crate): CLI argument parsing.
(Session::dump_tokenstream): Dump the tokenstream as a string in
the specified file.
* rust-session-manager.h (struct CompileOptions): Add
tokenstream dump option.
Add an option to output trailing commas depending on the configuration
of the TokenStream.
gcc/rust/ChangeLog:
* ast/rust-ast-tokenstream.cc (TokenStream::trailing_comma):
Output a trailing comma to the token stream according to the
configuration.
* ast/rust-ast-tokenstream.h: Add function prototype.
Add several functions to convert an ast back to a token stream. This may
be used later either to unify the ast dumping and modifications on ast
nodes in procedural macros.
gcc/rust/ChangeLog:
* Make-lang.in: Add rust-as-tokenstream to compile list.
* ast/rust-item.h: Add missing getter for location.
* ast/rust-ast-tokenstream.cc: Add ast visitor implementation.
* ast/rust-ast-tokenstream.h: New file.
* ast/rust-pattern.h: Add getter.
Philip Herron [Wed, 29 Mar 2023 15:28:11 +0000 (16:28 +0100)]
gccrs: handle bare function types with no specified return type
When we have a function with no return type this defaults to (), but we
need to be sure we generate a new implicit HirId for it otherwise it will
end up in a recursive reference chain.
Fixes #2042
gcc/rust/ChangeLog:
* typecheck/rust-hir-type-check-type.cc (TypeCheckType::visit):
Add implicit unit type as the return type when not specified
gcc/testsuite/ChangeLog:
* rust/compile/issue-2042.rs: New test.
Signed-off-by: Philip Herron <herron.philip@googlemail.com>
Philip Herron [Wed, 29 Mar 2023 14:57:56 +0000 (15:57 +0100)]
gccrs: fix ICE with bad match arm type
We expect to get some kind of ADT or Tuple type when computing this kind of
match arms this adds a new diagnostic to check for this case instead of
an assertion.
Fixes #2029
gcc/rust/ChangeLog:
* typecheck/rust-hir-type-check-expr.cc (TypeCheckExpr::visit): return early on bad type
* typecheck/rust-hir-type-check-pattern.cc (TypeCheckPattern::TypeCheckPattern):
remove assertion in favor of check
(TypeCheckPattern::visit): likewise
gcc/testsuite/ChangeLog:
* rust/compile/issue-2029.rs: New test.
Signed-off-by: Philip Herron <herron.philip@googlemail.com>
* libproc_macro/rust/bridge.rs: New file.
* libproc_macro/rust/bridge/group.rs: New file.
* libproc_macro/rust/bridge/ident.rs: New file.
* libproc_macro/rust/bridge/literal.rs: New file.
* libproc_macro/rust/bridge/punct.rs: New file.
* libproc_macro/rust/bridge/span.rs: New file.
* libproc_macro/rust/bridge/token_stream.rs: New file.
* libproc_macro/rust/group.rs: New file.
* libproc_macro/rust/ident.rs: New file.
* libproc_macro/rust/lib.rs: New file.
* libproc_macro/rust/literal.rs: New file.
* libproc_macro/rust/punct.rs: New file.
* libproc_macro/rust/span.rs: New file.
* libproc_macro/rust/token_stream.rs: New file.
Arthur Cohen [Tue, 28 Mar 2023 12:22:21 +0000 (14:22 +0200)]
gccrs: privacy: Check for pub(crate) when resolving visibility path.
This causes the function to return true and the checks to pass, but it
requires more thinking - how do we deal with pub(crate) in the current system?
Should we simply treat it as a pub item in the current crate, but export it as
a private item in the metadata?
gcc/rust/ChangeLog:
* ast/rust-item.h: Fix `Visibility::has_path()` implementation.
* checks/errors/privacy/rust-visibility-resolver.cc
(VisibilityResolver::resolve_module_path): Check if we are dealing with pub(crate) properly.
Arthur Cohen [Tue, 28 Mar 2023 11:55:19 +0000 (13:55 +0200)]
gccrs: resolver: Allow SimplePath to resolve to their root segment
SimplePath are allowed to resolve to their root segment even if it is `crate`
or `super` when checking visibilities. We now have to make sure that this
is allowed everywhere a SimplePath is kept.
gcc/rust/ChangeLog:
* resolve/rust-ast-resolve-path.cc (ResolvePath::resolve_path): Set
resolved_node_id when visiting root segment of SimplePath.
Arthur Cohen [Tue, 28 Mar 2023 09:34:27 +0000 (11:34 +0200)]
gccrs: ast: Fix location for pub(crate)
When creating a SimplePath from a string, we end up creating a path
which contains only one segment but no location. Take the location of
the one segment instead and use it as the simple path's location.
gcc/rust/ChangeLog:
* ast/rust-ast.h: Call SimplePath constructor differently in
SimplePath::from_str
Arthur Cohen [Tue, 21 Mar 2023 16:37:47 +0000 (17:37 +0100)]
gccrs: expand: Rename AttrVisitor -> CfgStrip
Since the expansion and stripping phase are now separated, it does not
make sense to keep AttrVisitor named AttrVisitor. Furthermore, the visitor
is already very complex, with a heavy mental load (erasing iterators,
rearranging them, performing cfg-expansion, etc) so further attribute
handling should probably happen in rust-attribute-checker.h
gcc/rust/ChangeLog:
* Make-lang.in: Rename rust-asttribute-visitor.o -> rust-cfg-strip.o
* expand/rust-attribute-visitor.cc: Moved to...
* expand/rust-cfg-strip.cc: ...here.
* expand/rust-attribute-visitor.h: Moved to...
* expand/rust-cfg-strip.h: ...here.
* expand/rust-macro-expand.cc: Fix include of rust-attribute-visitor.h
* expand/rust-macro-builtins.cc: Likewise.
* rust-session-manager.cc (Session::expansion): Call CfgStrip instead of
AttrVisitor.
Arthur Cohen [Mon, 20 Mar 2023 03:17:04 +0000 (04:17 +0100)]
gccrs: expand: Add new ExpandVisitor class
This class takes care of actually performing the macro expansion by calling
into the MacroExpander for each node of a given AST, leaving the job of
cfg-stripping nodes to the AttrVisitor.
gcc/rust/ChangeLog:
* Make-lang.in: Add new object file.
* expand/rust-attribute-visitor.cc (AttrVisitor::go): Visit all items of a
crate.
(AttrVisitor::expand_struct_fields): Do not perform macro expansion anymore.
(AttrVisitor::expand_function_params): Likewise.
(AttrVisitor::expand_generic_args): Likewise.
(AttrVisitor::expand_qualified_path_type): Likewise.
(AttrVisitor::expand_self_param): Likewise.
(AttrVisitor::expand_trait_function_decl): Likewise.
(AttrVisitor::expand_trait_method_decl): Likewise.
(AttrVisitor::visit): Likewise.
(AttrVisitor::maybe_expand_expr): Remove function.
(AttrVisitor::maybe_expand_type): Likewise.
* expand/rust-attribute-visitor.h: Do not keep MacroExpander inside AttrVisitor
anymore.
* expand/rust-macro-expand.h (struct MacroExpander): Turn ContextType into an
enum class for more type safety.
* expand/rust-macro-expand.cc (MacroExpander::expand_crate): Use new ContextType
API.
* rust-session-manager.cc (Session::expansion): Call into ExpandVisitor.
* expand/rust-expand-visitor.cc: New file.
* expand/rust-expand-visitor.h: New file.
gcc/testsuite/ChangeLog:
* rust/compile/macro49.rs: New test.
* rust/compile/macro50.rs: New test.
Arthur Cohen [Mon, 20 Mar 2023 02:10:26 +0000 (03:10 +0100)]
gccrs: expand: Move cfg-attrs related functions out of MacroExpander
The current situation where the MacroExpander and AttrVisitor recursively
call into each other is not great, and causes pains for macro imports. We
need to split this pass in two, with one pass being responsible for cfg-attr
checking and stripping, and one being responsible for macro expansion.
gcc/rust/ChangeLog:
* expand/rust-macro-expand.cc (MacroExpander::expand_crate): Do not cfg-attr
strip in MacroExpander
(MacroExpander::fails_cfg): Function moved...
(MacroExpander::fails_cfg_with_expand): Function moved...
(MacroExpander::expand_cfg_attrs): Function moved...
* expand/rust-attribute-visitor.cc (fails_cfg): ...here.
(fails_cfg_with_expand): ...here.
(expand_cfg_attrs): ...here.
(AttrVisitor::expand_struct_fields): Use new functions.
(AttrVisitor::expand_tuple_fields): Likewise.
(AttrVisitor::expand_function_params): Likewise.
(AttrVisitor::visit): Likewise.
(AttrVisitor::go): New function.
* expand/rust-attribute-visitor.h: Declare AttrVisitor::go.
* expand/rust-macro-expand.h (struct MacroExpander): Remove cfg-attr related
functions.
Arthur Cohen [Wed, 15 Mar 2023 16:30:25 +0000 (17:30 +0100)]
gccrs: mappings: Keep exported macro IDs
gcc/rust/ChangeLog:
* hir/rust-ast-lower-item.cc (ASTLoweringItem::visit): Add new visitor
for AST::MacroRulesDefinition.
* hir/rust-ast-lower-item.h: Declare above mentioned visitor.
* metadata/rust-export-metadata.cc (ExportContext::emit_macro): New function.
* metadata/rust-export-metadata.h: Declare it.
(PublicInterface::gather_export_data): Go through each exported macro.
* util/rust-hir-map.cc (Mappings::insert_exported_macro): New function.
(Mappings::get_exported_macros): New function.
* util/rust-hir-map.h: Add new mappings for exported macros.
Arthur Cohen [Wed, 15 Mar 2023 16:15:39 +0000 (17:15 +0100)]
gccrs: Revert "hir: Add ExportedMacro node and handling."
This reverts commit a1f940d193c6cdb13483690a4f4a7d501ad7040e. It is easier
and cleaner to store exported macros' NodeIds into our mappings rather than
create a new HIR kind of node.
Philip Herron [Fri, 24 Mar 2023 22:17:36 +0000 (22:17 +0000)]
gccrs: Fix ICE on parsing trait object missing dyn keyword
Trait objects are not required to use the 'dyn' keyword though it is
depreciated in later editions/version of Rustc. This patch handles the case
when we query the type for an HIR::Item which happens to be an HIR::Trait
and return a trait object or error.
Fixes #2037
gcc/rust/ChangeLog:
* typecheck/rust-hir-type-check-item.cc (TypeCheckItem::visit): return a TraitObject
gcc/testsuite/ChangeLog:
* rust/compile/issue-2037.rs: New test.
Signed-off-by: Philip Herron <herron.philip@googlemail.com>
Arthur Cohen [Mon, 20 Mar 2023 01:58:55 +0000 (02:58 +0100)]
gccrs: ast: Dump each MacroRule properly without the extra semicolon
Having the extra semicolon causes parsing errors when importing macros
exported using the AST dump, as no rule expects multiple tokens after a
single macro rule definition.
gcc/rust/ChangeLog:
* ast/rust-ast-dump.cc (Dump::visit): Remove extraneous semicolon when
dumping macro rules.
Philip Herron [Fri, 17 Mar 2023 22:46:37 +0000 (22:46 +0000)]
gccrs: support use declaration to write the type into the correct namespace
This builds upon the previous path resolution fixes so that it returns the
resolved_node_id or UNKNOWN_NODEID on failure to resolve the use-path.
It then exports the name to the current module namespace so it can be used.
Fixes #850 #855
gcc/rust/ChangeLog:
* ast/rust-ast.h: add const get_final_segment helper
* hir/rust-ast-lower-enumitem.h: dont add an item mapping for enumitems
* hir/rust-ast-lower-item.cc (ASTLoweringItem::visit): add enum to enum-items mappings
* hir/rust-ast-lower-stmt.cc (ASTLoweringStmt::visit): likewise
* hir/tree/rust-hir-item.h: add non const helper to get variants
* resolve/rust-ast-resolve-item.cc (ResolveItem::visit): resolve the use declaration
* resolve/rust-ast-resolve-path.cc (ResolvePath::resolve_path): handle self
* resolve/rust-ast-resolve-toplevel.h: add enum item mappings to module mappings
* typecheck/rust-hir-type-check-path.cc (TypeCheckExpr::resolve_root_path): ensure variant
(TypeCheckExpr::resolve_segments): likewise
* typecheck/rust-type-util.cc (query_type): lookup enum's
* util/rust-hir-map.cc (Mappings::insert_hir_enumitem): enum item mappings
(Mappings::lookup_hir_enumitem): likewise
* util/rust-hir-map.h: likewise
gcc/testsuite/ChangeLog:
* rust/compile/issue-850.rs: New test.
* rust/compile/issue-855.rs: New test.
Signed-off-by: Philip Herron <herron.philip@googlemail.com>
Philip Herron [Fri, 17 Mar 2023 22:46:20 +0000 (22:46 +0000)]
gccrs: Return resolved_node_id when possible
To enable use statement to export a name to a path such as the prelude
example: use option::Option::{self, None, Some} we need to get the resolved
node id of the self, None and Some in order to export the name of None and
Some to their respective NodeId.
Philip Herron [Fri, 17 Mar 2023 18:09:53 +0000 (18:09 +0000)]
gccrs: add new apply_primtiive_type_hint to inference variables
In the senario where you infer types via unify_site_and but choose to not
commit the result you can end up with coercion to infer the result later on
which does not get fully replaced resulting in a stray inference variable
that can be left alone as a general inference variable leading to missing
type context info. This patch gives support to add type hints to inference
variables so they can be defaulted correctly in more complex cases. The
old system relied on the unification result commiting and overriding the
inference variables so they dissapear out of the current typeing context.
This was needed to fix #1981 where it is valid to inject inference
variables here. This lead to a regression in a few of the complex generic
trait test cases such as execute/torture/traits9.rs which had the wrong
argument type and defaulted wrongly to i32 instead of isize.
gcc/rust/ChangeLog:
* typecheck/rust-hir-type-check-base.cc (TypeCheckBase::resolve_literal): fix ctor
* typecheck/rust-hir-type-check-stmt.cc (TypeCheckStmt::visit): likewise
* typecheck/rust-hir-type-check-type.cc (TypeCheckType::visit): likewise
* typecheck/rust-typecheck-context.cc (TypeCheckContext::push_new_loop_context): likewise
* typecheck/rust-tyty-util.cc (TyVar::get_implicit_infer_var): likewise
* typecheck/rust-tyty.cc (InferType::InferType): new ctor with type hint
(InferType::clone): fix ctor
(InferType::apply_primitive_type_hint): new function to apply possible hint
* typecheck/rust-tyty.h: update prototypes
* typecheck/rust-unify.cc (UnifyRules::expect_inference_variable): apply type hints
(UnifyRules::expect_bool): likewise
(UnifyRules::expect_char): likewise
(UnifyRules::expect_int): likewise
(UnifyRules::expect_uint): likewise
(UnifyRules::expect_float): likewise
(UnifyRules::expect_isize): likewise
(UnifyRules::expect_usize): likewise
Signed-off-by: Philip Herron <herron.philip@googlemail.com>
Mahmoud Mohamed [Wed, 15 Mar 2023 23:54:11 +0000 (02:54 +0300)]
gccrs: resolve: Add name resolution for AltPattern
The main changes in this commit can be logically split into two parts:
1) Pushing new pattern binding contexts in alt patterns. At the start
of each AltPattern we push an 'Or' context which represents the relation
between the bindings of different alts.
Before we resolve each alt arm we need to push a 'Product' context to represent
the relation between the bindings of this specific alt arm and each other.
This 'Product' context is removed after the alt arm visit and the its bindings
are pushed into the 'Or' context.
Eventually, the 'Or' context is removed as well after it fulfills its duty.
Its bindings are then pushed into the previous context similarly and so on.
2) Checking for consistent bindings between the alt arms which is handled
by check_bindings_consistency. The info necessary for this check is held
by binding_info_map inside PatternDeclaration class. The binding_info_map
only holds bindings for one alt arm at a time. After every alt arm visit,
these bindings info are pushed into a vec (that contains all alt arms info
and will eventually be passed to check_bindings_consistency) and emptied out
for the next alt arm visit. At the end, all the info from all the alt arms
are pushed again into binding_info in addition to the initial bindings that
were there before the alt arms visits, this way the binding_info will contain
all the binding_info present in the AltPattern (as it should).
In addition to that, some refactors were made (e.g. add_new_binding function)
and some errors emitted, no biggie.
gcc/rust/ChangeLog:
* resolve/rust-ast-resolve-pattern.cc (PatternDeclaration::go):
Print out consistency errors.
(PatternDeclaration::visit): Implement visit for AltPattern.
(PatternDeclaration::add_new_binding): New helper function for
adding a binding to identifier.
* resolve/rust-ast-resolve-pattern.h (struct BindingInfo):
New struct to facilitate checking for inconsistencies between bindings.
gcc/testsuite/ChangeLog:
* rust/compile/torture/alt_patterns1.rs: New test.
This is wrong because it turned into a double reference type becasuse this
bug was consistent bugs were not picked up but this is not correct. We now
reuse our type inference infrastructure to solve the parameters instead.
Philip Herron [Fri, 10 Mar 2023 18:13:01 +0000 (18:13 +0000)]
gccrs: tyty get rid of useless virtuals
This removes can_substitute and contains_type_parameters which were
confusing interfaces to act as a proxy to the SubstitionRef types. This
replaces them with a single base implementation which is much easier to
debug and follow.
projects / thirdparty / gcc.git / log
