diff options
Diffstat (limited to 'compiler/rustc_parse')
| -rw-r--r-- | compiler/rustc_parse/Cargo.toml | 9 | ||||
| -rw-r--r-- | compiler/rustc_parse/messages.ftl | 32 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/errors.rs | 143 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/lexer/mod.rs | 21 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/lib.rs | 33 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/attr.rs | 2 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/diagnostics.rs | 18 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/expr.rs | 24 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/generics.rs | 11 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/item.rs | 6 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/mod.rs | 96 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/path.rs | 4 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/token_type.rs | 6 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/ty.rs | 177 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/validate_attr.rs | 350 |
15 files changed, 334 insertions, 598 deletions
diff --git a/compiler/rustc_parse/Cargo.toml b/compiler/rustc_parse/Cargo.toml index 0ae0b613fa2..7cb4ae7ff5f 100644 --- a/compiler/rustc_parse/Cargo.toml +++ b/compiler/rustc_parse/Cargo.toml @@ -5,11 +5,10 @@ edition = "2024" [dependencies] # tidy-alphabetical-start -bitflags = "2.4.1" -rustc-literal-escaper = "0.0.5" +bitflags.workspace = true +rustc-literal-escaper.workspace = true rustc_ast = { path = "../rustc_ast" } rustc_ast_pretty = { path = "../rustc_ast_pretty" } -rustc_attr_parsing = { path = "../rustc_attr_parsing" } rustc_data_structures = { path = "../rustc_data_structures" } rustc_errors = { path = "../rustc_errors" } rustc_feature = { path = "../rustc_feature" } @@ -19,8 +18,8 @@ rustc_lexer = { path = "../rustc_lexer" } rustc_macros = { path = "../rustc_macros" } rustc_session = { path = "../rustc_session" } rustc_span = { path = "../rustc_span" } -thin-vec = "0.2.12" -tracing = "0.1" +thin-vec.workspace = true +tracing.workspace = true unicode-normalization = "0.1.11" unicode-width = "0.2.0" # tidy-alphabetical-end diff --git a/compiler/rustc_parse/messages.ftl b/compiler/rustc_parse/messages.ftl index 9e0075c21b9..4ca2f57bd87 100644 --- a/compiler/rustc_parse/messages.ftl +++ b/compiler/rustc_parse/messages.ftl @@ -359,6 +359,20 @@ parse_generics_in_path = unexpected generic arguments in path parse_help_set_edition_cargo = set `edition = "{$edition}"` in `Cargo.toml` parse_help_set_edition_standalone = pass `--edition {$edition}` to `rustc` + +parse_hidden_unicode_codepoints = unicode codepoint changing visible direction of text present in {$label} + .label = this {$label} contains {$count -> + [one] an invisible + *[other] invisible + } unicode text flow control {$count -> + [one] codepoint + *[other] codepoints + } + .note = these kind of unicode codepoints change the way text flows on applications that support them, but can cause confusion because they change the order of characters on the screen + .suggestion_remove = if their presence wasn't intentional, you can remove them + .suggestion_escape = if you want to keep them but make them visible in your source code, you can escape them + .no_suggestion_note_escape = if you want to keep them but make them visible in your source code, you can escape them: {$escaped} + parse_if_expression_missing_condition = missing condition for `if` expression .condition_label = expected condition here .block_label = if this block is the condition of the `if` expression, then it must be followed by another block @@ -436,11 +450,6 @@ parse_inner_doc_comment_not_permitted = expected outer doc comment .label_does_not_annotate_this = the inner doc comment doesn't annotate this {$item} .sugg_change_inner_to_outer = to annotate the {$item}, change the doc comment from inner to outer style -parse_invalid_attr_unsafe = `{$name}` is not an unsafe attribute - .label = this is not an unsafe attribute - .suggestion = remove the `unsafe(...)` - .note = extraneous unsafe is not allowed in attributes - parse_invalid_block_macro_segment = cannot use a `block` macro fragment here .label = the `block` fragment is within this context .suggestion = wrap this in another block @@ -468,9 +477,6 @@ parse_invalid_dyn_keyword = invalid `dyn` keyword parse_invalid_expression_in_let_else = a `{$operator}` expression cannot be directly assigned in `let...else` parse_invalid_identifier_with_leading_number = identifiers cannot start with a number -parse_invalid_label = - invalid label name `{$name}` - parse_invalid_literal_suffix_on_tuple_index = suffixes on a tuple index are invalid .label = invalid suffix `{$suffix}` .tuple_exception_line_1 = `{$suffix}` is *temporarily* accepted on tuple index fields as it was incorrectly accepted on stable for a few releases @@ -500,6 +506,8 @@ parse_invalid_unicode_escape = invalid unicode character escape parse_invalid_variable_declaration = invalid variable declaration +parse_keyword_label = labels cannot use keyword names + parse_keyword_lifetime = lifetimes cannot use keyword names @@ -601,7 +609,6 @@ parse_maybe_report_ambiguous_plus = ambiguous `+` in a type .suggestion = use parentheses to disambiguate -parse_meta_bad_delim = wrong meta list delimiters parse_meta_bad_delim_suggestion = the delimiters should be `(` and `)` parse_mismatched_closing_delimiter = mismatched closing delimiter: `{$delimiter}` @@ -748,9 +755,6 @@ parse_parentheses_with_struct_fields = invalid `struct` delimiters or `fn` call .suggestion_braces_for_struct = if `{$type}` is a struct, use braces as delimiters .suggestion_no_fields_for_fn = if `{$type}` is a function, use the arguments directly -parse_parenthesized_lifetime = parenthesized lifetime bounds are not supported -parse_parenthesized_lifetime_suggestion = remove the parentheses - parse_path_double_colon = path separator must be a double colon .suggestion = use a double colon instead @@ -993,10 +997,6 @@ parse_unmatched_angle_brackets = {$num_extra_brackets -> *[other] remove extra angle brackets } -parse_unsafe_attr_outside_unsafe = unsafe attribute used without unsafe - .label = usage of unsafe attribute -parse_unsafe_attr_outside_unsafe_suggestion = wrap the attribute in `unsafe(...)` - parse_unskipped_whitespace = whitespace symbol '{$ch}' is not skipped .label = {parse_unskipped_whitespace} diff --git a/compiler/rustc_parse/src/errors.rs b/compiler/rustc_parse/src/errors.rs index a105dd1909e..797d4830c2f 100644 --- a/compiler/rustc_parse/src/errors.rs +++ b/compiler/rustc_parse/src/errors.rs @@ -8,8 +8,9 @@ use rustc_ast::{Path, Visibility}; use rustc_errors::codes::*; use rustc_errors::{ Applicability, Diag, DiagCtxtHandle, Diagnostic, EmissionGuarantee, Level, Subdiagnostic, + SuggestionStyle, }; -use rustc_macros::{Diagnostic, Subdiagnostic}; +use rustc_macros::{Diagnostic, LintDiagnostic, Subdiagnostic}; use rustc_session::errors::ExprParenthesesNeeded; use rustc_span::edition::{Edition, LATEST_STABLE_EDITION}; use rustc_span::{Ident, Span, Symbol}; @@ -2228,11 +2229,10 @@ pub(crate) struct KeywordLifetime { } #[derive(Diagnostic)] -#[diag(parse_invalid_label)] -pub(crate) struct InvalidLabel { +#[diag(parse_keyword_label)] +pub(crate) struct KeywordLabel { #[primary_span] pub span: Span, - pub name: Symbol, } #[derive(Diagnostic)] @@ -3163,27 +3163,6 @@ pub(crate) struct ModifierLifetime { pub modifier: &'static str, } -#[derive(Subdiagnostic)] -#[multipart_suggestion( - parse_parenthesized_lifetime_suggestion, - applicability = "machine-applicable" -)] -pub(crate) struct RemoveParens { - #[suggestion_part(code = "")] - pub lo: Span, - #[suggestion_part(code = "")] - pub hi: Span, -} - -#[derive(Diagnostic)] -#[diag(parse_parenthesized_lifetime)] -pub(crate) struct ParenthesizedLifetime { - #[primary_span] - pub span: Span, - #[subdiagnostic] - pub sugg: RemoveParens, -} - #[derive(Diagnostic)] #[diag(parse_underscore_literal_suffix)] pub(crate) struct UnderscoreLiteralSuffix { @@ -3367,15 +3346,6 @@ pub(crate) struct KwBadCase<'a> { } #[derive(Diagnostic)] -#[diag(parse_meta_bad_delim)] -pub(crate) struct MetaBadDelim { - #[primary_span] - pub span: Span, - #[subdiagnostic] - pub sugg: MetaBadDelimSugg, -} - -#[derive(Diagnostic)] #[diag(parse_cfg_attr_bad_delim)] pub(crate) struct CfgAttrBadDelim { #[primary_span] @@ -3515,38 +3485,6 @@ pub(crate) struct DotDotRangeAttribute { } #[derive(Diagnostic)] -#[diag(parse_invalid_attr_unsafe)] -#[note] -pub(crate) struct InvalidAttrUnsafe { - #[primary_span] - #[label] - pub span: Span, - pub name: Path, -} - -#[derive(Diagnostic)] -#[diag(parse_unsafe_attr_outside_unsafe)] -pub(crate) struct UnsafeAttrOutsideUnsafe { - #[primary_span] - #[label] - pub span: Span, - #[subdiagnostic] - pub suggestion: UnsafeAttrOutsideUnsafeSuggestion, -} - -#[derive(Subdiagnostic)] -#[multipart_suggestion( - parse_unsafe_attr_outside_unsafe_suggestion, - applicability = "machine-applicable" -)] -pub(crate) struct UnsafeAttrOutsideUnsafeSuggestion { - #[suggestion_part(code = "unsafe(")] - pub left: Span, - #[suggestion_part(code = ")")] - pub right: Span, -} - -#[derive(Diagnostic)] #[diag(parse_binder_before_modifiers)] pub(crate) struct BinderBeforeModifiers { #[primary_span] @@ -3664,3 +3602,76 @@ pub(crate) struct ExpectedRegisterClassOrExplicitRegister { #[primary_span] pub(crate) span: Span, } + +#[derive(LintDiagnostic)] +#[diag(parse_hidden_unicode_codepoints)] +#[note] +pub(crate) struct HiddenUnicodeCodepointsDiag { + pub label: String, + pub count: usize, + #[label] + pub span_label: Span, + #[subdiagnostic] + pub labels: Option<HiddenUnicodeCodepointsDiagLabels>, + #[subdiagnostic] + pub sub: HiddenUnicodeCodepointsDiagSub, +} + +pub(crate) struct HiddenUnicodeCodepointsDiagLabels { + pub spans: Vec<(char, Span)>, +} + +impl Subdiagnostic for HiddenUnicodeCodepointsDiagLabels { + fn add_to_diag<G: EmissionGuarantee>(self, diag: &mut Diag<'_, G>) { + for (c, span) in self.spans { + diag.span_label(span, format!("{c:?}")); + } + } +} + +pub(crate) enum HiddenUnicodeCodepointsDiagSub { + Escape { spans: Vec<(char, Span)> }, + NoEscape { spans: Vec<(char, Span)> }, +} + +// Used because of multiple multipart_suggestion and note +impl Subdiagnostic for HiddenUnicodeCodepointsDiagSub { + fn add_to_diag<G: EmissionGuarantee>(self, diag: &mut Diag<'_, G>) { + match self { + HiddenUnicodeCodepointsDiagSub::Escape { spans } => { + diag.multipart_suggestion_with_style( + fluent::parse_suggestion_remove, + spans.iter().map(|(_, span)| (*span, "".to_string())).collect(), + Applicability::MachineApplicable, + SuggestionStyle::HideCodeAlways, + ); + diag.multipart_suggestion( + fluent::parse_suggestion_escape, + spans + .into_iter() + .map(|(c, span)| { + let c = format!("{c:?}"); + (span, c[1..c.len() - 1].to_string()) + }) + .collect(), + Applicability::MachineApplicable, + ); + } + HiddenUnicodeCodepointsDiagSub::NoEscape { spans } => { + // FIXME: in other suggestions we've reversed the inner spans of doc comments. We + // should do the same here to provide the same good suggestions as we do for + // literals above. + diag.arg( + "escaped", + spans + .into_iter() + .map(|(c, _)| format!("{c:?}")) + .collect::<Vec<String>>() + .join(", "), + ); + diag.note(fluent::parse_suggestion_remove); + diag.note(fluent::parse_no_suggestion_note_escape); + } + } + } +} diff --git a/compiler/rustc_parse/src/lexer/mod.rs b/compiler/rustc_parse/src/lexer/mod.rs index 85af5a615ae..9792240a548 100644 --- a/compiler/rustc_parse/src/lexer/mod.rs +++ b/compiler/rustc_parse/src/lexer/mod.rs @@ -49,6 +49,7 @@ pub(crate) fn lex_token_trees<'psess, 'src>( mut src: &'src str, mut start_pos: BytePos, override_span: Option<Span>, + frontmatter_allowed: FrontmatterAllowed, ) -> Result<TokenStream, Vec<Diag<'psess>>> { // Skip `#!`, if present. if let Some(shebang_len) = rustc_lexer::strip_shebang(src) { @@ -56,7 +57,7 @@ pub(crate) fn lex_token_trees<'psess, 'src>( start_pos = start_pos + BytePos::from_usize(shebang_len); } - let cursor = Cursor::new(src, FrontmatterAllowed::Yes); + let cursor = Cursor::new(src, frontmatter_allowed); let mut lexer = Lexer { psess, start_pos, @@ -542,21 +543,21 @@ impl<'psess, 'src> Lexer<'psess, 'src> { }) .collect(); + let label = label.to_string(); let count = spans.len(); - let labels = point_at_inner_spans.then_some(spans.clone()); + let labels = point_at_inner_spans + .then_some(errors::HiddenUnicodeCodepointsDiagLabels { spans: spans.clone() }); + let sub = if point_at_inner_spans && !spans.is_empty() { + errors::HiddenUnicodeCodepointsDiagSub::Escape { spans } + } else { + errors::HiddenUnicodeCodepointsDiagSub::NoEscape { spans } + }; self.psess.buffer_lint( TEXT_DIRECTION_CODEPOINT_IN_LITERAL, span, ast::CRATE_NODE_ID, - BuiltinLintDiag::HiddenUnicodeCodepoints { - label: label.to_string(), - count, - span_label: span, - labels, - escape: point_at_inner_spans && !spans.is_empty(), - spans, - }, + errors::HiddenUnicodeCodepointsDiag { label, count, span_label: span, labels, sub }, ); } diff --git a/compiler/rustc_parse/src/lib.rs b/compiler/rustc_parse/src/lib.rs index 0dd5d89491b..197333d942d 100644 --- a/compiler/rustc_parse/src/lib.rs +++ b/compiler/rustc_parse/src/lib.rs @@ -17,10 +17,11 @@ use std::str::Utf8Error; use std::sync::Arc; use rustc_ast as ast; -use rustc_ast::tokenstream::TokenStream; +use rustc_ast::tokenstream::{DelimSpan, TokenStream}; use rustc_ast::{AttrItem, Attribute, MetaItemInner, token}; use rustc_ast_pretty::pprust; use rustc_errors::{Diag, EmissionGuarantee, FatalError, PResult, pluralize}; +use rustc_lexer::FrontmatterAllowed; use rustc_session::parse::ParseSess; use rustc_span::source_map::SourceMap; use rustc_span::{FileName, SourceFile, Span}; @@ -31,8 +32,9 @@ pub const MACRO_ARGUMENTS: Option<&str> = Some("macro arguments"); #[macro_use] pub mod parser; use parser::Parser; +use rustc_ast::token::Delimiter; + pub mod lexer; -pub mod validate_attr; mod errors; @@ -147,7 +149,7 @@ fn new_parser_from_source_file( source_file: Arc<SourceFile>, ) -> Result<Parser<'_>, Vec<Diag<'_>>> { let end_pos = source_file.end_position(); - let stream = source_file_to_stream(psess, source_file, None)?; + let stream = source_file_to_stream(psess, source_file, None, FrontmatterAllowed::Yes)?; let mut parser = Parser::new(psess, stream, None); if parser.token == token::Eof { parser.token.span = Span::new(end_pos, end_pos, parser.token.span.ctxt(), None); @@ -162,7 +164,9 @@ pub fn source_str_to_stream( override_span: Option<Span>, ) -> Result<TokenStream, Vec<Diag<'_>>> { let source_file = psess.source_map().new_source_file(name, source); - source_file_to_stream(psess, source_file, override_span) + // used mainly for `proc_macro` and the likes, not for our parsing purposes, so don't parse + // frontmatters as frontmatters. + source_file_to_stream(psess, source_file, override_span, FrontmatterAllowed::No) } /// Given a source file, produces a sequence of token trees. Returns any buffered errors from @@ -171,6 +175,7 @@ fn source_file_to_stream<'psess>( psess: &'psess ParseSess, source_file: Arc<SourceFile>, override_span: Option<Span>, + frontmatter_allowed: FrontmatterAllowed, ) -> Result<TokenStream, Vec<Diag<'psess>>> { let src = source_file.src.as_ref().unwrap_or_else(|| { psess.dcx().bug(format!( @@ -179,7 +184,13 @@ fn source_file_to_stream<'psess>( )); }); - lexer::lex_token_trees(psess, src.as_str(), source_file.start_pos, override_span) + lexer::lex_token_trees( + psess, + src.as_str(), + source_file.start_pos, + override_span, + frontmatter_allowed, + ) } /// Runs the given subparser `f` on the tokens of the given `attr`'s item. @@ -226,7 +237,7 @@ pub fn parse_cfg_attr( ast::AttrArgs::Delimited(ast::DelimArgs { dspan, delim, ref tokens }) if !tokens.is_empty() => { - crate::validate_attr::check_cfg_attr_bad_delim(psess, dspan, delim); + check_cfg_attr_bad_delim(psess, dspan, delim); match parse_in(psess, tokens.clone(), "`cfg_attr` input", |p| p.parse_cfg_attr()) { Ok(r) => return Some(r), Err(e) => { @@ -245,3 +256,13 @@ pub fn parse_cfg_attr( } None } + +fn check_cfg_attr_bad_delim(psess: &ParseSess, span: DelimSpan, delim: Delimiter) { + if let Delimiter::Parenthesis = delim { + return; + } + psess.dcx().emit_err(errors::CfgAttrBadDelim { + span: span.entire(), + sugg: errors::MetaBadDelimSugg { open: span.open, close: span.close }, + }); +} diff --git a/compiler/rustc_parse/src/parser/attr.rs b/compiler/rustc_parse/src/parser/attr.rs index 7f6afeba28c..acd338156ce 100644 --- a/compiler/rustc_parse/src/parser/attr.rs +++ b/compiler/rustc_parse/src/parser/attr.rs @@ -399,7 +399,7 @@ impl<'a> Parser<'a> { } /// Matches `COMMASEP(meta_item_inner)`. - pub(crate) fn parse_meta_seq_top(&mut self) -> PResult<'a, ThinVec<ast::MetaItemInner>> { + pub fn parse_meta_seq_top(&mut self) -> PResult<'a, ThinVec<ast::MetaItemInner>> { // Presumably, the majority of the time there will only be one attr. let mut nmis = ThinVec::with_capacity(1); while self.token != token::Eof { diff --git a/compiler/rustc_parse/src/parser/diagnostics.rs b/compiler/rustc_parse/src/parser/diagnostics.rs index 220e4ac18fc..a28af7833c3 100644 --- a/compiler/rustc_parse/src/parser/diagnostics.rs +++ b/compiler/rustc_parse/src/parser/diagnostics.rs @@ -463,8 +463,8 @@ impl<'a> Parser<'a> { pub(super) fn expected_one_of_not_found( &mut self, - edible: &[ExpTokenPair<'_>], - inedible: &[ExpTokenPair<'_>], + edible: &[ExpTokenPair], + inedible: &[ExpTokenPair], ) -> PResult<'a, ErrorGuaranteed> { debug!("expected_one_of_not_found(edible: {:?}, inedible: {:?})", edible, inedible); fn tokens_to_string(tokens: &[TokenType]) -> String { @@ -1092,7 +1092,7 @@ impl<'a> Parser<'a> { /// Eats and discards tokens until one of `closes` is encountered. Respects token trees, /// passes through any errors encountered. Used for error recovery. - pub(super) fn eat_to_tokens(&mut self, closes: &[ExpTokenPair<'_>]) { + pub(super) fn eat_to_tokens(&mut self, closes: &[ExpTokenPair]) { if let Err(err) = self .parse_seq_to_before_tokens(closes, &[], SeqSep::none(), |p| Ok(p.parse_token_tree())) { @@ -1113,7 +1113,7 @@ impl<'a> Parser<'a> { pub(super) fn check_trailing_angle_brackets( &mut self, segment: &PathSegment, - end: &[ExpTokenPair<'_>], + end: &[ExpTokenPair], ) -> Option<ErrorGuaranteed> { if !self.may_recover() { return None; @@ -1196,7 +1196,7 @@ impl<'a> Parser<'a> { // second case. if self.look_ahead(position, |t| { trace!("check_trailing_angle_brackets: t={:?}", t); - end.iter().any(|exp| exp.tok == &t.kind) + end.iter().any(|exp| exp.tok == t.kind) }) { // Eat from where we started until the end token so that parsing can continue // as if we didn't have those extra angle brackets. @@ -2120,8 +2120,8 @@ impl<'a> Parser<'a> { pub(super) fn recover_seq_parse_error( &mut self, - open: ExpTokenPair<'_>, - close: ExpTokenPair<'_>, + open: ExpTokenPair, + close: ExpTokenPair, lo: Span, err: Diag<'a>, ) -> Box<Expr> { @@ -2386,8 +2386,8 @@ impl<'a> Parser<'a> { pub(super) fn consume_block( &mut self, - open: ExpTokenPair<'_>, - close: ExpTokenPair<'_>, + open: ExpTokenPair, + close: ExpTokenPair, consume_close: ConsumeClosingDelim, ) { let mut brace_depth = 0; diff --git a/compiler/rustc_parse/src/parser/expr.rs b/compiler/rustc_parse/src/parser/expr.rs index d0604f76317..7de4f6efd0b 100644 --- a/compiler/rustc_parse/src/parser/expr.rs +++ b/compiler/rustc_parse/src/parser/expr.rs @@ -1598,7 +1598,7 @@ impl<'a> Parser<'a> { self.maybe_recover_from_bad_qpath(expr) } - fn parse_expr_array_or_repeat(&mut self, close: ExpTokenPair<'_>) -> PResult<'a, Box<Expr>> { + fn parse_expr_array_or_repeat(&mut self, close: ExpTokenPair) -> PResult<'a, Box<Expr>> { let lo = self.token.span; self.bump(); // `[` or other open delim @@ -2156,7 +2156,7 @@ impl<'a> Parser<'a> { /// Keep this in sync with `Token::can_begin_literal_maybe_minus` and /// `Lit::from_token` (excluding unary negation). - fn eat_token_lit(&mut self) -> Option<token::Lit> { + pub fn eat_token_lit(&mut self) -> Option<token::Lit> { let check_expr = |expr: Box<Expr>| { if let ast::ExprKind::Lit(token_lit) = expr.kind { Some(token_lit) @@ -2397,20 +2397,24 @@ impl<'a> Parser<'a> { let before = self.prev_token; let binder = if self.check_keyword(exp!(For)) { let lo = self.token.span; - let (lifetime_defs, _) = self.parse_late_bound_lifetime_defs()?; + let (bound_vars, _) = self.parse_higher_ranked_binder()?; let span = lo.to(self.prev_token.span); self.psess.gated_spans.gate(sym::closure_lifetime_binder, span); - ClosureBinder::For { span, generic_params: lifetime_defs } + ClosureBinder::For { span, generic_params: bound_vars } } else { ClosureBinder::NotPresent }; let constness = self.parse_closure_constness(); - let movability = - if self.eat_keyword(exp!(Static)) { Movability::Static } else { Movability::Movable }; + let movability = if self.eat_keyword(exp!(Static)) { + self.psess.gated_spans.gate(sym::coroutines, self.prev_token.span); + Movability::Static + } else { + Movability::Movable + }; let coroutine_kind = if self.token_uninterpolated_span().at_least_rust_2018() { self.parse_coroutine_kind(Case::Sensitive) @@ -3090,7 +3094,7 @@ impl<'a> Parser<'a> { if let Some((ident, is_raw)) = self.token.lifetime() { // Disallow `'fn`, but with a better error message than `expect_lifetime`. if matches!(is_raw, IdentIsRaw::No) && ident.without_first_quote().is_reserved() { - self.dcx().emit_err(errors::InvalidLabel { span: ident.span, name: ident.name }); + self.dcx().emit_err(errors::KeywordLabel { span: ident.span }); } self.bump(); @@ -3657,7 +3661,7 @@ impl<'a> Parser<'a> { &mut self, pth: ast::Path, recover: bool, - close: ExpTokenPair<'_>, + close: ExpTokenPair, ) -> PResult< 'a, ( @@ -3676,8 +3680,8 @@ impl<'a> Parser<'a> { errors::HelpUseLatestEdition::new().add_to_diag(e); }; - while self.token != *close.tok { - if self.eat(exp!(DotDot)) || self.recover_struct_field_dots(close.tok) { + while self.token != close.tok { + if self.eat(exp!(DotDot)) || self.recover_struct_field_dots(&close.tok) { let exp_span = self.prev_token.span; // We permit `.. }` on the left-hand side of a destructuring assignment. if self.check(close) { diff --git a/compiler/rustc_parse/src/parser/generics.rs b/compiler/rustc_parse/src/parser/generics.rs index 4a8530a2b38..eb684c3a62f 100644 --- a/compiler/rustc_parse/src/parser/generics.rs +++ b/compiler/rustc_parse/src/parser/generics.rs @@ -172,8 +172,11 @@ impl<'a> Parser<'a> { }) } - /// Parses a (possibly empty) list of lifetime and type parameters, possibly including - /// a trailing comma and erroneous trailing attributes. + /// Parse a (possibly empty) list of generic (lifetime, type, const) parameters. + /// + /// ```ebnf + /// GenericParams = (GenericParam ("," GenericParam)* ","?)? + /// ``` pub(super) fn parse_generic_params(&mut self) -> PResult<'a, ThinVec<ast::GenericParam>> { let mut params = ThinVec::new(); let mut done = false; @@ -520,7 +523,7 @@ impl<'a> Parser<'a> { // * `for<'a> Trait1<'a>: Trait2<'a /* ok */>` // * `(for<'a> Trait1<'a>): Trait2<'a /* not ok */>` // * `for<'a> for<'b> Trait1<'a, 'b>: Trait2<'a /* ok */, 'b /* not ok */>` - let (lifetime_defs, _) = self.parse_late_bound_lifetime_defs()?; + let (bound_vars, _) = self.parse_higher_ranked_binder()?; // Parse type with mandatory colon and (possibly empty) bounds, // or with mandatory equality sign and the second type. @@ -528,7 +531,7 @@ impl<'a> Parser<'a> { if self.eat(exp!(Colon)) { let bounds = self.parse_generic_bounds()?; Ok(ast::WherePredicateKind::BoundPredicate(ast::WhereBoundPredicate { - bound_generic_params: lifetime_defs, + bound_generic_params: bound_vars, bounded_ty: ty, bounds, })) diff --git a/compiler/rustc_parse/src/parser/item.rs b/compiler/rustc_parse/src/parser/item.rs index ca89eb1e2cf..eb264f59fed 100644 --- a/compiler/rustc_parse/src/parser/item.rs +++ b/compiler/rustc_parse/src/parser/item.rs @@ -43,7 +43,7 @@ impl<'a> Parser<'a> { self.expect(exp!(OpenBrace))?; let (inner_attrs, items, inner_span) = self.parse_mod(exp!(CloseBrace))?; attrs.extend(inner_attrs); - ModKind::Loaded(items, Inline::Yes, inner_span, Ok(())) + ModKind::Loaded(items, Inline::Yes, inner_span) }; Ok(ItemKind::Mod(safety, ident, mod_kind)) } @@ -54,7 +54,7 @@ impl<'a> Parser<'a> { /// - `}` for mod items pub fn parse_mod( &mut self, - term: ExpTokenPair<'_>, + term: ExpTokenPair, ) -> PResult<'a, (AttrVec, ThinVec<Box<Item>>, ModSpans)> { let lo = self.token.span; let attrs = self.parse_inner_attributes()?; @@ -1201,7 +1201,7 @@ impl<'a> Parser<'a> { }?; let dash = exp!(Minus); - if self.token != *dash.tok { + if self.token != dash.tok { return Ok(ident); } diff --git a/compiler/rustc_parse/src/parser/mod.rs b/compiler/rustc_parse/src/parser/mod.rs index 41ed1f95a01..15598f32429 100644 --- a/compiler/rustc_parse/src/parser/mod.rs +++ b/compiler/rustc_parse/src/parser/mod.rs @@ -24,7 +24,7 @@ pub use diagnostics::AttemptLocalParseRecovery; pub(crate) use expr::ForbiddenLetReason; pub(crate) use item::{FnContext, FnParseMode}; pub use pat::{CommaRecoveryMode, RecoverColon, RecoverComma}; -use path::PathStyle; +pub use path::PathStyle; use rustc_ast::token::{ self, IdentIsRaw, InvisibleOrigin, MetaVarKind, NtExprKind, NtPatKind, Token, TokenKind, }; @@ -261,19 +261,19 @@ struct CaptureState { /// A sequence separator. #[derive(Debug)] -struct SeqSep<'a> { +struct SeqSep { /// The separator token. - sep: Option<ExpTokenPair<'a>>, + sep: Option<ExpTokenPair>, /// `true` if a trailing separator is allowed. trailing_sep_allowed: bool, } -impl<'a> SeqSep<'a> { - fn trailing_allowed(sep: ExpTokenPair<'a>) -> SeqSep<'a> { +impl SeqSep { + fn trailing_allowed(sep: ExpTokenPair) -> SeqSep { SeqSep { sep: Some(sep), trailing_sep_allowed: true } } - fn none() -> SeqSep<'a> { + fn none() -> SeqSep { SeqSep { sep: None, trailing_sep_allowed: false } } } @@ -285,7 +285,7 @@ pub enum FollowedByType { } #[derive(Copy, Clone, Debug)] -enum Trailing { +pub enum Trailing { No, Yes, } @@ -425,13 +425,13 @@ impl<'a> Parser<'a> { } /// Expects and consumes the token `t`. Signals an error if the next token is not `t`. - pub fn expect(&mut self, exp: ExpTokenPair<'_>) -> PResult<'a, Recovered> { + pub fn expect(&mut self, exp: ExpTokenPair) -> PResult<'a, Recovered> { if self.expected_token_types.is_empty() { - if self.token == *exp.tok { + if self.token == exp.tok { self.bump(); Ok(Recovered::No) } else { - self.unexpected_try_recover(exp.tok) + self.unexpected_try_recover(&exp.tok) } } else { self.expect_one_of(slice::from_ref(&exp), &[]) @@ -443,13 +443,13 @@ impl<'a> Parser<'a> { /// anything. Signal a fatal error if next token is unexpected. fn expect_one_of( &mut self, - edible: &[ExpTokenPair<'_>], - inedible: &[ExpTokenPair<'_>], + edible: &[ExpTokenPair], + inedible: &[ExpTokenPair], ) -> PResult<'a, Recovered> { - if edible.iter().any(|exp| exp.tok == &self.token.kind) { + if edible.iter().any(|exp| exp.tok == self.token.kind) { self.bump(); Ok(Recovered::No) - } else if inedible.iter().any(|exp| exp.tok == &self.token.kind) { + } else if inedible.iter().any(|exp| exp.tok == self.token.kind) { // leave it in the input Ok(Recovered::No) } else if self.token != token::Eof @@ -494,8 +494,8 @@ impl<'a> Parser<'a> { /// This method will automatically add `tok` to `expected_token_types` if `tok` is not /// encountered. #[inline] - fn check(&mut self, exp: ExpTokenPair<'_>) -> bool { - let is_present = self.token == *exp.tok; + pub fn check(&mut self, exp: ExpTokenPair) -> bool { + let is_present = self.token == exp.tok; if !is_present { self.expected_token_types.insert(exp.token_type); } @@ -542,7 +542,7 @@ impl<'a> Parser<'a> { /// Consumes a token 'tok' if it exists. Returns whether the given token was present. #[inline] #[must_use] - pub fn eat(&mut self, exp: ExpTokenPair<'_>) -> bool { + pub fn eat(&mut self, exp: ExpTokenPair) -> bool { let is_present = self.check(exp); if is_present { self.bump() @@ -633,7 +633,7 @@ impl<'a> Parser<'a> { } /// Consume a sequence produced by a metavar expansion, if present. - fn eat_metavar_seq<T>( + pub fn eat_metavar_seq<T>( &mut self, mv_kind: MetaVarKind, f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, @@ -745,13 +745,13 @@ impl<'a> Parser<'a> { /// Eats the expected token if it's present possibly breaking /// compound tokens like multi-character operators in process. /// Returns `true` if the token was eaten. - fn break_and_eat(&mut self, exp: ExpTokenPair<'_>) -> bool { - if self.token == *exp.tok { + fn break_and_eat(&mut self, exp: ExpTokenPair) -> bool { + if self.token == exp.tok { self.bump(); return true; } match self.token.kind.break_two_token_op(1) { - Some((first, second)) if first == *exp.tok => { + Some((first, second)) if first == exp.tok => { let first_span = self.psess.source_map().start_point(self.token.span); let second_span = self.token.span.with_lo(first_span.hi()); self.token = Token::new(first, first_span); @@ -826,7 +826,7 @@ impl<'a> Parser<'a> { /// Checks if the next token is contained within `closes`, and returns `true` if so. fn expect_any_with_type( &mut self, - closes_expected: &[ExpTokenPair<'_>], + closes_expected: &[ExpTokenPair], closes_not_expected: &[&TokenKind], ) -> bool { closes_expected.iter().any(|&close| self.check(close)) @@ -838,9 +838,9 @@ impl<'a> Parser<'a> { /// closing bracket. fn parse_seq_to_before_tokens<T>( &mut self, - closes_expected: &[ExpTokenPair<'_>], + closes_expected: &[ExpTokenPair], closes_not_expected: &[&TokenKind], - sep: SeqSep<'_>, + sep: SeqSep, mut f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, ) -> PResult<'a, (ThinVec<T>, Trailing, Recovered)> { let mut first = true; @@ -869,7 +869,7 @@ impl<'a> Parser<'a> { } Err(mut expect_err) => { let sp = self.prev_token.span.shrink_to_hi(); - let token_str = pprust::token_kind_to_string(exp.tok); + let token_str = pprust::token_kind_to_string(&exp.tok); match self.current_closure.take() { Some(closure_spans) if self.token == TokenKind::Semi => { @@ -1039,8 +1039,8 @@ impl<'a> Parser<'a> { /// closing bracket. fn parse_seq_to_before_end<T>( &mut self, - close: ExpTokenPair<'_>, - sep: SeqSep<'_>, + close: ExpTokenPair, + sep: SeqSep, f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, ) -> PResult<'a, (ThinVec<T>, Trailing, Recovered)> { self.parse_seq_to_before_tokens(&[close], &[], sep, f) @@ -1051,8 +1051,8 @@ impl<'a> Parser<'a> { /// closing bracket. fn parse_seq_to_end<T>( &mut self, - close: ExpTokenPair<'_>, - sep: SeqSep<'_>, + close: ExpTokenPair, + sep: SeqSep, f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, ) -> PResult<'a, (ThinVec<T>, Trailing)> { let (val, trailing, recovered) = self.parse_seq_to_before_end(close, sep, f)?; @@ -1070,9 +1070,9 @@ impl<'a> Parser<'a> { /// closing bracket. fn parse_unspanned_seq<T>( &mut self, - open: ExpTokenPair<'_>, - close: ExpTokenPair<'_>, - sep: SeqSep<'_>, + open: ExpTokenPair, + close: ExpTokenPair, + sep: SeqSep, f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, ) -> PResult<'a, (ThinVec<T>, Trailing)> { self.expect(open)?; @@ -1084,8 +1084,8 @@ impl<'a> Parser<'a> { /// closing bracket. fn parse_delim_comma_seq<T>( &mut self, - open: ExpTokenPair<'_>, - close: ExpTokenPair<'_>, + open: ExpTokenPair, + close: ExpTokenPair, f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, ) -> PResult<'a, (ThinVec<T>, Trailing)> { self.parse_unspanned_seq(open, close, SeqSep::trailing_allowed(exp!(Comma)), f) @@ -1094,7 +1094,7 @@ impl<'a> Parser<'a> { /// Parses a comma-separated sequence delimited by parentheses (e.g. `(x, y)`). /// The function `f` must consume tokens until reaching the next separator or /// closing bracket. - fn parse_paren_comma_seq<T>( + pub fn parse_paren_comma_seq<T>( &mut self, f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, ) -> PResult<'a, (ThinVec<T>, Trailing)> { @@ -1355,7 +1355,8 @@ impl<'a> Parser<'a> { AttrArgs::Delimited(args) } else if self.eat(exp!(Eq)) { let eq_span = self.prev_token.span; - AttrArgs::Eq { eq_span, expr: self.parse_expr_force_collect()? } + let expr = self.parse_expr_force_collect()?; + AttrArgs::Eq { eq_span, expr } } else { AttrArgs::Empty }) @@ -1388,15 +1389,26 @@ impl<'a> Parser<'a> { // matching `CloseDelim` we are *after* the delimited sequence, // i.e. at depth `d - 1`. let target_depth = self.token_cursor.stack.len() - 1; - loop { - // Advance one token at a time, so `TokenCursor::next()` - // can capture these tokens if necessary. + + if let Capturing::No = self.capture_state.capturing { + // We are not capturing tokens, so skip to the end of the + // delimited sequence. This is a perf win when dealing with + // declarative macros that pass large `tt` fragments through + // multiple rules, as seen in the uom-0.37.0 crate. + self.token_cursor.curr.bump_to_end(); self.bump(); - if self.token_cursor.stack.len() == target_depth { - debug_assert!(self.token.kind.close_delim().is_some()); - break; + debug_assert_eq!(self.token_cursor.stack.len(), target_depth); + } else { + loop { + // Advance one token at a time, so `TokenCursor::next()` + // can capture these tokens if necessary. + self.bump(); + if self.token_cursor.stack.len() == target_depth { + break; + } } } + debug_assert!(self.token.kind.close_delim().is_some()); // Consume close delimiter self.bump(); diff --git a/compiler/rustc_parse/src/parser/path.rs b/compiler/rustc_parse/src/parser/path.rs index 37fc723cd89..86045648859 100644 --- a/compiler/rustc_parse/src/parser/path.rs +++ b/compiler/rustc_parse/src/parser/path.rs @@ -26,7 +26,7 @@ use crate::parser::{ /// Specifies how to parse a path. #[derive(Copy, Clone, PartialEq)] -pub(super) enum PathStyle { +pub enum PathStyle { /// In some contexts, notably in expressions, paths with generic arguments are ambiguous /// with something else. For example, in expressions `segment < ....` can be interpreted /// as a comparison and `segment ( ....` can be interpreted as a function call. @@ -150,7 +150,7 @@ impl<'a> Parser<'a> { true } - pub(super) fn parse_path(&mut self, style: PathStyle) -> PResult<'a, Path> { + pub fn parse_path(&mut self, style: PathStyle) -> PResult<'a, Path> { self.parse_path_inner(style, None) } diff --git a/compiler/rustc_parse/src/parser/token_type.rs b/compiler/rustc_parse/src/parser/token_type.rs index b91548196a3..bd4bb368df0 100644 --- a/compiler/rustc_parse/src/parser/token_type.rs +++ b/compiler/rustc_parse/src/parser/token_type.rs @@ -416,8 +416,8 @@ impl TokenType { /// is always by used those methods. The second field is only used when the /// first field doesn't match. #[derive(Clone, Copy, Debug)] -pub struct ExpTokenPair<'a> { - pub tok: &'a TokenKind, +pub struct ExpTokenPair { + pub tok: TokenKind, pub token_type: TokenType, } @@ -444,7 +444,7 @@ macro_rules! exp { // `ExpTokenPair` helper rules. (@tok, $tok:ident) => { $crate::parser::token_type::ExpTokenPair { - tok: &rustc_ast::token::$tok, + tok: rustc_ast::token::$tok, token_type: $crate::parser::token_type::TokenType::$tok } }; diff --git a/compiler/rustc_parse/src/parser/ty.rs b/compiler/rustc_parse/src/parser/ty.rs index 290f0a440af..6168647183f 100644 --- a/compiler/rustc_parse/src/parser/ty.rs +++ b/compiler/rustc_parse/src/parser/ty.rs @@ -137,14 +137,37 @@ impl<'a> Parser<'a> { /// The difference from `parse_ty` is that this version allows `...` /// (`CVarArgs`) at the top level of the type. pub(super) fn parse_ty_for_param(&mut self) -> PResult<'a, Box<Ty>> { - self.parse_ty_common( + let ty = self.parse_ty_common( AllowPlus::Yes, AllowCVariadic::Yes, RecoverQPath::Yes, RecoverReturnSign::Yes, None, RecoverQuestionMark::Yes, - ) + )?; + + // Recover a trailing `= EXPR` if present. + if self.may_recover() + && self.check_noexpect(&token::Eq) + && self.look_ahead(1, |tok| tok.can_begin_expr()) + { + let snapshot = self.create_snapshot_for_diagnostic(); + self.bump(); + let eq_span = self.prev_token.span; + match self.parse_expr() { + Ok(e) => { + self.dcx() + .struct_span_err(eq_span.to(e.span), "parameter defaults are not supported") + .emit(); + } + Err(diag) => { + diag.cancel(); + self.restore_snapshot(snapshot); + } + } + } + + Ok(ty) } /// Parses a type in restricted contexts where `+` is not permitted. @@ -308,11 +331,11 @@ impl<'a> Parser<'a> { // Function pointer type or bound list (trait object type) starting with a poly-trait. // `for<'lt> [unsafe] [extern "ABI"] fn (&'lt S) -> T` // `for<'lt> Trait1<'lt> + Trait2 + 'a` - let (lifetime_defs, _) = self.parse_late_bound_lifetime_defs()?; + let (bound_vars, _) = self.parse_higher_ranked_binder()?; if self.check_fn_front_matter(false, Case::Sensitive) { self.parse_ty_fn_ptr( lo, - lifetime_defs, + bound_vars, Some(self.prev_token.span.shrink_to_lo()), recover_return_sign, )? @@ -326,7 +349,7 @@ impl<'a> Parser<'a> { let path = self.parse_path(PathStyle::Type)?; let parse_plus = allow_plus == AllowPlus::Yes && self.check_plus(); let kind = self.parse_remaining_bounds_path( - lifetime_defs, + bound_vars, path, lo, parse_plus, @@ -359,7 +382,7 @@ impl<'a> Parser<'a> { let path = self.parse_path(PathStyle::Type)?; let parse_plus = allow_plus == AllowPlus::Yes && self.check_plus(); self.parse_remaining_bounds_path( - lifetime_defs, + bound_vars, path, lo, parse_plus, @@ -443,7 +466,7 @@ impl<'a> Parser<'a> { let ty = ts.into_iter().next().unwrap(); let maybe_bounds = allow_plus == AllowPlus::Yes && self.token.is_like_plus(); match ty.kind { - // `(TY_BOUND_NOPAREN) + BOUND + ...`. + // `"(" BareTraitBound ")" "+" Bound "+" ...`. TyKind::Path(None, path) if maybe_bounds => self.parse_remaining_bounds_path( ThinVec::new(), path, @@ -853,10 +876,13 @@ impl<'a> Parser<'a> { Ok(TyKind::ImplTrait(ast::DUMMY_NODE_ID, bounds)) } - fn parse_precise_capturing_args( - &mut self, - ) -> PResult<'a, (ThinVec<PreciseCapturingArg>, Span)> { - let lo = self.token.span; + /// Parse a use-bound aka precise capturing list. + /// + /// ```ebnf + /// UseBound = "use" "<" (PreciseCapture ("," PreciseCapture)* ","?)? ">" + /// PreciseCapture = "Self" | Ident | Lifetime + /// ``` + fn parse_use_bound(&mut self, lo: Span, parens: ast::Parens) -> PResult<'a, GenericBound> { self.expect_lt()?; let (args, _, _) = self.parse_seq_to_before_tokens( &[exp!(Gt)], @@ -882,7 +908,13 @@ impl<'a> Parser<'a> { }, )?; self.expect_gt()?; - Ok((args, lo.to(self.prev_token.span))) + + if let ast::Parens::Yes = parens { + self.expect(exp!(CloseParen))?; + self.report_parenthesized_bound(lo, self.prev_token.span, "precise capturing lists"); + } + + Ok(GenericBound::Use(args, lo.to(self.prev_token.span))) } /// Is a `dyn B0 + ... + Bn` type allowed here? @@ -940,9 +972,10 @@ impl<'a> Parser<'a> { self.parse_generic_bounds_common(AllowPlus::Yes) } - /// Parses bounds of a type parameter `BOUND + BOUND + ...`, possibly with trailing `+`. + /// Parse generic bounds. /// - /// See `parse_generic_bound` for the `BOUND` grammar. + /// Only if `allow_plus` this parses a `+`-separated list of bounds (trailing `+` is admitted). + /// Otherwise, this only parses a single bound or none. fn parse_generic_bounds_common(&mut self, allow_plus: AllowPlus) -> PResult<'a, GenericBounds> { let mut bounds = Vec::new(); @@ -988,48 +1021,56 @@ impl<'a> Parser<'a> { || self.check_keyword(exp!(Use)) } - /// Parses a bound according to the grammar: + /// Parse a bound. + /// /// ```ebnf - /// BOUND = TY_BOUND | LT_BOUND + /// Bound = LifetimeBound | UseBound | TraitBound /// ``` fn parse_generic_bound(&mut self) -> PResult<'a, GenericBound> { - let lo = self.token.span; let leading_token = self.prev_token; + let lo = self.token.span; + + // We only admit parenthesized *trait* bounds. However, we want to gracefully recover from + // other kinds of parenthesized bounds, so parse the opening parenthesis *here*. + // + // In the future we might want to lift this syntactic restriction and + // introduce "`GenericBound::Paren(Box<GenericBound>)`". let parens = if self.eat(exp!(OpenParen)) { ast::Parens::Yes } else { ast::Parens::No }; - let bound = if self.token.is_lifetime() { - self.parse_generic_lt_bound(lo, parens)? + if self.token.is_lifetime() { + self.parse_lifetime_bound(lo, parens) } else if self.eat_keyword(exp!(Use)) { - // parse precise captures, if any. This is `use<'lt, 'lt, P, P>`; a list of - // lifetimes and ident params (including SelfUpper). These are validated later - // for order, duplication, and whether they actually reference params. - let use_span = self.prev_token.span; - let (args, args_span) = self.parse_precise_capturing_args()?; - GenericBound::Use(args, use_span.to(args_span)) + self.parse_use_bound(lo, parens) } else { - self.parse_generic_ty_bound(lo, parens, &leading_token)? - }; - - Ok(bound) + self.parse_trait_bound(lo, parens, &leading_token) + } } - /// Parses a lifetime ("outlives") bound, e.g. `'a`, according to: + /// Parse a lifetime-bound aka outlives-bound. + /// /// ```ebnf - /// LT_BOUND = LIFETIME + /// LifetimeBound = Lifetime /// ``` - fn parse_generic_lt_bound( - &mut self, - lo: Span, - parens: ast::Parens, - ) -> PResult<'a, GenericBound> { + fn parse_lifetime_bound(&mut self, lo: Span, parens: ast::Parens) -> PResult<'a, GenericBound> { let lt = self.expect_lifetime(); - let bound = GenericBound::Outlives(lt); + if let ast::Parens::Yes = parens { - // FIXME(Centril): Consider not erroring here and accepting `('lt)` instead, - // possibly introducing `GenericBound::Paren(Box<GenericBound>)`? - self.recover_paren_lifetime(lo)?; + self.expect(exp!(CloseParen))?; + self.report_parenthesized_bound(lo, self.prev_token.span, "lifetime bounds"); } - Ok(bound) + + Ok(GenericBound::Outlives(lt)) + } + + fn report_parenthesized_bound(&self, lo: Span, hi: Span, kind: &str) -> ErrorGuaranteed { + let mut diag = + self.dcx().struct_span_err(lo.to(hi), format!("{kind} may not be parenthesized")); + diag.multipart_suggestion( + "remove the parentheses", + vec![(lo, String::new()), (hi, String::new())], + Applicability::MachineApplicable, + ); + diag.emit() } /// Emits an error if any trait bound modifiers were present. @@ -1074,27 +1115,17 @@ impl<'a> Parser<'a> { unreachable!("lifetime bound intercepted in `parse_generic_ty_bound` but no modifiers?") } - /// Recover on `('lifetime)` with `(` already eaten. - fn recover_paren_lifetime(&mut self, lo: Span) -> PResult<'a, ()> { - self.expect(exp!(CloseParen))?; - let span = lo.to(self.prev_token.span); - let sugg = errors::RemoveParens { lo, hi: self.prev_token.span }; - - self.dcx().emit_err(errors::ParenthesizedLifetime { span, sugg }); - Ok(()) - } - /// Parses the modifiers that may precede a trait in a bound, e.g. `?Trait` or `[const] Trait`. /// /// If no modifiers are present, this does not consume any tokens. /// /// ```ebnf - /// CONSTNESS = [["["] "const" ["]"]] - /// ASYNCNESS = ["async"] - /// POLARITY = ["?" | "!"] + /// Constness = ("const" | "[" "const" "]")? + /// Asyncness = "async"? + /// Polarity = ("?" | "!")? /// ``` /// - /// See `parse_generic_ty_bound` for the complete grammar of trait bound modifiers. + /// See `parse_trait_bound` for more context. fn parse_trait_bound_modifiers(&mut self) -> PResult<'a, TraitBoundModifiers> { let modifier_lo = self.token.span; let constness = self.parse_bound_constness()?; @@ -1187,20 +1218,21 @@ impl<'a> Parser<'a> { }) } - /// Parses a type bound according to: + /// Parse a trait bound. + /// /// ```ebnf - /// TY_BOUND = TY_BOUND_NOPAREN | (TY_BOUND_NOPAREN) - /// TY_BOUND_NOPAREN = [for<GENERIC_PARAMS> CONSTNESS ASYNCNESS | POLARITY] SIMPLE_PATH + /// TraitBound = BareTraitBound | "(" BareTraitBound ")" + /// BareTraitBound = + /// (HigherRankedBinder Constness Asyncness | Polarity) + /// TypePath /// ``` - /// - /// For example, this grammar accepts `for<'a: 'b> [const] ?m::Trait<'a>`. - fn parse_generic_ty_bound( + fn parse_trait_bound( &mut self, lo: Span, parens: ast::Parens, leading_token: &Token, ) -> PResult<'a, GenericBound> { - let (mut lifetime_defs, binder_span) = self.parse_late_bound_lifetime_defs()?; + let (mut bound_vars, binder_span) = self.parse_higher_ranked_binder()?; let modifiers_lo = self.token.span; let modifiers = self.parse_trait_bound_modifiers()?; @@ -1223,11 +1255,11 @@ impl<'a> Parser<'a> { // e.g. `T: for<'a> 'a` or `T: [const] 'a`. if self.token.is_lifetime() { let _: ErrorGuaranteed = self.error_lt_bound_with_modifiers(modifiers, binder_span); - return self.parse_generic_lt_bound(lo, parens); + return self.parse_lifetime_bound(lo, parens); } - if let (more_lifetime_defs, Some(binder_span)) = self.parse_late_bound_lifetime_defs()? { - lifetime_defs.extend(more_lifetime_defs); + if let (more_bound_vars, Some(binder_span)) = self.parse_higher_ranked_binder()? { + bound_vars.extend(more_bound_vars); self.dcx().emit_err(errors::BinderBeforeModifiers { binder_span, modifiers_span }); } @@ -1287,7 +1319,7 @@ impl<'a> Parser<'a> { }; if self.may_recover() && self.token == TokenKind::OpenParen { - self.recover_fn_trait_with_lifetime_params(&mut path, &mut lifetime_defs)?; + self.recover_fn_trait_with_lifetime_params(&mut path, &mut bound_vars)?; } if let ast::Parens::Yes = parens { @@ -1310,7 +1342,7 @@ impl<'a> Parser<'a> { } let poly_trait = - PolyTraitRef::new(lifetime_defs, path, modifiers, lo.to(self.prev_token.span), parens); + PolyTraitRef::new(bound_vars, path, modifiers, lo.to(self.prev_token.span), parens); Ok(GenericBound::Trait(poly_trait)) } @@ -1349,8 +1381,12 @@ impl<'a> Parser<'a> { } } - /// Optionally parses `for<$generic_params>`. - pub(super) fn parse_late_bound_lifetime_defs( + /// Parse an optional higher-ranked binder. + /// + /// ```ebnf + /// HigherRankedBinder = ("for" "<" GenericParams ">")? + /// ``` + pub(super) fn parse_higher_ranked_binder( &mut self, ) -> PResult<'a, (ThinVec<GenericParam>, Option<Span>)> { if self.eat_keyword(exp!(For)) { @@ -1466,8 +1502,7 @@ impl<'a> Parser<'a> { pub(super) fn expect_lifetime(&mut self) -> Lifetime { if let Some((ident, is_raw)) = self.token.lifetime() { if matches!(is_raw, IdentIsRaw::No) - && ident.without_first_quote().is_reserved() - && ![kw::UnderscoreLifetime, kw::StaticLifetime].contains(&ident.name) + && ident.without_first_quote().is_reserved_lifetime() { self.dcx().emit_err(errors::KeywordLifetime { span: ident.span }); } diff --git a/compiler/rustc_parse/src/validate_attr.rs b/compiler/rustc_parse/src/validate_attr.rs deleted file mode 100644 index 68ef6d6f32c..00000000000 --- a/compiler/rustc_parse/src/validate_attr.rs +++ /dev/null @@ -1,350 +0,0 @@ -//! Meta-syntax validation logic of attributes for post-expansion. - -use std::slice; - -use rustc_ast::token::Delimiter; -use rustc_ast::tokenstream::DelimSpan; -use rustc_ast::{ - self as ast, AttrArgs, Attribute, DelimArgs, MetaItem, MetaItemInner, MetaItemKind, NodeId, - Path, Safety, -}; -use rustc_attr_parsing::{AttributeParser, Late}; -use rustc_errors::{Applicability, DiagCtxtHandle, FatalError, PResult}; -use rustc_feature::{AttributeSafety, AttributeTemplate, BUILTIN_ATTRIBUTE_MAP, BuiltinAttribute}; -use rustc_session::errors::report_lit_error; -use rustc_session::lint::BuiltinLintDiag; -use rustc_session::lint::builtin::{ILL_FORMED_ATTRIBUTE_INPUT, UNSAFE_ATTR_OUTSIDE_UNSAFE}; -use rustc_session::parse::ParseSess; -use rustc_span::{Span, Symbol, sym}; - -use crate::{errors, parse_in}; - -pub fn check_attr(psess: &ParseSess, attr: &Attribute, id: NodeId) { - if attr.is_doc_comment() || attr.has_name(sym::cfg_trace) || attr.has_name(sym::cfg_attr_trace) - { - return; - } - - let builtin_attr_info = attr.ident().and_then(|ident| BUILTIN_ATTRIBUTE_MAP.get(&ident.name)); - - let builtin_attr_safety = builtin_attr_info.map(|x| x.safety); - check_attribute_safety(psess, builtin_attr_safety, attr, id); - - // Check input tokens for built-in and key-value attributes. - match builtin_attr_info { - // `rustc_dummy` doesn't have any restrictions specific to built-in attributes. - Some(BuiltinAttribute { name, template, .. }) if *name != sym::rustc_dummy => { - match parse_meta(psess, attr) { - // Don't check safety again, we just did that - Ok(meta) => { - check_builtin_meta_item(psess, &meta, attr.style, *name, *template, false) - } - Err(err) => { - err.emit(); - } - } - } - _ => { - let attr_item = attr.get_normal_item(); - if let AttrArgs::Eq { .. } = attr_item.args { - // All key-value attributes are restricted to meta-item syntax. - match parse_meta(psess, attr) { - Ok(_) => {} - Err(err) => { - err.emit(); - } - } - } - } - } -} - -pub fn parse_meta<'a>(psess: &'a ParseSess, attr: &Attribute) -> PResult<'a, MetaItem> { - let item = attr.get_normal_item(); - Ok(MetaItem { - unsafety: item.unsafety, - span: attr.span, - path: item.path.clone(), - kind: match &item.args { - AttrArgs::Empty => MetaItemKind::Word, - AttrArgs::Delimited(DelimArgs { dspan, delim, tokens }) => { - check_meta_bad_delim(psess, *dspan, *delim); - let nmis = - parse_in(psess, tokens.clone(), "meta list", |p| p.parse_meta_seq_top())?; - MetaItemKind::List(nmis) - } - AttrArgs::Eq { expr, .. } => { - if let ast::ExprKind::Lit(token_lit) = expr.kind { - let res = ast::MetaItemLit::from_token_lit(token_lit, expr.span); - let res = match res { - Ok(lit) => { - if token_lit.suffix.is_some() { - let mut err = psess.dcx().struct_span_err( - expr.span, - "suffixed literals are not allowed in attributes", - ); - err.help( - "instead of using a suffixed literal (`1u8`, `1.0f32`, etc.), \ - use an unsuffixed version (`1`, `1.0`, etc.)", - ); - return Err(err); - } else { - MetaItemKind::NameValue(lit) - } - } - Err(err) => { - let guar = report_lit_error(psess, err, token_lit, expr.span); - let lit = ast::MetaItemLit { - symbol: token_lit.symbol, - suffix: token_lit.suffix, - kind: ast::LitKind::Err(guar), - span: expr.span, - }; - MetaItemKind::NameValue(lit) - } - }; - res - } else { - // Example cases: - // - `#[foo = 1+1]`: results in `ast::ExprKind::Binary`. - // - `#[foo = include_str!("nonexistent-file.rs")]`: - // results in `ast::ExprKind::Err`. In that case we delay - // the error because an earlier error will have already - // been reported. - let msg = "attribute value must be a literal"; - let mut err = psess.dcx().struct_span_err(expr.span, msg); - if let ast::ExprKind::Err(_) = expr.kind { - err.downgrade_to_delayed_bug(); - } - return Err(err); - } - } - }, - }) -} - -fn check_meta_bad_delim(psess: &ParseSess, span: DelimSpan, delim: Delimiter) { - if let Delimiter::Parenthesis = delim { - return; - } - psess.dcx().emit_err(errors::MetaBadDelim { - span: span.entire(), - sugg: errors::MetaBadDelimSugg { open: span.open, close: span.close }, - }); -} - -pub(super) fn check_cfg_attr_bad_delim(psess: &ParseSess, span: DelimSpan, delim: Delimiter) { - if let Delimiter::Parenthesis = delim { - return; - } - psess.dcx().emit_err(errors::CfgAttrBadDelim { - span: span.entire(), - sugg: errors::MetaBadDelimSugg { open: span.open, close: span.close }, - }); -} - -/// Checks that the given meta-item is compatible with this `AttributeTemplate`. -fn is_attr_template_compatible(template: &AttributeTemplate, meta: &ast::MetaItemKind) -> bool { - let is_one_allowed_subword = |items: &[MetaItemInner]| match items { - [item] => item.is_word() && template.one_of.iter().any(|&word| item.has_name(word)), - _ => false, - }; - match meta { - MetaItemKind::Word => template.word, - MetaItemKind::List(items) => template.list.is_some() || is_one_allowed_subword(items), - MetaItemKind::NameValue(lit) if lit.kind.is_str() => template.name_value_str.is_some(), - MetaItemKind::NameValue(..) => false, - } -} - -pub fn check_attribute_safety( - psess: &ParseSess, - builtin_attr_safety: Option<AttributeSafety>, - attr: &Attribute, - id: NodeId, -) { - let attr_item = attr.get_normal_item(); - match (builtin_attr_safety, attr_item.unsafety) { - // - Unsafe builtin attribute - // - User wrote `#[unsafe(..)]`, which is permitted on any edition - (Some(AttributeSafety::Unsafe { .. }), Safety::Unsafe(..)) => { - // OK - } - - // - Unsafe builtin attribute - // - User did not write `#[unsafe(..)]` - (Some(AttributeSafety::Unsafe { unsafe_since }), Safety::Default) => { - let path_span = attr_item.path.span; - - // If the `attr_item`'s span is not from a macro, then just suggest - // wrapping it in `unsafe(...)`. Otherwise, we suggest putting the - // `unsafe(`, `)` right after and right before the opening and closing - // square bracket respectively. - let diag_span = attr_item.span(); - - // Attributes can be safe in earlier editions, and become unsafe in later ones. - // - // Use the span of the attribute's name to determine the edition: the span of the - // attribute as a whole may be inaccurate if it was emitted by a macro. - // - // See https://github.com/rust-lang/rust/issues/142182. - let emit_error = match unsafe_since { - None => true, - Some(unsafe_since) => path_span.edition() >= unsafe_since, - }; - - if emit_error { - psess.dcx().emit_err(errors::UnsafeAttrOutsideUnsafe { - span: path_span, - suggestion: errors::UnsafeAttrOutsideUnsafeSuggestion { - left: diag_span.shrink_to_lo(), - right: diag_span.shrink_to_hi(), - }, - }); - } else { - psess.buffer_lint( - UNSAFE_ATTR_OUTSIDE_UNSAFE, - path_span, - id, - BuiltinLintDiag::UnsafeAttrOutsideUnsafe { - attribute_name_span: path_span, - sugg_spans: (diag_span.shrink_to_lo(), diag_span.shrink_to_hi()), - }, - ); - } - } - - // - Normal builtin attribute, or any non-builtin attribute - // - All non-builtin attributes are currently considered safe; writing `#[unsafe(..)]` is - // not permitted on non-builtin attributes or normal builtin attributes - (Some(AttributeSafety::Normal) | None, Safety::Unsafe(unsafe_span)) => { - psess.dcx().emit_err(errors::InvalidAttrUnsafe { - span: unsafe_span, - name: attr_item.path.clone(), - }); - } - - // - Normal builtin attribute - // - No explicit `#[unsafe(..)]` written. - (Some(AttributeSafety::Normal), Safety::Default) => { - // OK - } - - // - Non-builtin attribute - // - No explicit `#[unsafe(..)]` written. - (None, Safety::Default) => { - // OK - } - - ( - Some(AttributeSafety::Unsafe { .. } | AttributeSafety::Normal) | None, - Safety::Safe(..), - ) => { - psess.dcx().span_delayed_bug( - attr_item.span(), - "`check_attribute_safety` does not expect `Safety::Safe` on attributes", - ); - } - } -} - -// Called by `check_builtin_meta_item` and code that manually denies -// `unsafe(...)` in `cfg` -pub fn deny_builtin_meta_unsafety(diag: DiagCtxtHandle<'_>, unsafety: Safety, name: &Path) { - // This only supports denying unsafety right now - making builtin attributes - // support unsafety will requite us to thread the actual `Attribute` through - // for the nice diagnostics. - if let Safety::Unsafe(unsafe_span) = unsafety { - diag.emit_err(errors::InvalidAttrUnsafe { span: unsafe_span, name: name.clone() }); - } -} - -pub fn check_builtin_meta_item( - psess: &ParseSess, - meta: &MetaItem, - style: ast::AttrStyle, - name: Symbol, - template: AttributeTemplate, - deny_unsafety: bool, -) { - if !is_attr_template_compatible(&template, &meta.kind) { - // attrs with new parsers are locally validated so excluded here - if AttributeParser::<Late>::is_parsed_attribute(slice::from_ref(&name)) { - return; - } - emit_malformed_attribute(psess, style, meta.span, name, template); - } - - if deny_unsafety { - deny_builtin_meta_unsafety(psess.dcx(), meta.unsafety, &meta.path); - } -} - -fn emit_malformed_attribute( - psess: &ParseSess, - style: ast::AttrStyle, - span: Span, - name: Symbol, - template: AttributeTemplate, -) { - // Some of previously accepted forms were used in practice, - // report them as warnings for now. - let should_warn = |name| matches!(name, sym::doc | sym::link | sym::test | sym::bench); - - let error_msg = format!("malformed `{name}` attribute input"); - let mut suggestions = vec![]; - let inner = if style == ast::AttrStyle::Inner { "!" } else { "" }; - if template.word { - suggestions.push(format!("#{inner}[{name}]")); - } - if let Some(descr) = template.list { - for descr in descr { - suggestions.push(format!("#{inner}[{name}({descr})]")); - } - } - suggestions.extend(template.one_of.iter().map(|&word| format!("#{inner}[{name}({word})]"))); - if let Some(descr) = template.name_value_str { - for descr in descr { - suggestions.push(format!("#{inner}[{name} = \"{descr}\"]")); - } - } - if should_warn(name) { - psess.buffer_lint( - ILL_FORMED_ATTRIBUTE_INPUT, - span, - ast::CRATE_NODE_ID, - BuiltinLintDiag::IllFormedAttributeInput { - suggestions: suggestions.clone(), - docs: template.docs, - }, - ); - } else { - suggestions.sort(); - let mut err = psess.dcx().struct_span_err(span, error_msg).with_span_suggestions( - span, - if suggestions.len() == 1 { - "must be of the form" - } else { - "the following are the possible correct uses" - }, - suggestions, - Applicability::HasPlaceholders, - ); - if let Some(link) = template.docs { - err.note(format!("for more information, visit <{link}>")); - } - err.emit(); - } -} - -pub fn emit_fatal_malformed_builtin_attribute( - psess: &ParseSess, - attr: &Attribute, - name: Symbol, -) -> ! { - let template = BUILTIN_ATTRIBUTE_MAP.get(&name).expect("builtin attr defined").template; - emit_malformed_attribute(psess, attr.style, attr.span, name, template); - // This is fatal, otherwise it will likely cause a cascade of other errors - // (and an error here is expected to be very rare). - FatalError.raise() -} |