diff options
Diffstat (limited to 'compiler/rustc_parse/src')
| -rw-r--r-- | compiler/rustc_parse/src/errors.rs | 70 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/attr.rs | 6 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/attr_wrapper.rs | 162 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/diagnostics.rs | 58 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/expr.rs | 81 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/generics.rs | 2 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/item.rs | 118 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/mod.rs | 26 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/nonterminal.rs | 29 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/pat.rs | 3 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/path.rs | 27 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/tests.rs | 17 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/parser/ty.rs | 100 | ||||
| -rw-r--r-- | compiler/rustc_parse/src/validate_attr.rs | 75 |
14 files changed, 457 insertions, 317 deletions
diff --git a/compiler/rustc_parse/src/errors.rs b/compiler/rustc_parse/src/errors.rs index 46e15734853..3ae9b6dad99 100644 --- a/compiler/rustc_parse/src/errors.rs +++ b/compiler/rustc_parse/src/errors.rs @@ -1612,28 +1612,44 @@ pub(crate) struct DefaultNotFollowedByItem { #[derive(Diagnostic)] pub(crate) enum MissingKeywordForItemDefinition { + #[diag(parse_missing_enum_for_enum_definition)] + Enum { + #[primary_span] + span: Span, + #[suggestion(style = "verbose", applicability = "maybe-incorrect", code = "enum ")] + insert_span: Span, + ident: Ident, + }, + #[diag(parse_missing_enum_or_struct_for_item_definition)] + EnumOrStruct { + #[primary_span] + span: Span, + }, #[diag(parse_missing_struct_for_struct_definition)] Struct { #[primary_span] - #[suggestion(style = "short", applicability = "maybe-incorrect", code = " struct ")] span: Span, + #[suggestion(style = "verbose", applicability = "maybe-incorrect", code = "struct ")] + insert_span: Span, ident: Ident, }, #[diag(parse_missing_fn_for_function_definition)] Function { #[primary_span] - #[suggestion(style = "short", applicability = "maybe-incorrect", code = " fn ")] span: Span, + #[suggestion(style = "verbose", applicability = "maybe-incorrect", code = "fn ")] + insert_span: Span, ident: Ident, }, #[diag(parse_missing_fn_for_method_definition)] Method { #[primary_span] - #[suggestion(style = "short", applicability = "maybe-incorrect", code = " fn ")] span: Span, + #[suggestion(style = "verbose", applicability = "maybe-incorrect", code = "fn ")] + insert_span: Span, ident: Ident, }, - #[diag(parse_ambiguous_missing_keyword_for_item_definition)] + #[diag(parse_missing_fn_or_struct_for_item_definition)] Ambiguous { #[primary_span] span: Span, @@ -1644,7 +1660,12 @@ pub(crate) enum MissingKeywordForItemDefinition { #[derive(Subdiagnostic)] pub(crate) enum AmbiguousMissingKwForItemSub { - #[suggestion(parse_suggestion, applicability = "maybe-incorrect", code = "{snippet}!")] + #[suggestion( + parse_suggestion, + style = "verbose", + applicability = "maybe-incorrect", + code = "{snippet}!" + )] SuggestMacro { #[primary_span] span: Span, @@ -2568,14 +2589,6 @@ pub(crate) struct BadReturnTypeNotationOutput { } #[derive(Diagnostic)] -#[diag(parse_bad_return_type_notation_dotdot)] -pub(crate) struct BadReturnTypeNotationDotDot { - #[primary_span] - #[suggestion(code = "", applicability = "maybe-incorrect")] - pub span: Span, -} - -#[derive(Diagnostic)] #[diag(parse_bad_assoc_type_bounds)] pub(crate) struct BadAssocTypeBounds { #[primary_span] @@ -2997,3 +3010,34 @@ pub(crate) struct DotDotRangeAttribute { #[primary_span] pub span: Span, } + +#[derive(Diagnostic)] +#[diag(parse_invalid_attr_unsafe)] +#[note] +pub struct InvalidAttrUnsafe { + #[primary_span] + pub span: Span, + pub name: Path, +} + +#[derive(Diagnostic)] +#[diag(parse_unsafe_attr_outside_unsafe)] +pub 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 struct UnsafeAttrOutsideUnsafeSuggestion { + #[suggestion_part(code = "unsafe(")] + pub left: Span, + #[suggestion_part(code = ")")] + pub right: Span, +} diff --git a/compiler/rustc_parse/src/parser/attr.rs b/compiler/rustc_parse/src/parser/attr.rs index 58fef9b6c45..a8fe35f45b3 100644 --- a/compiler/rustc_parse/src/parser/attr.rs +++ b/compiler/rustc_parse/src/parser/attr.rs @@ -282,7 +282,7 @@ impl<'a> Parser<'a> { pub fn parse_inner_attributes(&mut self) -> PResult<'a, ast::AttrVec> { let mut attrs = ast::AttrVec::new(); loop { - let start_pos: u32 = self.num_bump_calls.try_into().unwrap(); + let start_pos = self.num_bump_calls; // Only try to parse if it is an inner attribute (has `!`). let attr = if self.check(&token::Pound) && self.look_ahead(1, |t| t == &token::Not) { Some(self.parse_attribute(InnerAttrPolicy::Permitted)?) @@ -303,7 +303,7 @@ impl<'a> Parser<'a> { None }; if let Some(attr) = attr { - let end_pos: u32 = self.num_bump_calls.try_into().unwrap(); + let end_pos = self.num_bump_calls; // If we are currently capturing tokens, mark the location of this inner attribute. // If capturing ends up creating a `LazyAttrTokenStream`, we will include // this replace range with it, removing the inner attribute from the final @@ -313,7 +313,7 @@ impl<'a> Parser<'a> { // corresponding macro). let range = start_pos..end_pos; if let Capturing::Yes = self.capture_state.capturing { - self.capture_state.inner_attr_ranges.insert(attr.id, (range, vec![])); + self.capture_state.inner_attr_ranges.insert(attr.id, (range, None)); } attrs.push(attr); } else { diff --git a/compiler/rustc_parse/src/parser/attr_wrapper.rs b/compiler/rustc_parse/src/parser/attr_wrapper.rs index b5480b6b7d2..38f18022e3c 100644 --- a/compiler/rustc_parse/src/parser/attr_wrapper.rs +++ b/compiler/rustc_parse/src/parser/attr_wrapper.rs @@ -1,6 +1,6 @@ use super::{Capturing, FlatToken, ForceCollect, Parser, ReplaceRange, TokenCursor, TrailingToken}; use rustc_ast::token::{self, Delimiter, Token, TokenKind}; -use rustc_ast::tokenstream::{AttrTokenStream, AttrTokenTree, AttributesData, DelimSpacing}; +use rustc_ast::tokenstream::{AttrTokenStream, AttrTokenTree, AttrsTarget, DelimSpacing}; use rustc_ast::tokenstream::{DelimSpan, LazyAttrTokenStream, Spacing, ToAttrTokenStream}; use rustc_ast::{self as ast}; use rustc_ast::{AttrVec, Attribute, HasAttrs, HasTokens}; @@ -8,7 +8,7 @@ use rustc_errors::PResult; use rustc_session::parse::ParseSess; use rustc_span::{sym, Span, DUMMY_SP}; -use std::ops::Range; +use std::{iter, mem}; /// A wrapper type to ensure that the parser handles outer attributes correctly. /// When we parse outer attributes, we need to ensure that we capture tokens @@ -29,15 +29,15 @@ pub struct AttrWrapper { // The start of the outer attributes in the token cursor. // This allows us to create a `ReplaceRange` for the entire attribute // target, including outer attributes. - start_pos: usize, + start_pos: u32, } impl AttrWrapper { - pub(super) fn new(attrs: AttrVec, start_pos: usize) -> AttrWrapper { + pub(super) fn new(attrs: AttrVec, start_pos: u32) -> AttrWrapper { AttrWrapper { attrs, start_pos } } pub fn empty() -> AttrWrapper { - AttrWrapper { attrs: AttrVec::new(), start_pos: usize::MAX } + AttrWrapper { attrs: AttrVec::new(), start_pos: u32::MAX } } pub(crate) fn take_for_recovery(self, psess: &ParseSess) -> AttrVec { @@ -53,7 +53,7 @@ impl AttrWrapper { // FIXME: require passing an NT to prevent misuse of this method pub(crate) fn prepend_to_nt_inner(self, attrs: &mut AttrVec) { let mut self_attrs = self.attrs; - std::mem::swap(attrs, &mut self_attrs); + mem::swap(attrs, &mut self_attrs); attrs.extend(self_attrs); } @@ -87,11 +87,10 @@ fn has_cfg_or_cfg_attr(attrs: &[Attribute]) -> bool { // // This also makes `Parser` very cheap to clone, since // there is no intermediate collection buffer to clone. -#[derive(Clone)] struct LazyAttrTokenStreamImpl { start_token: (Token, Spacing), cursor_snapshot: TokenCursor, - num_calls: usize, + num_calls: u32, break_last_token: bool, replace_ranges: Box<[ReplaceRange]>, } @@ -104,15 +103,16 @@ impl ToAttrTokenStream for LazyAttrTokenStreamImpl { // produce an empty `TokenStream` if no calls were made, and omit the // final token otherwise. let mut cursor_snapshot = self.cursor_snapshot.clone(); - let tokens = - std::iter::once((FlatToken::Token(self.start_token.0.clone()), self.start_token.1)) - .chain(std::iter::repeat_with(|| { - let token = cursor_snapshot.next(); - (FlatToken::Token(token.0), token.1) - })) - .take(self.num_calls); - - if !self.replace_ranges.is_empty() { + let tokens = iter::once((FlatToken::Token(self.start_token.0.clone()), self.start_token.1)) + .chain(iter::repeat_with(|| { + let token = cursor_snapshot.next(); + (FlatToken::Token(token.0), token.1) + })) + .take(self.num_calls as usize); + + if self.replace_ranges.is_empty() { + make_attr_token_stream(tokens, self.break_last_token) + } else { let mut tokens: Vec<_> = tokens.collect(); let mut replace_ranges = self.replace_ranges.to_vec(); replace_ranges.sort_by_key(|(range, _)| range.start); @@ -144,29 +144,26 @@ impl ToAttrTokenStream for LazyAttrTokenStreamImpl { // start position, we ensure that any replace range which encloses // another replace range will capture the *replaced* tokens for the inner // range, not the original tokens. - for (range, new_tokens) in replace_ranges.into_iter().rev() { + for (range, target) in replace_ranges.into_iter().rev() { assert!(!range.is_empty(), "Cannot replace an empty range: {range:?}"); - // Replace ranges are only allowed to decrease the number of tokens. - assert!( - range.len() >= new_tokens.len(), - "Range {range:?} has greater len than {new_tokens:?}" - ); - - // Replace any removed tokens with `FlatToken::Empty`. - // This keeps the total length of `tokens` constant throughout the - // replacement process, allowing us to use all of the `ReplaceRanges` entries - // without adjusting indices. - let filler = std::iter::repeat((FlatToken::Empty, Spacing::Alone)) - .take(range.len() - new_tokens.len()); + // Replace the tokens in range with zero or one `FlatToken::AttrsTarget`s, plus + // enough `FlatToken::Empty`s to fill up the rest of the range. This keeps the + // total length of `tokens` constant throughout the replacement process, allowing + // us to use all of the `ReplaceRanges` entries without adjusting indices. + let target_len = target.is_some() as usize; tokens.splice( (range.start as usize)..(range.end as usize), - new_tokens.into_iter().chain(filler), + target + .into_iter() + .map(|target| (FlatToken::AttrsTarget(target), Spacing::Alone)) + .chain( + iter::repeat((FlatToken::Empty, Spacing::Alone)) + .take(range.len() - target_len), + ), ); } - make_token_stream(tokens.into_iter(), self.break_last_token) - } else { - make_token_stream(tokens, self.break_last_token) + make_attr_token_stream(tokens.into_iter(), self.break_last_token) } } } @@ -218,24 +215,23 @@ impl<'a> Parser<'a> { let start_token = (self.token.clone(), self.token_spacing); let cursor_snapshot = self.token_cursor.clone(); let start_pos = self.num_bump_calls; - let has_outer_attrs = !attrs.attrs.is_empty(); - let prev_capturing = std::mem::replace(&mut self.capture_state.capturing, Capturing::Yes); let replace_ranges_start = self.capture_state.replace_ranges.len(); - let ret = f(self, attrs.attrs); - - self.capture_state.capturing = prev_capturing; - - let (mut ret, trailing) = ret?; + let (mut ret, trailing) = { + let prev_capturing = mem::replace(&mut self.capture_state.capturing, Capturing::Yes); + let ret_and_trailing = f(self, attrs.attrs); + self.capture_state.capturing = prev_capturing; + ret_and_trailing? + }; // When we're not in `capture-cfg` mode, then bail out early if: // 1. Our target doesn't support tokens at all (e.g we're parsing an `NtIdent`) // so there's nothing for us to do. // 2. Our target already has tokens set (e.g. we've parsed something - // like `#[my_attr] $item`. The actual parsing code takes care of prepending - // any attributes to the nonterminal, so we don't need to modify the - // already captured tokens. + // like `#[my_attr] $item`). The actual parsing code takes care of + // prepending any attributes to the nonterminal, so we don't need to + // modify the already captured tokens. // Note that this check is independent of `force_collect`- if we already // have tokens, or can't even store them, then there's never a need to // force collection of new tokens. @@ -276,37 +272,32 @@ impl<'a> Parser<'a> { let replace_ranges_end = self.capture_state.replace_ranges.len(); - let mut end_pos = self.num_bump_calls; - - let mut captured_trailing = false; - // Capture a trailing token if requested by the callback 'f' - match trailing { - TrailingToken::None => {} + let captured_trailing = match trailing { + TrailingToken::None => false, TrailingToken::Gt => { assert_eq!(self.token.kind, token::Gt); + false } TrailingToken::Semi => { assert_eq!(self.token.kind, token::Semi); - end_pos += 1; - captured_trailing = true; + true } - TrailingToken::MaybeComma => { - if self.token.kind == token::Comma { - end_pos += 1; - captured_trailing = true; - } - } - } + TrailingToken::MaybeComma => self.token.kind == token::Comma, + }; - // If we 'broke' the last token (e.g. breaking a '>>' token to two '>' tokens), - // then extend the range of captured tokens to include it, since the parser - // was not actually bumped past it. When the `LazyAttrTokenStream` gets converted - // into an `AttrTokenStream`, we will create the proper token. - if self.break_last_token { - assert!(!captured_trailing, "Cannot set break_last_token and have trailing token"); - end_pos += 1; - } + assert!( + !(self.break_last_token && captured_trailing), + "Cannot set break_last_token and have trailing token" + ); + + let end_pos = self.num_bump_calls + + captured_trailing as u32 + // If we 'broke' the last token (e.g. breaking a '>>' token to two '>' tokens), then + // extend the range of captured tokens to include it, since the parser was not actually + // bumped past it. When the `LazyAttrTokenStream` gets converted into an + // `AttrTokenStream`, we will create the proper token. + + self.break_last_token as u32; let num_calls = end_pos - start_pos; @@ -318,14 +309,11 @@ impl<'a> Parser<'a> { // Grab any replace ranges that occur *inside* the current AST node. // We will perform the actual replacement when we convert the `LazyAttrTokenStream` // to an `AttrTokenStream`. - let start_calls: u32 = start_pos.try_into().unwrap(); self.capture_state.replace_ranges[replace_ranges_start..replace_ranges_end] .iter() .cloned() .chain(inner_attr_replace_ranges.iter().cloned()) - .map(|(range, tokens)| { - ((range.start - start_calls)..(range.end - start_calls), tokens) - }) + .map(|(range, data)| ((range.start - start_pos)..(range.end - start_pos), data)) .collect() }; @@ -340,7 +328,7 @@ impl<'a> Parser<'a> { // If we support tokens at all if let Some(target_tokens) = ret.tokens_mut() { if target_tokens.is_none() { - // Store se our newly captured tokens into the AST node + // Store our newly captured tokens into the AST node. *target_tokens = Some(tokens.clone()); } } @@ -355,18 +343,14 @@ impl<'a> Parser<'a> { && matches!(self.capture_state.capturing, Capturing::Yes) && has_cfg_or_cfg_attr(final_attrs) { - let attr_data = AttributesData { attrs: final_attrs.iter().cloned().collect(), tokens }; + assert!(!self.break_last_token, "Should not have unglued last token with cfg attr"); - // Replace the entire AST node that we just parsed, including attributes, - // with a `FlatToken::AttrTarget`. If this AST node is inside an item - // that has `#[derive]`, then this will allow us to cfg-expand this - // AST node. + // Replace the entire AST node that we just parsed, including attributes, with + // `target`. If this AST node is inside an item that has `#[derive]`, then this will + // allow us to cfg-expand this AST node. let start_pos = if has_outer_attrs { attrs.start_pos } else { start_pos }; - let new_tokens = vec![(FlatToken::AttrTarget(attr_data), Spacing::Alone)]; - - assert!(!self.break_last_token, "Should not have unglued last token with cfg attr"); - let range: Range<u32> = (start_pos.try_into().unwrap())..(end_pos.try_into().unwrap()); - self.capture_state.replace_ranges.push((range, new_tokens)); + let target = AttrsTarget { attrs: final_attrs.iter().cloned().collect(), tokens }; + self.capture_state.replace_ranges.push((start_pos..end_pos, Some(target))); self.capture_state.replace_ranges.extend(inner_attr_replace_ranges); } @@ -382,10 +366,10 @@ impl<'a> Parser<'a> { } } -/// Converts a flattened iterator of tokens (including open and close delimiter tokens) -/// into a `TokenStream`, creating a `TokenTree::Delimited` for each matching pair -/// of open and close delims. -fn make_token_stream( +/// Converts a flattened iterator of tokens (including open and close delimiter tokens) into an +/// `AttrTokenStream`, creating an `AttrTokenTree::Delimited` for each matching pair of open and +/// close delims. +fn make_attr_token_stream( mut iter: impl Iterator<Item = (FlatToken, Spacing)>, break_last_token: bool, ) -> AttrTokenStream { @@ -428,11 +412,11 @@ fn make_token_stream( .expect("Bottom token frame is missing!") .inner .push(AttrTokenTree::Token(token, spacing)), - FlatToken::AttrTarget(data) => stack + FlatToken::AttrsTarget(target) => stack .last_mut() .expect("Bottom token frame is missing!") .inner - .push(AttrTokenTree::Attributes(data)), + .push(AttrTokenTree::AttrsTarget(target)), FlatToken::Empty => {} } token_and_spacing = iter.next(); @@ -464,6 +448,6 @@ mod size_asserts { use rustc_data_structures::static_assert_size; // tidy-alphabetical-start static_assert_size!(AttrWrapper, 16); - static_assert_size!(LazyAttrTokenStreamImpl, 104); + static_assert_size!(LazyAttrTokenStreamImpl, 96); // tidy-alphabetical-end } diff --git a/compiler/rustc_parse/src/parser/diagnostics.rs b/compiler/rustc_parse/src/parser/diagnostics.rs index c2b91488a11..81d5f0fca0e 100644 --- a/compiler/rustc_parse/src/parser/diagnostics.rs +++ b/compiler/rustc_parse/src/parser/diagnostics.rs @@ -2965,9 +2965,10 @@ impl<'a> Parser<'a> { /// This checks if this is a conflict marker, depending of the parameter passed. /// - /// * `>>>>>` - /// * `=====` - /// * `<<<<<` + /// * `<<<<<<<` + /// * `|||||||` + /// * `=======` + /// * `>>>>>>>` /// pub(super) fn is_vcs_conflict_marker( &mut self, @@ -2997,14 +2998,18 @@ impl<'a> Parser<'a> { } pub(crate) fn err_vcs_conflict_marker(&mut self) -> PResult<'a, ()> { + // <<<<<<< let Some(start) = self.conflict_marker(&TokenKind::BinOp(token::Shl), &TokenKind::Lt) else { return Ok(()); }; let mut spans = Vec::with_capacity(3); spans.push(start); + // ||||||| let mut middlediff3 = None; + // ======= let mut middle = None; + // >>>>>>> let mut end = None; loop { if self.token.kind == TokenKind::Eof { @@ -3025,29 +3030,50 @@ impl<'a> Parser<'a> { } self.bump(); } + let mut err = self.dcx().struct_span_err(spans, "encountered diff marker"); - err.span_label(start, "after this is the code before the merge"); - if let Some(middle) = middlediff3 { - err.span_label(middle, ""); - } + match middlediff3 { + // We're using diff3 + Some(middlediff3) => { + err.span_label( + start, + "between this marker and `|||||||` is the code that we're merging into", + ); + err.span_label(middlediff3, "between this marker and `=======` is the base code (what the two refs diverged from)"); + } + None => { + err.span_label( + start, + "between this marker and `=======` is the code that we're merging into", + ); + } + }; + if let Some(middle) = middle { - err.span_label(middle, ""); + err.span_label(middle, "between this marker and `>>>>>>>` is the incoming code"); } if let Some(end) = end { - err.span_label(end, "above this are the incoming code changes"); + err.span_label(end, "this marker concludes the conflict region"); } - err.help( - "if you're having merge conflicts after pulling new code, the top section is the code \ - you already had and the bottom section is the remote code", + err.note( + "conflict markers indicate that a merge was started but could not be completed due \ + to merge conflicts\n\ + to resolve a conflict, keep only the code you want and then delete the lines \ + containing conflict markers", ); err.help( - "if you're in the middle of a rebase, the top section is the code being rebased onto \ - and the bottom section is the code coming from the current commit being rebased", + "if you're having merge conflicts after pulling new code:\n\ + the top section is the code you already had and the bottom section is the remote code\n\ + if you're in the middle of a rebase:\n\ + the top section is the code being rebased onto and the bottom section is the code \ + coming from the current commit being rebased", ); + err.note( - "for an explanation on these markers from the `git` documentation, visit \ - <https://git-scm.com/book/en/v2/Git-Tools-Advanced-Merging#_checking_out_conflicts>", + "for an explanation on these markers from the `git` documentation:\n\ + visit <https://git-scm.com/book/en/v2/Git-Tools-Advanced-Merging#_checking_out_conflicts>", ); + Err(err) } diff --git a/compiler/rustc_parse/src/parser/expr.rs b/compiler/rustc_parse/src/parser/expr.rs index 09f706143fa..b2df9a14eb0 100644 --- a/compiler/rustc_parse/src/parser/expr.rs +++ b/compiler/rustc_parse/src/parser/expr.rs @@ -1,4 +1,5 @@ // ignore-tidy-filelength + use super::diagnostics::SnapshotParser; use super::pat::{CommaRecoveryMode, Expected, RecoverColon, RecoverComma}; use super::ty::{AllowPlus, RecoverQPath, RecoverReturnSign}; @@ -38,36 +39,6 @@ use rustc_span::{BytePos, ErrorGuaranteed, Pos, Span}; use thin_vec::{thin_vec, ThinVec}; use tracing::instrument; -/// Possibly accepts an `token::Interpolated` expression (a pre-parsed expression -/// dropped into the token stream, which happens while parsing the result of -/// macro expansion). Placement of these is not as complex as I feared it would -/// be. The important thing is to make sure that lookahead doesn't balk at -/// `token::Interpolated` tokens. -macro_rules! maybe_whole_expr { - ($p:expr) => { - if let token::Interpolated(nt) = &$p.token.kind { - match &**nt { - token::NtExpr(e) | token::NtLiteral(e) => { - let e = e.clone(); - $p.bump(); - return Ok(e); - } - token::NtPath(path) => { - let path = (**path).clone(); - $p.bump(); - return Ok($p.mk_expr($p.prev_token.span, ExprKind::Path(None, path))); - } - token::NtBlock(block) => { - let block = block.clone(); - $p.bump(); - return Ok($p.mk_expr($p.prev_token.span, ExprKind::Block(block, None))); - } - _ => {} - }; - } - }; -} - #[derive(Debug)] pub(super) enum LhsExpr { // Already parsed just the outer attributes. @@ -1420,7 +1391,27 @@ impl<'a> Parser<'a> { /// correctly if called from `parse_dot_or_call_expr()`. fn parse_expr_bottom(&mut self) -> PResult<'a, P<Expr>> { maybe_recover_from_interpolated_ty_qpath!(self, true); - maybe_whole_expr!(self); + + if let token::Interpolated(nt) = &self.token.kind { + match &**nt { + token::NtExpr(e) | token::NtLiteral(e) => { + let e = e.clone(); + self.bump(); + return Ok(e); + } + token::NtPath(path) => { + let path = (**path).clone(); + self.bump(); + return Ok(self.mk_expr(self.prev_token.span, ExprKind::Path(None, path))); + } + token::NtBlock(block) => { + let block = block.clone(); + self.bump(); + return Ok(self.mk_expr(self.prev_token.span, ExprKind::Block(block, None))); + } + _ => {} + }; + } // Outer attributes are already parsed and will be // added to the return value after the fact. @@ -2189,7 +2180,26 @@ impl<'a> Parser<'a> { /// Matches `'-' lit | lit` (cf. `ast_validation::AstValidator::check_expr_within_pat`). /// Keep this in sync with `Token::can_begin_literal_maybe_minus`. pub fn parse_literal_maybe_minus(&mut self) -> PResult<'a, P<Expr>> { - maybe_whole_expr!(self); + if let token::Interpolated(nt) = &self.token.kind { + match &**nt { + // FIXME(nnethercote) The `NtExpr` case should only match if + // `e` is an `ExprKind::Lit` or an `ExprKind::Unary` containing + // an `UnOp::Neg` and an `ExprKind::Lit`, like how + // `can_begin_literal_maybe_minus` works. But this method has + // been over-accepting for a long time, and to make that change + // here requires also changing some `parse_literal_maybe_minus` + // call sites to accept additional expression kinds. E.g. + // `ExprKind::Path` must be accepted when parsing range + // patterns. That requires some care. So for now, we continue + // being less strict here than we should be. + token::NtExpr(e) | token::NtLiteral(e) => { + let e = e.clone(); + self.bump(); + return Ok(e); + } + _ => {} + }; + } let lo = self.token.span; let minus_present = self.eat(&token::BinOp(token::Minus)); @@ -2317,7 +2327,7 @@ impl<'a> Parser<'a> { let before = self.prev_token.clone(); let binder = if self.check_keyword(kw::For) { let lo = self.token.span; - let lifetime_defs = self.parse_late_bound_lifetime_defs()?; + let (lifetime_defs, _) = self.parse_late_bound_lifetime_defs()?; let span = lo.to(self.prev_token.span); self.psess.gated_spans.gate(sym::closure_lifetime_binder, span); @@ -3422,8 +3432,9 @@ impl<'a> Parser<'a> { } } let capture_clause = self.parse_capture_clause()?; + let decl_span = lo.to(self.prev_token.span); let (attrs, body) = self.parse_inner_attrs_and_block()?; - let kind = ExprKind::Gen(capture_clause, body, kind); + let kind = ExprKind::Gen(capture_clause, body, kind, decl_span); Ok(self.mk_expr_with_attrs(lo.to(self.prev_token.span), kind, attrs)) } @@ -4012,7 +4023,7 @@ impl MutVisitor for CondChecker<'_> { | ExprKind::Match(_, _, _) | ExprKind::Closure(_) | ExprKind::Block(_, _) - | ExprKind::Gen(_, _, _) + | ExprKind::Gen(_, _, _, _) | ExprKind::TryBlock(_) | ExprKind::Underscore | ExprKind::Path(_, _) diff --git a/compiler/rustc_parse/src/parser/generics.rs b/compiler/rustc_parse/src/parser/generics.rs index fde16ac957d..10c7715c7dc 100644 --- a/compiler/rustc_parse/src/parser/generics.rs +++ b/compiler/rustc_parse/src/parser/generics.rs @@ -457,7 +457,7 @@ 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 (lifetime_defs, _) = self.parse_late_bound_lifetime_defs()?; // Parse type with mandatory colon and (possibly empty) bounds, // or with mandatory equality sign and the second type. diff --git a/compiler/rustc_parse/src/parser/item.rs b/compiler/rustc_parse/src/parser/item.rs index 2db777a9f70..f31e634f55c 100644 --- a/compiler/rustc_parse/src/parser/item.rs +++ b/compiler/rustc_parse/src/parser/item.rs @@ -239,6 +239,7 @@ impl<'a> Parser<'a> { self.recover_const_impl(const_span, attrs, def_())? } else { self.recover_const_mut(const_span); + self.recover_missing_kw_before_item()?; let (ident, generics, ty, expr) = self.parse_const_item()?; ( ident, @@ -311,6 +312,9 @@ impl<'a> Parser<'a> { Case::Insensitive, ); } else if macros_allowed && self.check_path() { + if self.isnt_macro_invocation() { + self.recover_missing_kw_before_item()?; + } // MACRO INVOCATION ITEM (Ident::empty(), ItemKind::MacCall(P(self.parse_item_macro(vis)?))) } else { @@ -374,25 +378,25 @@ impl<'a> Parser<'a> { self.check_ident() && self.look_ahead(1, |t| *t != token::Not && *t != token::PathSep) } - /// Recover on encountering a struct or method definition where the user - /// forgot to add the `struct` or `fn` keyword after writing `pub`: `pub S {}`. + /// Recover on encountering a struct, enum, or method definition where the user + /// forgot to add the `struct`, `enum`, or `fn` keyword fn recover_missing_kw_before_item(&mut self) -> PResult<'a, ()> { - // Space between `pub` keyword and the identifier - // - // pub S {} - // ^^^ `sp` points here - let sp = self.prev_token.span.between(self.token.span); - let full_sp = self.prev_token.span.to(self.token.span); - let ident_sp = self.token.span; - - let ident = if self.look_ahead(1, |t| { - [ - token::Lt, - token::OpenDelim(Delimiter::Brace), - token::OpenDelim(Delimiter::Parenthesis), - ] - .contains(&t.kind) - }) { + let is_pub = self.prev_token.is_keyword(kw::Pub); + let is_const = self.prev_token.is_keyword(kw::Const); + let ident_span = self.token.span; + let span = if is_pub { self.prev_token.span.to(ident_span) } else { ident_span }; + let insert_span = ident_span.shrink_to_lo(); + + let ident = if (!is_const + || self.look_ahead(1, |t| *t == token::OpenDelim(Delimiter::Parenthesis))) + && self.look_ahead(1, |t| { + [ + token::Lt, + token::OpenDelim(Delimiter::Brace), + token::OpenDelim(Delimiter::Parenthesis), + ] + .contains(&t.kind) + }) { self.parse_ident().unwrap() } else { return Ok(()); @@ -406,46 +410,56 @@ impl<'a> Parser<'a> { } let err = if self.check(&token::OpenDelim(Delimiter::Brace)) { - // possible public struct definition where `struct` was forgotten - Some(errors::MissingKeywordForItemDefinition::Struct { span: sp, ident }) + // possible struct or enum definition where `struct` or `enum` was forgotten + if self.look_ahead(1, |t| *t == token::CloseDelim(Delimiter::Brace)) { + // `S {}` could be unit enum or struct + Some(errors::MissingKeywordForItemDefinition::EnumOrStruct { span }) + } else if self.look_ahead(2, |t| *t == token::Colon) + || self.look_ahead(3, |t| *t == token::Colon) + { + // `S { f:` or `S { pub f:` + Some(errors::MissingKeywordForItemDefinition::Struct { span, insert_span, ident }) + } else { + Some(errors::MissingKeywordForItemDefinition::Enum { span, insert_span, ident }) + } } else if self.check(&token::OpenDelim(Delimiter::Parenthesis)) { - // possible public function or tuple struct definition where `fn`/`struct` was - // forgotten + // possible function or tuple struct definition where `fn` or `struct` was forgotten self.bump(); // `(` let is_method = self.recover_self_param(); self.consume_block(Delimiter::Parenthesis, ConsumeClosingDelim::Yes); - let err = - if self.check(&token::RArrow) || self.check(&token::OpenDelim(Delimiter::Brace)) { - self.eat_to_tokens(&[&token::OpenDelim(Delimiter::Brace)]); - self.bump(); // `{` - self.consume_block(Delimiter::Brace, ConsumeClosingDelim::Yes); - if is_method { - errors::MissingKeywordForItemDefinition::Method { span: sp, ident } - } else { - errors::MissingKeywordForItemDefinition::Function { span: sp, ident } - } - } else if self.check(&token::Semi) { - errors::MissingKeywordForItemDefinition::Struct { span: sp, ident } + let err = if self.check(&token::RArrow) + || self.check(&token::OpenDelim(Delimiter::Brace)) + { + self.eat_to_tokens(&[&token::OpenDelim(Delimiter::Brace)]); + self.bump(); // `{` + self.consume_block(Delimiter::Brace, ConsumeClosingDelim::Yes); + if is_method { + errors::MissingKeywordForItemDefinition::Method { span, insert_span, ident } } else { - errors::MissingKeywordForItemDefinition::Ambiguous { - span: sp, - subdiag: if found_generics { - None - } else if let Ok(snippet) = self.span_to_snippet(ident_sp) { - Some(errors::AmbiguousMissingKwForItemSub::SuggestMacro { - span: full_sp, - snippet, - }) - } else { - Some(errors::AmbiguousMissingKwForItemSub::HelpMacro) - }, - } - }; + errors::MissingKeywordForItemDefinition::Function { span, insert_span, ident } + } + } else if is_pub && self.check(&token::Semi) { + errors::MissingKeywordForItemDefinition::Struct { span, insert_span, ident } + } else { + errors::MissingKeywordForItemDefinition::Ambiguous { + span, + subdiag: if found_generics { + None + } else if let Ok(snippet) = self.span_to_snippet(ident_span) { + Some(errors::AmbiguousMissingKwForItemSub::SuggestMacro { + span: ident_span, + snippet, + }) + } else { + Some(errors::AmbiguousMissingKwForItemSub::HelpMacro) + }, + } + }; Some(err) } else if found_generics { - Some(errors::MissingKeywordForItemDefinition::Ambiguous { span: sp, subdiag: None }) + Some(errors::MissingKeywordForItemDefinition::Ambiguous { span, subdiag: None }) } else { None }; @@ -1228,7 +1242,7 @@ impl<'a> Parser<'a> { ident_span: ident.span, const_span, }); - ForeignItemKind::Static(Box::new(StaticForeignItem { + ForeignItemKind::Static(Box::new(StaticItem { ty, mutability: Mutability::Not, expr, @@ -1259,7 +1273,7 @@ impl<'a> Parser<'a> { self.token.is_keyword(kw::Unsafe) && self.is_keyword_ahead(1, &[kw::Extern]) && self.look_ahead( - 2 + self.look_ahead(2, |t| t.can_begin_literal_maybe_minus() as usize), + 2 + self.look_ahead(2, |t| t.can_begin_string_literal() as usize), |t| t.kind == token::OpenDelim(Delimiter::Brace), ) } @@ -2448,7 +2462,7 @@ impl<'a> Parser<'a> { }) // `extern ABI fn` || self.check_keyword_case(kw::Extern, case) - && self.look_ahead(1, |t| t.can_begin_literal_maybe_minus()) + && self.look_ahead(1, |t| t.can_begin_string_literal()) && (self.look_ahead(2, |t| t.is_keyword_case(kw::Fn, case)) || // this branch is only for better diagnostic in later, `pub` is not allowed here (self.may_recover() diff --git a/compiler/rustc_parse/src/parser/mod.rs b/compiler/rustc_parse/src/parser/mod.rs index 2f12459da57..45ca267fe5d 100644 --- a/compiler/rustc_parse/src/parser/mod.rs +++ b/compiler/rustc_parse/src/parser/mod.rs @@ -20,7 +20,7 @@ use path::PathStyle; use rustc_ast::ptr::P; use rustc_ast::token::{self, Delimiter, IdentIsRaw, Nonterminal, Token, TokenKind}; -use rustc_ast::tokenstream::{AttributesData, DelimSpacing, DelimSpan, Spacing}; +use rustc_ast::tokenstream::{AttrsTarget, DelimSpacing, DelimSpan, Spacing}; use rustc_ast::tokenstream::{TokenStream, TokenTree, TokenTreeCursor}; use rustc_ast::util::case::Case; use rustc_ast::{ @@ -153,7 +153,7 @@ pub struct Parser<'a> { expected_tokens: Vec<TokenType>, token_cursor: TokenCursor, // The number of calls to `bump`, i.e. the position in the token stream. - num_bump_calls: usize, + num_bump_calls: u32, // During parsing we may sometimes need to 'unglue' a glued token into two // component tokens (e.g. '>>' into '>' and '>), so the parser can consume // them one at a time. This process bypasses the normal capturing mechanism @@ -192,7 +192,7 @@ pub struct Parser<'a> { // This type is used a lot, e.g. it's cloned when matching many declarative macro rules with nonterminals. Make sure // it doesn't unintentionally get bigger. #[cfg(target_pointer_width = "64")] -rustc_data_structures::static_assert_size!(Parser<'_>, 264); +rustc_data_structures::static_assert_size!(Parser<'_>, 256); /// Stores span information about a closure. #[derive(Clone, Debug)] @@ -203,13 +203,13 @@ struct ClosureSpans { } /// Indicates a range of tokens that should be replaced by -/// the tokens in the provided vector. This is used in two +/// the tokens in the provided `AttrsTarget`. This is used in two /// places during token collection: /// /// 1. During the parsing of an AST node that may have a `#[derive]` /// attribute, we parse a nested AST node that has `#[cfg]` or `#[cfg_attr]` /// In this case, we use a `ReplaceRange` to replace the entire inner AST node -/// with `FlatToken::AttrTarget`, allowing us to perform eager cfg-expansion +/// with `FlatToken::AttrsTarget`, allowing us to perform eager cfg-expansion /// on an `AttrTokenStream`. /// /// 2. When we parse an inner attribute while collecting tokens. We @@ -219,7 +219,7 @@ struct ClosureSpans { /// the first macro inner attribute to invoke a proc-macro). /// When create a `TokenStream`, the inner attributes get inserted /// into the proper place in the token stream. -type ReplaceRange = (Range<u32>, Vec<(FlatToken, Spacing)>); +type ReplaceRange = (Range<u32>, Option<AttrsTarget>); /// Controls how we capture tokens. Capturing can be expensive, /// so we try to avoid performing capturing in cases where @@ -1214,6 +1214,9 @@ impl<'a> Parser<'a> { if self.eat_keyword_case(kw::Unsafe, case) { Safety::Unsafe(self.prev_token.uninterpolated_span()) } else if self.eat_keyword_case(kw::Safe, case) { + self.psess + .gated_spans + .gate(sym::unsafe_extern_blocks, self.prev_token.uninterpolated_span()); Safety::Safe(self.prev_token.uninterpolated_span()) } else { Safety::Default @@ -1569,7 +1572,7 @@ impl<'a> Parser<'a> { self.expected_tokens.clear(); } - pub fn approx_token_stream_pos(&self) -> usize { + pub fn approx_token_stream_pos(&self) -> u32 { self.num_bump_calls } } @@ -1605,11 +1608,10 @@ enum FlatToken { /// A token - this holds both delimiter (e.g. '{' and '}') /// and non-delimiter tokens Token(Token), - /// Holds the `AttributesData` for an AST node. The - /// `AttributesData` is inserted directly into the - /// constructed `AttrTokenStream` as - /// an `AttrTokenTree::Attributes`. - AttrTarget(AttributesData), + /// Holds the `AttrsTarget` for an AST node. The `AttrsTarget` is inserted + /// directly into the constructed `AttrTokenStream` as an + /// `AttrTokenTree::AttrsTarget`. + AttrsTarget(AttrsTarget), /// A special 'empty' token that is ignored during the conversion /// to an `AttrTokenStream`. This is used to simplify the /// handling of replace ranges. diff --git a/compiler/rustc_parse/src/parser/nonterminal.rs b/compiler/rustc_parse/src/parser/nonterminal.rs index a0b704aeea5..4a78b427832 100644 --- a/compiler/rustc_parse/src/parser/nonterminal.rs +++ b/compiler/rustc_parse/src/parser/nonterminal.rs @@ -1,5 +1,7 @@ use rustc_ast::ptr::P; -use rustc_ast::token::{self, Delimiter, Nonterminal::*, NonterminalKind, Token}; +use rustc_ast::token::{ + self, Delimiter, Nonterminal::*, NonterminalKind, NtExprKind::*, NtPatKind::*, Token, +}; use rustc_ast::HasTokens; use rustc_ast_pretty::pprust; use rustc_data_structures::sync::Lrc; @@ -36,18 +38,17 @@ impl<'a> Parser<'a> { } match kind { - NonterminalKind::Expr2021 => { + NonterminalKind::Expr(Expr2021 { .. }) => { token.can_begin_expr() // This exception is here for backwards compatibility. && !token.is_keyword(kw::Let) // This exception is here for backwards compatibility. && !token.is_keyword(kw::Const) } - NonterminalKind::Expr => { + NonterminalKind::Expr(Expr) => { token.can_begin_expr() // This exception is here for backwards compatibility. && !token.is_keyword(kw::Let) - && (!token.is_keyword(kw::Const) || token.span.edition().at_least_rust_2024()) } NonterminalKind::Ty => token.can_begin_type(), NonterminalKind::Ident => get_macro_ident(token).is_some(), @@ -75,7 +76,7 @@ impl<'a> Parser<'a> { token::Interpolated(nt) => may_be_ident(nt), _ => false, }, - NonterminalKind::PatParam { .. } | NonterminalKind::PatWithOr => match &token.kind { + NonterminalKind::Pat(pat_kind) => match &token.kind { // box, ref, mut, and other identifiers (can stricten) token::Ident(..) | token::NtIdent(..) | token::OpenDelim(Delimiter::Parenthesis) | // tuple pattern @@ -90,7 +91,7 @@ impl<'a> Parser<'a> { token::Lt | // path (UFCS constant) token::BinOp(token::Shl) => true, // path (double UFCS) // leading vert `|` or-pattern - token::BinOp(token::Or) => matches!(kind, NonterminalKind::PatWithOr), + token::BinOp(token::Or) => matches!(pat_kind, PatWithOr), token::Interpolated(nt) => may_be_ident(nt), _ => false, }, @@ -136,31 +137,25 @@ impl<'a> Parser<'a> { .create_err(UnexpectedNonterminal::Statement(self.token.span))); } }, - NonterminalKind::PatParam { .. } | NonterminalKind::PatWithOr => { - NtPat(self.collect_tokens_no_attrs(|this| match kind { - NonterminalKind::PatParam { .. } => this.parse_pat_no_top_alt(None, None), - NonterminalKind::PatWithOr => this.parse_pat_allow_top_alt( + NonterminalKind::Pat(pat_kind) => { + NtPat(self.collect_tokens_no_attrs(|this| match pat_kind { + PatParam { .. } => this.parse_pat_no_top_alt(None, None), + PatWithOr => this.parse_pat_allow_top_alt( None, RecoverComma::No, RecoverColon::No, CommaRecoveryMode::EitherTupleOrPipe, ), - _ => unreachable!(), })?) } - - NonterminalKind::Expr | NonterminalKind::Expr2021 => { - NtExpr(self.parse_expr_force_collect()?) - } + NonterminalKind::Expr(_) => NtExpr(self.parse_expr_force_collect()?), NonterminalKind::Literal => { // The `:literal` matcher does not support attributes NtLiteral(self.collect_tokens_no_attrs(|this| this.parse_literal_maybe_minus())?) } - NonterminalKind::Ty => { NtTy(self.collect_tokens_no_attrs(|this| this.parse_ty_no_question_mark_recover())?) } - // this could be handled like a token, since it is one NonterminalKind::Ident => { return if let Some((ident, is_raw)) = get_macro_ident(&self.token) { diff --git a/compiler/rustc_parse/src/parser/pat.rs b/compiler/rustc_parse/src/parser/pat.rs index 03aea0888d9..6f2b7177159 100644 --- a/compiler/rustc_parse/src/parser/pat.rs +++ b/compiler/rustc_parse/src/parser/pat.rs @@ -939,7 +939,8 @@ impl<'a> Parser<'a> { || self.look_ahead(dist, |t| { t.is_path_start() // e.g. `MY_CONST`; || t.kind == token::Dot // e.g. `.5` for recovery; - || t.can_begin_literal_maybe_minus() // e.g. `42`. + || matches!(t.kind, token::Literal(..) | token::BinOp(token::Minus)) + || t.is_bool_lit() || t.is_whole_expr() || t.is_lifetime() // recover `'a` instead of `'a'` || (self.may_recover() // recover leading `(` diff --git a/compiler/rustc_parse/src/parser/path.rs b/compiler/rustc_parse/src/parser/path.rs index da8d1194325..03c647dd527 100644 --- a/compiler/rustc_parse/src/parser/path.rs +++ b/compiler/rustc_parse/src/parser/path.rs @@ -353,18 +353,17 @@ impl<'a> Parser<'a> { })?; let span = lo.to(self.prev_token.span); AngleBracketedArgs { args, span }.into() - } else if self.may_recover() - && self.token.kind == token::OpenDelim(Delimiter::Parenthesis) + } else if self.token.kind == token::OpenDelim(Delimiter::Parenthesis) // FIXME(return_type_notation): Could also recover `...` here. && self.look_ahead(1, |tok| tok.kind == token::DotDot) { - self.bump(); - self.dcx() - .emit_err(errors::BadReturnTypeNotationDotDot { span: self.token.span }); - self.bump(); + self.bump(); // ( + self.bump(); // .. self.expect(&token::CloseDelim(Delimiter::Parenthesis))?; let span = lo.to(self.prev_token.span); + self.psess.gated_spans.gate(sym::return_type_notation, span); + if self.eat_noexpect(&token::RArrow) { let lo = self.prev_token.span; let ty = self.parse_ty()?; @@ -372,13 +371,7 @@ impl<'a> Parser<'a> { .emit_err(errors::BadReturnTypeNotationOutput { span: lo.to(ty.span) }); } - ParenthesizedArgs { - span, - inputs: ThinVec::new(), - inputs_span: span, - output: ast::FnRetTy::Default(self.prev_token.span.shrink_to_hi()), - } - .into() + P(ast::GenericArgs::ParenthesizedElided(span)) } else { // `(T, U) -> R` @@ -733,14 +726,6 @@ impl<'a> Parser<'a> { let span = lo.to(self.prev_token.span); - if let AssocItemConstraintKind::Bound { .. } = kind - && let Some(ast::GenericArgs::Parenthesized(args)) = &gen_args - && args.inputs.is_empty() - && let ast::FnRetTy::Default(..) = args.output - { - self.psess.gated_spans.gate(sym::return_type_notation, span); - } - let constraint = AssocItemConstraint { id: ast::DUMMY_NODE_ID, ident, gen_args, kind, span }; Ok(Some(AngleBracketedArg::Constraint(constraint))) diff --git a/compiler/rustc_parse/src/parser/tests.rs b/compiler/rustc_parse/src/parser/tests.rs index 3a4690670af..42392ad2163 100644 --- a/compiler/rustc_parse/src/parser/tests.rs +++ b/compiler/rustc_parse/src/parser/tests.rs @@ -322,9 +322,8 @@ error: foo --> test.rs:3:3 | 3 | X0 Y0 - | ___^__- - | |___| - | || + | ____^ - + | | ______| 4 | || X1 Y1 5 | || X2 Y2 | ||____^__- `Y` is a good letter too @@ -361,9 +360,8 @@ error: foo --> test.rs:3:3 | 3 | X0 Y0 - | ___^__- - | |___| - | || + | ____^ - + | | ______| 4 | || Y1 X1 | ||____-__^ `X` is a good letter | |____| @@ -445,10 +443,9 @@ error: foo --> test.rs:3:3 | 3 | X0 Y0 Z0 - | ___^__-__- - | |___|__| - | ||___| - | ||| + | _____^ - - + | | _______| | + | || _________| 4 | ||| X1 Y1 Z1 5 | ||| X2 Y2 Z2 | |||____^__-__- `Z` label diff --git a/compiler/rustc_parse/src/parser/ty.rs b/compiler/rustc_parse/src/parser/ty.rs index fcd623b477f..1e5b227aaa9 100644 --- a/compiler/rustc_parse/src/parser/ty.rs +++ b/compiler/rustc_parse/src/parser/ty.rs @@ -18,7 +18,7 @@ use rustc_ast::{ }; use rustc_errors::{Applicability, PResult}; use rustc_span::symbol::{kw, sym, Ident}; -use rustc_span::{Span, Symbol}; +use rustc_span::{ErrorGuaranteed, Span, Symbol}; use thin_vec::{thin_vec, ThinVec}; #[derive(Copy, Clone, PartialEq)] @@ -280,7 +280,7 @@ 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 (lifetime_defs, _) = self.parse_late_bound_lifetime_defs()?; if self.check_fn_front_matter(false, Case::Sensitive) { self.parse_ty_bare_fn( lo, @@ -608,7 +608,7 @@ impl<'a> Parser<'a> { self.dcx().emit_err(FnPointerCannotBeAsync { span: whole_span, qualifier: span }); } // FIXME(gen_blocks): emit a similar error for `gen fn()` - let decl_span = span_start.to(self.token.span); + let decl_span = span_start.to(self.prev_token.span); Ok(TyKind::BareFn(P(BareFnTy { ext, safety, generic_params: params, decl, decl_span }))) } @@ -833,12 +833,9 @@ impl<'a> Parser<'a> { let lo = self.token.span; let leading_token = self.prev_token.clone(); let has_parens = self.eat(&token::OpenDelim(Delimiter::Parenthesis)); - let inner_lo = self.token.span; - let modifiers = self.parse_trait_bound_modifiers()?; let bound = if self.token.is_lifetime() { - self.error_lt_bound_with_modifiers(modifiers); - self.parse_generic_lt_bound(lo, inner_lo, has_parens)? + self.parse_generic_lt_bound(lo, has_parens)? } else if self.eat_keyword(kw::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 @@ -848,7 +845,7 @@ impl<'a> Parser<'a> { let (args, args_span) = self.parse_precise_capturing_args()?; GenericBound::Use(args, use_span.to(args_span)) } else { - self.parse_generic_ty_bound(lo, has_parens, modifiers, &leading_token)? + self.parse_generic_ty_bound(lo, has_parens, &leading_token)? }; Ok(bound) @@ -858,50 +855,64 @@ impl<'a> Parser<'a> { /// ```ebnf /// LT_BOUND = LIFETIME /// ``` - fn parse_generic_lt_bound( - &mut self, - lo: Span, - inner_lo: Span, - has_parens: bool, - ) -> PResult<'a, GenericBound> { - let bound = GenericBound::Outlives(self.expect_lifetime()); + fn parse_generic_lt_bound(&mut self, lo: Span, has_parens: bool) -> PResult<'a, GenericBound> { + let lt = self.expect_lifetime(); + let bound = GenericBound::Outlives(lt); if has_parens { // FIXME(Centril): Consider not erroring here and accepting `('lt)` instead, // possibly introducing `GenericBound::Paren(P<GenericBound>)`? - self.recover_paren_lifetime(lo, inner_lo)?; + self.recover_paren_lifetime(lo, lt.ident.span)?; } Ok(bound) } /// Emits an error if any trait bound modifiers were present. - fn error_lt_bound_with_modifiers(&self, modifiers: TraitBoundModifiers) { - match modifiers.constness { + fn error_lt_bound_with_modifiers( + &self, + modifiers: TraitBoundModifiers, + binder_span: Option<Span>, + ) -> ErrorGuaranteed { + let TraitBoundModifiers { constness, asyncness, polarity } = modifiers; + + match constness { BoundConstness::Never => {} BoundConstness::Always(span) | BoundConstness::Maybe(span) => { - self.dcx().emit_err(errors::ModifierLifetime { - span, - modifier: modifiers.constness.as_str(), - }); + return self + .dcx() + .emit_err(errors::ModifierLifetime { span, modifier: constness.as_str() }); } } - match modifiers.polarity { + match polarity { BoundPolarity::Positive => {} BoundPolarity::Negative(span) | BoundPolarity::Maybe(span) => { - self.dcx().emit_err(errors::ModifierLifetime { - span, - modifier: modifiers.polarity.as_str(), - }); + return self + .dcx() + .emit_err(errors::ModifierLifetime { span, modifier: polarity.as_str() }); + } + } + + match asyncness { + BoundAsyncness::Normal => {} + BoundAsyncness::Async(span) => { + return self + .dcx() + .emit_err(errors::ModifierLifetime { span, modifier: asyncness.as_str() }); } } + + if let Some(span) = binder_span { + return self.dcx().emit_err(errors::ModifierLifetime { span, modifier: "for<...>" }); + } + + 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, inner_lo: Span) -> PResult<'a, ()> { - let inner_span = inner_lo.to(self.prev_token.span); + fn recover_paren_lifetime(&mut self, lo: Span, lt_span: Span) -> PResult<'a, ()> { self.expect(&token::CloseDelim(Delimiter::Parenthesis))?; let span = lo.to(self.prev_token.span); - let (sugg, snippet) = if let Ok(snippet) = self.span_to_snippet(inner_span) { + let (sugg, snippet) = if let Ok(snippet) = self.span_to_snippet(lt_span) { (Some(span), snippet) } else { (None, String::new()) @@ -916,7 +927,7 @@ impl<'a> Parser<'a> { /// If no modifiers are present, this does not consume any tokens. /// /// ```ebnf - /// TRAIT_BOUND_MODIFIERS = [["~"] "const"] ["?" | "!"] + /// TRAIT_BOUND_MODIFIERS = [["~"] "const"] ["async"] ["?" | "!"] /// ``` fn parse_trait_bound_modifiers(&mut self) -> PResult<'a, TraitBoundModifiers> { let constness = if self.eat(&token::Tilde) { @@ -970,15 +981,23 @@ impl<'a> Parser<'a> { /// TY_BOUND_NOPAREN = [TRAIT_BOUND_MODIFIERS] [for<LT_PARAM_DEFS>] SIMPLE_PATH /// ``` /// - /// For example, this grammar accepts `~const ?for<'a: 'b> m::Trait<'a>`. + /// For example, this grammar accepts `for<'a: 'b> ~const ?m::Trait<'a>`. fn parse_generic_ty_bound( &mut self, lo: Span, has_parens: bool, - modifiers: TraitBoundModifiers, leading_token: &Token, ) -> PResult<'a, GenericBound> { - let mut lifetime_defs = self.parse_late_bound_lifetime_defs()?; + let modifiers = self.parse_trait_bound_modifiers()?; + let (mut lifetime_defs, binder_span) = self.parse_late_bound_lifetime_defs()?; + + // Recover erroneous lifetime bound with modifiers or binder. + // 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, has_parens); + } + let mut path = if self.token.is_keyword(kw::Fn) && self.look_ahead(1, |tok| tok.kind == TokenKind::OpenDelim(Delimiter::Parenthesis)) && let Some(path) = self.recover_path_from_fn() @@ -1094,16 +1113,19 @@ impl<'a> Parser<'a> { } /// Optionally parses `for<$generic_params>`. - pub(super) fn parse_late_bound_lifetime_defs(&mut self) -> PResult<'a, ThinVec<GenericParam>> { + pub(super) fn parse_late_bound_lifetime_defs( + &mut self, + ) -> PResult<'a, (ThinVec<GenericParam>, Option<Span>)> { if self.eat_keyword(kw::For) { + let lo = self.token.span; self.expect_lt()?; let params = self.parse_generic_params()?; self.expect_gt()?; - // We rely on AST validation to rule out invalid cases: There must not be type - // parameters, and the lifetime parameters must not have bounds. - Ok(params) + // We rely on AST validation to rule out invalid cases: There must not be + // type or const parameters, and parameters must not have bounds. + Ok((params, Some(lo.to(self.prev_token.span)))) } else { - Ok(ThinVec::new()) + Ok((ThinVec::new(), None)) } } diff --git a/compiler/rustc_parse/src/validate_attr.rs b/compiler/rustc_parse/src/validate_attr.rs index 4ca52146039..3d5e6371f4c 100644 --- a/compiler/rustc_parse/src/validate_attr.rs +++ b/compiler/rustc_parse/src/validate_attr.rs @@ -4,22 +4,76 @@ use crate::{errors, parse_in}; use rustc_ast::token::Delimiter; use rustc_ast::tokenstream::DelimSpan; -use rustc_ast::MetaItemKind; -use rustc_ast::{self as ast, AttrArgs, AttrArgsEq, Attribute, DelimArgs, MetaItem}; +use rustc_ast::{ + self as ast, AttrArgs, AttrArgsEq, Attribute, DelimArgs, MetaItem, MetaItemKind, + NestedMetaItem, Safety, +}; use rustc_errors::{Applicability, FatalError, PResult}; -use rustc_feature::{AttributeTemplate, BuiltinAttribute, BUILTIN_ATTRIBUTE_MAP}; +use rustc_feature::{ + AttributeSafety, AttributeTemplate, BuiltinAttribute, Features, BUILTIN_ATTRIBUTE_MAP, +}; use rustc_session::errors::report_lit_error; -use rustc_session::lint::builtin::ILL_FORMED_ATTRIBUTE_INPUT; +use rustc_session::lint::builtin::{ILL_FORMED_ATTRIBUTE_INPUT, UNSAFE_ATTR_OUTSIDE_UNSAFE}; use rustc_session::lint::BuiltinLintDiag; use rustc_session::parse::ParseSess; -use rustc_span::{sym, Span, Symbol}; +use rustc_span::{sym, BytePos, Span, Symbol}; -pub fn check_attr(psess: &ParseSess, attr: &Attribute) { +pub fn check_attr(features: &Features, psess: &ParseSess, attr: &Attribute) { if attr.is_doc_comment() { return; } let attr_info = attr.ident().and_then(|ident| BUILTIN_ATTRIBUTE_MAP.get(&ident.name)); + let attr_item = attr.get_normal_item(); + + let is_unsafe_attr = attr_info.is_some_and(|attr| attr.safety == AttributeSafety::Unsafe); + + if features.unsafe_attributes { + if is_unsafe_attr { + if let ast::Safety::Default = attr_item.unsafety { + 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 = if attr_item.span().can_be_used_for_suggestions() { + attr_item.span() + } else { + attr.span + .with_lo(attr.span.lo() + BytePos(2)) + .with_hi(attr.span.hi() - BytePos(1)) + }; + + if attr.span.at_least_rust_2024() { + 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, + ast::CRATE_NODE_ID, + BuiltinLintDiag::UnsafeAttrOutsideUnsafe { + attribute_name_span: path_span, + sugg_spans: (diag_span.shrink_to_lo(), diag_span.shrink_to_hi()), + }, + ); + } + } + } else { + if let Safety::Unsafe(unsafe_span) = attr_item.unsafety { + psess.dcx().emit_err(errors::InvalidAttrUnsafe { + span: unsafe_span, + name: attr_item.path.clone(), + }); + } + } + } // Check input tokens for built-in and key-value attributes. match attr_info { @@ -32,7 +86,7 @@ pub fn check_attr(psess: &ParseSess, attr: &Attribute) { } } } - _ if let AttrArgs::Eq(..) = attr.get_normal_item().args => { + _ if let AttrArgs::Eq(..) = attr_item.args => { // All key-value attributes are restricted to meta-item syntax. match parse_meta(psess, attr) { Ok(_) => {} @@ -132,9 +186,13 @@ pub(super) fn check_cfg_attr_bad_delim(psess: &ParseSess, span: DelimSpan, delim /// 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: &[NestedMetaItem]| 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(..) => template.list.is_some(), + 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, } @@ -178,6 +236,7 @@ fn emit_malformed_attribute( if let Some(descr) = template.list { 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 { suggestions.push(format!("#{inner}[{name} = \"{descr}\"]")); } |