Auto merge of #56884 - euclio:codeblock-diagnostics, r=QuietMisdreavus
rustdoc: overhaul code block lexing errors
Fixes #53919.
This PR moves the reporting of code block lexing errors from rendering time to an early pass, so we can use the compiler's error reporting mechanisms. This dramatically improves the diagnostics in this situation: we now de-emphasize the lexing errors as a note under a warning that has a span and suggestion instead of just emitting errors at the top level.
Additionally, this PR generalizes the markdown -> source span calculation function, which should allow other rustdoc warnings to use better spans in the future.
Last, the PR makes sure that the code block is always emitted in the docs, even if it fails to highlight correctly.
Of note:
- The new pass unfortunately adds another pass over the docs to gather the doc blocks for syntax-checking. I wonder if this could be combined with the pass that looks for testable blocks? I'm not familiar with that code, so I don't know how feasible that is.
- `pulldown_cmark` doesn't make it easy to find the spans of the code blocks, so the code that calculates the spans is a little nasty. It works for all the test cases I threw at it, but I wouldn't be surprised if an edge case would break it. Should have a thorough review.
- This PR worsens the state of #56885, since those certain fatal lexing errors are now emitted before docs get generated at all.
diff --git a/src/librustdoc/html/highlight.rs b/src/librustdoc/html/highlight.rs
index 558ba1c..e43251b 100644
--- a/src/librustdoc/html/highlight.rs
+++ b/src/librustdoc/html/highlight.rs
@@ -25,40 +25,51 @@
tooltip: Option<(&str, &str)>,
) -> String {
debug!("highlighting: ================\n{}\n==============", src);
- let sess = parse::ParseSess::new(FilePathMapping::empty());
- let fm = sess.source_map().new_source_file(FileName::Custom("stdin".to_string()),
- src.to_string());
-
let mut out = Vec::new();
if let Some((tooltip, class)) = tooltip {
write!(out, "<div class='information'><div class='tooltip {}'>ⓘ<span \
class='tooltiptext'>{}</span></div></div>",
class, tooltip).unwrap();
}
- write_header(class, &mut out).unwrap();
- let lexer = match lexer::StringReader::new_without_err(&sess, fm, None, "Output from rustc:") {
- Ok(l) => l,
- Err(_) => {
- let first_line = src.lines().next().unwrap_or_else(|| "");
- let mut err = sess.span_diagnostic
- .struct_warn(&format!("Invalid doc comment starting with: `{}`\n\
- (Ignoring this codeblock)",
- first_line));
- err.emit();
- return String::new();
+ let sess = parse::ParseSess::new(FilePathMapping::empty());
+ let fm = sess.source_map().new_source_file(
+ FileName::Custom(String::from("rustdoc-highlighting")),
+ src.to_owned(),
+ );
+ let highlight_result =
+ lexer::StringReader::new_or_buffered_errs(&sess, fm, None).and_then(|lexer| {
+ let mut classifier = Classifier::new(lexer, sess.source_map());
+
+ let mut highlighted_source = vec![];
+ if classifier.write_source(&mut highlighted_source).is_err() {
+ Err(classifier.lexer.buffer_fatal_errors())
+ } else {
+ Ok(String::from_utf8_lossy(&highlighted_source).into_owned())
+ }
+ });
+
+ match highlight_result {
+ Ok(highlighted_source) => {
+ write_header(class, &mut out).unwrap();
+ write!(out, "{}", highlighted_source).unwrap();
+ if let Some(extension) = extension {
+ write!(out, "{}", extension).unwrap();
+ }
+ write_footer(&mut out).unwrap();
}
- };
- let mut classifier = Classifier::new(lexer, sess.source_map());
- if classifier.write_source(&mut out).is_err() {
- classifier.lexer.emit_fatal_errors();
- return format!("<pre>{}</pre>", src);
+ Err(errors) => {
+ // If errors are encountered while trying to highlight, cancel the errors and just emit
+ // the unhighlighted source. The errors will have already been reported in the
+ // `check-code-block-syntax` pass.
+ for mut error in errors {
+ error.cancel();
+ }
+
+ write!(out, "<pre><code>{}</code></pre>", src).unwrap();
+ }
}
- if let Some(extension) = extension {
- write!(out, "{}", extension).unwrap();
- }
- write_footer(&mut out).unwrap();
String::from_utf8_lossy(&out[..]).into_owned()
}
@@ -151,6 +162,17 @@
}
}
+enum HighlightError {
+ LexError,
+ IoError(io::Error),
+}
+
+impl From<io::Error> for HighlightError {
+ fn from(err: io::Error) -> Self {
+ HighlightError::IoError(err)
+ }
+}
+
impl<'a> Classifier<'a> {
fn new(lexer: lexer::StringReader<'a>, source_map: &'a SourceMap) -> Classifier<'a> {
Classifier {
@@ -162,17 +184,11 @@
}
}
- /// Gets the next token out of the lexer, emitting fatal errors if lexing fails.
- fn try_next_token(&mut self) -> io::Result<TokenAndSpan> {
+ /// Gets the next token out of the lexer.
+ fn try_next_token(&mut self) -> Result<TokenAndSpan, HighlightError> {
match self.lexer.try_next_token() {
Ok(tas) => Ok(tas),
- Err(_) => {
- let mut err = self.lexer.sess.span_diagnostic
- .struct_warn("Backing out of syntax highlighting");
- err.note("You probably did not intend to render this as a rust code-block");
- err.emit();
- Err(io::Error::new(io::ErrorKind::Other, ""))
- }
+ Err(_) => Err(HighlightError::LexError),
}
}
@@ -185,7 +201,7 @@
/// source.
fn write_source<W: Writer>(&mut self,
out: &mut W)
- -> io::Result<()> {
+ -> Result<(), HighlightError> {
loop {
let next = self.try_next_token()?;
if next.tok == token::Eof {
@@ -202,7 +218,7 @@
fn write_token<W: Writer>(&mut self,
out: &mut W,
tas: TokenAndSpan)
- -> io::Result<()> {
+ -> Result<(), HighlightError> {
let klass = match tas.tok {
token::Shebang(s) => {
out.string(Escape(&s.as_str()), Class::None)?;
@@ -341,7 +357,9 @@
// Anything that didn't return above is the simple case where we the
// class just spans a single token, so we can use the `string` method.
- out.string(Escape(&self.snip(tas.sp)), klass)
+ out.string(Escape(&self.snip(tas.sp)), klass)?;
+
+ Ok(())
}
// Helper function to get a snippet from the source_map.
diff --git a/src/librustdoc/html/markdown.rs b/src/librustdoc/html/markdown.rs
index 05a9a2d..6b7f540 100644
--- a/src/librustdoc/html/markdown.rs
+++ b/src/librustdoc/html/markdown.rs
@@ -919,6 +919,115 @@
links
}
+#[derive(Debug)]
+crate struct RustCodeBlock {
+ /// The range in the markdown that the code block occupies. Note that this includes the fences
+ /// for fenced code blocks.
+ pub range: Range<usize>,
+ /// The range in the markdown that the code within the code block occupies.
+ pub code: Range<usize>,
+ pub is_fenced: bool,
+ pub syntax: Option<String>,
+}
+
+/// Returns a range of bytes for each code block in the markdown that is tagged as `rust` or
+/// untagged (and assumed to be rust).
+crate fn rust_code_blocks(md: &str) -> Vec<RustCodeBlock> {
+ let mut code_blocks = vec![];
+
+ if md.is_empty() {
+ return code_blocks;
+ }
+
+ let mut opts = Options::empty();
+ opts.insert(OPTION_ENABLE_TABLES);
+ opts.insert(OPTION_ENABLE_FOOTNOTES);
+ let mut p = Parser::new_ext(md, opts);
+
+ let mut code_block_start = 0;
+ let mut code_start = 0;
+ let mut is_fenced = false;
+ let mut previous_offset = 0;
+ let mut in_rust_code_block = false;
+ while let Some(event) = p.next() {
+ let offset = p.get_offset();
+
+ match event {
+ Event::Start(Tag::CodeBlock(syntax)) => {
+ let lang_string = if syntax.is_empty() {
+ LangString::all_false()
+ } else {
+ LangString::parse(&*syntax, ErrorCodes::Yes)
+ };
+
+ if lang_string.rust {
+ in_rust_code_block = true;
+
+ code_start = offset;
+ code_block_start = match md[previous_offset..offset].find("```") {
+ Some(fence_idx) => {
+ is_fenced = true;
+ previous_offset + fence_idx
+ }
+ None => offset,
+ };
+ }
+ }
+ Event::End(Tag::CodeBlock(syntax)) if in_rust_code_block => {
+ in_rust_code_block = false;
+
+ let code_block_end = if is_fenced {
+ let fence_str = &md[previous_offset..offset]
+ .chars()
+ .rev()
+ .collect::<String>();
+ fence_str
+ .find("```")
+ .map(|fence_idx| offset - fence_idx)
+ .unwrap_or_else(|| offset)
+ } else if md
+ .as_bytes()
+ .get(offset)
+ .map(|b| *b == b'\n')
+ .unwrap_or_default()
+ {
+ offset - 1
+ } else {
+ offset
+ };
+
+ let code_end = if is_fenced {
+ previous_offset
+ } else {
+ code_block_end
+ };
+
+ code_blocks.push(RustCodeBlock {
+ is_fenced,
+ range: Range {
+ start: code_block_start,
+ end: code_block_end,
+ },
+ code: Range {
+ start: code_start,
+ end: code_end,
+ },
+ syntax: if !syntax.is_empty() {
+ Some(syntax.into_owned())
+ } else {
+ None
+ },
+ });
+ }
+ _ => (),
+ }
+
+ previous_offset = offset;
+ }
+
+ code_blocks
+}
+
#[derive(Clone, Default, Debug)]
pub struct IdMap {
map: FxHashMap<String, usize>,
diff --git a/src/librustdoc/lib.rs b/src/librustdoc/lib.rs
index 4bbc01d..f4149b5 100644
--- a/src/librustdoc/lib.rs
+++ b/src/librustdoc/lib.rs
@@ -3,6 +3,7 @@
html_root_url = "https://doc.rust-lang.org/nightly/",
html_playground_url = "https://play.rust-lang.org/")]
+#![feature(bind_by_move_pattern_guards)]
#![feature(rustc_private)]
#![feature(box_patterns)]
#![feature(box_syntax)]
diff --git a/src/librustdoc/passes/check_code_block_syntax.rs b/src/librustdoc/passes/check_code_block_syntax.rs
new file mode 100644
index 0000000..a013cc3
--- /dev/null
+++ b/src/librustdoc/passes/check_code_block_syntax.rs
@@ -0,0 +1,109 @@
+use errors::Applicability;
+use syntax::parse::lexer::{TokenAndSpan, StringReader as Lexer};
+use syntax::parse::{ParseSess, token};
+use syntax::source_map::FilePathMapping;
+use syntax_pos::FileName;
+
+use clean;
+use core::DocContext;
+use fold::DocFolder;
+use html::markdown::{self, RustCodeBlock};
+use passes::Pass;
+
+pub const CHECK_CODE_BLOCK_SYNTAX: Pass =
+ Pass::early("check-code-block-syntax", check_code_block_syntax,
+ "validates syntax inside Rust code blocks");
+
+pub fn check_code_block_syntax(krate: clean::Crate, cx: &DocContext) -> clean::Crate {
+ SyntaxChecker { cx }.fold_crate(krate)
+}
+
+struct SyntaxChecker<'a, 'tcx: 'a, 'rcx: 'a> {
+ cx: &'a DocContext<'a, 'tcx, 'rcx>,
+}
+
+impl<'a, 'tcx, 'rcx> SyntaxChecker<'a, 'tcx, 'rcx> {
+ fn check_rust_syntax(&self, item: &clean::Item, dox: &str, code_block: RustCodeBlock) {
+ let sess = ParseSess::new(FilePathMapping::empty());
+ let source_file = sess.source_map().new_source_file(
+ FileName::Custom(String::from("doctest")),
+ dox[code_block.code].to_owned(),
+ );
+
+ let errors = Lexer::new_or_buffered_errs(&sess, source_file, None).and_then(|mut lexer| {
+ while let Ok(TokenAndSpan { tok, .. }) = lexer.try_next_token() {
+ if tok == token::Eof {
+ break;
+ }
+ }
+
+ let errors = lexer.buffer_fatal_errors();
+
+ if !errors.is_empty() {
+ Err(errors)
+ } else {
+ Ok(())
+ }
+ });
+
+ if let Err(errors) = errors {
+ let mut diag = if let Some(sp) =
+ super::source_span_for_markdown_range(self.cx, &dox, &code_block.range, &item.attrs)
+ {
+ let mut diag = self
+ .cx
+ .sess()
+ .struct_span_warn(sp, "could not parse code block as Rust code");
+
+ for mut err in errors {
+ diag.note(&format!("error from rustc: {}", err.message()));
+ err.cancel();
+ }
+
+ if code_block.syntax.is_none() && code_block.is_fenced {
+ let sp = sp.from_inner_byte_pos(0, 3);
+ diag.span_suggestion_with_applicability(
+ sp,
+ "mark blocks that do not contain Rust code as text",
+ String::from("```text"),
+ Applicability::MachineApplicable,
+ );
+ }
+
+ diag
+ } else {
+ // We couldn't calculate the span of the markdown block that had the error, so our
+ // diagnostics are going to be a bit lacking.
+ let mut diag = self.cx.sess().struct_span_warn(
+ super::span_of_attrs(&item.attrs),
+ "doc comment contains an invalid Rust code block",
+ );
+
+ for mut err in errors {
+ // Don't bother reporting the error, because we can't show where it happened.
+ err.cancel();
+ }
+
+ if code_block.syntax.is_none() && code_block.is_fenced {
+ diag.help("mark blocks that do not contain Rust code as text: ```text");
+ }
+
+ diag
+ };
+
+ diag.emit();
+ }
+ }
+}
+
+impl<'a, 'tcx, 'rcx> DocFolder for SyntaxChecker<'a, 'tcx, 'rcx> {
+ fn fold_item(&mut self, item: clean::Item) -> Option<clean::Item> {
+ if let Some(dox) = &item.attrs.collapsed_doc_value() {
+ for code_block in markdown::rust_code_blocks(&dox) {
+ self.check_rust_syntax(&item, &dox, code_block);
+ }
+ }
+
+ self.fold_item_recur(item)
+ }
+}
diff --git a/src/librustdoc/passes/collect_intra_doc_links.rs b/src/librustdoc/passes/collect_intra_doc_links.rs
index fdc1c06..3d6096b 100644
--- a/src/librustdoc/passes/collect_intra_doc_links.rs
+++ b/src/librustdoc/passes/collect_intra_doc_links.rs
@@ -6,7 +6,7 @@
use syntax::ast::{self, Ident, NodeId};
use syntax::feature_gate::UnstableFeatures;
use syntax::symbol::Symbol;
-use syntax_pos::{self, DUMMY_SP};
+use syntax_pos::DUMMY_SP;
use std::ops::Range;
@@ -16,6 +16,7 @@
use clean::*;
use passes::{look_for_tests, Pass};
+use super::span_of_attrs;
pub const COLLECT_INTRA_DOC_LINKS: Pass =
Pass::early("collect-intra-doc-links", collect_intra_doc_links,
@@ -440,28 +441,11 @@
None
}
-pub fn span_of_attrs(attrs: &Attributes) -> syntax_pos::Span {
- if attrs.doc_strings.is_empty() {
- return DUMMY_SP;
- }
- let start = attrs.doc_strings[0].span();
- let end = attrs.doc_strings.last().expect("No doc strings provided").span();
- start.to(end)
-}
-
/// Reports a resolution failure diagnostic.
///
-/// Ideally we can report the diagnostic with the actual span in the source where the link failure
-/// occurred. However, there's a mismatch between the span in the source code and the span in the
-/// markdown, so we have to do a bit of work to figure out the correspondence.
-///
-/// It's not too hard to find the span for sugared doc comments (`///` and `/**`), because the
-/// source will match the markdown exactly, excluding the comment markers. However, it's much more
-/// difficult to calculate the spans for unsugared docs, because we have to deal with escaping and
-/// other source features. So, we attempt to find the exact source span of the resolution failure
-/// in sugared docs, but use the span of the documentation attributes themselves for unsugared
-/// docs. Because this span might be overly large, we display the markdown line containing the
-/// failure as a note.
+/// If we cannot find the exact source span of the resolution failure, we use the span of the
+/// documentation attributes themselves. This is a little heavy-handed, so we display the markdown
+/// line containing the failure as a note as well.
fn resolution_failure(
cx: &DocContext,
attrs: &Attributes,
@@ -473,54 +457,7 @@
let msg = format!("`[{}]` cannot be resolved, ignoring it...", path_str);
let mut diag = if let Some(link_range) = link_range {
- let src = cx.sess().source_map().span_to_snippet(sp);
- let is_all_sugared_doc = attrs.doc_strings.iter().all(|frag| match frag {
- DocFragment::SugaredDoc(..) => true,
- _ => false,
- });
-
- if let (Ok(src), true) = (src, is_all_sugared_doc) {
- // The number of markdown lines up to and including the resolution failure.
- let num_lines = dox[..link_range.start].lines().count();
-
- // We use `split_terminator('\n')` instead of `lines()` when counting bytes to ensure
- // that DOS-style line endings do not cause the spans to be calculated incorrectly.
- let mut src_lines = src.split_terminator('\n');
- let mut md_lines = dox.split_terminator('\n').take(num_lines).peekable();
-
- // The number of bytes from the start of the source span to the resolution failure that
- // are *not* part of the markdown, like comment markers.
- let mut extra_src_bytes = 0;
-
- while let Some(md_line) = md_lines.next() {
- loop {
- let source_line = src_lines
- .next()
- .expect("could not find markdown line in source");
-
- match source_line.find(md_line) {
- Some(offset) => {
- extra_src_bytes += if md_lines.peek().is_some() {
- source_line.len() - md_line.len()
- } else {
- offset
- };
- break;
- }
- None => {
- // Since this is a source line that doesn't include a markdown line,
- // we have to count the newline that we split from earlier.
- extra_src_bytes += source_line.len() + 1;
- }
- }
- }
- }
-
- let sp = sp.from_inner_byte_pos(
- link_range.start + extra_src_bytes,
- link_range.end + extra_src_bytes,
- );
-
+ if let Some(sp) = super::source_span_for_markdown_range(cx, dox, &link_range, attrs) {
let mut diag = cx.tcx.struct_span_lint_node(
lint::builtin::INTRA_DOC_LINK_RESOLUTION_FAILURE,
NodeId::from_u32(0),
diff --git a/src/librustdoc/passes/mod.rs b/src/librustdoc/passes/mod.rs
index e897b9a..c9a3a2c 100644
--- a/src/librustdoc/passes/mod.rs
+++ b/src/librustdoc/passes/mod.rs
@@ -8,6 +8,8 @@
use std::mem;
use std::fmt;
use syntax::ast::NodeId;
+use syntax_pos::{DUMMY_SP, Span};
+use std::ops::Range;
use clean::{self, GetDefId, Item};
use core::{DocContext, DocAccessLevels};
@@ -16,8 +18,6 @@
use html::markdown::{find_testable_code, ErrorCodes, LangString};
-use self::collect_intra_doc_links::span_of_attrs;
-
mod collapse_docs;
pub use self::collapse_docs::COLLAPSE_DOCS;
@@ -45,6 +45,9 @@
mod collect_trait_impls;
pub use self::collect_trait_impls::COLLECT_TRAIT_IMPLS;
+mod check_code_block_syntax;
+pub use self::check_code_block_syntax::CHECK_CODE_BLOCK_SYNTAX;
+
/// Represents a single pass.
#[derive(Copy, Clone)]
pub enum Pass {
@@ -135,6 +138,7 @@
STRIP_PRIV_IMPORTS,
PROPAGATE_DOC_CFG,
COLLECT_INTRA_DOC_LINKS,
+ CHECK_CODE_BLOCK_SYNTAX,
COLLECT_TRAIT_IMPLS,
];
@@ -145,6 +149,7 @@
"strip-hidden",
"strip-private",
"collect-intra-doc-links",
+ "check-code-block-syntax",
"collapse-docs",
"unindent-comments",
"propagate-doc-cfg",
@@ -156,6 +161,7 @@
"check-private-items-doc-tests",
"strip-priv-imports",
"collect-intra-doc-links",
+ "check-code-block-syntax",
"collapse-docs",
"unindent-comments",
"propagate-doc-cfg",
@@ -396,3 +402,94 @@
}
}
}
+
+/// Return a span encompassing all the given attributes.
+crate fn span_of_attrs(attrs: &clean::Attributes) -> Span {
+ if attrs.doc_strings.is_empty() {
+ return DUMMY_SP;
+ }
+ let start = attrs.doc_strings[0].span();
+ let end = attrs.doc_strings.last().expect("No doc strings provided").span();
+ start.to(end)
+}
+
+/// Attempts to match a range of bytes from parsed markdown to a `Span` in the source code.
+///
+/// This method will return `None` if we cannot construct a span from the source map or if the
+/// attributes are not all sugared doc comments. It's difficult to calculate the correct span in
+/// that case due to escaping and other source features.
+crate fn source_span_for_markdown_range(
+ cx: &DocContext,
+ markdown: &str,
+ md_range: &Range<usize>,
+ attrs: &clean::Attributes,
+) -> Option<Span> {
+ let is_all_sugared_doc = attrs.doc_strings.iter().all(|frag| match frag {
+ clean::DocFragment::SugaredDoc(..) => true,
+ _ => false,
+ });
+
+ if !is_all_sugared_doc {
+ return None;
+ }
+
+ let snippet = cx
+ .sess()
+ .source_map()
+ .span_to_snippet(span_of_attrs(attrs))
+ .ok()?;
+
+ let starting_line = markdown[..md_range.start].lines().count() - 1;
+ let ending_line = markdown[..md_range.end].lines().count() - 1;
+
+ // We use `split_terminator('\n')` instead of `lines()` when counting bytes so that we only
+ // we can treat CRLF and LF line endings the same way.
+ let mut src_lines = snippet.split_terminator('\n');
+ let md_lines = markdown.split_terminator('\n');
+
+ // The number of bytes from the source span to the markdown span that are not part
+ // of the markdown, like comment markers.
+ let mut start_bytes = 0;
+ let mut end_bytes = 0;
+
+ 'outer: for (line_no, md_line) in md_lines.enumerate() {
+ loop {
+ let source_line = src_lines.next().expect("could not find markdown in source");
+ match source_line.find(md_line) {
+ Some(offset) => {
+ if line_no == starting_line {
+ start_bytes += offset;
+
+ if starting_line == ending_line {
+ break 'outer;
+ }
+ } else if line_no == ending_line {
+ end_bytes += offset;
+ break 'outer;
+ } else if line_no < starting_line {
+ start_bytes += source_line.len() - md_line.len();
+ } else {
+ end_bytes += source_line.len() - md_line.len();
+ }
+ break;
+ }
+ None => {
+ // Since this is a source line that doesn't include a markdown line,
+ // we have to count the newline that we split from earlier.
+ if line_no <= starting_line {
+ start_bytes += source_line.len() + 1;
+ } else {
+ end_bytes += source_line.len() + 1;
+ }
+ }
+ }
+ }
+ }
+
+ let sp = span_of_attrs(attrs).from_inner_byte_pos(
+ md_range.start + start_bytes,
+ md_range.end + start_bytes + end_bytes,
+ );
+
+ Some(sp)
+}
diff --git a/src/libsyntax/parse/lexer/mod.rs b/src/libsyntax/parse/lexer/mod.rs
index cf51d3e..8827e04 100644
--- a/src/libsyntax/parse/lexer/mod.rs
+++ b/src/libsyntax/parse/lexer/mod.rs
@@ -238,19 +238,6 @@
sr
}
- pub fn new_without_err(sess: &'a ParseSess,
- source_file: Lrc<syntax_pos::SourceFile>,
- override_span: Option<Span>,
- prepend_error_text: &str) -> Result<Self, ()> {
- let mut sr = StringReader::new_raw(sess, source_file, override_span);
- if sr.advance_token().is_err() {
- eprintln!("{}", prepend_error_text);
- sr.emit_fatal_errors();
- return Err(());
- }
- Ok(sr)
- }
-
pub fn new_or_buffered_errs(sess: &'a ParseSess,
source_file: Lrc<syntax_pos::SourceFile>,
override_span: Option<Span>) -> Result<Self, Vec<Diagnostic>> {
diff --git a/src/test/rustdoc-ui/invalid-syntax.rs b/src/test/rustdoc-ui/invalid-syntax.rs
index 537816b..924e038 100644
--- a/src/test/rustdoc-ui/invalid-syntax.rs
+++ b/src/test/rustdoc-ui/invalid-syntax.rs
@@ -1,7 +1,66 @@
// compile-pass
-// compile-flags: --error-format=human
/// ```
/// \__________pkt->size___________/ \_result->size_/ \__pkt->size__/
/// ```
pub fn foo() {}
+
+/// ```
+/// |
+/// LL | use foobar::Baz;
+/// | ^^^^^^ did you mean `baz::foobar`?
+/// ```
+pub fn bar() {}
+
+/// ```
+/// valid
+/// ```
+///
+/// ```
+/// \_
+/// ```
+///
+/// ```text
+/// "invalid
+/// ```
+pub fn valid_and_invalid() {}
+
+/// This is a normal doc comment, but...
+///
+/// There's a code block with bad syntax in it:
+///
+/// ```rust
+/// \_
+/// ```
+///
+/// Good thing we tested it!
+pub fn baz() {}
+
+/// Indented block start
+///
+/// code with bad syntax
+/// \_
+///
+/// Indented block end
+pub fn quux() {}
+
+/// Unclosed fence
+///
+/// ```
+/// slkdjf
+pub fn xyzzy() {}
+
+/// Indented code that contains a fence
+///
+/// ```
+pub fn blah() {}
+
+/// ```edition2018
+/// \_
+/// ```
+pub fn blargh() {}
+
+#[doc = "```"]
+/// \_
+#[doc = "```"]
+pub fn crazy_attrs() {}
diff --git a/src/test/rustdoc-ui/invalid-syntax.stderr b/src/test/rustdoc-ui/invalid-syntax.stderr
index b566133..1080038 100644
--- a/src/test/rustdoc-ui/invalid-syntax.stderr
+++ b/src/test/rustdoc-ui/invalid-syntax.stderr
@@ -1,10 +1,97 @@
-Output from rustc:
-error: unknown start of token: /
- --> <stdin>:1:1
- |
-1 | /__________pkt->size___________/ /_result->size_/ /__pkt->size__/
- | ^
+warning: could not parse code block as Rust code
+ --> $DIR/invalid-syntax.rs:3:5
+ |
+LL | /// ```
+ | _____^
+LL | | /// /__________pkt->size___________/ /_result->size_/ /__pkt->size__/
+LL | | /// ```
+ | |_______^
+ |
+ = note: error from rustc: unknown start of token: /
+help: mark blocks that do not contain Rust code as text
+ |
+LL | /// ```text
+ | ^^^^^^^
-warning: Invalid doc comment starting with: `/__________pkt->size___________/ /_result->size_/ /__pkt->size__/`
-(Ignoring this codeblock)
+warning: could not parse code block as Rust code
+ --> $DIR/invalid-syntax.rs:8:5
+ |
+LL | /// ```
+ | _____^
+LL | | /// |
+LL | | /// LL | use foobar::Baz;
+LL | | /// | ^^^^^^ did you mean `baz::foobar`?
+LL | | /// ```
+ | |_______^
+ |
+ = note: error from rustc: unknown start of token: `
+help: mark blocks that do not contain Rust code as text
+ |
+LL | /// ```text
+ | ^^^^^^^
+
+warning: could not parse code block as Rust code
+ --> $DIR/invalid-syntax.rs:19:5
+ |
+LL | /// ```
+ | _____^
+LL | | /// /_
+LL | | /// ```
+ | |_______^
+ |
+ = note: error from rustc: unknown start of token: /
+help: mark blocks that do not contain Rust code as text
+ |
+LL | /// ```text
+ | ^^^^^^^
+
+warning: could not parse code block as Rust code
+ --> $DIR/invalid-syntax.rs:32:5
+ |
+LL | /// ```rust
+ | _____^
+LL | | /// /_
+LL | | /// ```
+ | |_______^
+ |
+ = note: error from rustc: unknown start of token: /
+
+warning: could not parse code block as Rust code
+ --> $DIR/invalid-syntax.rs:41:9
+ |
+LL | /// code with bad syntax
+ | _________^
+LL | | /// /_
+ | |__________^
+ |
+ = note: error from rustc: unknown start of token: /
+
+warning: could not parse code block as Rust code
+ --> $DIR/invalid-syntax.rs:55:9
+ |
+LL | /// ```
+ | ^^^
+ |
+ = note: error from rustc: unknown start of token: `
+
+warning: could not parse code block as Rust code
+ --> $DIR/invalid-syntax.rs:58:5
+ |
+LL | /// ```edition2018
+ | _____^
+LL | | /// /_
+LL | | /// ```
+ | |_______^
+ |
+ = note: error from rustc: unknown start of token: /
+
+warning: doc comment contains an invalid Rust code block
+ --> $DIR/invalid-syntax.rs:63:1
+ |
+LL | / #[doc = "```"]
+LL | | /// /_
+LL | | #[doc = "```"]
+ | |______________^
+ |
+ = help: mark blocks that do not contain Rust code as text: ```text
diff --git a/src/test/rustdoc/bad-codeblock-syntax.rs b/src/test/rustdoc/bad-codeblock-syntax.rs
new file mode 100644
index 0000000..0ab2f68
--- /dev/null
+++ b/src/test/rustdoc/bad-codeblock-syntax.rs
@@ -0,0 +1,27 @@
+// @has bad_codeblock_syntax/fn.foo.html
+// @has - '//*[@class="docblock"]/pre/code' '\_'
+/// ```
+/// \_
+/// ```
+pub fn foo() {}
+
+// @has bad_codeblock_syntax/fn.bar.html
+// @has - '//*[@class="docblock"]/pre/code' '`baz::foobar`'
+/// ```
+/// `baz::foobar`
+/// ```
+pub fn bar() {}
+
+// @has bad_codeblock_syntax/fn.quux.html
+// @has - '//*[@class="docblock"]/pre/code' '\_'
+/// ```rust
+/// \_
+/// ```
+pub fn quux() {}
+
+// @has bad_codeblock_syntax/fn.ok.html
+// @has - '//*[@class="docblock"]/pre/code[@class="language-text"]' '\_'
+/// ```text
+/// \_
+/// ```
+pub fn ok() {}
