From 6887202ea3a1d3e3df0c88c07c754defd87b9712 Mon Sep 17 00:00:00 2001
From: mitaa <mitaa.ceb@gmail.com>
Date: Mon, 25 Apr 2016 17:20:32 +0200
Subject: [PATCH] Make some fatal lexer errors recoverable

---
 src/librustdoc/html/highlight.rs   |  26 +++--
 src/libsyntax/errors/mod.rs        |   2 +
 src/libsyntax/ext/tt/transcribe.rs |   4 +-
 src/libsyntax/parse/lexer/mod.rs   | 169 ++++++++++++++++++-----------
 src/test/rustdoc/issue-12834.rs    |  21 ++++
 5 files changed, 148 insertions(+), 74 deletions(-)
 create mode 100644 src/test/rustdoc/issue-12834.rs

diff --git a/src/librustdoc/html/highlight.rs b/src/librustdoc/html/highlight.rs
index 2f0ae540fbf07..789239c3ab020 100644
--- a/src/librustdoc/html/highlight.rs
+++ b/src/librustdoc/html/highlight.rs
@@ -29,9 +29,11 @@ pub fn render_with_highlighting(src: &str, class: Option<&str>, id: Option<&str>
 
     let mut out = Vec::new();
     write_header(class, id, &mut out).unwrap();
-    write_source(&sess,
-                 lexer::StringReader::new(&sess.span_diagnostic, fm),
-                 &mut out).unwrap();
+    if let Err(_) = write_source(&sess,
+                                 lexer::StringReader::new(&sess.span_diagnostic, fm),
+                                 &mut out) {
+        return format!("<pre>{}</pre>", src)
+    }
     write_footer(&mut out).unwrap();
     String::from_utf8_lossy(&out[..]).into_owned()
 }
@@ -39,15 +41,15 @@ pub fn render_with_highlighting(src: &str, class: Option<&str>, id: Option<&str>
 /// Highlights `src`, returning the HTML output. Returns only the inner html to
 /// be inserted into an element. C.f., `render_with_highlighting` which includes
 /// an enclosing `<pre>` block.
-pub fn render_inner_with_highlighting(src: &str) -> String {
+pub fn render_inner_with_highlighting(src: &str) -> io::Result<String> {
     let sess = parse::ParseSess::new();
     let fm = sess.codemap().new_filemap("<stdin>".to_string(), src.to_string());
 
     let mut out = Vec::new();
     write_source(&sess,
                  lexer::StringReader::new(&sess.span_diagnostic, fm),
-                 &mut out).unwrap();
-    String::from_utf8_lossy(&out[..]).into_owned()
+                 &mut out)?;
+    Ok(String::from_utf8_lossy(&out[..]).into_owned())
 }
 
 /// Exhausts the `lexer` writing the output into `out`.
@@ -65,7 +67,17 @@ fn write_source(sess: &parse::ParseSess,
     let mut is_macro = false;
     let mut is_macro_nonterminal = false;
     loop {
-        let next = lexer.next_token();
+        let next = match lexer.try_next_token() {
+            Ok(tok) => tok,
+            Err(_) => {
+                lexer.emit_fatal_errors();
+                lexer.span_diagnostic.struct_warn("Backing out of syntax highlighting")
+                                     .note("You probably did not intend to render this \
+                                            as a rust code-block")
+                                     .emit();
+                return Err(io::Error::new(io::ErrorKind::Other, ""))
+            },
+        };
 
         let snip = |sp| sess.codemap().span_to_snippet(sp).unwrap();
 
diff --git a/src/libsyntax/errors/mod.rs b/src/libsyntax/errors/mod.rs
index acf65fe15b154..792828b3054e9 100644
--- a/src/libsyntax/errors/mod.rs
+++ b/src/libsyntax/errors/mod.rs
@@ -177,6 +177,7 @@ impl error::Error for ExplicitBug {
 
 /// Used for emitting structured error messages and other diagnostic information.
 #[must_use]
+#[derive(Clone)]
 pub struct DiagnosticBuilder<'a> {
     emitter: &'a RefCell<Box<Emitter>>,
     level: Level,
@@ -187,6 +188,7 @@ pub struct DiagnosticBuilder<'a> {
 }
 
 /// For example a note attached to an error.
+#[derive(Clone)]
 struct SubDiagnostic {
     level: Level,
     message: String,
diff --git a/src/libsyntax/ext/tt/transcribe.rs b/src/libsyntax/ext/tt/transcribe.rs
index 7f53d0f412cca..6b3b5ce9de914 100644
--- a/src/libsyntax/ext/tt/transcribe.rs
+++ b/src/libsyntax/ext/tt/transcribe.rs
@@ -12,7 +12,7 @@ use self::LockstepIterSize::*;
 use ast;
 use ast::{TokenTree, Ident, Name};
 use codemap::{Span, DUMMY_SP};
-use errors::Handler;
+use errors::{Handler, DiagnosticBuilder};
 use ext::tt::macro_parser::{NamedMatch, MatchedSeq, MatchedNonterminal};
 use parse::token::{DocComment, MatchNt, SubstNt};
 use parse::token::{Token, NtIdent, SpecialMacroVar};
@@ -50,6 +50,7 @@ pub struct TtReader<'a> {
     pub cur_span: Span,
     /// Transform doc comments. Only useful in macro invocations
     pub desugar_doc_comments: bool,
+    pub fatal_errs: Vec<DiagnosticBuilder<'a>>,
 }
 
 /// This can do Macro-By-Example transcription. On the other hand, if
@@ -99,6 +100,7 @@ pub fn new_tt_reader_with_doc_flag(sp_diag: &Handler,
         /* dummy values, never read: */
         cur_tok: token::Eof,
         cur_span: DUMMY_SP,
+        fatal_errs: Vec::new(),
     };
     tt_next_token(&mut r); /* get cur_tok and cur_span set up */
     r
diff --git a/src/libsyntax/parse/lexer/mod.rs b/src/libsyntax/parse/lexer/mod.rs
index 2eda13adcb580..881663a056c78 100644
--- a/src/libsyntax/parse/lexer/mod.rs
+++ b/src/libsyntax/parse/lexer/mod.rs
@@ -29,24 +29,42 @@ mod unicode_chars;
 
 pub trait Reader {
     fn is_eof(&self) -> bool;
-    fn next_token(&mut self) -> TokenAndSpan;
+    fn try_next_token(&mut self) -> Result<TokenAndSpan, ()>;
+    fn next_token(&mut self) -> TokenAndSpan where Self: Sized {
+        let res = self.try_next_token();
+        self.unwrap_or_abort(res)
+    }
     /// Report a fatal error with the current span.
     fn fatal(&self, &str) -> FatalError;
     /// Report a non-fatal error with the current span.
     fn err(&self, &str);
+    fn emit_fatal_errors(&mut self);
+    fn unwrap_or_abort(&mut self, res: Result<TokenAndSpan, ()>) -> TokenAndSpan {
+        match res {
+            Ok(tok) => tok,
+            Err(_) => {
+                self.emit_fatal_errors();
+                panic!(FatalError);
+            }
+        }
+    }
     fn peek(&self) -> TokenAndSpan;
     /// Get a token the parser cares about.
-    fn real_token(&mut self) -> TokenAndSpan {
-        let mut t = self.next_token();
+    fn try_real_token(&mut self) -> Result<TokenAndSpan, ()> {
+        let mut t = self.try_next_token()?;
         loop {
             match t.tok {
                 token::Whitespace | token::Comment | token::Shebang(_) => {
-                    t = self.next_token();
+                    t = self.try_next_token()?;
                 }
                 _ => break,
             }
         }
-        t
+        Ok(t)
+    }
+    fn real_token(&mut self) -> TokenAndSpan {
+        let res = self.try_real_token();
+        self.unwrap_or_abort(res)
     }
 }
 
@@ -70,7 +88,7 @@ pub struct StringReader<'a> {
     // cached:
     pub peek_tok: token::Token,
     pub peek_span: Span,
-
+    pub fatal_errs: Vec<DiagnosticBuilder<'a>>,
     // cache a direct reference to the source text, so that we don't have to
     // retrieve it via `self.filemap.src.as_ref().unwrap()` all the time.
     source_text: Rc<String>,
@@ -81,13 +99,14 @@ impl<'a> Reader for StringReader<'a> {
         self.curr.is_none()
     }
     /// Return the next token. EFFECT: advances the string_reader.
-    fn next_token(&mut self) -> TokenAndSpan {
+    fn try_next_token(&mut self) -> Result<TokenAndSpan, ()> {
+        assert!(self.fatal_errs.is_empty());
         let ret_val = TokenAndSpan {
             tok: replace(&mut self.peek_tok, token::Underscore),
             sp: self.peek_span,
         };
-        self.advance_token();
-        ret_val
+        self.advance_token()?;
+        Ok(ret_val)
     }
     fn fatal(&self, m: &str) -> FatalError {
         self.fatal_span(self.peek_span, m)
@@ -95,6 +114,12 @@ impl<'a> Reader for StringReader<'a> {
     fn err(&self, m: &str) {
         self.err_span(self.peek_span, m)
     }
+    fn emit_fatal_errors(&mut self) {
+        for err in &mut self.fatal_errs {
+            err.emit();
+        }
+        self.fatal_errs.clear();
+    }
     fn peek(&self) -> TokenAndSpan {
         // FIXME(pcwalton): Bad copy!
         TokenAndSpan {
@@ -108,10 +133,11 @@ impl<'a> Reader for TtReader<'a> {
     fn is_eof(&self) -> bool {
         self.cur_tok == token::Eof
     }
-    fn next_token(&mut self) -> TokenAndSpan {
+    fn try_next_token(&mut self) -> Result<TokenAndSpan, ()> {
+        assert!(self.fatal_errs.is_empty());
         let r = tt_next_token(self);
         debug!("TtReader: r={:?}", r);
-        r
+        Ok(r)
     }
     fn fatal(&self, m: &str) -> FatalError {
         self.sp_diag.span_fatal(self.cur_span, m)
@@ -119,6 +145,12 @@ impl<'a> Reader for TtReader<'a> {
     fn err(&self, m: &str) {
         self.sp_diag.span_err(self.cur_span, m);
     }
+    fn emit_fatal_errors(&mut self) {
+        for err in &mut self.fatal_errs {
+            err.emit();
+        }
+        self.fatal_errs.clear();
+    }
     fn peek(&self) -> TokenAndSpan {
         TokenAndSpan {
             tok: self.cur_tok.clone(),
@@ -151,6 +183,7 @@ impl<'a> StringReader<'a> {
             peek_tok: token::Eof,
             peek_span: codemap::DUMMY_SP,
             source_text: source_text,
+            fatal_errs: Vec::new(),
         };
         sr.bump();
         sr
@@ -160,7 +193,10 @@ impl<'a> StringReader<'a> {
                    filemap: Rc<codemap::FileMap>)
                    -> StringReader<'b> {
         let mut sr = StringReader::new_raw(span_diagnostic, filemap);
-        sr.advance_token();
+        if let Err(_) = sr.advance_token() {
+            sr.emit_fatal_errors();
+            panic!(FatalError);
+        }
         sr
     }
 
@@ -249,7 +285,7 @@ impl<'a> StringReader<'a> {
 
     /// Advance peek_tok and peek_span to refer to the next token, and
     /// possibly update the interner.
-    fn advance_token(&mut self) {
+    fn advance_token(&mut self) -> Result<(), ()> {
         match self.scan_whitespace_or_comment() {
             Some(comment) => {
                 self.peek_span = comment.sp;
@@ -261,11 +297,12 @@ impl<'a> StringReader<'a> {
                     self.peek_span = codemap::mk_sp(self.filemap.end_pos, self.filemap.end_pos);
                 } else {
                     let start_bytepos = self.last_pos;
-                    self.peek_tok = self.next_token_inner();
+                    self.peek_tok = self.next_token_inner()?;
                     self.peek_span = codemap::mk_sp(start_bytepos, self.last_pos);
                 };
             }
         }
+        Ok(())
     }
 
     fn byte_offset(&self, pos: BytePos) -> BytePos {
@@ -1013,7 +1050,7 @@ impl<'a> StringReader<'a> {
 
     /// Return the next token from the string, advances the input past that
     /// token, and updates the interner
-    fn next_token_inner(&mut self) -> token::Token {
+    fn next_token_inner(&mut self) -> Result<token::Token, ()> {
         let c = self.curr;
         if ident_start(c) &&
            match (c.unwrap(), self.nextch(), self.nextnextch()) {
@@ -1033,32 +1070,32 @@ impl<'a> StringReader<'a> {
                 self.bump();
             }
 
-            return self.with_str_from(start, |string| {
+            return Ok(self.with_str_from(start, |string| {
                 if string == "_" {
                     token::Underscore
                 } else {
                     // FIXME: perform NFKC normalization here. (Issue #2253)
                     token::Ident(str_to_ident(string))
                 }
-            });
+            }));
         }
 
         if is_dec_digit(c) {
             let num = self.scan_number(c.unwrap());
             let suffix = self.scan_optional_raw_name();
             debug!("next_token_inner: scanned number {:?}, {:?}", num, suffix);
-            return token::Literal(num, suffix);
+            return Ok(token::Literal(num, suffix));
         }
 
         match c.expect("next_token_inner called at EOF") {
             // One-byte tokens.
             ';' => {
                 self.bump();
-                return token::Semi;
+                return Ok(token::Semi);
             }
             ',' => {
                 self.bump();
-                return token::Comma;
+                return Ok(token::Comma);
             }
             '.' => {
                 self.bump();
@@ -1066,67 +1103,67 @@ impl<'a> StringReader<'a> {
                     self.bump();
                     if self.curr_is('.') {
                         self.bump();
-                        token::DotDotDot
+                        Ok(token::DotDotDot)
                     } else {
-                        token::DotDot
+                        Ok(token::DotDot)
                     }
                 } else {
-                    token::Dot
+                    Ok(token::Dot)
                 };
             }
             '(' => {
                 self.bump();
-                return token::OpenDelim(token::Paren);
+                return Ok(token::OpenDelim(token::Paren));
             }
             ')' => {
                 self.bump();
-                return token::CloseDelim(token::Paren);
+                return Ok(token::CloseDelim(token::Paren));
             }
             '{' => {
                 self.bump();
-                return token::OpenDelim(token::Brace);
+                return Ok(token::OpenDelim(token::Brace));
             }
             '}' => {
                 self.bump();
-                return token::CloseDelim(token::Brace);
+                return Ok(token::CloseDelim(token::Brace));
             }
             '[' => {
                 self.bump();
-                return token::OpenDelim(token::Bracket);
+                return Ok(token::OpenDelim(token::Bracket));
             }
             ']' => {
                 self.bump();
-                return token::CloseDelim(token::Bracket);
+                return Ok(token::CloseDelim(token::Bracket));
             }
             '@' => {
                 self.bump();
-                return token::At;
+                return Ok(token::At);
             }
             '#' => {
                 self.bump();
-                return token::Pound;
+                return Ok(token::Pound);
             }
             '~' => {
                 self.bump();
-                return token::Tilde;
+                return Ok(token::Tilde);
             }
             '?' => {
                 self.bump();
-                return token::Question;
+                return Ok(token::Question);
             }
             ':' => {
                 self.bump();
                 if self.curr_is(':') {
                     self.bump();
-                    return token::ModSep;
+                    return Ok(token::ModSep);
                 } else {
-                    return token::Colon;
+                    return Ok(token::Colon);
                 }
             }
 
             '$' => {
                 self.bump();
-                return token::Dollar;
+                return Ok(token::Dollar);
             }
 
             // Multi-byte tokens.
@@ -1134,21 +1171,21 @@ impl<'a> StringReader<'a> {
                 self.bump();
                 if self.curr_is('=') {
                     self.bump();
-                    return token::EqEq;
+                    return Ok(token::EqEq);
                 } else if self.curr_is('>') {
                     self.bump();
-                    return token::FatArrow;
+                    return Ok(token::FatArrow);
                 } else {
-                    return token::Eq;
+                    return Ok(token::Eq);
                 }
             }
             '!' => {
                 self.bump();
                 if self.curr_is('=') {
                     self.bump();
-                    return token::Ne;
+                    return Ok(token::Ne);
                 } else {
-                    return token::Not;
+                    return Ok(token::Not);
                 }
             }
             '<' => {
@@ -1156,21 +1193,21 @@ impl<'a> StringReader<'a> {
                 match self.curr.unwrap_or('\x00') {
                     '=' => {
                         self.bump();
-                        return token::Le;
+                        return Ok(token::Le);
                     }
                     '<' => {
-                        return self.binop(token::Shl);
+                        return Ok(self.binop(token::Shl));
                     }
                     '-' => {
                         self.bump();
                         match self.curr.unwrap_or('\x00') {
                             _ => {
-                                return token::LArrow;
+                                return Ok(token::LArrow);
                             }
                         }
                     }
                     _ => {
-                        return token::Lt;
+                        return Ok(token::Lt);
                     }
                 }
             }
@@ -1179,13 +1216,13 @@ impl<'a> StringReader<'a> {
                 match self.curr.unwrap_or('\x00') {
                     '=' => {
                         self.bump();
-                        return token::Ge;
+                        return Ok(token::Ge);
                     }
                     '>' => {
-                        return self.binop(token::Shr);
+                        return Ok(self.binop(token::Shr));
                     }
                     _ => {
-                        return token::Gt;
+                        return Ok(token::Gt);
                     }
                 }
             }
@@ -1233,7 +1270,7 @@ impl<'a> StringReader<'a> {
                         self.err_span_(start, last_bpos, "lifetimes cannot use keyword names");
                     }
 
-                    return token::Lifetime(ident);
+                    return Ok(token::Lifetime(ident));
                 }
 
                 let valid = self.scan_char_or_byte(start,
@@ -1255,7 +1292,7 @@ impl<'a> StringReader<'a> {
                 };
                 self.bump(); // advance curr past token
                 let suffix = self.scan_optional_raw_name();
-                return token::Literal(token::Char(id), suffix);
+                return Ok(token::Literal(token::Char(id), suffix));
             }
             'b' => {
                 self.bump();
@@ -1266,7 +1303,7 @@ impl<'a> StringReader<'a> {
                     _ => unreachable!(),  // Should have been a token::Ident above.
                 };
                 let suffix = self.scan_optional_raw_name();
-                return token::Literal(lit, suffix);
+                return Ok(token::Literal(lit, suffix));
             }
             '"' => {
                 let start_bpos = self.last_pos;
@@ -1297,7 +1334,7 @@ impl<'a> StringReader<'a> {
                 };
                 self.bump();
                 let suffix = self.scan_optional_raw_name();
-                return token::Literal(token::Str_(id), suffix);
+                return Ok(token::Literal(token::Str_(id), suffix));
             }
             'r' => {
                 let start_bpos = self.last_pos;
@@ -1368,24 +1405,24 @@ impl<'a> StringReader<'a> {
                     token::intern("??")
                 };
                 let suffix = self.scan_optional_raw_name();
-                return token::Literal(token::StrRaw(id, hash_count), suffix);
+                return Ok(token::Literal(token::StrRaw(id, hash_count), suffix));
             }
             '-' => {
                 if self.nextch_is('>') {
                     self.bump();
                     self.bump();
-                    return token::RArrow;
+                    return Ok(token::RArrow);
                 } else {
-                    return self.binop(token::Minus);
+                    return Ok(self.binop(token::Minus));
                 }
             }
             '&' => {
                 if self.nextch_is('&') {
                     self.bump();
                     self.bump();
-                    return token::AndAnd;
+                    return Ok(token::AndAnd);
                 } else {
-                    return self.binop(token::And);
+                    return Ok(self.binop(token::And));
                 }
             }
             '|' => {
@@ -1393,27 +1430,27 @@ impl<'a> StringReader<'a> {
                     Some('|') => {
                         self.bump();
                         self.bump();
-                        return token::OrOr;
+                        return Ok(token::OrOr);
                     }
                     _ => {
-                        return self.binop(token::Or);
+                        return Ok(self.binop(token::Or));
                     }
                 }
             }
             '+' => {
-                return self.binop(token::Plus);
+                return Ok(self.binop(token::Plus));
             }
             '*' => {
-                return self.binop(token::Star);
+                return Ok(self.binop(token::Star));
             }
             '/' => {
-                return self.binop(token::Slash);
+                return Ok(self.binop(token::Slash));
             }
             '^' => {
-                return self.binop(token::Caret);
+                return Ok(self.binop(token::Caret));
             }
             '%' => {
-                return self.binop(token::Percent);
+                return Ok(self.binop(token::Percent));
             }
             c => {
                 let last_bpos = self.last_pos;
@@ -1423,8 +1460,8 @@ impl<'a> StringReader<'a> {
                                                           "unknown start of token",
                                                           c);
                 unicode_chars::check_for_substitution(&self, c, &mut err);
-                err.emit();
-                panic!(FatalError);
+                self.fatal_errs.push(err);
+                Err(())
             }
         }
     }
diff --git a/src/test/rustdoc/issue-12834.rs b/src/test/rustdoc/issue-12834.rs
new file mode 100644
index 0000000000000..30dce27e73894
--- /dev/null
+++ b/src/test/rustdoc/issue-12834.rs
@@ -0,0 +1,21 @@
+// Copyright 2016 The Rust Project Developers. See the COPYRIGHT
+// file at the top-level directory of this distribution and at
+// http://rust-lang.org/COPYRIGHT.
+//
+// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
+// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
+// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
+// option. This file may not be copied, modified, or distributed
+// except according to those terms.
+
+// Tests that failing to syntax highlight a rust code-block doesn't cause
+// rustdoc to fail, while still rendering the code-block (without highlighting).
+
+
+// @has issue_12834/fn.foo.html
+// @has - //pre 'a + b '
+
+/// ```
+/// a + b ∈ Self ∀ a, b ∈ Self
+/// ```
+pub fn foo() {}