about summary refs log tree commit diff
path: root/compiler
diff options
context:
space:
mode:
authorEsteban Küber <esteban@kuber.com.ar>2024-07-09 17:00:19 +0000
committerEsteban Küber <esteban@kuber.com.ar>2024-07-18 20:08:38 +0000
commit2d7795dfb942d49d447837ef43db89216de15696 (patch)
treef1ef42e40392a1a0e1b7114662492507309e526e /compiler
parent89f273f40dafb693139496ed6f914872b6533fa6 (diff)
downloadrust-2d7795dfb942d49d447837ef43db89216de15696.tar.gz
rust-2d7795dfb942d49d447837ef43db89216de15696.zip
Be more accurate about calculating `display_col` from a `BytePos`
No longer track "zero-width" chars in `SourceMap`, read directly from the line when calculating the `display_col` of a `BytePos`. Move `char_width` to `rustc_span` and use it from the emitter.

This change allows the following to properly align in terminals (depending on the font, the replaced control codepoints are rendered as 1 or 2 width, on my terminal they are rendered as 1, on VSCode text they are rendered as 2):

```
error: this file contains an unclosed delimiter
  --> $DIR/issue-68629.rs:5:17
   |
LL | ␜␟ts␀![{i
   |       -- unclosed delimiter
   |       |
   |       unclosed delimiter
LL | ␀␀  fn rݻoa>rݻm
   |                ^
```
Diffstat (limited to 'compiler')
-rw-r--r--compiler/rustc_errors/Cargo.toml1
-rw-r--r--compiler/rustc_errors/src/emitter.rs17
-rw-r--r--compiler/rustc_metadata/src/rmeta/decoder.rs2
-rw-r--r--compiler/rustc_query_system/src/ich/impls_syntax.rs6
-rw-r--r--compiler/rustc_span/src/analyze_source_file.rs40
-rw-r--r--compiler/rustc_span/src/analyze_source_file/tests.rs24
-rw-r--r--compiler/rustc_span/src/lib.rs133
-rw-r--r--compiler/rustc_span/src/source_map.rs2
-rw-r--r--compiler/rustc_span/src/source_map/tests.rs2
9 files changed, 42 insertions, 185 deletions
diff --git a/compiler/rustc_errors/Cargo.toml b/compiler/rustc_errors/Cargo.toml
index cc114fdcd8c..2fff9f2de50 100644
--- a/compiler/rustc_errors/Cargo.toml
+++ b/compiler/rustc_errors/Cargo.toml
@@ -26,7 +26,6 @@ serde_json = "1.0.59"
 termcolor = "1.2.0"
 termize = "0.1.1"
 tracing = "0.1"
-unicode-width = "0.1.4"
 # tidy-alphabetical-end
 
 [target.'cfg(windows)'.dependencies.windows]
diff --git a/compiler/rustc_errors/src/emitter.rs b/compiler/rustc_errors/src/emitter.rs
index 95e1b5348b7..16fa0ff7a2d 100644
--- a/compiler/rustc_errors/src/emitter.rs
+++ b/compiler/rustc_errors/src/emitter.rs
@@ -8,7 +8,7 @@
 //! The output types are defined in `rustc_session::config::ErrorOutputType`.
 
 use rustc_span::source_map::SourceMap;
-use rustc_span::{FileLines, FileName, SourceFile, Span};
+use rustc_span::{char_width, FileLines, FileName, SourceFile, Span};
 
 use crate::snippet::{
     Annotation, AnnotationColumn, AnnotationType, Line, MultilineAnnotation, Style, StyledString,
@@ -2614,21 +2614,6 @@ fn normalize_whitespace(str: &str) -> String {
     s
 }
 
-fn char_width(ch: char) -> usize {
-    // FIXME: `unicode_width` sometimes disagrees with terminals on how wide a `char` is. For now,
-    // just accept that sometimes the code line will be longer than desired.
-    match ch {
-        '\t' => 4,
-        '\u{0000}' | '\u{0001}' | '\u{0002}' | '\u{0003}' | '\u{0004}' | '\u{0005}'
-        | '\u{0006}' | '\u{0007}' | '\u{0008}' | '\u{000B}' | '\u{000C}' | '\u{000D}'
-        | '\u{000E}' | '\u{000F}' | '\u{0010}' | '\u{0011}' | '\u{0012}' | '\u{0013}'
-        | '\u{0014}' | '\u{0015}' | '\u{0016}' | '\u{0017}' | '\u{0018}' | '\u{0019}'
-        | '\u{001A}' | '\u{001B}' | '\u{001C}' | '\u{001D}' | '\u{001E}' | '\u{001F}'
-        | '\u{007F}' => 1,
-        _ => unicode_width::UnicodeWidthChar::width(ch).unwrap_or(1),
-    }
-}
-
 fn draw_col_separator(buffer: &mut StyledBuffer, line: usize, col: usize) {
     buffer.puts(line, col, "| ", Style::LineNumber);
 }
diff --git a/compiler/rustc_metadata/src/rmeta/decoder.rs b/compiler/rustc_metadata/src/rmeta/decoder.rs
index 0ba9b940eed..9f2fe254753 100644
--- a/compiler/rustc_metadata/src/rmeta/decoder.rs
+++ b/compiler/rustc_metadata/src/rmeta/decoder.rs
@@ -1727,7 +1727,6 @@ impl<'a, 'tcx> CrateMetadataRef<'a> {
                     source_len,
                     lines,
                     multibyte_chars,
-                    non_narrow_chars,
                     normalized_pos,
                     stable_id,
                     ..
@@ -1779,7 +1778,6 @@ impl<'a, 'tcx> CrateMetadataRef<'a> {
                     self.cnum,
                     lines,
                     multibyte_chars,
-                    non_narrow_chars,
                     normalized_pos,
                     source_file_index,
                 );
diff --git a/compiler/rustc_query_system/src/ich/impls_syntax.rs b/compiler/rustc_query_system/src/ich/impls_syntax.rs
index 5bd4fe04848..39da5e395c4 100644
--- a/compiler/rustc_query_system/src/ich/impls_syntax.rs
+++ b/compiler/rustc_query_system/src/ich/impls_syntax.rs
@@ -73,7 +73,6 @@ impl<'a> HashStable<StableHashingContext<'a>> for SourceFile {
             source_len: _,
             lines: _,
             ref multibyte_chars,
-            ref non_narrow_chars,
             ref normalized_pos,
         } = *self;
 
@@ -98,11 +97,6 @@ impl<'a> HashStable<StableHashingContext<'a>> for SourceFile {
             char_pos.hash_stable(hcx, hasher);
         }
 
-        non_narrow_chars.len().hash_stable(hcx, hasher);
-        for &char_pos in non_narrow_chars.iter() {
-            char_pos.hash_stable(hcx, hasher);
-        }
-
         normalized_pos.len().hash_stable(hcx, hasher);
         for &char_pos in normalized_pos.iter() {
             char_pos.hash_stable(hcx, hasher);
diff --git a/compiler/rustc_span/src/analyze_source_file.rs b/compiler/rustc_span/src/analyze_source_file.rs
index d9e1ebaf0bc..ba7e0cec5bd 100644
--- a/compiler/rustc_span/src/analyze_source_file.rs
+++ b/compiler/rustc_span/src/analyze_source_file.rs
@@ -1,5 +1,4 @@
 use super::*;
-use unicode_width::UnicodeWidthChar;
 
 #[cfg(test)]
 mod tests;
@@ -9,15 +8,12 @@ mod tests;
 ///
 /// This function will use an SSE2 enhanced implementation if hardware support
 /// is detected at runtime.
-pub fn analyze_source_file(
-    src: &str,
-) -> (Vec<RelativeBytePos>, Vec<MultiByteChar>, Vec<NonNarrowChar>) {
+pub fn analyze_source_file(src: &str) -> (Vec<RelativeBytePos>, Vec<MultiByteChar>) {
     let mut lines = vec![RelativeBytePos::from_u32(0)];
     let mut multi_byte_chars = vec![];
-    let mut non_narrow_chars = vec![];
 
     // Calls the right implementation, depending on hardware support available.
-    analyze_source_file_dispatch(src, &mut lines, &mut multi_byte_chars, &mut non_narrow_chars);
+    analyze_source_file_dispatch(src, &mut lines, &mut multi_byte_chars);
 
     // The code above optimistically registers a new line *after* each \n
     // it encounters. If that point is already outside the source_file, remove
@@ -30,7 +26,7 @@ pub fn analyze_source_file(
         }
     }
 
-    (lines, multi_byte_chars, non_narrow_chars)
+    (lines, multi_byte_chars)
 }
 
 cfg_match! {
@@ -39,11 +35,10 @@ cfg_match! {
             src: &str,
             lines: &mut Vec<RelativeBytePos>,
             multi_byte_chars: &mut Vec<MultiByteChar>,
-            non_narrow_chars: &mut Vec<NonNarrowChar>,
         ) {
             if is_x86_feature_detected!("sse2") {
                 unsafe {
-                    analyze_source_file_sse2(src, lines, multi_byte_chars, non_narrow_chars);
+                    analyze_source_file_sse2(src, lines, multi_byte_chars);
                 }
             } else {
                 analyze_source_file_generic(
@@ -52,7 +47,6 @@ cfg_match! {
                     RelativeBytePos::from_u32(0),
                     lines,
                     multi_byte_chars,
-                    non_narrow_chars,
                 );
             }
         }
@@ -66,7 +60,6 @@ cfg_match! {
             src: &str,
             lines: &mut Vec<RelativeBytePos>,
             multi_byte_chars: &mut Vec<MultiByteChar>,
-            non_narrow_chars: &mut Vec<NonNarrowChar>,
         ) {
             #[cfg(target_arch = "x86")]
             use std::arch::x86::*;
@@ -159,7 +152,6 @@ cfg_match! {
                     RelativeBytePos::from_usize(scan_start),
                     lines,
                     multi_byte_chars,
-                    non_narrow_chars,
                 );
             }
 
@@ -172,7 +164,6 @@ cfg_match! {
                     RelativeBytePos::from_usize(tail_start),
                     lines,
                     multi_byte_chars,
-                    non_narrow_chars,
                 );
             }
         }
@@ -183,7 +174,6 @@ cfg_match! {
             src: &str,
             lines: &mut Vec<RelativeBytePos>,
             multi_byte_chars: &mut Vec<MultiByteChar>,
-            non_narrow_chars: &mut Vec<NonNarrowChar>,
         ) {
             analyze_source_file_generic(
                 src,
@@ -191,7 +181,6 @@ cfg_match! {
                 RelativeBytePos::from_u32(0),
                 lines,
                 multi_byte_chars,
-                non_narrow_chars,
             );
         }
     }
@@ -205,7 +194,6 @@ fn analyze_source_file_generic(
     output_offset: RelativeBytePos,
     lines: &mut Vec<RelativeBytePos>,
     multi_byte_chars: &mut Vec<MultiByteChar>,
-    non_narrow_chars: &mut Vec<NonNarrowChar>,
 ) -> usize {
     assert!(src.len() >= scan_len);
     let mut i = 0;
@@ -227,16 +215,8 @@ fn analyze_source_file_generic(
 
             let pos = RelativeBytePos::from_usize(i) + output_offset;
 
-            match byte {
-                b'\n' => {
-                    lines.push(pos + RelativeBytePos(1));
-                }
-                b'\t' => {
-                    non_narrow_chars.push(NonNarrowChar::Tab(pos));
-                }
-                _ => {
-                    non_narrow_chars.push(NonNarrowChar::ZeroWidth(pos));
-                }
+            if let b'\n' = byte {
+                lines.push(pos + RelativeBytePos(1));
             }
         } else if byte >= 127 {
             // The slow path:
@@ -252,14 +232,6 @@ fn analyze_source_file_generic(
                 let mbc = MultiByteChar { pos, bytes: char_len as u8 };
                 multi_byte_chars.push(mbc);
             }
-
-            // Assume control characters are zero width.
-            // FIXME: How can we decide between `width` and `width_cjk`?
-            let char_width = UnicodeWidthChar::width(c).unwrap_or(0);
-
-            if char_width != 1 {
-                non_narrow_chars.push(NonNarrowChar::new(pos, char_width));
-            }
         }
 
         i += char_len;
diff --git a/compiler/rustc_span/src/analyze_source_file/tests.rs b/compiler/rustc_span/src/analyze_source_file/tests.rs
index 0c77d080c17..e4a24239d8e 100644
--- a/compiler/rustc_span/src/analyze_source_file/tests.rs
+++ b/compiler/rustc_span/src/analyze_source_file/tests.rs
@@ -4,11 +4,10 @@ macro_rules! test {
     (case: $test_name:ident,
      text: $text:expr,
      lines: $lines:expr,
-     multi_byte_chars: $multi_byte_chars:expr,
-     non_narrow_chars: $non_narrow_chars:expr,) => {
+     multi_byte_chars: $multi_byte_chars:expr,) => {
         #[test]
         fn $test_name() {
-            let (lines, multi_byte_chars, non_narrow_chars) = analyze_source_file($text);
+            let (lines, multi_byte_chars) = analyze_source_file($text);
 
             let expected_lines: Vec<RelativeBytePos> =
                 $lines.into_iter().map(RelativeBytePos).collect();
@@ -21,13 +20,6 @@ macro_rules! test {
                 .collect();
 
             assert_eq!(multi_byte_chars, expected_mbcs);
-
-            let expected_nncs: Vec<NonNarrowChar> = $non_narrow_chars
-                .into_iter()
-                .map(|(pos, width)| NonNarrowChar::new(RelativeBytePos(pos), width))
-                .collect();
-
-            assert_eq!(non_narrow_chars, expected_nncs);
         }
     };
 }
@@ -37,7 +29,6 @@ test!(
     text: "",
     lines: vec![],
     multi_byte_chars: vec![],
-    non_narrow_chars: vec![],
 );
 
 test!(
@@ -45,7 +36,6 @@ test!(
     text: "a\nc",
     lines: vec![0, 2],
     multi_byte_chars: vec![],
-    non_narrow_chars: vec![],
 );
 
 test!(
@@ -53,7 +43,6 @@ test!(
     text: "012345678\nabcdef012345678\na",
     lines: vec![0, 10, 26],
     multi_byte_chars: vec![],
-    non_narrow_chars: vec![],
 );
 
 test!(
@@ -61,7 +50,6 @@ test!(
     text: "01234β789\nbcdef0123456789abcdef",
     lines: vec![0, 11],
     multi_byte_chars: vec![(5, 2)],
-    non_narrow_chars: vec![],
 );
 
 test!(
@@ -69,7 +57,6 @@ test!(
     text: "01234\u{07}6789\nbcdef0123456789abcdef",
     lines: vec![0, 11],
     multi_byte_chars: vec![],
-    non_narrow_chars: vec![(5, 0)],
 );
 
 test!(
@@ -77,7 +64,6 @@ test!(
     text: "aβc",
     lines: vec![0],
     multi_byte_chars: vec![(1, 2)],
-    non_narrow_chars: vec![],
 );
 
 test!(
@@ -85,7 +71,6 @@ test!(
     text: "0123456789abcΔf012345β",
     lines: vec![0],
     multi_byte_chars: vec![(13, 2), (22, 2)],
-    non_narrow_chars: vec![],
 );
 
 test!(
@@ -93,7 +78,6 @@ test!(
     text: "0123456789abcdeΔ123456789abcdef01234",
     lines: vec![0],
     multi_byte_chars: vec![(15, 2)],
-    non_narrow_chars: vec![],
 );
 
 test!(
@@ -101,7 +85,6 @@ test!(
     text: "0123456789abcdeΔ....",
     lines: vec![0],
     multi_byte_chars: vec![(15, 2)],
-    non_narrow_chars: vec![],
 );
 
 test!(
@@ -109,7 +92,6 @@ test!(
     text: "0\t2",
     lines: vec![0],
     multi_byte_chars: vec![],
-    non_narrow_chars: vec![(1, 4)],
 );
 
 test!(
@@ -117,7 +99,6 @@ test!(
     text: "01\t3456789abcdef01234567\u{07}9",
     lines: vec![0],
     multi_byte_chars: vec![],
-    non_narrow_chars: vec![(2, 4), (24, 0)],
 );
 
 test!(
@@ -125,5 +106,4 @@ test!(
     text: "01\t345\n789abcΔf01234567\u{07}9\nbcΔf",
     lines: vec![0, 7, 27],
     multi_byte_chars: vec![(13, 2), (29, 2)],
-    non_narrow_chars: vec![(2, 4), (24, 0)],
 );
diff --git a/compiler/rustc_span/src/lib.rs b/compiler/rustc_span/src/lib.rs
index 266956d63d7..ea57c1ce818 100644
--- a/compiler/rustc_span/src/lib.rs
+++ b/compiler/rustc_span/src/lib.rs
@@ -1345,68 +1345,6 @@ pub struct MultiByteChar {
     pub bytes: u8,
 }
 
-/// Identifies an offset of a non-narrow character in a `SourceFile`.
-#[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug, HashStable_Generic)]
-pub enum NonNarrowChar {
-    /// Represents a zero-width character.
-    ZeroWidth(RelativeBytePos),
-    /// Represents a wide (full-width) character.
-    Wide(RelativeBytePos),
-    /// Represents a tab character, represented visually with a width of 4 characters.
-    Tab(RelativeBytePos),
-}
-
-impl NonNarrowChar {
-    fn new(pos: RelativeBytePos, width: usize) -> Self {
-        match width {
-            0 => NonNarrowChar::ZeroWidth(pos),
-            2 => NonNarrowChar::Wide(pos),
-            4 => NonNarrowChar::Tab(pos),
-            _ => panic!("width {width} given for non-narrow character"),
-        }
-    }
-
-    /// Returns the relative offset of the character in the `SourceFile`.
-    pub fn pos(&self) -> RelativeBytePos {
-        match *self {
-            NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
-        }
-    }
-
-    /// Returns the width of the character, 0 (zero-width) or 2 (wide).
-    pub fn width(&self) -> usize {
-        match *self {
-            NonNarrowChar::ZeroWidth(_) => 0,
-            NonNarrowChar::Wide(_) => 2,
-            NonNarrowChar::Tab(_) => 4,
-        }
-    }
-}
-
-impl Add<RelativeBytePos> for NonNarrowChar {
-    type Output = Self;
-
-    fn add(self, rhs: RelativeBytePos) -> Self {
-        match self {
-            NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
-            NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
-            NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
-        }
-    }
-}
-
-impl Sub<RelativeBytePos> for NonNarrowChar {
-    type Output = Self;
-
-    fn sub(self, rhs: RelativeBytePos) -> Self {
-        match self {
-            NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
-            NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
-            NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
-        }
-    }
-}
-
 /// Identifies an offset of a character that was normalized away from `SourceFile`.
 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug, HashStable_Generic)]
 pub struct NormalizedPos {
@@ -1581,8 +1519,6 @@ pub struct SourceFile {
     pub lines: FreezeLock<SourceFileLines>,
     /// Locations of multi-byte characters in the source code.
     pub multibyte_chars: Vec<MultiByteChar>,
-    /// Width of characters that are not narrow in the source code.
-    pub non_narrow_chars: Vec<NonNarrowChar>,
     /// Locations of characters removed during normalization.
     pub normalized_pos: Vec<NormalizedPos>,
     /// A hash of the filename & crate-id, used for uniquely identifying source
@@ -1604,7 +1540,6 @@ impl Clone for SourceFile {
             source_len: self.source_len,
             lines: self.lines.clone(),
             multibyte_chars: self.multibyte_chars.clone(),
-            non_narrow_chars: self.non_narrow_chars.clone(),
             normalized_pos: self.normalized_pos.clone(),
             stable_id: self.stable_id,
             cnum: self.cnum,
@@ -1679,7 +1614,6 @@ impl<S: SpanEncoder> Encodable<S> for SourceFile {
         }
 
         self.multibyte_chars.encode(s);
-        self.non_narrow_chars.encode(s);
         self.stable_id.encode(s);
         self.normalized_pos.encode(s);
         self.cnum.encode(s);
@@ -1706,7 +1640,6 @@ impl<D: SpanDecoder> Decodable<D> for SourceFile {
             }
         };
         let multibyte_chars: Vec<MultiByteChar> = Decodable::decode(d);
-        let non_narrow_chars: Vec<NonNarrowChar> = Decodable::decode(d);
         let stable_id = Decodable::decode(d);
         let normalized_pos: Vec<NormalizedPos> = Decodable::decode(d);
         let cnum: CrateNum = Decodable::decode(d);
@@ -1721,7 +1654,6 @@ impl<D: SpanDecoder> Decodable<D> for SourceFile {
             external_src: FreezeLock::frozen(ExternalSource::Unneeded),
             lines: FreezeLock::new(lines),
             multibyte_chars,
-            non_narrow_chars,
             normalized_pos,
             stable_id,
             cnum,
@@ -1809,8 +1741,7 @@ impl SourceFile {
         let source_len = src.len();
         let source_len = u32::try_from(source_len).map_err(|_| OffsetOverflowError)?;
 
-        let (lines, multibyte_chars, non_narrow_chars) =
-            analyze_source_file::analyze_source_file(&src);
+        let (lines, multibyte_chars) = analyze_source_file::analyze_source_file(&src);
 
         Ok(SourceFile {
             name,
@@ -1821,7 +1752,6 @@ impl SourceFile {
             source_len: RelativeBytePos::from_u32(source_len),
             lines: FreezeLock::frozen(SourceFileLines::Lines(lines)),
             multibyte_chars,
-            non_narrow_chars,
             normalized_pos,
             stable_id,
             cnum: LOCAL_CRATE,
@@ -2130,41 +2060,44 @@ impl SourceFile {
         let pos = self.relative_position(pos);
         let (line, col_or_chpos) = self.lookup_file_pos(pos);
         if line > 0 {
-            let col = col_or_chpos;
-            let linebpos = self.lines()[line - 1];
-            let col_display = {
-                let start_width_idx = self
-                    .non_narrow_chars
-                    .binary_search_by_key(&linebpos, |x| x.pos())
-                    .unwrap_or_else(|x| x);
-                let end_width_idx = self
-                    .non_narrow_chars
-                    .binary_search_by_key(&pos, |x| x.pos())
-                    .unwrap_or_else(|x| x);
-                let special_chars = end_width_idx - start_width_idx;
-                let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
-                    .iter()
-                    .map(|x| x.width())
-                    .sum();
-                col.0 - special_chars + non_narrow
+            let Some(code) = self.get_line(line - 1) else {
+                // If we don't have the code available, it is ok as a fallback to return the bytepos
+                // instead of the "display" column, which is only used to properly show underlines
+                // in the terminal.
+                // FIXME: we'll want better handling of this in the future for the sake of tools
+                // that want to use the display col instead of byte offsets to modify Rust code, but
+                // that is a problem for another day, the previous code was already incorrect for
+                // both displaying *and* third party tools using the json output naïvely.
+                tracing::info!("couldn't find line {line} {:?}", self.name);
+                return (line, col_or_chpos, col_or_chpos.0);
             };
-            (line, col, col_display)
+            let display_col = code.chars().take(col_or_chpos.0).map(|ch| char_width(ch)).sum();
+            (line, col_or_chpos, display_col)
         } else {
-            let chpos = col_or_chpos;
-            let col_display = {
-                let end_width_idx = self
-                    .non_narrow_chars
-                    .binary_search_by_key(&pos, |x| x.pos())
-                    .unwrap_or_else(|x| x);
-                let non_narrow: usize =
-                    self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
-                chpos.0 - end_width_idx + non_narrow
-            };
-            (0, chpos, col_display)
+            // This is never meant to happen?
+            (0, col_or_chpos, col_or_chpos.0)
         }
     }
 }
 
+pub fn char_width(ch: char) -> usize {
+    // FIXME: `unicode_width` sometimes disagrees with terminals on how wide a `char` is. For now,
+    // just accept that sometimes the code line will be longer than desired.
+    match ch {
+        '\t' => 4,
+        // Keep the following list in sync with `rustc_errors::emitter::OUTPUT_REPLACEMENTS`. These
+        // are control points that we replace before printing with a visible codepoint for the sake
+        // of being able to point at them with underlines.
+        '\u{0000}' | '\u{0001}' | '\u{0002}' | '\u{0003}' | '\u{0004}' | '\u{0005}'
+        | '\u{0006}' | '\u{0007}' | '\u{0008}' | '\u{000B}' | '\u{000C}' | '\u{000D}'
+        | '\u{000E}' | '\u{000F}' | '\u{0010}' | '\u{0011}' | '\u{0012}' | '\u{0013}'
+        | '\u{0014}' | '\u{0015}' | '\u{0016}' | '\u{0017}' | '\u{0018}' | '\u{0019}'
+        | '\u{001A}' | '\u{001B}' | '\u{001C}' | '\u{001D}' | '\u{001E}' | '\u{001F}'
+        | '\u{007F}' => 1,
+        _ => unicode_width::UnicodeWidthChar::width(ch).unwrap_or(1),
+    }
+}
+
 /// Normalizes the source code and records the normalizations.
 fn normalize_src(src: &mut String) -> Vec<NormalizedPos> {
     let mut normalized_pos = vec![];
diff --git a/compiler/rustc_span/src/source_map.rs b/compiler/rustc_span/src/source_map.rs
index fb212d67997..14c157a0111 100644
--- a/compiler/rustc_span/src/source_map.rs
+++ b/compiler/rustc_span/src/source_map.rs
@@ -330,7 +330,6 @@ impl SourceMap {
         cnum: CrateNum,
         file_local_lines: FreezeLock<SourceFileLines>,
         multibyte_chars: Vec<MultiByteChar>,
-        non_narrow_chars: Vec<NonNarrowChar>,
         normalized_pos: Vec<NormalizedPos>,
         metadata_index: u32,
     ) -> Lrc<SourceFile> {
@@ -348,7 +347,6 @@ impl SourceMap {
             source_len,
             lines: file_local_lines,
             multibyte_chars,
-            non_narrow_chars,
             normalized_pos,
             stable_id,
             cnum,
diff --git a/compiler/rustc_span/src/source_map/tests.rs b/compiler/rustc_span/src/source_map/tests.rs
index dcb02da3719..0c818b94b85 100644
--- a/compiler/rustc_span/src/source_map/tests.rs
+++ b/compiler/rustc_span/src/source_map/tests.rs
@@ -232,7 +232,6 @@ fn t10() {
         source_len,
         lines,
         multibyte_chars,
-        non_narrow_chars,
         normalized_pos,
         stable_id,
         ..
@@ -246,7 +245,6 @@ fn t10() {
         CrateNum::ZERO,
         FreezeLock::new(lines.read().clone()),
         multibyte_chars,
-        non_narrow_chars,
         normalized_pos,
         0,
     );