Auto merge of #140695 - Zalathar:rollup-i32gzbo, r=Zalathar

Rollup of 12 pull requests

Successful merges:

 - #139550 (Fix `-Zremap-path-scope` rmeta handling)
 - #139764 (Consistent trait bounds for ExtractIf Debug impls)
 - #139773 (Implement `Iterator::last` for `vec::IntoIter`)
 - #140035 (Implement RFC 3503: frontmatters)
 - #140251 (coverage-dump: Resolve global file IDs to filenames)
 - #140393 (std: get rid of `sys_common::process`)
 - #140532 (Fix RustAnalyzer discovery of rustc's `stable_mir` crate)
 - #140598 (Steer docs to `utf8_chunks` and `Iterator::take`)
 - #140634 (Use more accurate ELF flags on MIPS)
 - #140673 (Clean rustdoc tests folder)
 - #140678 (Be a bit more relaxed about not yet constrained infer vars in closure upvar analysis)
 - #140687 (Update mdbook to 0.4.49)

r? `@ghost`
`@rustbot` modify labels: rollup

30 files changed, 476 insertions, 120 deletions

diff --git a/src/bootstrap/src/core/build_steps/check.rs b/src/bootstrap/src/core/build_steps/check.rs
index ae9511b7867..fa848c492b4 100644
--- a/src/bootstrap/src/core/build_steps/check.rs
+++ b/src/bootstrap/src/core/build_steps/check.rs
@@ -527,3 +527,70 @@ tool_check_step!(Bootstrap { path: "src/bootstrap", default: false });
 // `run-make-support` will be built as part of suitable run-make compiletest test steps, but support
 // check to make it easier to work on.
 tool_check_step!(RunMakeSupport { path: "src/tools/run-make-support", default: false });
+
+/// Check step for the `coverage-dump` bootstrap tool. The coverage-dump tool
+/// is used internally by coverage tests.
+///
+/// FIXME(Zalathar): This is temporarily separate from the other tool check
+/// steps so that it can use the stage 0 compiler instead of `top_stage`,
+/// without introducing conflicts with the stage 0 redesign (#119899).
+///
+/// After the stage 0 redesign lands, we can look into using the stage 0
+/// compiler to check all bootstrap tools (#139170).
+#[derive(Debug, Clone, PartialEq, Eq, Hash)]
+pub(crate) struct CoverageDump;
+
+impl CoverageDump {
+    const PATH: &str = "src/tools/coverage-dump";
+}
+
+impl Step for CoverageDump {
+    type Output = ();
+
+    /// Most contributors won't care about coverage-dump, so don't make their
+    /// check builds slower unless they opt in and check it explicitly.
+    const DEFAULT: bool = false;
+    const ONLY_HOSTS: bool = true;
+
+    fn should_run(run: ShouldRun<'_>) -> ShouldRun<'_> {
+        run.path(Self::PATH)
+    }
+
+    fn make_run(run: RunConfig<'_>) {
+        run.builder.ensure(Self {});
+    }
+
+    fn run(self, builder: &Builder<'_>) -> Self::Output {
+        // Make sure we haven't forgotten any fields, if there are any.
+        let Self {} = self;
+        let display_name = "coverage-dump";
+        let host = builder.config.build;
+        let target = host;
+        let mode = Mode::ToolBootstrap;
+
+        let compiler = builder.compiler(0, host);
+        let cargo = prepare_tool_cargo(
+            builder,
+            compiler,
+            mode,
+            target,
+            builder.kind,
+            Self::PATH,
+            SourceType::InTree,
+            &[],
+        );
+
+        let stamp = BuildStamp::new(&builder.cargo_out(compiler, mode, target))
+            .with_prefix(&format!("{display_name}-check"));
+
+        let _guard = builder.msg_tool(
+            builder.kind,
+            mode,
+            display_name,
+            compiler.stage,
+            &compiler.host,
+            &target,
+        );
+        run_cargo(builder, cargo, builder.config.free_args.clone(), &stamp, vec![], true, false);
+    }
+}
diff --git a/src/bootstrap/src/core/build_steps/run.rs b/src/bootstrap/src/core/build_steps/run.rs
index 5cacd5b9914..7ff38505294 100644
--- a/src/bootstrap/src/core/build_steps/run.rs
+++ b/src/bootstrap/src/core/build_steps/run.rs
@@ -392,3 +392,31 @@ impl Step for CyclicStep {
         builder.ensure(CyclicStep { n: self.n.saturating_sub(1) })
     }
 }
+
+/// Step to manually run the coverage-dump tool (`./x run coverage-dump`).
+///
+/// The coverage-dump tool is an internal detail of coverage tests, so this run
+/// step is only needed when testing coverage-dump manually.
+#[derive(Debug, PartialOrd, Ord, Clone, Hash, PartialEq, Eq)]
+pub struct CoverageDump;
+
+impl Step for CoverageDump {
+    type Output = ();
+
+    const DEFAULT: bool = false;
+    const ONLY_HOSTS: bool = true;
+
+    fn should_run(run: ShouldRun<'_>) -> ShouldRun<'_> {
+        run.path("src/tools/coverage-dump")
+    }
+
+    fn make_run(run: RunConfig<'_>) {
+        run.builder.ensure(Self {});
+    }
+
+    fn run(self, builder: &Builder<'_>) {
+        let mut cmd = builder.tool_cmd(Tool::CoverageDump);
+        cmd.args(&builder.config.free_args);
+        cmd.run(builder);
+    }
+}
diff --git a/src/bootstrap/src/core/build_steps/test.rs b/src/bootstrap/src/core/build_steps/test.rs
index a7a3b5a878c..29fb576f574 100644
--- a/src/bootstrap/src/core/build_steps/test.rs
+++ b/src/bootstrap/src/core/build_steps/test.rs
@@ -54,6 +54,7 @@ impl Step for CrateBootstrap {
         run.path("src/tools/jsondoclint")
             .path("src/tools/suggest-tests")
             .path("src/tools/replace-version-placeholder")
+            .path("src/tools/coverage-dump")
             // We want `./x test tidy` to _run_ the tidy tool, not its tests.
             // So we need a separate alias to test the tidy tool itself.
             .alias("tidyselftest")
diff --git a/src/bootstrap/src/core/builder/mod.rs b/src/bootstrap/src/core/builder/mod.rs
index 6469bb5f272..15dc3380a39 100644
--- a/src/bootstrap/src/core/builder/mod.rs
+++ b/src/bootstrap/src/core/builder/mod.rs
@@ -961,6 +961,7 @@ impl<'a> Builder<'a> {
                 check::RunMakeSupport,
                 check::Compiletest,
                 check::FeaturesStatusDump,
+                check::CoverageDump,
             ),
             Kind::Test => describe!(
                 crate::core::build_steps::toolstate::ToolStateCheck,
@@ -1114,6 +1115,7 @@ impl<'a> Builder<'a> {
                 run::UnicodeTableGenerator,
                 run::FeaturesStatusDump,
                 run::CyclicStep,
+                run::CoverageDump,
             ),
             Kind::Setup => {
                 describe!(setup::Profile, setup::Hook, setup::Link, setup::Editor)
diff --git a/src/doc/unstable-book/src/language-features/arbitrary-self-types.md b/src/doc/unstable-book/src/language-features/arbitrary-self-types.md
index 2f8b52d4043..d660dd13fe4 100644
--- a/src/doc/unstable-book/src/language-features/arbitrary-self-types.md
+++ b/src/doc/unstable-book/src/language-features/arbitrary-self-types.md
@@ -2,7 +2,7 @@
 
 The tracking issue for this feature is: [#44874]
 
-[#38788]: https://github.com/rust-lang/rust/issues/44874
+[#44874]: https://github.com/rust-lang/rust/issues/44874
 
 ------------------------
 
diff --git a/src/doc/unstable-book/src/language-features/f128.md b/src/doc/unstable-book/src/language-features/f128.md
index 0cc5f677230..b523ffe10f2 100644
--- a/src/doc/unstable-book/src/language-features/f128.md
+++ b/src/doc/unstable-book/src/language-features/f128.md
@@ -6,4 +6,4 @@ The tracking issue for this feature is: [#116909]
 
 ---
 
-Enable the `f128` type for  IEEE 128-bit floating numbers (quad precision).
+Enable the `f128` type for IEEE 128-bit floating numbers (quad precision).
diff --git a/src/doc/unstable-book/src/language-features/f16.md b/src/doc/unstable-book/src/language-features/f16.md
index efb07a5146d..5f31dcbb06c 100644
--- a/src/doc/unstable-book/src/language-features/f16.md
+++ b/src/doc/unstable-book/src/language-features/f16.md
@@ -6,4 +6,4 @@ The tracking issue for this feature is: [#116909]
 
 ---
 
-Enable the `f16` type for  IEEE 16-bit floating numbers (half precision).
+Enable the `f16` type for IEEE 16-bit floating numbers (half precision).
diff --git a/src/doc/unstable-book/src/language-features/frontmatter.md b/src/doc/unstable-book/src/language-features/frontmatter.md
new file mode 100644
index 00000000000..1d5b4feb6ac
--- /dev/null
+++ b/src/doc/unstable-book/src/language-features/frontmatter.md
@@ -0,0 +1,25 @@
+# `frontmatter`
+
+The tracking issue for this feature is: [#136889]
+
+------
+
+The `frontmatter` feature allows an extra metadata block at the top of files for consumption by
+external tools. For example, it can be used by [`cargo-script`] files to specify dependencies.
+
+```rust
+#!/usr/bin/env -S cargo -Zscript
+---
+[dependencies]
+libc = "0.2.172"
+---
+#![feature(frontmatter)]
+# mod libc { pub type c_int = i32; }
+
+fn main() {
+    let x: libc::c_int = 1i32;
+}
+```
+
+[#136889]: https://github.com/rust-lang/rust/issues/136889
+[`cargo-script`]: https://rust-lang.github.io/rfcs/3502-cargo-script.html
diff --git a/src/librustdoc/html/highlight.rs b/src/librustdoc/html/highlight.rs
index c943d3ad4d0..2db1ea8450c 100644
--- a/src/librustdoc/html/highlight.rs
+++ b/src/librustdoc/html/highlight.rs
@@ -9,7 +9,7 @@ use std::collections::VecDeque;
 use std::fmt::{Display, Write};
 
 use rustc_data_structures::fx::FxIndexMap;
-use rustc_lexer::{Cursor, LiteralKind, TokenKind};
+use rustc_lexer::{Cursor, FrontmatterAllowed, LiteralKind, TokenKind};
 use rustc_span::edition::Edition;
 use rustc_span::symbol::Symbol;
 use rustc_span::{BytePos, DUMMY_SP, Span};
@@ -638,7 +638,8 @@ impl<'src> Classifier<'src> {
     /// Takes as argument the source code to HTML-ify, the rust edition to use and the source code
     /// file span which will be used later on by the `span_correspondence_map`.
     fn new(src: &'src str, file_span: Span, decoration_info: Option<&DecorationInfo>) -> Self {
-        let tokens = PeekIter::new(TokenIter { src, cursor: Cursor::new(src) });
+        let tokens =
+            PeekIter::new(TokenIter { src, cursor: Cursor::new(src, FrontmatterAllowed::Yes) });
         let decorations = decoration_info.map(Decorations::new);
         Classifier {
             tokens,
@@ -884,6 +885,7 @@ impl<'src> Classifier<'src> {
             | TokenKind::At
             | TokenKind::Tilde
             | TokenKind::Colon
+            | TokenKind::Frontmatter { .. }
             | TokenKind::Unknown => return no_highlight(sink),
 
             TokenKind::Question => Class::QuestionMark,
diff --git a/src/tools/clippy/tests/ui/double_ended_iterator_last.fixed b/src/tools/clippy/tests/ui/double_ended_iterator_last.fixed
index 2ce0c04c301..be31ee5fb48 100644
--- a/src/tools/clippy/tests/ui/double_ended_iterator_last.fixed
+++ b/src/tools/clippy/tests/ui/double_ended_iterator_last.fixed
@@ -84,6 +84,19 @@ fn issue_14139() {
 }
 
 fn drop_order() {
+    struct DropDeIterator(std::vec::IntoIter<S>);
+    impl Iterator for DropDeIterator {
+        type Item = S;
+        fn next(&mut self) -> Option<Self::Item> {
+            self.0.next()
+        }
+    }
+    impl DoubleEndedIterator for DropDeIterator {
+        fn next_back(&mut self) -> Option<Self::Item> {
+            self.0.next_back()
+        }
+    }
+
     struct S(&'static str);
     impl std::ops::Drop for S {
         fn drop(&mut self) {
@@ -92,7 +105,7 @@ fn drop_order() {
     }
 
     let v = vec![S("one"), S("two"), S("three")];
-    let mut v = v.into_iter();
+    let mut v = DropDeIterator(v.into_iter());
     println!("Last element is {}", v.next_back().unwrap().0);
     //~^ ERROR: called `Iterator::last` on a `DoubleEndedIterator`
     println!("Done");
diff --git a/src/tools/clippy/tests/ui/double_ended_iterator_last.rs b/src/tools/clippy/tests/ui/double_ended_iterator_last.rs
index a4eb9b3337b..30864e15bce 100644
--- a/src/tools/clippy/tests/ui/double_ended_iterator_last.rs
+++ b/src/tools/clippy/tests/ui/double_ended_iterator_last.rs
@@ -84,6 +84,19 @@ fn issue_14139() {
 }
 
 fn drop_order() {
+    struct DropDeIterator(std::vec::IntoIter<S>);
+    impl Iterator for DropDeIterator {
+        type Item = S;
+        fn next(&mut self) -> Option<Self::Item> {
+            self.0.next()
+        }
+    }
+    impl DoubleEndedIterator for DropDeIterator {
+        fn next_back(&mut self) -> Option<Self::Item> {
+            self.0.next_back()
+        }
+    }
+
     struct S(&'static str);
     impl std::ops::Drop for S {
         fn drop(&mut self) {
@@ -92,7 +105,7 @@ fn drop_order() {
     }
 
     let v = vec![S("one"), S("two"), S("three")];
-    let v = v.into_iter();
+    let v = DropDeIterator(v.into_iter());
     println!("Last element is {}", v.last().unwrap().0);
     //~^ ERROR: called `Iterator::last` on a `DoubleEndedIterator`
     println!("Done");
diff --git a/src/tools/clippy/tests/ui/double_ended_iterator_last.stderr b/src/tools/clippy/tests/ui/double_ended_iterator_last.stderr
index fe8cf2dcb25..72a6ead47a9 100644
--- a/src/tools/clippy/tests/ui/double_ended_iterator_last.stderr
+++ b/src/tools/clippy/tests/ui/double_ended_iterator_last.stderr
@@ -18,7 +18,7 @@ LL |     let _ = DeIterator.last();
    |                        help: try: `next_back()`
 
 error: called `Iterator::last` on a `DoubleEndedIterator`; this will needlessly iterate the entire iterator
-  --> tests/ui/double_ended_iterator_last.rs:96:36
+  --> tests/ui/double_ended_iterator_last.rs:109:36
    |
 LL |     println!("Last element is {}", v.last().unwrap().0);
    |                                    ^^^^^^^^
@@ -26,7 +26,7 @@ LL |     println!("Last element is {}", v.last().unwrap().0);
    = note: this change will alter drop order which may be undesirable
 help: try
    |
-LL ~     let mut v = v.into_iter();
+LL ~     let mut v = DropDeIterator(v.into_iter());
 LL ~     println!("Last element is {}", v.next_back().unwrap().0);
    |
 
diff --git a/src/tools/clippy/tests/ui/double_ended_iterator_last_unfixable.rs b/src/tools/clippy/tests/ui/double_ended_iterator_last_unfixable.rs
index 7c5de8832d6..e9218bbb409 100644
--- a/src/tools/clippy/tests/ui/double_ended_iterator_last_unfixable.rs
+++ b/src/tools/clippy/tests/ui/double_ended_iterator_last_unfixable.rs
@@ -11,6 +11,19 @@ fn main() {
 }
 
 fn drop_order() {
+    struct DropDeIterator(std::vec::IntoIter<S>);
+    impl Iterator for DropDeIterator {
+        type Item = S;
+        fn next(&mut self) -> Option<Self::Item> {
+            self.0.next()
+        }
+    }
+    impl DoubleEndedIterator for DropDeIterator {
+        fn next_back(&mut self) -> Option<Self::Item> {
+            self.0.next_back()
+        }
+    }
+
     struct S(&'static str);
     impl std::ops::Drop for S {
         fn drop(&mut self) {
@@ -19,7 +32,7 @@ fn drop_order() {
     }
 
     let v = vec![S("one"), S("two"), S("three")];
-    let v = (v.into_iter(), 42);
+    let v = (DropDeIterator(v.into_iter()), 42);
     println!("Last element is {}", v.0.last().unwrap().0);
     //~^ ERROR: called `Iterator::last` on a `DoubleEndedIterator`
     println!("Done");
diff --git a/src/tools/clippy/tests/ui/double_ended_iterator_last_unfixable.stderr b/src/tools/clippy/tests/ui/double_ended_iterator_last_unfixable.stderr
index 845afc11f04..e330a22a354 100644
--- a/src/tools/clippy/tests/ui/double_ended_iterator_last_unfixable.stderr
+++ b/src/tools/clippy/tests/ui/double_ended_iterator_last_unfixable.stderr
@@ -1,5 +1,5 @@
 error: called `Iterator::last` on a `DoubleEndedIterator`; this will needlessly iterate the entire iterator
-  --> tests/ui/double_ended_iterator_last_unfixable.rs:23:36
+  --> tests/ui/double_ended_iterator_last_unfixable.rs:36:36
    |
 LL |     println!("Last element is {}", v.0.last().unwrap().0);
    |                                    ^^^^------
@@ -8,7 +8,7 @@ LL |     println!("Last element is {}", v.0.last().unwrap().0);
    |
    = note: this change will alter drop order which may be undesirable
 note: this must be made mutable to use `.next_back()`
-  --> tests/ui/double_ended_iterator_last_unfixable.rs:23:36
+  --> tests/ui/double_ended_iterator_last_unfixable.rs:36:36
    |
 LL |     println!("Last element is {}", v.0.last().unwrap().0);
    |                                    ^^^
diff --git a/src/tools/coverage-dump/Cargo.toml b/src/tools/coverage-dump/Cargo.toml
index 7f14286b5d0..6f92ac50d96 100644
--- a/src/tools/coverage-dump/Cargo.toml
+++ b/src/tools/coverage-dump/Cargo.toml
@@ -7,6 +7,7 @@ edition = "2021"
 
 [dependencies]
 anyhow = "1.0.71"
+itertools = "0.12"
 leb128 = "0.2.5"
 md5 = { package = "md-5" , version = "0.10.5" }
 miniz_oxide = "0.7.1"
diff --git a/src/tools/coverage-dump/src/covfun.rs b/src/tools/coverage-dump/src/covfun.rs
index 82ebd33d0d1..1cc9f4dc5d6 100644
--- a/src/tools/coverage-dump/src/covfun.rs
+++ b/src/tools/coverage-dump/src/covfun.rs
@@ -1,23 +1,33 @@
 use std::collections::HashMap;
 use std::fmt::{self, Debug, Write as _};
-use std::sync::OnceLock;
+use std::sync::LazyLock;
 
-use anyhow::{Context, anyhow};
+use anyhow::{Context, anyhow, bail, ensure};
+use itertools::Itertools;
 use regex::Regex;
 
-use crate::parser::{Parser, unescape_llvm_string_contents};
+use crate::covmap::FilenameTables;
+use crate::llvm_utils::unescape_llvm_string_contents;
+use crate::parser::Parser;
+
+#[cfg(test)]
+mod tests;
 
 pub(crate) fn dump_covfun_mappings(
     llvm_ir: &str,
+    filename_tables: &FilenameTables,
     function_names: &HashMap<u64, String>,
 ) -> anyhow::Result<()> {
     // Extract function coverage entries from the LLVM IR assembly, and associate
     // each entry with its (demangled) name.
     let mut covfun_entries = llvm_ir
         .lines()
-        .filter_map(covfun_line_data)
-        .map(|line_data| (function_names.get(&line_data.name_hash).map(String::as_str), line_data))
-        .collect::<Vec<_>>();
+        .filter(|line| is_covfun_line(line))
+        .map(parse_covfun_line)
+        .map_ok(|line_data| {
+            (function_names.get(&line_data.name_hash).map(String::as_str), line_data)
+        })
+        .collect::<Result<Vec<_>, _>>()?;
     covfun_entries.sort_by(|a, b| {
         // Sort entries primarily by name, to help make the order consistent
         // across platforms and relatively insensitive to changes.
@@ -41,8 +51,12 @@ pub(crate) fn dump_covfun_mappings(
         println!("Number of files: {num_files}");
 
         for i in 0..num_files {
-            let global_file_id = parser.read_uleb128_u32()?;
-            println!("- file {i} => global file {global_file_id}");
+            let global_file_id = parser.read_uleb128_usize()?;
+            let &CovfunLineData { filenames_hash, .. } = line_data;
+            let Some(filename) = filename_tables.lookup(filenames_hash, global_file_id) else {
+                bail!("couldn't resolve global file: {filenames_hash}, {global_file_id}");
+            };
+            println!("- file {i} => {filename}");
         }
 
         let num_expressions = parser.read_uleb128_u32()?;
@@ -107,36 +121,50 @@ pub(crate) fn dump_covfun_mappings(
     Ok(())
 }
 
+#[derive(Debug, PartialEq, Eq)]
 struct CovfunLineData {
-    name_hash: u64,
     is_used: bool,
+    name_hash: u64,
+    filenames_hash: u64,
     payload: Vec<u8>,
 }
 
-/// Checks a line of LLVM IR assembly to see if it contains an `__llvm_covfun`
-/// entry, and if so extracts relevant data in a `CovfunLineData`.
-fn covfun_line_data(line: &str) -> Option<CovfunLineData> {
-    let re = {
-        // We cheat a little bit and match variable names `@__covrec_[HASH]u`
-        // rather than the section name, because the section name is harder to
-        // extract and differs across Linux/Windows/macOS. We also extract the
-        // symbol name hash from the variable name rather than the data, since
-        // it's easier and both should match.
-        static RE: OnceLock<Regex> = OnceLock::new();
-        RE.get_or_init(|| {
-            Regex::new(
-                r#"^@__covrec_(?<name_hash>[0-9A-Z]+)(?<is_used>u)? = .*\[[0-9]+ x i8\] c"(?<payload>[^"]*)".*$"#,
-            )
-            .unwrap()
-        })
-    };
+fn is_covfun_line(line: &str) -> bool {
+    line.starts_with("@__covrec_")
+}
 
-    let captures = re.captures(line)?;
-    let name_hash = u64::from_str_radix(&captures["name_hash"], 16).unwrap();
+/// Given a line of LLVM IR assembly that should contain an `__llvm_covfun`
+/// entry, parses it to extract relevant data in a `CovfunLineData`.
+fn parse_covfun_line(line: &str) -> anyhow::Result<CovfunLineData> {
+    ensure!(is_covfun_line(line));
+
+    // We cheat a little bit and match variable names `@__covrec_[HASH]u`
+    // rather than the section name, because the section name is harder to
+    // extract and differs across Linux/Windows/macOS.
+    const RE_STRING: &str = r#"(?x)^
+        @__covrec_[0-9A-Z]+(?<is_used>u)?
+        \ = \ # (trailing space)
+        .*
+        <\{
+            \ i64 \ (?<name_hash> -? [0-9]+),
+            \ i32 \ -? [0-9]+, # (length of payload; currently unused)
+            \ i64 \ -? [0-9]+, # (source hash; currently unused)
+            \ i64 \ (?<filenames_hash> -? [0-9]+),
+            \ \[ [0-9]+ \ x \ i8 \] \ c"(?<payload>[^"]*)"
+            \ # (trailing space)
+        }>
+        .*$
+    "#;
+    static RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(RE_STRING).unwrap());
+
+    let captures =
+        RE.captures(line).with_context(|| format!("couldn't parse covfun line: {line:?}"))?;
     let is_used = captures.name("is_used").is_some();
+    let name_hash = i64::from_str_radix(&captures["name_hash"], 10).unwrap() as u64;
+    let filenames_hash = i64::from_str_radix(&captures["filenames_hash"], 10).unwrap() as u64;
     let payload = unescape_llvm_string_contents(&captures["payload"]);
 
-    Some(CovfunLineData { name_hash, is_used, payload })
+    Ok(CovfunLineData { is_used, name_hash, filenames_hash, payload })
 }
 
 // Extra parser methods only needed when parsing `covfun` payloads.
diff --git a/src/tools/coverage-dump/src/covfun/tests.rs b/src/tools/coverage-dump/src/covfun/tests.rs
new file mode 100644
index 00000000000..1ce833784bd
--- /dev/null
+++ b/src/tools/coverage-dump/src/covfun/tests.rs
@@ -0,0 +1,53 @@
+use super::{CovfunLineData, parse_covfun_line};
+
+/// Integers in LLVM IR are not inherently signed/unsigned, and the text format tends
+/// to emit them in signed form, so this helper function converts `i64` to `u64`.
+fn as_u64(x: i64) -> u64 {
+    x as u64
+}
+
+#[test]
+fn parse_covfun_line_data() {
+    struct Case {
+        line: &'static str,
+        expected: CovfunLineData,
+    }
+    let cases = &[
+        // Copied from `trivial.ll`:
+        Case {
+            line: r#"@__covrec_49A9BAAE5F896E81u = linkonce_odr hidden constant <{ i64, i32, i64, i64, [9 x i8] }> <{ i64 5307978893922758273, i32 9, i64 445092354169400020, i64 6343436898695299756, [9 x i8] c"\01\01\00\01\01\03\01\00\0D" }>, section "__LLVM_COV,__llvm_covfun", align 8"#,
+            expected: CovfunLineData {
+                is_used: true,
+                name_hash: as_u64(5307978893922758273),
+                filenames_hash: as_u64(6343436898695299756),
+                payload: b"\x01\x01\x00\x01\x01\x03\x01\x00\x0D".to_vec(),
+            },
+        },
+        // Copied from `on-off-sandwich.ll`:
+        Case {
+            line: r#"@__covrec_D0CE53C5E64F319Au = linkonce_odr hidden constant <{ i64, i32, i64, i64, [14 x i8] }> <{ i64 -3400688559180533350, i32 14, i64 7307957714577672185, i64 892196767019953100, [14 x i8] c"\01\01\00\02\01\10\05\02\10\01\07\05\00\06" }>, section "__LLVM_COV,__llvm_covfun", align 8"#,
+            expected: CovfunLineData {
+                is_used: true,
+                name_hash: as_u64(-3400688559180533350),
+                filenames_hash: as_u64(892196767019953100),
+                payload: b"\x01\x01\x00\x02\x01\x10\x05\x02\x10\x01\x07\x05\x00\x06".to_vec(),
+            },
+        },
+        // Copied from `no-core.ll`:
+        Case {
+            line: r#"@__covrec_F8016FC82D46106u = linkonce_odr hidden constant <{ i64, i32, i64, i64, [9 x i8] }> <{ i64 1116917981370409222, i32 9, i64 -8857254680411629915, i64 -3625186110715410276, [9 x i8] c"\01\01\00\01\01\0C\01\00\0D" }>, section "__LLVM_COV,__llvm_covfun", align 8"#,
+            expected: CovfunLineData {
+                is_used: true,
+                name_hash: as_u64(1116917981370409222),
+                filenames_hash: as_u64(-3625186110715410276),
+                payload: b"\x01\x01\x00\x01\x01\x0C\x01\x00\x0D".to_vec(),
+            },
+        },
+    ];
+
+    for &Case { line, ref expected } in cases {
+        println!("- {line}");
+        let line_data = parse_covfun_line(line).map_err(|e| e.to_string());
+        assert_eq!(line_data.as_ref(), Ok(expected));
+    }
+}
diff --git a/src/tools/coverage-dump/src/covmap.rs b/src/tools/coverage-dump/src/covmap.rs
new file mode 100644
index 00000000000..2246ca2d575
--- /dev/null
+++ b/src/tools/coverage-dump/src/covmap.rs
@@ -0,0 +1,75 @@
+use std::collections::HashMap;
+use std::sync::LazyLock;
+
+use anyhow::{Context, ensure};
+use regex::Regex;
+
+use crate::llvm_utils::{truncated_md5, unescape_llvm_string_contents};
+use crate::parser::Parser;
+
+#[derive(Debug, Default)]
+pub(crate) struct FilenameTables {
+    map: HashMap<u64, Vec<String>>,
+}
+
+impl FilenameTables {
+    pub(crate) fn lookup(&self, filenames_hash: u64, global_file_id: usize) -> Option<&str> {
+        let table = self.map.get(&filenames_hash)?;
+        let filename = table.get(global_file_id)?;
+        Some(filename)
+    }
+}
+
+struct CovmapLineData {
+    payload: Vec<u8>,
+}
+
+pub(crate) fn make_filename_tables(llvm_ir: &str) -> anyhow::Result<FilenameTables> {
+    let mut map = HashMap::default();
+
+    for line in llvm_ir.lines().filter(|line| is_covmap_line(line)) {
+        let CovmapLineData { payload } = parse_covmap_line(line)?;
+
+        let mut parser = Parser::new(&payload);
+        let n_filenames = parser.read_uleb128_usize()?;
+        let uncompressed_bytes = parser.read_chunk_to_uncompressed_bytes()?;
+        parser.ensure_empty()?;
+
+        let mut filenames_table = vec![];
+
+        let mut parser = Parser::new(&uncompressed_bytes);
+        for _ in 0..n_filenames {
+            let len = parser.read_uleb128_usize()?;
+            let bytes = parser.read_n_bytes(len)?;
+            let filename = str::from_utf8(bytes)?;
+            filenames_table.push(filename.to_owned());
+        }
+
+        let filenames_hash = truncated_md5(&payload);
+        map.insert(filenames_hash, filenames_table);
+    }
+
+    Ok(FilenameTables { map })
+}
+
+fn is_covmap_line(line: &str) -> bool {
+    line.starts_with("@__llvm_coverage_mapping ")
+}
+
+fn parse_covmap_line(line: &str) -> anyhow::Result<CovmapLineData> {
+    ensure!(is_covmap_line(line));
+
+    const RE_STRING: &str = r#"(?x)^
+        @__llvm_coverage_mapping \ =
+        .*
+        \[ [0-9]+ \ x \ i8 \] \ c"(?<payload>[^"]*)"
+        .*$
+    "#;
+    static RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(RE_STRING).unwrap());
+
+    let captures =
+        RE.captures(line).with_context(|| format!("couldn't parse covmap line: {line:?}"))?;
+    let payload = unescape_llvm_string_contents(&captures["payload"]);
+
+    Ok(CovmapLineData { payload })
+}
diff --git a/src/tools/coverage-dump/src/llvm_utils.rs b/src/tools/coverage-dump/src/llvm_utils.rs
new file mode 100644
index 00000000000..92322b256a8
--- /dev/null
+++ b/src/tools/coverage-dump/src/llvm_utils.rs
@@ -0,0 +1,85 @@
+use std::borrow::Cow;
+use std::sync::OnceLock;
+
+use anyhow::{anyhow, ensure};
+use regex::bytes;
+
+use crate::parser::Parser;
+
+#[cfg(test)]
+mod tests;
+
+/// Given the raw contents of a string literal in LLVM IR assembly, decodes any
+/// backslash escapes and returns a vector containing the resulting byte string.
+pub(crate) fn unescape_llvm_string_contents(contents: &str) -> Vec<u8> {
+    let escape_re = {
+        static RE: OnceLock<bytes::Regex> = OnceLock::new();
+        // LLVM IR supports two string escapes: `\\` and `\xx`.
+        RE.get_or_init(|| bytes::Regex::new(r"\\\\|\\([0-9A-Za-z]{2})").unwrap())
+    };
+
+    fn u8_from_hex_digits(digits: &[u8]) -> u8 {
+        // We know that the input contains exactly 2 hex digits, so these calls
+        // should never fail.
+        assert_eq!(digits.len(), 2);
+        let digits = std::str::from_utf8(digits).unwrap();
+        u8::from_str_radix(digits, 16).unwrap()
+    }
+
+    escape_re
+        .replace_all(contents.as_bytes(), |captures: &bytes::Captures<'_>| {
+            let byte = match captures.get(1) {
+                None => b'\\',
+                Some(hex_digits) => u8_from_hex_digits(hex_digits.as_bytes()),
+            };
+            [byte]
+        })
+        .into_owned()
+}
+
+/// LLVM's profiler/coverage metadata often uses an MD5 hash truncated to
+/// 64 bits as a way to associate data stored in different tables/sections.
+pub(crate) fn truncated_md5(bytes: &[u8]) -> u64 {
+    use md5::{Digest, Md5};
+    let mut hasher = Md5::new();
+    hasher.update(bytes);
+    let hash: [u8; 8] = hasher.finalize().as_slice()[..8].try_into().unwrap();
+    // The truncated hash is explicitly little-endian, regardless of host
+    // or target platform. (See `MD5Result::low` in LLVM's `MD5.h`.)
+    u64::from_le_bytes(hash)
+}
+
+impl<'a> Parser<'a> {
+    /// Reads a sequence of:
+    /// - Length of uncompressed data in bytes, as ULEB128
+    /// - Length of compressed data in bytes (or 0), as ULEB128
+    /// - The indicated number of compressed or uncompressed bytes
+    ///
+    /// If the number of compressed bytes is 0, the subsequent bytes are
+    /// uncompressed. Otherwise, the subsequent bytes are compressed, and will
+    /// be decompressed.
+    ///
+    /// Returns the uncompressed bytes that were read directly or decompressed.
+    pub(crate) fn read_chunk_to_uncompressed_bytes(&mut self) -> anyhow::Result<Cow<'a, [u8]>> {
+        let uncompressed_len = self.read_uleb128_usize()?;
+        let compressed_len = self.read_uleb128_usize()?;
+
+        if compressed_len == 0 {
+            // The bytes are uncompressed, so read them directly.
+            let uncompressed_bytes = self.read_n_bytes(uncompressed_len)?;
+            Ok(Cow::Borrowed(uncompressed_bytes))
+        } else {
+            // The bytes are compressed, so read and decompress them.
+            let compressed_bytes = self.read_n_bytes(compressed_len)?;
+
+            let uncompressed_bytes = miniz_oxide::inflate::decompress_to_vec_zlib_with_limit(
+                compressed_bytes,
+                uncompressed_len,
+            )
+            .map_err(|e| anyhow!("{e:?}"))?;
+            ensure!(uncompressed_bytes.len() == uncompressed_len);
+
+            Ok(Cow::Owned(uncompressed_bytes))
+        }
+    }
+}
diff --git a/src/tools/coverage-dump/src/parser/tests.rs b/src/tools/coverage-dump/src/llvm_utils/tests.rs
index a673606b9c4..506b0a6200b 100644
--- a/src/tools/coverage-dump/src/parser/tests.rs
+++ b/src/tools/coverage-dump/src/llvm_utils/tests.rs
@@ -1,9 +1,5 @@
 use super::unescape_llvm_string_contents;
 
-// WARNING: These tests don't necessarily run in CI, and were mainly used to
-// help track down problems when originally developing this tool.
-// (The tool is still tested indirectly by snapshot tests that rely on it.)
-
 // Tests for `unescape_llvm_string_contents`:
 
 #[test]
diff --git a/src/tools/coverage-dump/src/main.rs b/src/tools/coverage-dump/src/main.rs
index b21e3e292f2..2c76d2f2460 100644
--- a/src/tools/coverage-dump/src/main.rs
+++ b/src/tools/coverage-dump/src/main.rs
@@ -1,4 +1,6 @@
 mod covfun;
+mod covmap;
+mod llvm_utils;
 mod parser;
 mod prf_names;
 
@@ -17,8 +19,9 @@ fn main() -> anyhow::Result<()> {
     let llvm_ir_path = args.get(1).context("LLVM IR file not specified")?;
     let llvm_ir = std::fs::read_to_string(llvm_ir_path).context("couldn't read LLVM IR file")?;
 
+    let filename_tables = covmap::make_filename_tables(&llvm_ir)?;
     let function_names = crate::prf_names::make_function_names_table(&llvm_ir)?;
-    crate::covfun::dump_covfun_mappings(&llvm_ir, &function_names)?;
+    crate::covfun::dump_covfun_mappings(&llvm_ir, &filename_tables, &function_names)?;
 
     Ok(())
 }
diff --git a/src/tools/coverage-dump/src/parser.rs b/src/tools/coverage-dump/src/parser.rs
index 0bd4abdae3e..f26a57b43b3 100644
--- a/src/tools/coverage-dump/src/parser.rs
+++ b/src/tools/coverage-dump/src/parser.rs
@@ -1,38 +1,4 @@
-#[cfg(test)]
-mod tests;
-
-use std::sync::OnceLock;
-
 use anyhow::ensure;
-use regex::bytes;
-
-/// Given the raw contents of a string literal in LLVM IR assembly, decodes any
-/// backslash escapes and returns a vector containing the resulting byte string.
-pub(crate) fn unescape_llvm_string_contents(contents: &str) -> Vec<u8> {
-    let escape_re = {
-        static RE: OnceLock<bytes::Regex> = OnceLock::new();
-        // LLVM IR supports two string escapes: `\\` and `\xx`.
-        RE.get_or_init(|| bytes::Regex::new(r"\\\\|\\([0-9A-Za-z]{2})").unwrap())
-    };
-
-    fn u8_from_hex_digits(digits: &[u8]) -> u8 {
-        // We know that the input contains exactly 2 hex digits, so these calls
-        // should never fail.
-        assert_eq!(digits.len(), 2);
-        let digits = std::str::from_utf8(digits).unwrap();
-        u8::from_str_radix(digits, 16).unwrap()
-    }
-
-    escape_re
-        .replace_all(contents.as_bytes(), |captures: &bytes::Captures<'_>| {
-            let byte = match captures.get(1) {
-                None => b'\\',
-                Some(hex_digits) => u8_from_hex_digits(hex_digits.as_bytes()),
-            };
-            [byte]
-        })
-        .into_owned()
-}
 
 pub(crate) struct Parser<'a> {
     rest: &'a [u8],
diff --git a/src/tools/coverage-dump/src/prf_names.rs b/src/tools/coverage-dump/src/prf_names.rs
index 96d097c79a3..f9ab35deba5 100644
--- a/src/tools/coverage-dump/src/prf_names.rs
+++ b/src/tools/coverage-dump/src/prf_names.rs
@@ -1,10 +1,10 @@
 use std::collections::HashMap;
 use std::sync::OnceLock;
 
-use anyhow::{anyhow, ensure};
 use regex::Regex;
 
-use crate::parser::{Parser, unescape_llvm_string_contents};
+use crate::llvm_utils::{truncated_md5, unescape_llvm_string_contents};
+use crate::parser::Parser;
 
 /// Scans through the contents of an LLVM IR assembly file to find `__llvm_prf_names`
 /// entries, decodes them, and creates a table that maps name hash values to
@@ -25,18 +25,6 @@ pub(crate) fn make_function_names_table(llvm_ir: &str) -> anyhow::Result<HashMap
         Some(payload)
     }
 
-    /// LLVM's profiler/coverage metadata often uses an MD5 hash truncated to
-    /// 64 bits as a way to associate data stored in different tables/sections.
-    fn truncated_md5(bytes: &[u8]) -> u64 {
-        use md5::{Digest, Md5};
-        let mut hasher = Md5::new();
-        hasher.update(bytes);
-        let hash: [u8; 8] = hasher.finalize().as_slice()[..8].try_into().unwrap();
-        // The truncated hash is explicitly little-endian, regardless of host
-        // or target platform. (See `MD5Result::low` in LLVM's `MD5.h`.)
-        u64::from_le_bytes(hash)
-    }
-
     fn demangle_if_able(symbol_name_bytes: &[u8]) -> anyhow::Result<String> {
         // In practice, raw symbol names should always be ASCII.
         let symbol_name_str = std::str::from_utf8(symbol_name_bytes)?;
@@ -54,26 +42,8 @@ pub(crate) fn make_function_names_table(llvm_ir: &str) -> anyhow::Result<HashMap
     for payload in llvm_ir.lines().filter_map(prf_names_payload).map(unescape_llvm_string_contents)
     {
         let mut parser = Parser::new(&payload);
-        let uncompressed_len = parser.read_uleb128_usize()?;
-        let compressed_len = parser.read_uleb128_usize()?;
-
-        let uncompressed_bytes_vec;
-        let uncompressed_bytes: &[u8] = if compressed_len == 0 {
-            // The symbol name bytes are uncompressed, so read them directly.
-            parser.read_n_bytes(uncompressed_len)?
-        } else {
-            // The symbol name bytes are compressed, so read and decompress them.
-            let compressed_bytes = parser.read_n_bytes(compressed_len)?;
-
-            uncompressed_bytes_vec = miniz_oxide::inflate::decompress_to_vec_zlib_with_limit(
-                compressed_bytes,
-                uncompressed_len,
-            )
-            .map_err(|e| anyhow!("{e:?}"))?;
-            ensure!(uncompressed_bytes_vec.len() == uncompressed_len);
-
-            &uncompressed_bytes_vec
-        };
+        let uncompressed_bytes = parser.read_chunk_to_uncompressed_bytes()?;
+        parser.ensure_empty()?;
 
         // Symbol names in the payload are separated by `0x01` bytes.
         for raw_name in uncompressed_bytes.split(|&b| b == 0x01) {
@@ -81,8 +51,6 @@ pub(crate) fn make_function_names_table(llvm_ir: &str) -> anyhow::Result<HashMap
             let demangled = demangle_if_able(raw_name)?;
             map.insert(hash, demangled);
         }
-
-        parser.ensure_empty()?;
     }
 
     Ok(map)
diff --git a/src/tools/rust-analyzer/crates/parser/src/lexed_str.rs b/src/tools/rust-analyzer/crates/parser/src/lexed_str.rs
index 585e7ffb1ae..0a5c16dc4c4 100644
--- a/src/tools/rust-analyzer/crates/parser/src/lexed_str.rs
+++ b/src/tools/rust-analyzer/crates/parser/src/lexed_str.rs
@@ -179,6 +179,15 @@ impl<'a> Converter<'a> {
                     COMMENT
                 }
 
+                rustc_lexer::TokenKind::Frontmatter  { has_invalid_preceding_whitespace, invalid_infostring } => {
+                    if *has_invalid_preceding_whitespace {
+                        err = "invalid preceding whitespace for frontmatter opening"
+                    } else if *invalid_infostring {
+                        err = "invalid infostring for frontmatter"
+                    }
+                    FRONTMATTER
+                }
+
                 rustc_lexer::TokenKind::Whitespace => WHITESPACE,
 
                 rustc_lexer::TokenKind::Ident if token_text == "_" => UNDERSCORE,
diff --git a/src/tools/rust-analyzer/crates/parser/src/syntax_kind/generated.rs b/src/tools/rust-analyzer/crates/parser/src/syntax_kind/generated.rs
index e6f93a1fbda..b1727509b13 100644
--- a/src/tools/rust-analyzer/crates/parser/src/syntax_kind/generated.rs
+++ b/src/tools/rust-analyzer/crates/parser/src/syntax_kind/generated.rs
@@ -150,6 +150,7 @@ pub enum SyntaxKind {
     STRING,
     COMMENT,
     ERROR,
+    FRONTMATTER,
     IDENT,
     LIFETIME_IDENT,
     NEWLINE,
@@ -483,6 +484,7 @@ impl SyntaxKind {
             | YIELD_EXPR
             | COMMENT
             | ERROR
+            | FRONTMATTER
             | IDENT
             | LIFETIME_IDENT
             | NEWLINE
@@ -994,7 +996,7 @@ impl SyntaxKind {
     }
 }
 #[macro_export]
-macro_rules ! T_ { [$] => { $ crate :: SyntaxKind :: DOLLAR } ; [;] => { $ crate :: SyntaxKind :: SEMICOLON } ; [,] => { $ crate :: SyntaxKind :: COMMA } ; ['('] => { $ crate :: SyntaxKind :: L_PAREN } ; [')'] => { $ crate :: SyntaxKind :: R_PAREN } ; ['{'] => { $ crate :: SyntaxKind :: L_CURLY } ; ['}'] => { $ crate :: SyntaxKind :: R_CURLY } ; ['['] => { $ crate :: SyntaxKind :: L_BRACK } ; [']'] => { $ crate :: SyntaxKind :: R_BRACK } ; [<] => { $ crate :: SyntaxKind :: L_ANGLE } ; [>] => { $ crate :: SyntaxKind :: R_ANGLE } ; [@] => { $ crate :: SyntaxKind :: AT } ; [#] => { $ crate :: SyntaxKind :: POUND } ; [~] => { $ crate :: SyntaxKind :: TILDE } ; [?] => { $ crate :: SyntaxKind :: QUESTION } ; [&] => { $ crate :: SyntaxKind :: AMP } ; [|] => { $ crate :: SyntaxKind :: PIPE } ; [+] => { $ crate :: SyntaxKind :: PLUS } ; [*] => { $ crate :: SyntaxKind :: STAR } ; [/] => { $ crate :: SyntaxKind :: SLASH } ; [^] => { $ crate :: SyntaxKind :: CARET } ; [%] => { $ crate :: SyntaxKind :: PERCENT } ; [_] => { $ crate :: SyntaxKind :: UNDERSCORE } ; [.] => { $ crate :: SyntaxKind :: DOT } ; [..] => { $ crate :: SyntaxKind :: DOT2 } ; [...] => { $ crate :: SyntaxKind :: DOT3 } ; [..=] => { $ crate :: SyntaxKind :: DOT2EQ } ; [:] => { $ crate :: SyntaxKind :: COLON } ; [::] => { $ crate :: SyntaxKind :: COLON2 } ; [=] => { $ crate :: SyntaxKind :: EQ } ; [==] => { $ crate :: SyntaxKind :: EQ2 } ; [=>] => { $ crate :: SyntaxKind :: FAT_ARROW } ; [!] => { $ crate :: SyntaxKind :: BANG } ; [!=] => { $ crate :: SyntaxKind :: NEQ } ; [-] => { $ crate :: SyntaxKind :: MINUS } ; [->] => { $ crate :: SyntaxKind :: THIN_ARROW } ; [<=] => { $ crate :: SyntaxKind :: LTEQ } ; [>=] => { $ crate :: SyntaxKind :: GTEQ } ; [+=] => { $ crate :: SyntaxKind :: PLUSEQ } ; [-=] => { $ crate :: SyntaxKind :: MINUSEQ } ; [|=] => { $ crate :: SyntaxKind :: PIPEEQ } ; [&=] => { $ crate :: SyntaxKind :: AMPEQ } ; [^=] => { $ crate :: SyntaxKind :: CARETEQ } ; [/=] => { $ crate :: SyntaxKind :: SLASHEQ } ; [*=] => { $ crate :: SyntaxKind :: STAREQ } ; [%=] => { $ crate :: SyntaxKind :: PERCENTEQ } ; [&&] => { $ crate :: SyntaxKind :: AMP2 } ; [||] => { $ crate :: SyntaxKind :: PIPE2 } ; [<<] => { $ crate :: SyntaxKind :: SHL } ; [>>] => { $ crate :: SyntaxKind :: SHR } ; [<<=] => { $ crate :: SyntaxKind :: SHLEQ } ; [>>=] => { $ crate :: SyntaxKind :: SHREQ } ; [Self] => { $ crate :: SyntaxKind :: SELF_TYPE_KW } ; [abstract] => { $ crate :: SyntaxKind :: ABSTRACT_KW } ; [as] => { $ crate :: SyntaxKind :: AS_KW } ; [become] => { $ crate :: SyntaxKind :: BECOME_KW } ; [box] => { $ crate :: SyntaxKind :: BOX_KW } ; [break] => { $ crate :: SyntaxKind :: BREAK_KW } ; [const] => { $ crate :: SyntaxKind :: CONST_KW } ; [continue] => { $ crate :: SyntaxKind :: CONTINUE_KW } ; [crate] => { $ crate :: SyntaxKind :: CRATE_KW } ; [do] => { $ crate :: SyntaxKind :: DO_KW } ; [else] => { $ crate :: SyntaxKind :: ELSE_KW } ; [enum] => { $ crate :: SyntaxKind :: ENUM_KW } ; [extern] => { $ crate :: SyntaxKind :: EXTERN_KW } ; [false] => { $ crate :: SyntaxKind :: FALSE_KW } ; [final] => { $ crate :: SyntaxKind :: FINAL_KW } ; [fn] => { $ crate :: SyntaxKind :: FN_KW } ; [for] => { $ crate :: SyntaxKind :: FOR_KW } ; [if] => { $ crate :: SyntaxKind :: IF_KW } ; [impl] => { $ crate :: SyntaxKind :: IMPL_KW } ; [in] => { $ crate :: SyntaxKind :: IN_KW } ; [let] => { $ crate :: SyntaxKind :: LET_KW } ; [loop] => { $ crate :: SyntaxKind :: LOOP_KW } ; [macro] => { $ crate :: SyntaxKind :: MACRO_KW } ; [match] => { $ crate :: SyntaxKind :: MATCH_KW } ; [mod] => { $ crate :: SyntaxKind :: MOD_KW } ; [move] => { $ crate :: SyntaxKind :: MOVE_KW } ; [mut] => { $ crate :: SyntaxKind :: MUT_KW } ; [override] => { $ crate :: SyntaxKind :: OVERRIDE_KW } ; [priv] => { $ crate :: SyntaxKind :: PRIV_KW } ; [pub] => { $ crate :: SyntaxKind :: PUB_KW } ; [ref] => { $ crate :: SyntaxKind :: REF_KW } ; [return] => { $ crate :: SyntaxKind :: RETURN_KW } ; [self] => { $ crate :: SyntaxKind :: SELF_KW } ; [static] => { $ crate :: SyntaxKind :: STATIC_KW } ; [struct] => { $ crate :: SyntaxKind :: STRUCT_KW } ; [super] => { $ crate :: SyntaxKind :: SUPER_KW } ; [trait] => { $ crate :: SyntaxKind :: TRAIT_KW } ; [true] => { $ crate :: SyntaxKind :: TRUE_KW } ; [type] => { $ crate :: SyntaxKind :: TYPE_KW } ; [typeof] => { $ crate :: SyntaxKind :: TYPEOF_KW } ; [unsafe] => { $ crate :: SyntaxKind :: UNSAFE_KW } ; [unsized] => { $ crate :: SyntaxKind :: UNSIZED_KW } ; [use] => { $ crate :: SyntaxKind :: USE_KW } ; [virtual] => { $ crate :: SyntaxKind :: VIRTUAL_KW } ; [where] => { $ crate :: SyntaxKind :: WHERE_KW } ; [while] => { $ crate :: SyntaxKind :: WHILE_KW } ; [yield] => { $ crate :: SyntaxKind :: YIELD_KW } ; [asm] => { $ crate :: SyntaxKind :: ASM_KW } ; [att_syntax] => { $ crate :: SyntaxKind :: ATT_SYNTAX_KW } ; [auto] => { $ crate :: SyntaxKind :: AUTO_KW } ; [builtin] => { $ crate :: SyntaxKind :: BUILTIN_KW } ; [clobber_abi] => { $ crate :: SyntaxKind :: CLOBBER_ABI_KW } ; [default] => { $ crate :: SyntaxKind :: DEFAULT_KW } ; [dyn] => { $ crate :: SyntaxKind :: DYN_KW } ; [format_args] => { $ crate :: SyntaxKind :: FORMAT_ARGS_KW } ; [inlateout] => { $ crate :: SyntaxKind :: INLATEOUT_KW } ; [inout] => { $ crate :: SyntaxKind :: INOUT_KW } ; [label] => { $ crate :: SyntaxKind :: LABEL_KW } ; [lateout] => { $ crate :: SyntaxKind :: LATEOUT_KW } ; [macro_rules] => { $ crate :: SyntaxKind :: MACRO_RULES_KW } ; [may_unwind] => { $ crate :: SyntaxKind :: MAY_UNWIND_KW } ; [nomem] => { $ crate :: SyntaxKind :: NOMEM_KW } ; [noreturn] => { $ crate :: SyntaxKind :: NORETURN_KW } ; [nostack] => { $ crate :: SyntaxKind :: NOSTACK_KW } ; [offset_of] => { $ crate :: SyntaxKind :: OFFSET_OF_KW } ; [options] => { $ crate :: SyntaxKind :: OPTIONS_KW } ; [out] => { $ crate :: SyntaxKind :: OUT_KW } ; [preserves_flags] => { $ crate :: SyntaxKind :: PRESERVES_FLAGS_KW } ; [pure] => { $ crate :: SyntaxKind :: PURE_KW } ; [raw] => { $ crate :: SyntaxKind :: RAW_KW } ; [readonly] => { $ crate :: SyntaxKind :: READONLY_KW } ; [safe] => { $ crate :: SyntaxKind :: SAFE_KW } ; [sym] => { $ crate :: SyntaxKind :: SYM_KW } ; [union] => { $ crate :: SyntaxKind :: UNION_KW } ; [yeet] => { $ crate :: SyntaxKind :: YEET_KW } ; [async] => { $ crate :: SyntaxKind :: ASYNC_KW } ; [await] => { $ crate :: SyntaxKind :: AWAIT_KW } ; [dyn] => { $ crate :: SyntaxKind :: DYN_KW } ; [gen] => { $ crate :: SyntaxKind :: GEN_KW } ; [try] => { $ crate :: SyntaxKind :: TRY_KW } ; [lifetime_ident] => { $ crate :: SyntaxKind :: LIFETIME_IDENT } ; [int_number] => { $ crate :: SyntaxKind :: INT_NUMBER } ; [ident] => { $ crate :: SyntaxKind :: IDENT } ; [string] => { $ crate :: SyntaxKind :: STRING } ; [shebang] => { $ crate :: SyntaxKind :: SHEBANG } ; }
+macro_rules ! T_ { [$] => { $ crate :: SyntaxKind :: DOLLAR } ; [;] => { $ crate :: SyntaxKind :: SEMICOLON } ; [,] => { $ crate :: SyntaxKind :: COMMA } ; ['('] => { $ crate :: SyntaxKind :: L_PAREN } ; [')'] => { $ crate :: SyntaxKind :: R_PAREN } ; ['{'] => { $ crate :: SyntaxKind :: L_CURLY } ; ['}'] => { $ crate :: SyntaxKind :: R_CURLY } ; ['['] => { $ crate :: SyntaxKind :: L_BRACK } ; [']'] => { $ crate :: SyntaxKind :: R_BRACK } ; [<] => { $ crate :: SyntaxKind :: L_ANGLE } ; [>] => { $ crate :: SyntaxKind :: R_ANGLE } ; [@] => { $ crate :: SyntaxKind :: AT } ; [#] => { $ crate :: SyntaxKind :: POUND } ; [~] => { $ crate :: SyntaxKind :: TILDE } ; [?] => { $ crate :: SyntaxKind :: QUESTION } ; [&] => { $ crate :: SyntaxKind :: AMP } ; [|] => { $ crate :: SyntaxKind :: PIPE } ; [+] => { $ crate :: SyntaxKind :: PLUS } ; [*] => { $ crate :: SyntaxKind :: STAR } ; [/] => { $ crate :: SyntaxKind :: SLASH } ; [^] => { $ crate :: SyntaxKind :: CARET } ; [%] => { $ crate :: SyntaxKind :: PERCENT } ; [_] => { $ crate :: SyntaxKind :: UNDERSCORE } ; [.] => { $ crate :: SyntaxKind :: DOT } ; [..] => { $ crate :: SyntaxKind :: DOT2 } ; [...] => { $ crate :: SyntaxKind :: DOT3 } ; [..=] => { $ crate :: SyntaxKind :: DOT2EQ } ; [:] => { $ crate :: SyntaxKind :: COLON } ; [::] => { $ crate :: SyntaxKind :: COLON2 } ; [=] => { $ crate :: SyntaxKind :: EQ } ; [==] => { $ crate :: SyntaxKind :: EQ2 } ; [=>] => { $ crate :: SyntaxKind :: FAT_ARROW } ; [!] => { $ crate :: SyntaxKind :: BANG } ; [!=] => { $ crate :: SyntaxKind :: NEQ } ; [-] => { $ crate :: SyntaxKind :: MINUS } ; [->] => { $ crate :: SyntaxKind :: THIN_ARROW } ; [<=] => { $ crate :: SyntaxKind :: LTEQ } ; [>=] => { $ crate :: SyntaxKind :: GTEQ } ; [+=] => { $ crate :: SyntaxKind :: PLUSEQ } ; [-=] => { $ crate :: SyntaxKind :: MINUSEQ } ; [|=] => { $ crate :: SyntaxKind :: PIPEEQ } ; [&=] => { $ crate :: SyntaxKind :: AMPEQ } ; [^=] => { $ crate :: SyntaxKind :: CARETEQ } ; [/=] => { $ crate :: SyntaxKind :: SLASHEQ } ; [*=] => { $ crate :: SyntaxKind :: STAREQ } ; [%=] => { $ crate :: SyntaxKind :: PERCENTEQ } ; [&&] => { $ crate :: SyntaxKind :: AMP2 } ; [||] => { $ crate :: SyntaxKind :: PIPE2 } ; [<<] => { $ crate :: SyntaxKind :: SHL } ; [>>] => { $ crate :: SyntaxKind :: SHR } ; [<<=] => { $ crate :: SyntaxKind :: SHLEQ } ; [>>=] => { $ crate :: SyntaxKind :: SHREQ } ; [Self] => { $ crate :: SyntaxKind :: SELF_TYPE_KW } ; [abstract] => { $ crate :: SyntaxKind :: ABSTRACT_KW } ; [as] => { $ crate :: SyntaxKind :: AS_KW } ; [become] => { $ crate :: SyntaxKind :: BECOME_KW } ; [box] => { $ crate :: SyntaxKind :: BOX_KW } ; [break] => { $ crate :: SyntaxKind :: BREAK_KW } ; [const] => { $ crate :: SyntaxKind :: CONST_KW } ; [continue] => { $ crate :: SyntaxKind :: CONTINUE_KW } ; [crate] => { $ crate :: SyntaxKind :: CRATE_KW } ; [do] => { $ crate :: SyntaxKind :: DO_KW } ; [else] => { $ crate :: SyntaxKind :: ELSE_KW } ; [enum] => { $ crate :: SyntaxKind :: ENUM_KW } ; [extern] => { $ crate :: SyntaxKind :: EXTERN_KW } ; [false] => { $ crate :: SyntaxKind :: FALSE_KW } ; [final] => { $ crate :: SyntaxKind :: FINAL_KW } ; [fn] => { $ crate :: SyntaxKind :: FN_KW } ; [for] => { $ crate :: SyntaxKind :: FOR_KW } ; [if] => { $ crate :: SyntaxKind :: IF_KW } ; [impl] => { $ crate :: SyntaxKind :: IMPL_KW } ; [in] => { $ crate :: SyntaxKind :: IN_KW } ; [let] => { $ crate :: SyntaxKind :: LET_KW } ; [loop] => { $ crate :: SyntaxKind :: LOOP_KW } ; [macro] => { $ crate :: SyntaxKind :: MACRO_KW } ; [match] => { $ crate :: SyntaxKind :: MATCH_KW } ; [mod] => { $ crate :: SyntaxKind :: MOD_KW } ; [move] => { $ crate :: SyntaxKind :: MOVE_KW } ; [mut] => { $ crate :: SyntaxKind :: MUT_KW } ; [override] => { $ crate :: SyntaxKind :: OVERRIDE_KW } ; [priv] => { $ crate :: SyntaxKind :: PRIV_KW } ; [pub] => { $ crate :: SyntaxKind :: PUB_KW } ; [ref] => { $ crate :: SyntaxKind :: REF_KW } ; [return] => { $ crate :: SyntaxKind :: RETURN_KW } ; [self] => { $ crate :: SyntaxKind :: SELF_KW } ; [static] => { $ crate :: SyntaxKind :: STATIC_KW } ; [struct] => { $ crate :: SyntaxKind :: STRUCT_KW } ; [super] => { $ crate :: SyntaxKind :: SUPER_KW } ; [trait] => { $ crate :: SyntaxKind :: TRAIT_KW } ; [true] => { $ crate :: SyntaxKind :: TRUE_KW } ; [type] => { $ crate :: SyntaxKind :: TYPE_KW } ; [typeof] => { $ crate :: SyntaxKind :: TYPEOF_KW } ; [unsafe] => { $ crate :: SyntaxKind :: UNSAFE_KW } ; [unsized] => { $ crate :: SyntaxKind :: UNSIZED_KW } ; [use] => { $ crate :: SyntaxKind :: USE_KW } ; [virtual] => { $ crate :: SyntaxKind :: VIRTUAL_KW } ; [where] => { $ crate :: SyntaxKind :: WHERE_KW } ; [while] => { $ crate :: SyntaxKind :: WHILE_KW } ; [yield] => { $ crate :: SyntaxKind :: YIELD_KW } ; [asm] => { $ crate :: SyntaxKind :: ASM_KW } ; [att_syntax] => { $ crate :: SyntaxKind :: ATT_SYNTAX_KW } ; [auto] => { $ crate :: SyntaxKind :: AUTO_KW } ; [builtin] => { $ crate :: SyntaxKind :: BUILTIN_KW } ; [clobber_abi] => { $ crate :: SyntaxKind :: CLOBBER_ABI_KW } ; [default] => { $ crate :: SyntaxKind :: DEFAULT_KW } ; [dyn] => { $ crate :: SyntaxKind :: DYN_KW } ; [format_args] => { $ crate :: SyntaxKind :: FORMAT_ARGS_KW } ; [inlateout] => { $ crate :: SyntaxKind :: INLATEOUT_KW } ; [inout] => { $ crate :: SyntaxKind :: INOUT_KW } ; [label] => { $ crate :: SyntaxKind :: LABEL_KW } ; [lateout] => { $ crate :: SyntaxKind :: LATEOUT_KW } ; [macro_rules] => { $ crate :: SyntaxKind :: MACRO_RULES_KW } ; [may_unwind] => { $ crate :: SyntaxKind :: MAY_UNWIND_KW } ; [nomem] => { $ crate :: SyntaxKind :: NOMEM_KW } ; [noreturn] => { $ crate :: SyntaxKind :: NORETURN_KW } ; [nostack] => { $ crate :: SyntaxKind :: NOSTACK_KW } ; [offset_of] => { $ crate :: SyntaxKind :: OFFSET_OF_KW } ; [options] => { $ crate :: SyntaxKind :: OPTIONS_KW } ; [out] => { $ crate :: SyntaxKind :: OUT_KW } ; [preserves_flags] => { $ crate :: SyntaxKind :: PRESERVES_FLAGS_KW } ; [pure] => { $ crate :: SyntaxKind :: PURE_KW } ; [raw] => { $ crate :: SyntaxKind :: RAW_KW } ; [readonly] => { $ crate :: SyntaxKind :: READONLY_KW } ; [safe] => { $ crate :: SyntaxKind :: SAFE_KW } ; [sym] => { $ crate :: SyntaxKind :: SYM_KW } ; [union] => { $ crate :: SyntaxKind :: UNION_KW } ; [yeet] => { $ crate :: SyntaxKind :: YEET_KW } ; [async] => { $ crate :: SyntaxKind :: ASYNC_KW } ; [await] => { $ crate :: SyntaxKind :: AWAIT_KW } ; [dyn] => { $ crate :: SyntaxKind :: DYN_KW } ; [gen] => { $ crate :: SyntaxKind :: GEN_KW } ; [try] => { $ crate :: SyntaxKind :: TRY_KW } ; [lifetime_ident] => { $ crate :: SyntaxKind :: LIFETIME_IDENT } ; [int_number] => { $ crate :: SyntaxKind :: INT_NUMBER } ; [ident] => { $ crate :: SyntaxKind :: IDENT } ; [string] => { $ crate :: SyntaxKind :: STRING } ; [shebang] => { $ crate :: SyntaxKind :: SHEBANG } ; [frontmatter] => { $ crate :: SyntaxKind :: FRONTMATTER } ; }
 impl ::core::marker::Copy for SyntaxKind {}
 impl ::core::clone::Clone for SyntaxKind {
     #[inline]
diff --git a/src/tools/rust-analyzer/crates/syntax/rust.ungram b/src/tools/rust-analyzer/crates/syntax/rust.ungram
index a0ae0d68581..10abca7d35d 100644
--- a/src/tools/rust-analyzer/crates/syntax/rust.ungram
+++ b/src/tools/rust-analyzer/crates/syntax/rust.ungram
@@ -133,6 +133,7 @@ Meta =
 
 SourceFile =
   '#shebang'?
+  '#frontmatter'?
   Attr*
   Item*
 
diff --git a/src/tools/rust-analyzer/crates/syntax/src/ast/generated/nodes.rs b/src/tools/rust-analyzer/crates/syntax/src/ast/generated/nodes.rs
index 1243f6418fe..cd9f4dba890 100644
--- a/src/tools/rust-analyzer/crates/syntax/src/ast/generated/nodes.rs
+++ b/src/tools/rust-analyzer/crates/syntax/src/ast/generated/nodes.rs
@@ -1525,6 +1525,10 @@ impl ast::HasDocComments for SourceFile {}
 impl ast::HasModuleItem for SourceFile {}
 impl SourceFile {
     #[inline]
+    pub fn frontmatter_token(&self) -> Option<SyntaxToken> {
+        support::token(&self.syntax, T![frontmatter])
+    }
+    #[inline]
     pub fn shebang_token(&self) -> Option<SyntaxToken> { support::token(&self.syntax, T![shebang]) }
 }
 pub struct Static {
diff --git a/src/tools/rust-analyzer/xtask/src/codegen/grammar.rs b/src/tools/rust-analyzer/xtask/src/codegen/grammar.rs
index 4b9c6edbe30..824b38fc872 100644
--- a/src/tools/rust-analyzer/xtask/src/codegen/grammar.rs
+++ b/src/tools/rust-analyzer/xtask/src/codegen/grammar.rs
@@ -670,6 +670,7 @@ fn generate_syntax_kinds(grammar: KindsSrc) -> String {
             [ident] => { $crate::SyntaxKind::IDENT };
             [string] => { $crate::SyntaxKind::STRING };
             [shebang] => { $crate::SyntaxKind::SHEBANG };
+            [frontmatter] => { $crate::SyntaxKind::FRONTMATTER };
         }
 
         impl ::core::marker::Copy for SyntaxKind {}
diff --git a/src/tools/rustbook/Cargo.lock b/src/tools/rustbook/Cargo.lock
index 81e0f6d572b..3b5b31bed14 100644
--- a/src/tools/rustbook/Cargo.lock
+++ b/src/tools/rustbook/Cargo.lock
@@ -938,9 +938,9 @@ dependencies = [
 
 [[package]]
 name = "mdbook"
-version = "0.4.48"
+version = "0.4.49"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8b6fbb4ac2d9fd7aa987c3510309ea3c80004a968d063c42f0d34fea070817c1"
+checksum = "d1daacee059634081dee4250d2814763a365b92dfe14bfdef964bc27835209d4"
 dependencies = [
  "ammonia",
  "anyhow",
diff --git a/src/tools/rustbook/Cargo.toml b/src/tools/rustbook/Cargo.toml
index a0b220c3557..10fde31306d 100644
--- a/src/tools/rustbook/Cargo.toml
+++ b/src/tools/rustbook/Cargo.toml
@@ -15,6 +15,6 @@ mdbook-i18n-helpers = "0.3.3"
 mdbook-spec = { path = "../../doc/reference/mdbook-spec" }
 
 [dependencies.mdbook]
-version = "0.4.48"
+version = "0.4.49"
 default-features = false
 features = ["search"]