Auto merge of rust-lang#95458 - calebcartwright:sync-rustfmt-subtree, r=calebcartwright

Sync rustfmt subtree

r? `@calebcartwright`

Author: bors

src/tools/rustfmt/.github/workflows/linux.yml

@@ -40,6 +40,10 @@ jobs:
         rustc -Vv
         cargo -V
         cargo build
+      env:
+        RUSTFLAGS: '-D warnings'
 
     - name: test
       run: cargo test
+      env:
+        RUSTFLAGS: '-D warnings'

src/tools/rustfmt/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## [Unreleased]
 
+### Fixed
+
+- Fixes issue where wrapped strings would be incorrectly indented in macro defs when `format_strings` was enabled [#4036](https://github.com/rust-lang/rustfmt/issues/4036)
+
 ## [1.4.38] 2021-10-20
 
 ### Changed
@@ -57,6 +61,7 @@ Note this hit the rustup distributions prior to the v1.4.38 release as part of a
 
 - New `One` variant added to `imports_granularity` configuration option which can be used to reformat all imports into a single use statement [#4669](https://github.com/rust-lang/rustfmt/issues/4669)
 - rustfmt will now skip files that are annotated with `@generated` at the top of the file [#3958](https://github.com/rust-lang/rustfmt/issues/3958)
+  if `format_generated_files` option is set to `false` (by default `@generated` files are formatted)
 - New configuration option `hex_literal_case` that allows user to control the casing utilized for hex literals [PR #4903](https://github.com/rust-lang/rustfmt/pull/4903)
 
 See the section on the configuration site for more information

src/tools/rustfmt/Cargo.lock

@@ -275,9 +275,9 @@ dependencies = [
 
 [[package]]
 name = "itertools"
-version = "0.9.0"
+version = "0.10.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "284f18f85651fe11e8a991b2adb42cb078325c996ed026d994719efcfca1d54b"
+checksum = "a9a9d19fa1e79b6215ff29b9d6880b706147f16e9b1dbb1e4e5947b5b02bc5e3"
 dependencies = [
  "either",
 ]

src/tools/rustfmt/Configurations.md

@@ -930,6 +930,8 @@ fn add_one(x: i32) -> i32 {
 
 Format generated files. A file is considered generated
 if any of the first five lines contain a `@generated` comment marker.
+By default, generated files are reformatted, i. e. `@generated` marker is ignored.
+This option is currently ignored for stdin (`@generated` in stdin is ignored.)
 
 - **Default value**: `true`
 - **Possible values**: `true`, `false`
@@ -2198,13 +2200,47 @@ specific version of rustfmt is used in your CI, use this option.
 - **Possible values**: any published version (e.g. `"0.3.8"`)
 - **Stable**: No (tracking issue: [#3386](https://github.com/rust-lang/rustfmt/issues/3386))
 
+## `short_array_element_width_threshold`
+
+The width threshold for an array element to be considered "short".
+
+The layout of an array is dependent on the length of each of its elements. 
+If the length of every element in an array is below this threshold (all elements are "short") then the array can be formatted in the mixed/compressed style, but if any one element has a length that exceeds this threshold then the array elements will have to be formatted vertically.
+
+- **Default value**: `10`
+- **Possible values**: any positive integer that is less than or equal to the value specified for [`max_width`](#max_width)
+- **Stable**: Yes
+
+#### `10` (default):
+```rust
+fn main() {
+    pub const FORMAT_TEST: [u64; 5] = [
+        0x0000000000000000,
+        0xaaaaaaaaaaaaaaaa,
+        0xbbbbbbbbbbbbbbbb,
+        0xcccccccccccccccc,
+        0xdddddddddddddddd,
+    ];
+}
+```
+#### `20`:
+```rust
+fn main() {
+    pub const FORMAT_TEST: [u64; 5] = [
+        0x0000000000000000, 0xaaaaaaaaaaaaaaaa, 0xbbbbbbbbbbbbbbbb, 0xcccccccccccccccc,
+        0xdddddddddddddddd,
+    ];
+}
+```
+See also [`max_width`](#max_width).
+
 ## `skip_children`
 
 Don't reformat out of line modules
 
 - **Default value**: `false`
 - **Possible values**: `true`, `false`
-- **Stable**: No (tracking issue: [#3389](https://github.com/rust-lang/rustfmt/issues/3386))
+- **Stable**: No (tracking issue: [#3389](https://github.com/rust-lang/rustfmt/issues/3389))
 
 ## `single_line_if_else_max_width`
 

src/tools/rustfmt/config_proc_macro/src/utils.rs

@@ -22,10 +22,10 @@ pub fn is_unit(v: &syn::Variant) -> bool {
 #[cfg(feature = "debug-with-rustfmt")]
 /// Pretty-print the output of proc macro using rustfmt.
 pub fn debug_with_rustfmt(input: &TokenStream) {
-    use std::io::Write;
-    use std::process::{Command, Stdio};
     use std::env;
     use std::ffi::OsStr;
+    use std::io::Write;
+    use std::process::{Command, Stdio};
 
     let rustfmt_var = env::var_os("RUSTFMT");
     let rustfmt = match &rustfmt_var {

src/tools/rustfmt/rust-toolchain

@@ -1,3 +1,3 @@
 [toolchain]
-channel = "nightly-2021-12-29"
+channel = "nightly-2022-03-27"
 components = ["rustc-dev"]

src/tools/rustfmt/src/attr.rs

@@ -389,6 +389,10 @@ impl Rewrite for [ast::Attribute] {
         let mut attrs = self;
         let mut result = String::new();
 
+        // Determine if the source text is annotated with `#[rustfmt::skip::attributes(derive)]`
+        // or `#![rustfmt::skip::attributes(derive)]`
+        let skip_derives = context.skip_context.skip_attribute("derive");
+
         // This is not just a simple map because we need to handle doc comments
         // (where we take as many doc comment attributes as possible) and possibly
         // merging derives into a single attribute.
@@ -431,7 +435,7 @@ impl Rewrite for [ast::Attribute] {
             }
 
             // Handle derives if we will merge them.
-            if context.config.merge_derives() && is_derive(&attrs[0]) {
+            if !skip_derives && context.config.merge_derives() && is_derive(&attrs[0]) {
                 let derives = take_while_with_pred(context, attrs, is_derive);
                 let derive_str = format_derive(derives, shape, context)?;
                 result.push_str(&derive_str);

src/tools/rustfmt/src/bin/main.rs

@@ -26,7 +26,7 @@ fn main() {
     let exit_code = match execute(&opts) {
         Ok(code) => code,
         Err(e) => {
-            eprintln!("{}", e);
+            eprintln!("{:#}", e);
             1
         }
     };
@@ -74,14 +74,10 @@ pub enum OperationError {
     /// An io error during reading or writing.
     #[error("{0}")]
     IoError(IoError),
-    /// Attempt to use --check with stdin, which isn't currently
-    /// supported.
-    #[error("The `--check` option is not supported with standard input.")]
-    CheckWithStdin,
-    /// Attempt to use --emit=json with stdin, which isn't currently
-    /// supported.
-    #[error("Using `--emit` other than stdout is not supported with standard input.")]
-    EmitWithStdin,
+    /// Attempt to use --emit with a mode which is not currently
+    /// supported with stdandard input.
+    #[error("Emit mode {0} not supported with standard output.")]
+    StdinBadEmit(EmitMode),
 }
 
 impl From<IoError> for OperationError {
@@ -255,15 +251,20 @@ fn format_string(input: String, options: GetOptsOptions) -> Result<i32> {
     let (mut config, _) = load_config(Some(Path::new(".")), Some(options.clone()))?;
 
     if options.check {
-        return Err(OperationError::CheckWithStdin.into());
-    }
-    if let Some(emit_mode) = options.emit_mode {
-        if emit_mode != EmitMode::Stdout {
-            return Err(OperationError::EmitWithStdin.into());
+        config.set().emit_mode(EmitMode::Diff);
+    } else {
+        match options.emit_mode {
+            // Emit modes which work with standard input
+            // None means default, which is Stdout.
+            None | Some(EmitMode::Stdout) | Some(EmitMode::Checkstyle) | Some(EmitMode::Json) => {}
+            Some(emit_mode) => {
+                return Err(OperationError::StdinBadEmit(emit_mode).into());
+            }
         }
+        config
+            .set()
+            .emit_mode(options.emit_mode.unwrap_or(EmitMode::Stdout));
     }
-    // emit mode is always Stdout for Stdin.
-    config.set().emit_mode(EmitMode::Stdout);
     config.set().verbose(Verbosity::Quiet);
 
     // parse file_lines
@@ -393,9 +394,8 @@ fn print_usage_to_stdout(opts: &Options, reason: &str) {
         format!("{}\n\n", reason)
     };
     let msg = format!(
-        "{}Format Rust code\n\nusage: {} [options] <file>...",
-        sep,
-        env::args_os().next().unwrap().to_string_lossy()
+        "{}Format Rust code\n\nusage: rustfmt [options] <file>...",
+        sep
     );
     println!("{}", opts.usage(&msg));
 }

src/tools/rustfmt/src/cargo-fmt/main.rs

@@ -387,8 +387,7 @@ fn get_targets_root_only(
                         .unwrap_or_default()
                         == current_dir_manifest
             })
-            .map(|p| p.targets)
-            .flatten()
+            .flat_map(|p| p.targets)
             .collect(),
     };
 

src/tools/rustfmt/src/comment.rs

@@ -432,36 +432,49 @@ impl CodeBlockAttribute {
 
 /// Block that is formatted as an item.
 ///
-/// An item starts with either a star `*` or a dash `-`. Different level of indentation are
-/// handled by shrinking the shape accordingly.
+/// An item starts with either a star `*` a dash `-` or a greater-than `>`.
+/// Different level of indentation are handled by shrinking the shape accordingly.
 struct ItemizedBlock {
     /// the lines that are identified as part of an itemized block
     lines: Vec<String>,
-    /// the number of whitespaces up to the item sigil
+    /// the number of characters (typically whitespaces) up to the item sigil
     indent: usize,
     /// the string that marks the start of an item
     opener: String,
-    /// sequence of whitespaces to prefix new lines that are part of the item
+    /// sequence of characters (typically whitespaces) to prefix new lines that are part of the item
     line_start: String,
 }
 
 impl ItemizedBlock {
     /// Returns `true` if the line is formatted as an item
     fn is_itemized_line(line: &str) -> bool {
         let trimmed = line.trim_start();
-        trimmed.starts_with("* ") || trimmed.starts_with("- ")
+        trimmed.starts_with("* ") || trimmed.starts_with("- ") || trimmed.starts_with("> ")
     }
 
     /// Creates a new ItemizedBlock described with the given line.
     /// The `is_itemized_line` needs to be called first.
     fn new(line: &str) -> ItemizedBlock {
         let space_to_sigil = line.chars().take_while(|c| c.is_whitespace()).count();
-        let indent = space_to_sigil + 2;
+        // +2 = '* ', which will add the appropriate amount of whitespace to keep itemized
+        // content formatted correctly.
+        let mut indent = space_to_sigil + 2;
+        let mut line_start = " ".repeat(indent);
+
+        // Markdown blockquote start with a "> "
+        if line.trim_start().starts_with(">") {
+            // remove the original +2 indent because there might be multiple nested block quotes
+            // and it's easier to reason about the final indent by just taking the length
+            // of th new line_start. We update the indent because it effects the max width
+            // of each formatted line.
+            line_start = itemized_block_quote_start(line, line_start, 2);
+            indent = line_start.len();
+        }
         ItemizedBlock {
             lines: vec![line[indent..].to_string()],
             indent,
             opener: line[..indent].to_string(),
-            line_start: " ".repeat(indent),
+            line_start,
         }
     }
 
@@ -504,6 +517,25 @@ impl ItemizedBlock {
     }
 }
 
+/// Determine the line_start when formatting markdown block quotes.
+/// The original line_start likely contains indentation (whitespaces), which we'd like to
+/// replace with '> ' characters.
+fn itemized_block_quote_start(line: &str, mut line_start: String, remove_indent: usize) -> String {
+    let quote_level = line
+        .chars()
+        .take_while(|c| !c.is_alphanumeric())
+        .fold(0, |acc, c| if c == '>' { acc + 1 } else { acc });
+
+    for _ in 0..remove_indent {
+        line_start.pop();
+    }
+
+    for _ in 0..quote_level {
+        line_start.push_str("> ")
+    }
+    line_start
+}
+
 struct CommentRewrite<'a> {
     result: String,
     code_block_buffer: String,
@@ -651,6 +683,7 @@ impl<'a> CommentRewrite<'a> {
         i: usize,
         line: &'a str,
         has_leading_whitespace: bool,
+        is_doc_comment: bool,
     ) -> bool {
         let num_newlines = count_newlines(orig);
         let is_last = i == num_newlines;
@@ -757,10 +790,19 @@ impl<'a> CommentRewrite<'a> {
             }
         }
 
-        if self.fmt.config.wrap_comments()
+        let is_markdown_header_doc_comment = is_doc_comment && line.starts_with("#");
+
+        // We only want to wrap the comment if:
+        // 1) wrap_comments = true is configured
+        // 2) The comment is not the start of a markdown header doc comment
+        // 3) The comment width exceeds the shape's width
+        // 4) No URLS were found in the commnet
+        let should_wrap_comment = self.fmt.config.wrap_comments()
+            && !is_markdown_header_doc_comment
             && unicode_str_width(line) > self.fmt.shape.width
-            && !has_url(line)
-        {
+            && !has_url(line);
+
+        if should_wrap_comment {
             match rewrite_string(line, &self.fmt, self.max_width) {
                 Some(ref s) => {
                     self.is_prev_line_multi_line = s.contains('\n');
@@ -850,7 +892,7 @@ fn rewrite_comment_inner(
         });
 
     for (i, (line, has_leading_whitespace)) in lines.enumerate() {
-        if rewriter.handle_line(orig, i, line, has_leading_whitespace) {
+        if rewriter.handle_line(orig, i, line, has_leading_whitespace, is_doc_comment) {
             break;
         }
     }