src/tools/tidy/src/style.rs

   1 //! Tidy check to enforce various stylistic guidelines on the Rust codebase.
   2 //!
   3 //! Example checks are:
   4 //!
   5 //! * No lines over 100 characters (in non-Rust files).
   6 //! * No files with over 3000 lines (in non-Rust files).
   7 //! * No tabs.
   8 //! * No trailing whitespace.
   9 //! * No CR characters.
  10 //! * No `TODO` or `XXX` directives.
  11 //! * No unexplained ` ```ignore ` or ` ```rust,ignore ` doc tests.
  12 //!
  13 //! Note that some of these rules are excluded from Rust files because we enforce rustfmt. It is
  14 //! preferable to be formatted rather than tidy-clean.
  15 //!
  16 //! A number of these checks can be opted-out of with various directives of the form:
  17 //! `// ignore-tidy-CHECK-NAME`.
  18
  19 use regex::Regex;
  20 use std::path::Path;
  21
  22 /// Error code markdown is restricted to 80 columns because they can be
  23 /// displayed on the console with --example.
  24 const ERROR_CODE_COLS: usize = 80;
  25 const COLS: usize = 100;
  26
  27 const LINES: usize = 3000;
  28
  29 const UNEXPLAINED_IGNORE_DOCTEST_INFO: &str = r#"unexplained "```ignore" doctest; try one:
  30
  31 * make the test actually pass, by adding necessary imports and declarations, or
  32 * use "```text", if the code is not Rust code, or
  33 * use "```compile_fail,Ennnn", if the code is expected to fail at compile time, or
  34 * use "```should_panic", if the code is expected to fail at run time, or
  35 * use "```no_run", if the code should type-check but not necessary linkable/runnable, or
  36 * explain it like "```ignore (cannot-test-this-because-xxxx)", if the annotation cannot be avoided.
  37
  38 "#;
  39
  40 const LLVM_UNREACHABLE_INFO: &str = r"\
  41 C++ code used llvm_unreachable, which triggers undefined behavior
  42 when executed when assertions are disabled.
  43 Use llvm::report_fatal_error for increased robustness.";
  44
  45 const ANNOTATIONS_TO_IGNORE: &[&str] = &[
  46     "// @!has",
  47     "// @has",
  48     "// @matches",
  49     "// CHECK",
  50     "// EMIT_MIR",
  51     "// compile-flags",
  52     "// error-pattern",
  53     "// gdb",
  54     "// lldb",
  55     "// cdb",
  56     "// normalize-stderr-test",
  57 ];
  58
  59 // Intentionally written in decimal rather than hex
  60 const PROBLEMATIC_CONSTS: &[u32] = &[
  61     184594741, 2880289470, 2881141438, 2965027518, 2976579765, 3203381950, 3405691582, 3405697037,
  62     3735927486, 4027431614, 4276992702,
  63 ];
  64
  65 /// Parser states for `line_is_url`.
  66 #[derive(Clone, Copy, PartialEq)]
  67 #[allow(non_camel_case_types)]
  68 enum LIUState {
  69     EXP_COMMENT_START,
  70     EXP_LINK_LABEL_OR_URL,
  71     EXP_URL,
  72     EXP_END,
  73 }
  74
  75 /// Returns `true` if `line` appears to be a line comment containing a URL,
  76 /// possibly with a Markdown link label in front, and nothing else.
  77 /// The Markdown link label, if present, may not contain whitespace.
  78 /// Lines of this form are allowed to be overlength, because Markdown
  79 /// offers no way to split a line in the middle of a URL, and the lengths
  80 /// of URLs to external references are beyond our control.
  81 fn line_is_url(is_error_code: bool, columns: usize, line: &str) -> bool {
  82     // more basic check for markdown, to avoid complexity in implementing two state machines
  83     if is_error_code {
  84         return line.starts_with('[') && line.contains("]:") && line.contains("http");
  85     }
  86
  87     use self::LIUState::*;
  88     let mut state: LIUState = EXP_COMMENT_START;
  89     let is_url = |w: &str| w.starts_with("http://") || w.starts_with("https://");
  90
  91     for tok in line.split_whitespace() {
  92         match (state, tok) {
  93             (EXP_COMMENT_START, "//") | (EXP_COMMENT_START, "///") | (EXP_COMMENT_START, "//!") => {
  94                 state = EXP_LINK_LABEL_OR_URL
  95             }
  96
  97             (EXP_LINK_LABEL_OR_URL, w)
  98                 if w.len() >= 4 && w.starts_with('[') && w.ends_with("]:") =>
  99             {
 100                 state = EXP_URL
 101             }
 102
 103             (EXP_LINK_LABEL_OR_URL, w) if is_url(w) => state = EXP_END,
 104
 105             (EXP_URL, w) if is_url(w) || w.starts_with("../") => state = EXP_END,
 106
 107             (_, w) if w.len() > columns && is_url(w) => state = EXP_END,
 108
 109             (_, _) => {}
 110         }
 111     }
 112
 113     state == EXP_END
 114 }
 115
 116 /// Returns `true` if `line` can be ignored. This is the case when it contains
 117 /// an annotation that is explicitly ignored.
 118 fn should_ignore(line: &str) -> bool {
 119     // Matches test annotations like `//~ ERROR text`.
 120     // This mirrors the regex in src/tools/compiletest/src/runtest.rs, please
 121     // update both if either are changed.
 122     let re = Regex::new("\\s*//(\\[.*\\])?~.*").unwrap();
 123     re.is_match(line) || ANNOTATIONS_TO_IGNORE.iter().any(|a| line.contains(a))
 124 }
 125
 126 /// Returns `true` if `line` is allowed to be longer than the normal limit.
 127 fn long_line_is_ok(extension: &str, is_error_code: bool, max_columns: usize, line: &str) -> bool {
 128     match extension {
 129         // fluent files are allowed to be any length
 130         "ftl" => true,
 131         // non-error code markdown is allowed to be any length
 132         "md" if !is_error_code => true,
 133         _ => line_is_url(is_error_code, max_columns, line) || should_ignore(line),
 134     }
 135 }
 136
 137 enum Directive {
 138     /// By default, tidy always warns against style issues.
 139     Deny,
 140
 141     /// `Ignore(false)` means that an `ignore-tidy-*` directive
 142     /// has been provided, but is unnecessary. `Ignore(true)`
 143     /// means that it is necessary (i.e. a warning would be
 144     /// produced if `ignore-tidy-*` was not present).
 145     Ignore(bool),
 146 }
 147
 148 fn contains_ignore_directive(can_contain: bool, contents: &str, check: &str) -> Directive {
 149     if !can_contain {
 150         return Directive::Deny;
 151     }
 152     // Update `can_contain` when changing this
 153     if contents.contains(&format!("// ignore-tidy-{check}"))
 154         || contents.contains(&format!("# ignore-tidy-{check}"))
 155         || contents.contains(&format!("/* ignore-tidy-{check} */"))
 156     {
 157         Directive::Ignore(false)
 158     } else {
 159         Directive::Deny
 160     }
 161 }
 162
 163 macro_rules! suppressible_tidy_err {
 164     ($err:ident, $skip:ident, $msg:expr) => {
 165         if let Directive::Deny = $skip {
 166             $err($msg);
 167         } else {
 168             $skip = Directive::Ignore(true);
 169         }
 170     };
 171 }
 172
 173 pub fn is_in(full_path: &Path, parent_folder_to_find: &str, folder_to_find: &str) -> bool {
 174     if let Some(parent) = full_path.parent() {
 175         if parent.file_name().map_or_else(
 176             || false,
 177             |f| {
 178                 f.to_string_lossy() == folder_to_find
 179                     && parent
 180                         .parent()
 181                         .and_then(|f| f.file_name())
 182                         .map_or_else(|| false, |f| f == parent_folder_to_find)
 183             },
 184         ) {
 185             true
 186         } else {
 187             is_in(parent, parent_folder_to_find, folder_to_find)
 188         }
 189     } else {
 190         false
 191     }
 192 }
 193
 194 fn skip_markdown_path(path: &Path) -> bool {
 195     // These aren't ready for tidy.
 196     const SKIP_MD: &[&str] = &[
 197         "src/doc/edition-guide",
 198         "src/doc/embedded-book",
 199         "src/doc/nomicon",
 200         "src/doc/reference",
 201         "src/doc/rust-by-example",
 202         "src/doc/rustc-dev-guide",
 203     ];
 204     SKIP_MD.iter().any(|p| path.ends_with(p))
 205 }
 206
 207 fn is_unexplained_ignore(extension: &str, line: &str) -> bool {
 208     if !line.ends_with("```ignore") && !line.ends_with("```rust,ignore") {
 209         return false;
 210     }
 211     if extension == "md" && line.trim().starts_with("//") {
 212         // Markdown examples may include doc comments with ignore inside a
 213         // code block.
 214         return false;
 215     }
 216     true
 217 }
 218
 219 pub fn check(path: &Path, bad: &mut bool) {
 220     fn skip(path: &Path) -> bool {
 221         super::filter_dirs(path) || skip_markdown_path(path)
 222     }
 223     let problematic_consts_strings: Vec<String> = (PROBLEMATIC_CONSTS.iter().map(u32::to_string))
 224         .chain(PROBLEMATIC_CONSTS.iter().map(|v| format!("{:x}", v)))
 225         .chain(PROBLEMATIC_CONSTS.iter().map(|v| format!("{:X}", v)))
 226         .collect();
 227     super::walk(path, &mut skip, &mut |entry, contents| {
 228         let file = entry.path();
 229         let filename = file.file_name().unwrap().to_string_lossy();
 230         let extensions = [".rs", ".py", ".js", ".sh", ".c", ".cpp", ".h", ".md", ".css", ".ftl"];
 231         if extensions.iter().all(|e| !filename.ends_with(e)) || filename.starts_with(".#") {
 232             return;
 233         }
 234
 235         let is_style_file = filename.ends_with(".css");
 236         let under_rustfmt = filename.ends_with(".rs") &&
 237             // This list should ideally be sourced from rustfmt.toml but we don't want to add a toml
 238             // parser to tidy.
 239             !file.ancestors().any(|a| {
 240                 a.ends_with("src/test") ||
 241                     a.ends_with("src/doc/book")
 242             });
 243
 244         if is_style_file && !is_in(file, "src", "librustdoc") {
 245             // We only check CSS files in rustdoc.
 246             return;
 247         }
 248
 249         if contents.is_empty() {
 250             tidy_error!(bad, "{}: empty file", file.display());
 251         }
 252
 253         let extension = file.extension().unwrap().to_string_lossy();
 254         let is_error_code = extension == "md" && is_in(file, "src", "error_codes");
 255
 256         let max_columns = if is_error_code { ERROR_CODE_COLS } else { COLS };
 257
 258         let can_contain = contents.contains("// ignore-tidy-")
 259             || contents.contains("# ignore-tidy-")
 260             || contents.contains("/* ignore-tidy-");
 261         // Enable testing ICE's that require specific (untidy)
 262         // file formats easily eg. `issue-1234-ignore-tidy.rs`
 263         if filename.contains("ignore-tidy") {
 264             return;
 265         }
 266         let mut skip_cr = contains_ignore_directive(can_contain, &contents, "cr");
 267         let mut skip_undocumented_unsafe =
 268             contains_ignore_directive(can_contain, &contents, "undocumented-unsafe");
 269         let mut skip_tab = contains_ignore_directive(can_contain, &contents, "tab");
 270         let mut skip_line_length = contains_ignore_directive(can_contain, &contents, "linelength");
 271         let mut skip_file_length = contains_ignore_directive(can_contain, &contents, "filelength");
 272         let mut skip_end_whitespace =
 273             contains_ignore_directive(can_contain, &contents, "end-whitespace");
 274         let mut skip_trailing_newlines =
 275             contains_ignore_directive(can_contain, &contents, "trailing-newlines");
 276         let mut skip_leading_newlines =
 277             contains_ignore_directive(can_contain, &contents, "leading-newlines");
 278         let mut skip_copyright = contains_ignore_directive(can_contain, &contents, "copyright");
 279         let mut leading_new_lines = false;
 280         let mut trailing_new_lines = 0;
 281         let mut lines = 0;
 282         let mut last_safety_comment = false;
 283         for (i, line) in contents.split('\n').enumerate() {
 284             let mut err = |msg: &str| {
 285                 tidy_error!(bad, "{}:{}: {}", file.display(), i + 1, msg);
 286             };
 287             if !under_rustfmt
 288                 && line.chars().count() > max_columns
 289                 && !long_line_is_ok(&extension, is_error_code, max_columns, line)
 290             {
 291                 suppressible_tidy_err!(
 292                     err,
 293                     skip_line_length,
 294                     &format!("line longer than {max_columns} chars")
 295                 );
 296             }
 297             if !is_style_file && line.contains('\t') {
 298                 suppressible_tidy_err!(err, skip_tab, "tab character");
 299             }
 300             if line.ends_with(' ') || line.ends_with('\t') {
 301                 suppressible_tidy_err!(err, skip_end_whitespace, "trailing whitespace");
 302             }
 303             if is_style_file && line.starts_with(' ') {
 304                 err("CSS files use tabs for indent");
 305             }
 306             if line.contains('\r') {
 307                 suppressible_tidy_err!(err, skip_cr, "CR character");
 308             }
 309             if filename != "style.rs" {
 310                 if line.contains("TODO") {
 311                     err("TODO is deprecated; use FIXME")
 312                 }
 313                 if line.contains("//") && line.contains(" XXX") {
 314                     err("XXX is deprecated; use FIXME")
 315                 }
 316                 for s in problematic_consts_strings.iter() {
 317                     if line.contains(s) {
 318                         err("Don't use magic numbers that spell things (consider 0x12345678)");
 319                     }
 320                 }
 321             }
 322             let is_test = || file.components().any(|c| c.as_os_str() == "tests");
 323             // for now we just check libcore
 324             if line.contains("unsafe {") && !line.trim().starts_with("//") && !last_safety_comment {
 325                 if file.components().any(|c| c.as_os_str() == "core") && !is_test() {
 326                     suppressible_tidy_err!(err, skip_undocumented_unsafe, "undocumented unsafe");
 327                 }
 328             }
 329             if line.contains("// SAFETY:") {
 330                 last_safety_comment = true;
 331             } else if line.trim().starts_with("//") || line.trim().is_empty() {
 332                 // keep previous value
 333             } else {
 334                 last_safety_comment = false;
 335             }
 336             if (line.starts_with("// Copyright")
 337                 || line.starts_with("# Copyright")
 338                 || line.starts_with("Copyright"))
 339                 && (line.contains("Rust Developers") || line.contains("Rust Project Developers"))
 340             {
 341                 suppressible_tidy_err!(
 342                     err,
 343                     skip_copyright,
 344                     "copyright notices attributed to the Rust Project Developers are deprecated"
 345                 );
 346             }
 347             if is_unexplained_ignore(&extension, line) {
 348                 err(UNEXPLAINED_IGNORE_DOCTEST_INFO);
 349             }
 350             if filename.ends_with(".cpp") && line.contains("llvm_unreachable") {
 351                 err(LLVM_UNREACHABLE_INFO);
 352             }
 353             if line.is_empty() {
 354                 if i == 0 {
 355                     leading_new_lines = true;
 356                 }
 357                 trailing_new_lines += 1;
 358             } else {
 359                 trailing_new_lines = 0;
 360             }
 361
 362             if !line.trim().starts_with("//") {
 363                 lines += 1;
 364             }
 365         }
 366         if leading_new_lines {
 367             let mut err = |_| {
 368                 tidy_error!(bad, "{}: leading newline", file.display());
 369             };
 370             suppressible_tidy_err!(err, skip_leading_newlines, "mising leading newline");
 371         }
 372         let mut err = |msg: &str| {
 373             tidy_error!(bad, "{}: {}", file.display(), msg);
 374         };
 375         match trailing_new_lines {
 376             0 => suppressible_tidy_err!(err, skip_trailing_newlines, "missing trailing newline"),
 377             1 => {}
 378             n => suppressible_tidy_err!(
 379                 err,
 380                 skip_trailing_newlines,
 381                 &format!("too many trailing newlines ({n})")
 382             ),
 383         };
 384         if lines > LINES {
 385             let mut err = |_| {
 386                 tidy_error!(
 387                     bad,
 388                     "{}: too many lines ({}) (add `// \
 389                      ignore-tidy-filelength` to the file to suppress this error)",
 390                     file.display(),
 391                     lines
 392                 );
 393             };
 394             suppressible_tidy_err!(err, skip_file_length, "");
 395         }
 396
 397         if let Directive::Ignore(false) = skip_cr {
 398             tidy_error!(bad, "{}: ignoring CR characters unnecessarily", file.display());
 399         }
 400         if let Directive::Ignore(false) = skip_tab {
 401             tidy_error!(bad, "{}: ignoring tab characters unnecessarily", file.display());
 402         }
 403         if let Directive::Ignore(false) = skip_end_whitespace {
 404             tidy_error!(bad, "{}: ignoring trailing whitespace unnecessarily", file.display());
 405         }
 406         if let Directive::Ignore(false) = skip_trailing_newlines {
 407             tidy_error!(bad, "{}: ignoring trailing newlines unnecessarily", file.display());
 408         }
 409         if let Directive::Ignore(false) = skip_leading_newlines {
 410             tidy_error!(bad, "{}: ignoring leading newlines unnecessarily", file.display());
 411         }
 412         if let Directive::Ignore(false) = skip_copyright {
 413             tidy_error!(bad, "{}: ignoring copyright unnecessarily", file.display());
 414         }
 415         // We deliberately do not warn about these being unnecessary,
 416         // that would just lead to annoying churn.
 417         let _unused = skip_line_length;
 418         let _unused = skip_file_length;
 419     })
 420 }