| 1 | use core::convert::TryFrom; |
| 2 | use core::{char, fmt, iter, mem, str}; |
| 3 | |
| 4 | #[allow (unused_macros)] |
| 5 | macro_rules! write { |
| 6 | ($($ignored:tt)*) => { |
| 7 | compile_error!( |
| 8 | "use `self.print(value)` or `fmt::Trait::fmt(&value, self.out)`, \ |
| 9 | instead of `write!(self.out, \"{...} \", value)`" |
| 10 | ) |
| 11 | }; |
| 12 | } |
| 13 | |
| 14 | // Maximum recursion depth when parsing symbols before we just bail out saying |
| 15 | // "this symbol is invalid" |
| 16 | const MAX_DEPTH: u32 = 500; |
| 17 | |
| 18 | /// Representation of a demangled symbol name. |
| 19 | pub struct Demangle<'a> { |
| 20 | inner: &'a str, |
| 21 | } |
| 22 | |
| 23 | #[derive (PartialEq, Eq, Debug)] |
| 24 | pub enum ParseError { |
| 25 | /// Symbol doesn't match the expected `v0` grammar. |
| 26 | Invalid, |
| 27 | |
| 28 | /// Parsing the symbol crossed the recursion limit (see `MAX_DEPTH`). |
| 29 | RecursedTooDeep, |
| 30 | } |
| 31 | |
| 32 | /// De-mangles a Rust symbol into a more readable version |
| 33 | /// |
| 34 | /// This function will take a **mangled** symbol and return a value. When printed, |
| 35 | /// the de-mangled version will be written. If the symbol does not look like |
| 36 | /// a mangled symbol, the original value will be written instead. |
| 37 | pub fn demangle(s: &str) -> Result<(Demangle, &str), ParseError> { |
| 38 | // First validate the symbol. If it doesn't look like anything we're |
| 39 | // expecting, we just print it literally. Note that we must handle non-Rust |
| 40 | // symbols because we could have any function in the backtrace. |
| 41 | let inner; |
| 42 | if s.len() > 2 && s.starts_with("_R" ) { |
| 43 | inner = &s[2..]; |
| 44 | } else if s.len() > 1 && s.starts_with('R' ) { |
| 45 | // On Windows, dbghelp strips leading underscores, so we accept "R..." |
| 46 | // form too. |
| 47 | inner = &s[1..]; |
| 48 | } else if s.len() > 3 && s.starts_with("__R" ) { |
| 49 | // On OSX, symbols are prefixed with an extra _ |
| 50 | inner = &s[3..]; |
| 51 | } else { |
| 52 | return Err(ParseError::Invalid); |
| 53 | } |
| 54 | |
| 55 | // Paths always start with uppercase characters. |
| 56 | match inner.as_bytes()[0] { |
| 57 | b'A' ..=b'Z' => {} |
| 58 | _ => return Err(ParseError::Invalid), |
| 59 | } |
| 60 | |
| 61 | // only work with ascii text |
| 62 | if inner.bytes().any(|c| c & 0x80 != 0) { |
| 63 | return Err(ParseError::Invalid); |
| 64 | } |
| 65 | |
| 66 | // Verify that the symbol is indeed a valid path. |
| 67 | let try_parse_path = |parser| { |
| 68 | let mut dummy_printer = Printer { |
| 69 | parser: Ok(parser), |
| 70 | out: None, |
| 71 | bound_lifetime_depth: 0, |
| 72 | }; |
| 73 | dummy_printer |
| 74 | .print_path(false) |
| 75 | .expect("`fmt::Error`s should be impossible without a `fmt::Formatter`" ); |
| 76 | dummy_printer.parser |
| 77 | }; |
| 78 | let mut parser = Parser { |
| 79 | sym: inner, |
| 80 | next: 0, |
| 81 | depth: 0, |
| 82 | }; |
| 83 | parser = try_parse_path(parser)?; |
| 84 | |
| 85 | // Instantiating crate (paths always start with uppercase characters). |
| 86 | if let Some(&(b'A' ..=b'Z' )) = parser.sym.as_bytes().get(parser.next) { |
| 87 | parser = try_parse_path(parser)?; |
| 88 | } |
| 89 | |
| 90 | Ok((Demangle { inner }, &parser.sym[parser.next..])) |
| 91 | } |
| 92 | |
| 93 | impl<'s> fmt::Display for Demangle<'s> { |
| 94 | fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { |
| 95 | let mut printer: Printer<'_, '_, '_> = Printer { |
| 96 | parser: Ok(Parser { |
| 97 | sym: self.inner, |
| 98 | next: 0, |
| 99 | depth: 0, |
| 100 | }), |
| 101 | out: Some(f), |
| 102 | bound_lifetime_depth: 0, |
| 103 | }; |
| 104 | printer.print_path(in_value:true) |
| 105 | } |
| 106 | } |
| 107 | |
| 108 | struct Ident<'s> { |
| 109 | /// ASCII part of the identifier. |
| 110 | ascii: &'s str, |
| 111 | /// Punycode insertion codes for Unicode codepoints, if any. |
| 112 | punycode: &'s str, |
| 113 | } |
| 114 | |
| 115 | const SMALL_PUNYCODE_LEN: usize = 128; |
| 116 | |
| 117 | impl<'s> Ident<'s> { |
| 118 | /// Attempt to decode punycode on the stack (allocation-free), |
| 119 | /// and pass the char slice to the closure, if successful. |
| 120 | /// This supports up to `SMALL_PUNYCODE_LEN` characters. |
| 121 | fn try_small_punycode_decode<F: FnOnce(&[char]) -> R, R>(&self, f: F) -> Option<R> { |
| 122 | let mut out = [' \0' ; SMALL_PUNYCODE_LEN]; |
| 123 | let mut out_len = 0; |
| 124 | let r = self.punycode_decode(|i, c| { |
| 125 | // Check there's space left for another character. |
| 126 | out.get(out_len).ok_or(())?; |
| 127 | |
| 128 | // Move the characters after the insert position. |
| 129 | let mut j = out_len; |
| 130 | out_len += 1; |
| 131 | |
| 132 | while j > i { |
| 133 | out[j] = out[j - 1]; |
| 134 | j -= 1; |
| 135 | } |
| 136 | |
| 137 | // Insert the new character. |
| 138 | out[i] = c; |
| 139 | |
| 140 | Ok(()) |
| 141 | }); |
| 142 | if r.is_ok() { |
| 143 | Some(f(&out[..out_len])) |
| 144 | } else { |
| 145 | None |
| 146 | } |
| 147 | } |
| 148 | |
| 149 | /// Decode punycode as insertion positions and characters |
| 150 | /// and pass them to the closure, which can return `Err(())` |
| 151 | /// to stop the decoding process. |
| 152 | fn punycode_decode<F: FnMut(usize, char) -> Result<(), ()>>( |
| 153 | &self, |
| 154 | mut insert: F, |
| 155 | ) -> Result<(), ()> { |
| 156 | let mut punycode_bytes = self.punycode.bytes().peekable(); |
| 157 | if punycode_bytes.peek().is_none() { |
| 158 | return Err(()); |
| 159 | } |
| 160 | |
| 161 | let mut len = 0; |
| 162 | |
| 163 | // Populate initial output from ASCII fragment. |
| 164 | for c in self.ascii.chars() { |
| 165 | insert(len, c)?; |
| 166 | len += 1; |
| 167 | } |
| 168 | |
| 169 | // Punycode parameters and initial state. |
| 170 | let base = 36; |
| 171 | let t_min = 1; |
| 172 | let t_max = 26; |
| 173 | let skew = 38; |
| 174 | let mut damp = 700; |
| 175 | let mut bias = 72; |
| 176 | let mut i: usize = 0; |
| 177 | let mut n: usize = 0x80; |
| 178 | |
| 179 | loop { |
| 180 | // Read one delta value. |
| 181 | let mut delta: usize = 0; |
| 182 | let mut w = 1; |
| 183 | let mut k: usize = 0; |
| 184 | loop { |
| 185 | use core::cmp::{max, min}; |
| 186 | |
| 187 | k += base; |
| 188 | let t = min(max(k.saturating_sub(bias), t_min), t_max); |
| 189 | |
| 190 | let d = match punycode_bytes.next() { |
| 191 | Some(d @ b'a' ..=b'z' ) => d - b'a' , |
| 192 | Some(d @ b'0' ..=b'9' ) => 26 + (d - b'0' ), |
| 193 | _ => return Err(()), |
| 194 | }; |
| 195 | let d = d as usize; |
| 196 | delta = delta.checked_add(d.checked_mul(w).ok_or(())?).ok_or(())?; |
| 197 | if d < t { |
| 198 | break; |
| 199 | } |
| 200 | w = w.checked_mul(base - t).ok_or(())?; |
| 201 | } |
| 202 | |
| 203 | // Compute the new insert position and character. |
| 204 | len += 1; |
| 205 | i = i.checked_add(delta).ok_or(())?; |
| 206 | n = n.checked_add(i / len).ok_or(())?; |
| 207 | i %= len; |
| 208 | |
| 209 | let n_u32 = n as u32; |
| 210 | let c = if n_u32 as usize == n { |
| 211 | char::from_u32(n_u32).ok_or(())? |
| 212 | } else { |
| 213 | return Err(()); |
| 214 | }; |
| 215 | |
| 216 | // Insert the new character and increment the insert position. |
| 217 | insert(i, c)?; |
| 218 | i += 1; |
| 219 | |
| 220 | // If there are no more deltas, decoding is complete. |
| 221 | if punycode_bytes.peek().is_none() { |
| 222 | return Ok(()); |
| 223 | } |
| 224 | |
| 225 | // Perform bias adaptation. |
| 226 | delta /= damp; |
| 227 | damp = 2; |
| 228 | |
| 229 | delta += delta / len; |
| 230 | let mut k = 0; |
| 231 | while delta > ((base - t_min) * t_max) / 2 { |
| 232 | delta /= base - t_min; |
| 233 | k += base; |
| 234 | } |
| 235 | bias = k + ((base - t_min + 1) * delta) / (delta + skew); |
| 236 | } |
| 237 | } |
| 238 | } |
| 239 | |
| 240 | impl<'s> fmt::Display for Ident<'s> { |
| 241 | fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { |
| 242 | self.try_small_punycode_decode(|chars| { |
| 243 | for &c in chars { |
| 244 | c.fmt(f)?; |
| 245 | } |
| 246 | Ok(()) |
| 247 | }) |
| 248 | .unwrap_or_else(|| { |
| 249 | if !self.punycode.is_empty() { |
| 250 | f.write_str("punycode{" )?; |
| 251 | |
| 252 | // Reconstruct a standard Punycode encoding, |
| 253 | // by using `-` as the separator. |
| 254 | if !self.ascii.is_empty() { |
| 255 | f.write_str(self.ascii)?; |
| 256 | f.write_str("-" )?; |
| 257 | } |
| 258 | f.write_str(self.punycode)?; |
| 259 | |
| 260 | f.write_str("}" ) |
| 261 | } else { |
| 262 | f.write_str(self.ascii) |
| 263 | } |
| 264 | }) |
| 265 | } |
| 266 | } |
| 267 | |
| 268 | /// Sequence of lowercase hexadecimal nibbles (`0-9a-f`), used by leaf consts. |
| 269 | struct HexNibbles<'s> { |
| 270 | nibbles: &'s str, |
| 271 | } |
| 272 | |
| 273 | impl<'s> HexNibbles<'s> { |
| 274 | /// Decode an integer value (with the "most significant nibble" first), |
| 275 | /// returning `None` if it can't fit in an `u64`. |
| 276 | // FIXME(eddyb) should this "just" use `u128` instead? |
| 277 | fn try_parse_uint(&self) -> Option<u64> { |
| 278 | let nibbles = self.nibbles.trim_start_matches("0" ); |
| 279 | |
| 280 | if nibbles.len() > 16 { |
| 281 | return None; |
| 282 | } |
| 283 | |
| 284 | let mut v = 0; |
| 285 | for nibble in nibbles.chars() { |
| 286 | v = (v << 4) | (nibble.to_digit(16).unwrap() as u64); |
| 287 | } |
| 288 | Some(v) |
| 289 | } |
| 290 | |
| 291 | /// Decode a UTF-8 byte sequence (with each byte using a pair of nibbles) |
| 292 | /// into individual `char`s, returning `None` for invalid UTF-8. |
| 293 | fn try_parse_str_chars(&self) -> Option<impl Iterator<Item = char> + 's> { |
| 294 | if self.nibbles.len() % 2 != 0 { |
| 295 | return None; |
| 296 | } |
| 297 | |
| 298 | // FIXME(eddyb) use `array_chunks` instead, when that becomes stable. |
| 299 | let mut bytes = self |
| 300 | .nibbles |
| 301 | .as_bytes() |
| 302 | .chunks_exact(2) |
| 303 | .map(|slice| match slice { |
| 304 | [a, b] => [a, b], |
| 305 | _ => unreachable!(), |
| 306 | }) |
| 307 | .map(|[&hi, &lo]| { |
| 308 | let half = |nibble: u8| (nibble as char).to_digit(16).unwrap() as u8; |
| 309 | (half(hi) << 4) | half(lo) |
| 310 | }); |
| 311 | |
| 312 | let chars = iter::from_fn(move || { |
| 313 | // As long as there are any bytes left, there's at least one more |
| 314 | // UTF-8-encoded `char` to decode (or the possibility of error). |
| 315 | bytes.next().map(|first_byte| -> Result<char, ()> { |
| 316 | // FIXME(eddyb) this `enum` and `fn` should be somewhere in `core`. |
| 317 | enum Utf8FirstByteError { |
| 318 | ContinuationByte, |
| 319 | TooLong, |
| 320 | } |
| 321 | fn utf8_len_from_first_byte(byte: u8) -> Result<usize, Utf8FirstByteError> { |
| 322 | match byte { |
| 323 | 0x00..=0x7f => Ok(1), |
| 324 | 0x80..=0xbf => Err(Utf8FirstByteError::ContinuationByte), |
| 325 | 0xc0..=0xdf => Ok(2), |
| 326 | 0xe0..=0xef => Ok(3), |
| 327 | 0xf0..=0xf7 => Ok(4), |
| 328 | 0xf8..=0xff => Err(Utf8FirstByteError::TooLong), |
| 329 | } |
| 330 | } |
| 331 | |
| 332 | // Collect the appropriate amount of bytes (up to 4), according |
| 333 | // to the UTF-8 length implied by the first byte. |
| 334 | let utf8_len = utf8_len_from_first_byte(first_byte).map_err(|_| ())?; |
| 335 | let utf8 = &mut [first_byte, 0, 0, 0][..utf8_len]; |
| 336 | for i in 1..utf8_len { |
| 337 | utf8[i] = bytes.next().ok_or(())?; |
| 338 | } |
| 339 | |
| 340 | // Fully validate the UTF-8 sequence. |
| 341 | let s = str::from_utf8(utf8).map_err(|_| ())?; |
| 342 | |
| 343 | // Since we included exactly one UTF-8 sequence, and validation |
| 344 | // succeeded, `str::chars` should return exactly one `char`. |
| 345 | let mut chars = s.chars(); |
| 346 | match (chars.next(), chars.next()) { |
| 347 | (Some(c), None) => Ok(c), |
| 348 | _ => unreachable!( |
| 349 | "str::from_utf8( {:?}) = {:?} was expected to have 1 char, \ |
| 350 | but {} chars were found" , |
| 351 | utf8, |
| 352 | s, |
| 353 | s.chars().count() |
| 354 | ), |
| 355 | } |
| 356 | }) |
| 357 | }); |
| 358 | |
| 359 | // HACK(eddyb) doing a separate validation iteration like this might be |
| 360 | // wasteful, but it's easier to avoid starting to print a string literal |
| 361 | // in the first place, than to abort it mid-string. |
| 362 | if chars.clone().any(|r| r.is_err()) { |
| 363 | None |
| 364 | } else { |
| 365 | Some(chars.map(Result::unwrap)) |
| 366 | } |
| 367 | } |
| 368 | } |
| 369 | |
| 370 | fn basic_type(tag: u8) -> Option<&'static str> { |
| 371 | Some(match tag { |
| 372 | b'b' => "bool" , |
| 373 | b'c' => "char" , |
| 374 | b'e' => "str" , |
| 375 | b'u' => "()" , |
| 376 | b'a' => "i8" , |
| 377 | b's' => "i16" , |
| 378 | b'l' => "i32" , |
| 379 | b'x' => "i64" , |
| 380 | b'n' => "i128" , |
| 381 | b'i' => "isize" , |
| 382 | b'h' => "u8" , |
| 383 | b't' => "u16" , |
| 384 | b'm' => "u32" , |
| 385 | b'y' => "u64" , |
| 386 | b'o' => "u128" , |
| 387 | b'j' => "usize" , |
| 388 | b'f' => "f32" , |
| 389 | b'd' => "f64" , |
| 390 | b'z' => "!" , |
| 391 | b'p' => "_" , |
| 392 | b'v' => "..." , |
| 393 | |
| 394 | _ => return None, |
| 395 | }) |
| 396 | } |
| 397 | |
| 398 | struct Parser<'s> { |
| 399 | sym: &'s str, |
| 400 | next: usize, |
| 401 | depth: u32, |
| 402 | } |
| 403 | |
| 404 | impl<'s> Parser<'s> { |
| 405 | fn push_depth(&mut self) -> Result<(), ParseError> { |
| 406 | self.depth += 1; |
| 407 | if self.depth > MAX_DEPTH { |
| 408 | Err(ParseError::RecursedTooDeep) |
| 409 | } else { |
| 410 | Ok(()) |
| 411 | } |
| 412 | } |
| 413 | |
| 414 | fn pop_depth(&mut self) { |
| 415 | self.depth -= 1; |
| 416 | } |
| 417 | |
| 418 | fn peek(&self) -> Option<u8> { |
| 419 | self.sym.as_bytes().get(self.next).cloned() |
| 420 | } |
| 421 | |
| 422 | fn eat(&mut self, b: u8) -> bool { |
| 423 | if self.peek() == Some(b) { |
| 424 | self.next += 1; |
| 425 | true |
| 426 | } else { |
| 427 | false |
| 428 | } |
| 429 | } |
| 430 | |
| 431 | fn next(&mut self) -> Result<u8, ParseError> { |
| 432 | let b = self.peek().ok_or(ParseError::Invalid)?; |
| 433 | self.next += 1; |
| 434 | Ok(b) |
| 435 | } |
| 436 | |
| 437 | fn hex_nibbles(&mut self) -> Result<HexNibbles<'s>, ParseError> { |
| 438 | let start = self.next; |
| 439 | loop { |
| 440 | match self.next()? { |
| 441 | b'0' ..=b'9' | b'a' ..=b'f' => {} |
| 442 | b'_' => break, |
| 443 | _ => return Err(ParseError::Invalid), |
| 444 | } |
| 445 | } |
| 446 | Ok(HexNibbles { |
| 447 | nibbles: &self.sym[start..self.next - 1], |
| 448 | }) |
| 449 | } |
| 450 | |
| 451 | fn digit_10(&mut self) -> Result<u8, ParseError> { |
| 452 | let d = match self.peek() { |
| 453 | Some(d @ b'0' ..=b'9' ) => d - b'0' , |
| 454 | _ => return Err(ParseError::Invalid), |
| 455 | }; |
| 456 | self.next += 1; |
| 457 | Ok(d) |
| 458 | } |
| 459 | |
| 460 | fn digit_62(&mut self) -> Result<u8, ParseError> { |
| 461 | let d = match self.peek() { |
| 462 | Some(d @ b'0' ..=b'9' ) => d - b'0' , |
| 463 | Some(d @ b'a' ..=b'z' ) => 10 + (d - b'a' ), |
| 464 | Some(d @ b'A' ..=b'Z' ) => 10 + 26 + (d - b'A' ), |
| 465 | _ => return Err(ParseError::Invalid), |
| 466 | }; |
| 467 | self.next += 1; |
| 468 | Ok(d) |
| 469 | } |
| 470 | |
| 471 | fn integer_62(&mut self) -> Result<u64, ParseError> { |
| 472 | if self.eat(b'_' ) { |
| 473 | return Ok(0); |
| 474 | } |
| 475 | |
| 476 | let mut x: u64 = 0; |
| 477 | while !self.eat(b'_' ) { |
| 478 | let d = self.digit_62()? as u64; |
| 479 | x = x.checked_mul(62).ok_or(ParseError::Invalid)?; |
| 480 | x = x.checked_add(d).ok_or(ParseError::Invalid)?; |
| 481 | } |
| 482 | x.checked_add(1).ok_or(ParseError::Invalid) |
| 483 | } |
| 484 | |
| 485 | fn opt_integer_62(&mut self, tag: u8) -> Result<u64, ParseError> { |
| 486 | if !self.eat(tag) { |
| 487 | return Ok(0); |
| 488 | } |
| 489 | self.integer_62()?.checked_add(1).ok_or(ParseError::Invalid) |
| 490 | } |
| 491 | |
| 492 | fn disambiguator(&mut self) -> Result<u64, ParseError> { |
| 493 | self.opt_integer_62(b's' ) |
| 494 | } |
| 495 | |
| 496 | fn namespace(&mut self) -> Result<Option<char>, ParseError> { |
| 497 | match self.next()? { |
| 498 | // Special namespaces, like closures and shims. |
| 499 | ns @ b'A' ..=b'Z' => Ok(Some(ns as char)), |
| 500 | |
| 501 | // Implementation-specific/unspecified namespaces. |
| 502 | b'a' ..=b'z' => Ok(None), |
| 503 | |
| 504 | _ => Err(ParseError::Invalid), |
| 505 | } |
| 506 | } |
| 507 | |
| 508 | fn backref(&mut self) -> Result<Parser<'s>, ParseError> { |
| 509 | let s_start = self.next - 1; |
| 510 | let i = self.integer_62()?; |
| 511 | if i >= s_start as u64 { |
| 512 | return Err(ParseError::Invalid); |
| 513 | } |
| 514 | let mut new_parser = Parser { |
| 515 | sym: self.sym, |
| 516 | next: i as usize, |
| 517 | depth: self.depth, |
| 518 | }; |
| 519 | new_parser.push_depth()?; |
| 520 | Ok(new_parser) |
| 521 | } |
| 522 | |
| 523 | fn ident(&mut self) -> Result<Ident<'s>, ParseError> { |
| 524 | let is_punycode = self.eat(b'u' ); |
| 525 | let mut len = self.digit_10()? as usize; |
| 526 | if len != 0 { |
| 527 | while let Ok(d) = self.digit_10() { |
| 528 | len = len.checked_mul(10).ok_or(ParseError::Invalid)?; |
| 529 | len = len.checked_add(d as usize).ok_or(ParseError::Invalid)?; |
| 530 | } |
| 531 | } |
| 532 | |
| 533 | // Skip past the optional `_` separator. |
| 534 | self.eat(b'_' ); |
| 535 | |
| 536 | let start = self.next; |
| 537 | self.next = self.next.checked_add(len).ok_or(ParseError::Invalid)?; |
| 538 | if self.next > self.sym.len() { |
| 539 | return Err(ParseError::Invalid); |
| 540 | } |
| 541 | |
| 542 | let ident = &self.sym[start..self.next]; |
| 543 | |
| 544 | if is_punycode { |
| 545 | let ident = match ident.bytes().rposition(|b| b == b'_' ) { |
| 546 | Some(i) => Ident { |
| 547 | ascii: &ident[..i], |
| 548 | punycode: &ident[i + 1..], |
| 549 | }, |
| 550 | None => Ident { |
| 551 | ascii: "" , |
| 552 | punycode: ident, |
| 553 | }, |
| 554 | }; |
| 555 | if ident.punycode.is_empty() { |
| 556 | return Err(ParseError::Invalid); |
| 557 | } |
| 558 | Ok(ident) |
| 559 | } else { |
| 560 | Ok(Ident { |
| 561 | ascii: ident, |
| 562 | punycode: "" , |
| 563 | }) |
| 564 | } |
| 565 | } |
| 566 | } |
| 567 | |
| 568 | struct Printer<'a, 'b: 'a, 's> { |
| 569 | /// The input parser to demangle from, or `Err` if any (parse) error was |
| 570 | /// encountered (in order to disallow further likely-incorrect demangling). |
| 571 | /// |
| 572 | /// See also the documentation on the `invalid!` and `parse!` macros below. |
| 573 | parser: Result<Parser<'s>, ParseError>, |
| 574 | |
| 575 | /// The output formatter to demangle to, or `None` while skipping printing. |
| 576 | out: Option<&'a mut fmt::Formatter<'b>>, |
| 577 | |
| 578 | /// Cumulative number of lifetimes bound by `for<...>` binders ('G'), |
| 579 | /// anywhere "around" the current entity (e.g. type) being demangled. |
| 580 | /// This value is not tracked while skipping printing, as it'd be unused. |
| 581 | /// |
| 582 | /// See also the documentation on the `Printer::in_binder` method. |
| 583 | bound_lifetime_depth: u32, |
| 584 | } |
| 585 | |
| 586 | impl ParseError { |
| 587 | /// Snippet to print when the error is initially encountered. |
| 588 | fn message(&self) -> &str { |
| 589 | match self { |
| 590 | ParseError::Invalid => "{invalid syntax}" , |
| 591 | ParseError::RecursedTooDeep => "{recursion limit reached}" , |
| 592 | } |
| 593 | } |
| 594 | } |
| 595 | |
| 596 | /// Mark the parser as errored (with `ParseError::Invalid`), print the |
| 597 | /// appropriate message (see `ParseError::message`) and return early. |
| 598 | macro_rules! invalid { |
| 599 | ($printer:ident) => {{ |
| 600 | let err = ParseError::Invalid; |
| 601 | $printer.print(err.message())?; |
| 602 | $printer.parser = Err(err); |
| 603 | return Ok(()); |
| 604 | }}; |
| 605 | } |
| 606 | |
| 607 | /// Call a parser method (if the parser hasn't errored yet), |
| 608 | /// and mark the parser as errored if it returns `Err`. |
| 609 | /// |
| 610 | /// If the parser errored, before or now, this returns early, |
| 611 | /// from the current function, after printing either: |
| 612 | /// * for a new error, the appropriate message (see `ParseError::message`) |
| 613 | /// * for an earlier error, only `?` - this allows callers to keep printing |
| 614 | /// the approximate syntax of the path/type/const, despite having errors, |
| 615 | /// e.g. `Vec<[(A, ?); ?]>` instead of `Vec<[(A, ?` |
| 616 | macro_rules! parse { |
| 617 | ($printer:ident, $method:ident $(($($arg:expr),*))*) => { |
| 618 | match $printer.parser { |
| 619 | Ok(ref mut parser) => match parser.$method($($($arg),*)*) { |
| 620 | Ok(x) => x, |
| 621 | Err(err) => { |
| 622 | $printer.print(err.message())?; |
| 623 | $printer.parser = Err(err); |
| 624 | return Ok(()); |
| 625 | } |
| 626 | } |
| 627 | Err(_) => return $printer.print("?" ), |
| 628 | } |
| 629 | }; |
| 630 | } |
| 631 | |
| 632 | impl<'a, 'b, 's> Printer<'a, 'b, 's> { |
| 633 | /// Eat the given character from the parser, |
| 634 | /// returning `false` if the parser errored. |
| 635 | fn eat(&mut self, b: u8) -> bool { |
| 636 | self.parser.as_mut().map(|p| p.eat(b)) == Ok(true) |
| 637 | } |
| 638 | |
| 639 | /// Skip printing (i.e. `self.out` will be `None`) for the duration of the |
| 640 | /// given closure. This should not change parsing behavior, only disable the |
| 641 | /// output, but there may be optimizations (such as not traversing backrefs). |
| 642 | fn skipping_printing<F>(&mut self, f: F) |
| 643 | where |
| 644 | F: FnOnce(&mut Self) -> fmt::Result, |
| 645 | { |
| 646 | let orig_out = self.out.take(); |
| 647 | f(self).expect("`fmt::Error`s should be impossible without a `fmt::Formatter`" ); |
| 648 | self.out = orig_out; |
| 649 | } |
| 650 | |
| 651 | /// Print the target of a backref, using the given closure. |
| 652 | /// When printing is being skipped, the backref will only be parsed, |
| 653 | /// ignoring the backref's target completely. |
| 654 | fn print_backref<F>(&mut self, f: F) -> fmt::Result |
| 655 | where |
| 656 | F: FnOnce(&mut Self) -> fmt::Result, |
| 657 | { |
| 658 | let backref_parser = parse!(self, backref); |
| 659 | |
| 660 | if self.out.is_none() { |
| 661 | return Ok(()); |
| 662 | } |
| 663 | |
| 664 | let orig_parser = mem::replace(&mut self.parser, Ok(backref_parser)); |
| 665 | let r = f(self); |
| 666 | self.parser = orig_parser; |
| 667 | r |
| 668 | } |
| 669 | |
| 670 | fn pop_depth(&mut self) { |
| 671 | if let Ok(ref mut parser) = self.parser { |
| 672 | parser.pop_depth(); |
| 673 | } |
| 674 | } |
| 675 | |
| 676 | /// Output the given value to `self.out` (using `fmt::Display` formatting), |
| 677 | /// if printing isn't being skipped. |
| 678 | fn print(&mut self, x: impl fmt::Display) -> fmt::Result { |
| 679 | if let Some(out) = &mut self.out { |
| 680 | fmt::Display::fmt(&x, out)?; |
| 681 | } |
| 682 | Ok(()) |
| 683 | } |
| 684 | |
| 685 | /// Output the given `char`s (escaped using `char::escape_debug`), with the |
| 686 | /// whole sequence wrapped in quotes, for either a `char` or `&str` literal, |
| 687 | /// if printing isn't being skipped. |
| 688 | fn print_quoted_escaped_chars( |
| 689 | &mut self, |
| 690 | quote: char, |
| 691 | chars: impl Iterator<Item = char>, |
| 692 | ) -> fmt::Result { |
| 693 | if let Some(out) = &mut self.out { |
| 694 | use core::fmt::Write; |
| 695 | |
| 696 | out.write_char(quote)?; |
| 697 | for c in chars { |
| 698 | // Special-case not escaping a single/double quote, when |
| 699 | // inside the opposite kind of quote. |
| 700 | if matches!((quote, c), (' \'' , '"' ) | ('"' , ' \'' )) { |
| 701 | out.write_char(c)?; |
| 702 | continue; |
| 703 | } |
| 704 | |
| 705 | for escaped in c.escape_debug() { |
| 706 | out.write_char(escaped)?; |
| 707 | } |
| 708 | } |
| 709 | out.write_char(quote)?; |
| 710 | } |
| 711 | Ok(()) |
| 712 | } |
| 713 | |
| 714 | /// Print the lifetime according to the previously decoded index. |
| 715 | /// An index of `0` always refers to `'_`, but starting with `1`, |
| 716 | /// indices refer to late-bound lifetimes introduced by a binder. |
| 717 | fn print_lifetime_from_index(&mut self, lt: u64) -> fmt::Result { |
| 718 | // Bound lifetimes aren't tracked when skipping printing. |
| 719 | if self.out.is_none() { |
| 720 | return Ok(()); |
| 721 | } |
| 722 | |
| 723 | self.print("'" )?; |
| 724 | if lt == 0 { |
| 725 | return self.print("_" ); |
| 726 | } |
| 727 | match (self.bound_lifetime_depth as u64).checked_sub(lt) { |
| 728 | Some(depth) => { |
| 729 | // Try to print lifetimes alphabetically first. |
| 730 | if depth < 26 { |
| 731 | let c = (b'a' + depth as u8) as char; |
| 732 | self.print(c) |
| 733 | } else { |
| 734 | // Use `'_123` after running out of letters. |
| 735 | self.print("_" )?; |
| 736 | self.print(depth) |
| 737 | } |
| 738 | } |
| 739 | None => invalid!(self), |
| 740 | } |
| 741 | } |
| 742 | |
| 743 | /// Optionally enter a binder ('G') for late-bound lifetimes, |
| 744 | /// printing e.g. `for<'a, 'b> ` before calling the closure, |
| 745 | /// and make those lifetimes visible to it (via depth level). |
| 746 | fn in_binder<F>(&mut self, f: F) -> fmt::Result |
| 747 | where |
| 748 | F: FnOnce(&mut Self) -> fmt::Result, |
| 749 | { |
| 750 | let bound_lifetimes = parse!(self, opt_integer_62(b'G' )); |
| 751 | |
| 752 | // Don't track bound lifetimes when skipping printing. |
| 753 | if self.out.is_none() { |
| 754 | return f(self); |
| 755 | } |
| 756 | |
| 757 | if bound_lifetimes > 0 { |
| 758 | self.print("for<" )?; |
| 759 | for i in 0..bound_lifetimes { |
| 760 | if i > 0 { |
| 761 | self.print(", " )?; |
| 762 | } |
| 763 | self.bound_lifetime_depth += 1; |
| 764 | self.print_lifetime_from_index(1)?; |
| 765 | } |
| 766 | self.print("> " )?; |
| 767 | } |
| 768 | |
| 769 | let r = f(self); |
| 770 | |
| 771 | // Restore `bound_lifetime_depth` to the previous value. |
| 772 | self.bound_lifetime_depth -= bound_lifetimes as u32; |
| 773 | |
| 774 | r |
| 775 | } |
| 776 | |
| 777 | /// Print list elements using the given closure and separator, |
| 778 | /// until the end of the list ('E') is found, or the parser errors. |
| 779 | /// Returns the number of elements printed. |
| 780 | fn print_sep_list<F>(&mut self, f: F, sep: &str) -> Result<usize, fmt::Error> |
| 781 | where |
| 782 | F: Fn(&mut Self) -> fmt::Result, |
| 783 | { |
| 784 | let mut i = 0; |
| 785 | while self.parser.is_ok() && !self.eat(b'E' ) { |
| 786 | if i > 0 { |
| 787 | self.print(sep)?; |
| 788 | } |
| 789 | f(self)?; |
| 790 | i += 1; |
| 791 | } |
| 792 | Ok(i) |
| 793 | } |
| 794 | |
| 795 | fn print_path(&mut self, in_value: bool) -> fmt::Result { |
| 796 | parse!(self, push_depth); |
| 797 | |
| 798 | let tag = parse!(self, next); |
| 799 | match tag { |
| 800 | b'C' => { |
| 801 | let dis = parse!(self, disambiguator); |
| 802 | let name = parse!(self, ident); |
| 803 | |
| 804 | self.print(name)?; |
| 805 | if let Some(out) = &mut self.out { |
| 806 | if !out.alternate() && dis != 0 { |
| 807 | out.write_str("[" )?; |
| 808 | fmt::LowerHex::fmt(&dis, out)?; |
| 809 | out.write_str("]" )?; |
| 810 | } |
| 811 | } |
| 812 | } |
| 813 | b'N' => { |
| 814 | let ns = parse!(self, namespace); |
| 815 | |
| 816 | self.print_path(in_value)?; |
| 817 | |
| 818 | // HACK(eddyb) if the parser is already marked as having errored, |
| 819 | // `parse!` below will print a `?` without its preceding `::` |
| 820 | // (because printing the `::` is skipped in certain conditions, |
| 821 | // i.e. a lowercase namespace with an empty identifier), |
| 822 | // so in order to get `::?`, the `::` has to be printed here. |
| 823 | if self.parser.is_err() { |
| 824 | self.print("::" )?; |
| 825 | } |
| 826 | |
| 827 | let dis = parse!(self, disambiguator); |
| 828 | let name = parse!(self, ident); |
| 829 | |
| 830 | match ns { |
| 831 | // Special namespaces, like closures and shims. |
| 832 | Some(ns) => { |
| 833 | self.print("::{" )?; |
| 834 | match ns { |
| 835 | 'C' => self.print("closure" )?, |
| 836 | 'S' => self.print("shim" )?, |
| 837 | _ => self.print(ns)?, |
| 838 | } |
| 839 | if !name.ascii.is_empty() || !name.punycode.is_empty() { |
| 840 | self.print(":" )?; |
| 841 | self.print(name)?; |
| 842 | } |
| 843 | self.print("#" )?; |
| 844 | self.print(dis)?; |
| 845 | self.print("}" )?; |
| 846 | } |
| 847 | |
| 848 | // Implementation-specific/unspecified namespaces. |
| 849 | None => { |
| 850 | if !name.ascii.is_empty() || !name.punycode.is_empty() { |
| 851 | self.print("::" )?; |
| 852 | self.print(name)?; |
| 853 | } |
| 854 | } |
| 855 | } |
| 856 | } |
| 857 | b'M' | b'X' | b'Y' => { |
| 858 | if tag != b'Y' { |
| 859 | // Ignore the `impl`'s own path. |
| 860 | parse!(self, disambiguator); |
| 861 | self.skipping_printing(|this| this.print_path(false)); |
| 862 | } |
| 863 | |
| 864 | self.print("<" )?; |
| 865 | self.print_type()?; |
| 866 | if tag != b'M' { |
| 867 | self.print(" as " )?; |
| 868 | self.print_path(false)?; |
| 869 | } |
| 870 | self.print(">" )?; |
| 871 | } |
| 872 | b'I' => { |
| 873 | self.print_path(in_value)?; |
| 874 | if in_value { |
| 875 | self.print("::" )?; |
| 876 | } |
| 877 | self.print("<" )?; |
| 878 | self.print_sep_list(Self::print_generic_arg, ", " )?; |
| 879 | self.print(">" )?; |
| 880 | } |
| 881 | b'B' => { |
| 882 | self.print_backref(|this| this.print_path(in_value))?; |
| 883 | } |
| 884 | _ => invalid!(self), |
| 885 | } |
| 886 | |
| 887 | self.pop_depth(); |
| 888 | Ok(()) |
| 889 | } |
| 890 | |
| 891 | fn print_generic_arg(&mut self) -> fmt::Result { |
| 892 | if self.eat(b'L' ) { |
| 893 | let lt = parse!(self, integer_62); |
| 894 | self.print_lifetime_from_index(lt) |
| 895 | } else if self.eat(b'K' ) { |
| 896 | self.print_const(false) |
| 897 | } else { |
| 898 | self.print_type() |
| 899 | } |
| 900 | } |
| 901 | |
| 902 | fn print_type(&mut self) -> fmt::Result { |
| 903 | let tag = parse!(self, next); |
| 904 | |
| 905 | if let Some(ty) = basic_type(tag) { |
| 906 | return self.print(ty); |
| 907 | } |
| 908 | |
| 909 | parse!(self, push_depth); |
| 910 | |
| 911 | match tag { |
| 912 | b'R' | b'Q' => { |
| 913 | self.print("&" )?; |
| 914 | if self.eat(b'L' ) { |
| 915 | let lt = parse!(self, integer_62); |
| 916 | if lt != 0 { |
| 917 | self.print_lifetime_from_index(lt)?; |
| 918 | self.print(" " )?; |
| 919 | } |
| 920 | } |
| 921 | if tag != b'R' { |
| 922 | self.print("mut " )?; |
| 923 | } |
| 924 | self.print_type()?; |
| 925 | } |
| 926 | |
| 927 | b'P' | b'O' => { |
| 928 | self.print("*" )?; |
| 929 | if tag != b'P' { |
| 930 | self.print("mut " )?; |
| 931 | } else { |
| 932 | self.print("const " )?; |
| 933 | } |
| 934 | self.print_type()?; |
| 935 | } |
| 936 | |
| 937 | b'A' | b'S' => { |
| 938 | self.print("[" )?; |
| 939 | self.print_type()?; |
| 940 | if tag == b'A' { |
| 941 | self.print("; " )?; |
| 942 | self.print_const(true)?; |
| 943 | } |
| 944 | self.print("]" )?; |
| 945 | } |
| 946 | b'T' => { |
| 947 | self.print("(" )?; |
| 948 | let count = self.print_sep_list(Self::print_type, ", " )?; |
| 949 | if count == 1 { |
| 950 | self.print("," )?; |
| 951 | } |
| 952 | self.print(")" )?; |
| 953 | } |
| 954 | b'F' => self.in_binder(|this| { |
| 955 | let is_unsafe = this.eat(b'U' ); |
| 956 | let abi = if this.eat(b'K' ) { |
| 957 | if this.eat(b'C' ) { |
| 958 | Some("C" ) |
| 959 | } else { |
| 960 | let abi = parse!(this, ident); |
| 961 | if abi.ascii.is_empty() || !abi.punycode.is_empty() { |
| 962 | invalid!(this); |
| 963 | } |
| 964 | Some(abi.ascii) |
| 965 | } |
| 966 | } else { |
| 967 | None |
| 968 | }; |
| 969 | |
| 970 | if is_unsafe { |
| 971 | this.print("unsafe " )?; |
| 972 | } |
| 973 | |
| 974 | if let Some(abi) = abi { |
| 975 | this.print("extern \"" )?; |
| 976 | |
| 977 | // If the ABI had any `-`, they were replaced with `_`, |
| 978 | // so the parts between `_` have to be re-joined with `-`. |
| 979 | let mut parts = abi.split('_' ); |
| 980 | this.print(parts.next().unwrap())?; |
| 981 | for part in parts { |
| 982 | this.print("-" )?; |
| 983 | this.print(part)?; |
| 984 | } |
| 985 | |
| 986 | this.print(" \" " )?; |
| 987 | } |
| 988 | |
| 989 | this.print("fn(" )?; |
| 990 | this.print_sep_list(Self::print_type, ", " )?; |
| 991 | this.print(")" )?; |
| 992 | |
| 993 | if this.eat(b'u' ) { |
| 994 | // Skip printing the return type if it's 'u', i.e. `()`. |
| 995 | } else { |
| 996 | this.print(" -> " )?; |
| 997 | this.print_type()?; |
| 998 | } |
| 999 | |
| 1000 | Ok(()) |
| 1001 | })?, |
| 1002 | b'D' => { |
| 1003 | self.print("dyn " )?; |
| 1004 | self.in_binder(|this| { |
| 1005 | this.print_sep_list(Self::print_dyn_trait, " + " )?; |
| 1006 | Ok(()) |
| 1007 | })?; |
| 1008 | |
| 1009 | if !self.eat(b'L' ) { |
| 1010 | invalid!(self); |
| 1011 | } |
| 1012 | let lt = parse!(self, integer_62); |
| 1013 | if lt != 0 { |
| 1014 | self.print(" + " )?; |
| 1015 | self.print_lifetime_from_index(lt)?; |
| 1016 | } |
| 1017 | } |
| 1018 | b'B' => { |
| 1019 | self.print_backref(Self::print_type)?; |
| 1020 | } |
| 1021 | _ => { |
| 1022 | // Go back to the tag, so `print_path` also sees it. |
| 1023 | let _ = self.parser.as_mut().map(|p| p.next -= 1); |
| 1024 | self.print_path(false)?; |
| 1025 | } |
| 1026 | } |
| 1027 | |
| 1028 | self.pop_depth(); |
| 1029 | Ok(()) |
| 1030 | } |
| 1031 | |
| 1032 | /// A trait in a trait object may have some "existential projections" |
| 1033 | /// (i.e. associated type bindings) after it, which should be printed |
| 1034 | /// in the `<...>` of the trait, e.g. `dyn Trait<T, U, Assoc=X>`. |
| 1035 | /// To this end, this method will keep the `<...>` of an 'I' path |
| 1036 | /// open, by omitting the `>`, and return `Ok(true)` in that case. |
| 1037 | fn print_path_maybe_open_generics(&mut self) -> Result<bool, fmt::Error> { |
| 1038 | if self.eat(b'B' ) { |
| 1039 | // NOTE(eddyb) the closure may not run if printing is being skipped, |
| 1040 | // but in that case the returned boolean doesn't matter. |
| 1041 | let mut open = false; |
| 1042 | self.print_backref(|this| { |
| 1043 | open = this.print_path_maybe_open_generics()?; |
| 1044 | Ok(()) |
| 1045 | })?; |
| 1046 | Ok(open) |
| 1047 | } else if self.eat(b'I' ) { |
| 1048 | self.print_path(false)?; |
| 1049 | self.print("<" )?; |
| 1050 | self.print_sep_list(Self::print_generic_arg, ", " )?; |
| 1051 | Ok(true) |
| 1052 | } else { |
| 1053 | self.print_path(false)?; |
| 1054 | Ok(false) |
| 1055 | } |
| 1056 | } |
| 1057 | |
| 1058 | fn print_dyn_trait(&mut self) -> fmt::Result { |
| 1059 | let mut open = self.print_path_maybe_open_generics()?; |
| 1060 | |
| 1061 | while self.eat(b'p' ) { |
| 1062 | if !open { |
| 1063 | self.print("<" )?; |
| 1064 | open = true; |
| 1065 | } else { |
| 1066 | self.print(", " )?; |
| 1067 | } |
| 1068 | |
| 1069 | let name = parse!(self, ident); |
| 1070 | self.print(name)?; |
| 1071 | self.print(" = " )?; |
| 1072 | self.print_type()?; |
| 1073 | } |
| 1074 | |
| 1075 | if open { |
| 1076 | self.print(">" )?; |
| 1077 | } |
| 1078 | |
| 1079 | Ok(()) |
| 1080 | } |
| 1081 | |
| 1082 | fn print_const(&mut self, in_value: bool) -> fmt::Result { |
| 1083 | let tag = parse!(self, next); |
| 1084 | |
| 1085 | parse!(self, push_depth); |
| 1086 | |
| 1087 | // Only literals (and the names of `const` generic parameters, but they |
| 1088 | // don't get mangled at all), can appear in generic argument position |
| 1089 | // without any disambiguation, all other expressions require braces. |
| 1090 | // To avoid duplicating the mapping between `tag` and what syntax gets |
| 1091 | // used (especially any special-casing), every case that needs braces |
| 1092 | // has to call `open_brace(self)?` (and the closing brace is automatic). |
| 1093 | let mut opened_brace = false; |
| 1094 | let mut open_brace_if_outside_expr = |this: &mut Self| { |
| 1095 | // If this expression is nested in another, braces aren't required. |
| 1096 | if in_value { |
| 1097 | return Ok(()); |
| 1098 | } |
| 1099 | |
| 1100 | opened_brace = true; |
| 1101 | this.print("{" ) |
| 1102 | }; |
| 1103 | |
| 1104 | match tag { |
| 1105 | b'p' => self.print("_" )?, |
| 1106 | |
| 1107 | // Primitive leaves with hex-encoded values (see `basic_type`). |
| 1108 | b'h' | b't' | b'm' | b'y' | b'o' | b'j' => self.print_const_uint(tag)?, |
| 1109 | b'a' | b's' | b'l' | b'x' | b'n' | b'i' => { |
| 1110 | if self.eat(b'n' ) { |
| 1111 | self.print("-" )?; |
| 1112 | } |
| 1113 | |
| 1114 | self.print_const_uint(tag)?; |
| 1115 | } |
| 1116 | b'b' => match parse!(self, hex_nibbles).try_parse_uint() { |
| 1117 | Some(0) => self.print("false" )?, |
| 1118 | Some(1) => self.print("true" )?, |
| 1119 | _ => invalid!(self), |
| 1120 | }, |
| 1121 | b'c' => { |
| 1122 | let valid_char = parse!(self, hex_nibbles) |
| 1123 | .try_parse_uint() |
| 1124 | .and_then(|v| u32::try_from(v).ok()) |
| 1125 | .and_then(char::from_u32); |
| 1126 | match valid_char { |
| 1127 | Some(c) => self.print_quoted_escaped_chars(' \'' , iter::once(c))?, |
| 1128 | None => invalid!(self), |
| 1129 | } |
| 1130 | } |
| 1131 | b'e' => { |
| 1132 | // NOTE(eddyb) a string literal `"..."` has type `&str`, so |
| 1133 | // to get back the type `str`, `*"..."` syntax is needed |
| 1134 | // (even if that may not be valid in Rust itself). |
| 1135 | open_brace_if_outside_expr(self)?; |
| 1136 | self.print("*" )?; |
| 1137 | |
| 1138 | self.print_const_str_literal()?; |
| 1139 | } |
| 1140 | |
| 1141 | b'R' | b'Q' => { |
| 1142 | // NOTE(eddyb) this prints `"..."` instead of `&*"..."`, which |
| 1143 | // is what `Re..._` would imply (see comment for `str` above). |
| 1144 | if tag == b'R' && self.eat(b'e' ) { |
| 1145 | self.print_const_str_literal()?; |
| 1146 | } else { |
| 1147 | open_brace_if_outside_expr(self)?; |
| 1148 | self.print("&" )?; |
| 1149 | if tag != b'R' { |
| 1150 | self.print("mut " )?; |
| 1151 | } |
| 1152 | self.print_const(true)?; |
| 1153 | } |
| 1154 | } |
| 1155 | b'A' => { |
| 1156 | open_brace_if_outside_expr(self)?; |
| 1157 | self.print("[" )?; |
| 1158 | self.print_sep_list(|this| this.print_const(true), ", " )?; |
| 1159 | self.print("]" )?; |
| 1160 | } |
| 1161 | b'T' => { |
| 1162 | open_brace_if_outside_expr(self)?; |
| 1163 | self.print("(" )?; |
| 1164 | let count = self.print_sep_list(|this| this.print_const(true), ", " )?; |
| 1165 | if count == 1 { |
| 1166 | self.print("," )?; |
| 1167 | } |
| 1168 | self.print(")" )?; |
| 1169 | } |
| 1170 | b'V' => { |
| 1171 | open_brace_if_outside_expr(self)?; |
| 1172 | self.print_path(true)?; |
| 1173 | match parse!(self, next) { |
| 1174 | b'U' => {} |
| 1175 | b'T' => { |
| 1176 | self.print("(" )?; |
| 1177 | self.print_sep_list(|this| this.print_const(true), ", " )?; |
| 1178 | self.print(")" )?; |
| 1179 | } |
| 1180 | b'S' => { |
| 1181 | self.print(" { " )?; |
| 1182 | self.print_sep_list( |
| 1183 | |this| { |
| 1184 | parse!(this, disambiguator); |
| 1185 | let name = parse!(this, ident); |
| 1186 | this.print(name)?; |
| 1187 | this.print(": " )?; |
| 1188 | this.print_const(true) |
| 1189 | }, |
| 1190 | ", " , |
| 1191 | )?; |
| 1192 | self.print(" }" )?; |
| 1193 | } |
| 1194 | _ => invalid!(self), |
| 1195 | } |
| 1196 | } |
| 1197 | b'B' => { |
| 1198 | self.print_backref(|this| this.print_const(in_value))?; |
| 1199 | } |
| 1200 | _ => invalid!(self), |
| 1201 | } |
| 1202 | |
| 1203 | if opened_brace { |
| 1204 | self.print("}" )?; |
| 1205 | } |
| 1206 | |
| 1207 | self.pop_depth(); |
| 1208 | Ok(()) |
| 1209 | } |
| 1210 | |
| 1211 | fn print_const_uint(&mut self, ty_tag: u8) -> fmt::Result { |
| 1212 | let hex = parse!(self, hex_nibbles); |
| 1213 | |
| 1214 | match hex.try_parse_uint() { |
| 1215 | Some(v) => self.print(v)?, |
| 1216 | |
| 1217 | // Print anything that doesn't fit in `u64` verbatim. |
| 1218 | None => { |
| 1219 | self.print("0x" )?; |
| 1220 | self.print(hex.nibbles)?; |
| 1221 | } |
| 1222 | } |
| 1223 | |
| 1224 | if let Some(out) = &mut self.out { |
| 1225 | if !out.alternate() { |
| 1226 | let ty = basic_type(ty_tag).unwrap(); |
| 1227 | self.print(ty)?; |
| 1228 | } |
| 1229 | } |
| 1230 | |
| 1231 | Ok(()) |
| 1232 | } |
| 1233 | |
| 1234 | fn print_const_str_literal(&mut self) -> fmt::Result { |
| 1235 | match parse!(self, hex_nibbles).try_parse_str_chars() { |
| 1236 | Some(chars) => self.print_quoted_escaped_chars('"' , chars), |
| 1237 | None => invalid!(self), |
| 1238 | } |
| 1239 | } |
| 1240 | } |
| 1241 | |
| 1242 | #[cfg (test)] |
| 1243 | mod tests { |
| 1244 | use std::prelude::v1::*; |
| 1245 | |
| 1246 | macro_rules! t { |
| 1247 | ($a:expr, $b:expr) => {{ |
| 1248 | assert_eq!(format!("{}" , ::demangle($a)), $b); |
| 1249 | }}; |
| 1250 | } |
| 1251 | macro_rules! t_nohash { |
| 1252 | ($a:expr, $b:expr) => {{ |
| 1253 | assert_eq!(format!("{:#}" , ::demangle($a)), $b); |
| 1254 | }}; |
| 1255 | } |
| 1256 | macro_rules! t_nohash_type { |
| 1257 | ($a:expr, $b:expr) => { |
| 1258 | t_nohash!(concat!("_RMC0" , $a), concat!("<" , $b, ">" )) |
| 1259 | }; |
| 1260 | } |
| 1261 | macro_rules! t_const { |
| 1262 | ($mangled:expr, $value:expr) => { |
| 1263 | t_nohash!( |
| 1264 | concat!("_RIC0K" , $mangled, "E" ), |
| 1265 | concat!("::<" , $value, ">" ) |
| 1266 | ) |
| 1267 | }; |
| 1268 | } |
| 1269 | macro_rules! t_const_suffixed { |
| 1270 | ($mangled:expr, $value:expr, $value_ty_suffix:expr) => {{ |
| 1271 | t_const!($mangled, $value); |
| 1272 | t!( |
| 1273 | concat!("_RIC0K" , $mangled, "E" ), |
| 1274 | concat!("::<" , $value, $value_ty_suffix, ">" ) |
| 1275 | ); |
| 1276 | }}; |
| 1277 | } |
| 1278 | |
| 1279 | #[test ] |
| 1280 | fn demangle_crate_with_leading_digit() { |
| 1281 | t_nohash!("_RNvC6_123foo3bar" , "123foo::bar" ); |
| 1282 | } |
| 1283 | |
| 1284 | #[test ] |
| 1285 | fn demangle_crate_with_zero_disambiguator() { |
| 1286 | t!("_RC4f128" , "f128" ); |
| 1287 | t_nohash!("_RC4f128" , "f128" ); |
| 1288 | } |
| 1289 | |
| 1290 | #[test ] |
| 1291 | fn demangle_utf8_idents() { |
| 1292 | t_nohash!( |
| 1293 | "_RNqCs4fqI2P2rA04_11utf8_identsu30____7hkackfecea1cbdathfdh9hlq6y" , |
| 1294 | "utf8_idents::საჭმელად_გემრიელი_სადილი" |
| 1295 | ); |
| 1296 | } |
| 1297 | |
| 1298 | #[test ] |
| 1299 | fn demangle_closure() { |
| 1300 | t_nohash!( |
| 1301 | "_RNCNCNgCs6DXkGYLi8lr_2cc5spawn00B5_" , |
| 1302 | "cc::spawn::{closure#0}::{closure#0}" |
| 1303 | ); |
| 1304 | t_nohash!( |
| 1305 | "_RNCINkXs25_NgCsbmNqQUJIY6D_4core5sliceINyB9_4IterhENuNgNoBb_4iter8iterator8Iterator9rpositionNCNgNpB9_6memchr7memrchrs_0E0Bb_" , |
| 1306 | "<core::slice::Iter<u8> as core::iter::iterator::Iterator>::rposition::<core::slice::memchr::memrchr::{closure#1}>::{closure#0}" |
| 1307 | ); |
| 1308 | } |
| 1309 | |
| 1310 | #[test ] |
| 1311 | fn demangle_dyn_trait() { |
| 1312 | t_nohash!( |
| 1313 | "_RINbNbCskIICzLVDPPb_5alloc5alloc8box_freeDINbNiB4_5boxed5FnBoxuEp6OutputuEL_ECs1iopQbuBiw2_3std" , |
| 1314 | "alloc::alloc::box_free::<dyn alloc::boxed::FnBox<(), Output = ()>>" |
| 1315 | ); |
| 1316 | } |
| 1317 | |
| 1318 | #[test ] |
| 1319 | fn demangle_const_generics_preview() { |
| 1320 | // NOTE(eddyb) this was hand-written, before rustc had working |
| 1321 | // const generics support (but the mangling format did include them). |
| 1322 | t_nohash_type!( |
| 1323 | "INtC8arrayvec8ArrayVechKj7b_E" , |
| 1324 | "arrayvec::ArrayVec<u8, 123>" |
| 1325 | ); |
| 1326 | t_const_suffixed!("j7b_" , "123" , "usize" ); |
| 1327 | } |
| 1328 | |
| 1329 | #[test ] |
| 1330 | fn demangle_min_const_generics() { |
| 1331 | t_const!("p" , "_" ); |
| 1332 | t_const_suffixed!("hb_" , "11" , "u8" ); |
| 1333 | t_const_suffixed!("off00ff00ff00ff00ff_" , "0xff00ff00ff00ff00ff" , "u128" ); |
| 1334 | t_const_suffixed!("s98_" , "152" , "i16" ); |
| 1335 | t_const_suffixed!("anb_" , "-11" , "i8" ); |
| 1336 | t_const!("b0_" , "false" ); |
| 1337 | t_const!("b1_" , "true" ); |
| 1338 | t_const!("c76_" , "'v'" ); |
| 1339 | t_const!("c22_" , r#"'"'"# ); |
| 1340 | t_const!("ca_" , "' \\n'" ); |
| 1341 | t_const!("c2202_" , "'∂'" ); |
| 1342 | } |
| 1343 | |
| 1344 | #[test ] |
| 1345 | fn demangle_const_str() { |
| 1346 | t_const!("e616263_" , "{* \"abc \"}" ); |
| 1347 | t_const!("e27_" , r#"{*"'"}"# ); |
| 1348 | t_const!("e090a_" , "{* \"\\t \\n \"}" ); |
| 1349 | t_const!("ee28882c3bc_" , "{* \"∂ü \"}" ); |
| 1350 | t_const!( |
| 1351 | "ee183a1e18390e183ade1839be18394e1839ae18390e183935fe18392e18394e1839b\ |
| 1352 | e183a0e18398e18394e1839ae183985fe183a1e18390e18393e18398e1839ae18398_" , |
| 1353 | "{* \"საჭმელად_გემრიელი_სადილი \"}" |
| 1354 | ); |
| 1355 | t_const!( |
| 1356 | "ef09f908af09fa688f09fa686f09f90ae20c2a720f09f90b6f09f9192e298\ |
| 1357 | 95f09f94a520c2a720f09fa7a1f09f929bf09f929af09f9299f09f929c_" , |
| 1358 | "{* \"🐊🦈🦆🐮 § 🐶👒☕🔥 § 🧡💛💚💙💜 \"}" |
| 1359 | ); |
| 1360 | } |
| 1361 | |
| 1362 | // NOTE(eddyb) this uses the same strings as `demangle_const_str` and should |
| 1363 | // be kept in sync with it - while a macro could be used to generate both |
| 1364 | // `str` and `&str` tests, from a single list of strings, this seems clearer. |
| 1365 | #[test ] |
| 1366 | fn demangle_const_ref_str() { |
| 1367 | t_const!("Re616263_" , " \"abc \"" ); |
| 1368 | t_const!("Re27_" , r#""'""# ); |
| 1369 | t_const!("Re090a_" , " \"\\t \\n \"" ); |
| 1370 | t_const!("Ree28882c3bc_" , " \"∂ü \"" ); |
| 1371 | t_const!( |
| 1372 | "Ree183a1e18390e183ade1839be18394e1839ae18390e183935fe18392e18394e1839b\ |
| 1373 | e183a0e18398e18394e1839ae183985fe183a1e18390e18393e18398e1839ae18398_" , |
| 1374 | " \"საჭმელად_გემრიელი_სადილი \"" |
| 1375 | ); |
| 1376 | t_const!( |
| 1377 | "Ref09f908af09fa688f09fa686f09f90ae20c2a720f09f90b6f09f9192e298\ |
| 1378 | 95f09f94a520c2a720f09fa7a1f09f929bf09f929af09f9299f09f929c_" , |
| 1379 | " \"🐊🦈🦆🐮 § 🐶👒☕🔥 § 🧡💛💚💙💜 \"" |
| 1380 | ); |
| 1381 | } |
| 1382 | |
| 1383 | #[test ] |
| 1384 | fn demangle_const_ref() { |
| 1385 | t_const!("Rp" , "{&_}" ); |
| 1386 | t_const!("Rh7b_" , "{&123}" ); |
| 1387 | t_const!("Rb0_" , "{&false}" ); |
| 1388 | t_const!("Rc58_" , "{&'X'}" ); |
| 1389 | t_const!("RRRh0_" , "{&&&0}" ); |
| 1390 | t_const!("RRRe_" , "{&& \"\"}" ); |
| 1391 | t_const!("QAE" , "{&mut []}" ); |
| 1392 | } |
| 1393 | |
| 1394 | #[test ] |
| 1395 | fn demangle_const_array() { |
| 1396 | t_const!("AE" , "{[]}" ); |
| 1397 | t_const!("Aj0_E" , "{[0]}" ); |
| 1398 | t_const!("Ah1_h2_h3_E" , "{[1, 2, 3]}" ); |
| 1399 | t_const!("ARe61_Re62_Re63_E" , "{[ \"a \", \"b \", \"c \"]}" ); |
| 1400 | t_const!("AAh1_h2_EAh3_h4_EE" , "{[[1, 2], [3, 4]]}" ); |
| 1401 | } |
| 1402 | |
| 1403 | #[test ] |
| 1404 | fn demangle_const_tuple() { |
| 1405 | t_const!("TE" , "{()}" ); |
| 1406 | t_const!("Tj0_E" , "{(0,)}" ); |
| 1407 | t_const!("Th1_b0_E" , "{(1, false)}" ); |
| 1408 | t_const!( |
| 1409 | "TRe616263_c78_RAh1_h2_h3_EE" , |
| 1410 | "{( \"abc \", 'x', &[1, 2, 3])}" |
| 1411 | ); |
| 1412 | } |
| 1413 | |
| 1414 | #[test ] |
| 1415 | fn demangle_const_adt() { |
| 1416 | t_const!( |
| 1417 | "VNvINtNtC4core6option6OptionjE4NoneU" , |
| 1418 | "{core::option::Option::<usize>::None}" |
| 1419 | ); |
| 1420 | t_const!( |
| 1421 | "VNvINtNtC4core6option6OptionjE4SomeTj0_E" , |
| 1422 | "{core::option::Option::<usize>::Some(0)}" |
| 1423 | ); |
| 1424 | t_const!( |
| 1425 | "VNtC3foo3BarS1sRe616263_2chc78_5sliceRAh1_h2_h3_EE" , |
| 1426 | "{foo::Bar { s: \"abc \", ch: 'x', slice: &[1, 2, 3] }}" |
| 1427 | ); |
| 1428 | } |
| 1429 | |
| 1430 | #[test ] |
| 1431 | fn demangle_exponential_explosion() { |
| 1432 | // NOTE(eddyb) because of the prefix added by `t_nohash_type!` is |
| 1433 | // 3 bytes long, `B2_` refers to the start of the type, not `B_`. |
| 1434 | // 6 backrefs (`B8_E` through `B3_E`) result in 2^6 = 64 copies of `_`. |
| 1435 | // Also, because the `p` (`_`) type is after all of the starts of the |
| 1436 | // backrefs, it can be replaced with any other type, independently. |
| 1437 | t_nohash_type!( |
| 1438 | concat!("TTTTTT" , "p" , "B8_E" , "B7_E" , "B6_E" , "B5_E" , "B4_E" , "B3_E" ), |
| 1439 | "((((((_, _), (_, _)), ((_, _), (_, _))), (((_, _), (_, _)), ((_, _), (_, _)))), \ |
| 1440 | ((((_, _), (_, _)), ((_, _), (_, _))), (((_, _), (_, _)), ((_, _), (_, _))))), \ |
| 1441 | (((((_, _), (_, _)), ((_, _), (_, _))), (((_, _), (_, _)), ((_, _), (_, _)))), \ |
| 1442 | ((((_, _), (_, _)), ((_, _), (_, _))), (((_, _), (_, _)), ((_, _), (_, _))))))" |
| 1443 | ); |
| 1444 | } |
| 1445 | |
| 1446 | #[test ] |
| 1447 | fn demangle_thinlto() { |
| 1448 | t_nohash!("_RC3foo.llvm.9D1C9369" , "foo" ); |
| 1449 | t_nohash!("_RC3foo.llvm.9D1C9369@@16" , "foo" ); |
| 1450 | t_nohash!("_RNvC9backtrace3foo.llvm.A5310EB9" , "backtrace::foo" ); |
| 1451 | } |
| 1452 | |
| 1453 | #[test ] |
| 1454 | fn demangle_extra_suffix() { |
| 1455 | // From alexcrichton/rustc-demangle#27: |
| 1456 | t_nohash!( |
| 1457 | "_RNvNtNtNtNtCs92dm3009vxr_4rand4rngs7adapter9reseeding4fork23FORK_HANDLER_REGISTERED.0.0" , |
| 1458 | "rand::rngs::adapter::reseeding::fork::FORK_HANDLER_REGISTERED.0.0" |
| 1459 | ); |
| 1460 | } |
| 1461 | |
| 1462 | #[test ] |
| 1463 | fn demangling_limits() { |
| 1464 | // Stress tests found via fuzzing. |
| 1465 | |
| 1466 | for sym in include_str!("v0-large-test-symbols/early-recursion-limit" ) |
| 1467 | .lines() |
| 1468 | .filter(|line| !line.is_empty() && !line.starts_with('#' )) |
| 1469 | { |
| 1470 | assert_eq!( |
| 1471 | super::demangle(sym).map(|_| ()), |
| 1472 | Err(super::ParseError::RecursedTooDeep) |
| 1473 | ); |
| 1474 | } |
| 1475 | |
| 1476 | assert_contains!( |
| 1477 | ::demangle( |
| 1478 | "RIC20tRYIMYNRYFG05_EB5_B_B6_RRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRR\ |
| 1479 | RRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRB_E" , |
| 1480 | ) |
| 1481 | .to_string(), |
| 1482 | "{recursion limit reached}" |
| 1483 | ); |
| 1484 | } |
| 1485 | |
| 1486 | #[test ] |
| 1487 | fn recursion_limit_leaks() { |
| 1488 | // NOTE(eddyb) this test checks that both paths and types support the |
| 1489 | // recursion limit correctly, i.e. matching `push_depth` and `pop_depth`, |
| 1490 | // and don't leak "recursion levels" and trip the limit. |
| 1491 | // The test inputs are generated on the fly, using a repeated pattern, |
| 1492 | // as hardcoding the actual strings would be too verbose. |
| 1493 | // Also, `MAX_DEPTH` can be directly used, instead of assuming its value. |
| 1494 | for &(sym_leaf, expected_leaf) in &[("p" , "_" ), ("Rp" , "&_" ), ("C1x" , "x" )] { |
| 1495 | let mut sym = format!("_RIC0p" ); |
| 1496 | let mut expected = format!("::<_" ); |
| 1497 | for _ in 0..(super::MAX_DEPTH * 2) { |
| 1498 | sym.push_str(sym_leaf); |
| 1499 | expected.push_str(", " ); |
| 1500 | expected.push_str(expected_leaf); |
| 1501 | } |
| 1502 | sym.push('E' ); |
| 1503 | expected.push('>' ); |
| 1504 | |
| 1505 | t_nohash!(&sym, expected); |
| 1506 | } |
| 1507 | } |
| 1508 | |
| 1509 | #[test ] |
| 1510 | fn recursion_limit_backref_free_bypass() { |
| 1511 | // NOTE(eddyb) this test checks that long symbols cannot bypass the |
| 1512 | // recursion limit by not using backrefs, and cause a stack overflow. |
| 1513 | |
| 1514 | // This value was chosen to be high enough that stack overflows were |
| 1515 | // observed even with `cargo test --release`. |
| 1516 | let depth = 100_000; |
| 1517 | |
| 1518 | // In order to hide the long mangling from the initial "shallow" parse, |
| 1519 | // it's nested in an identifier (crate name), preceding its use. |
| 1520 | let mut sym = format!("_RIC{}" , depth); |
| 1521 | let backref_start = sym.len() - 2; |
| 1522 | for _ in 0..depth { |
| 1523 | sym.push('R' ); |
| 1524 | } |
| 1525 | |
| 1526 | // Write a backref to just after the length of the identifier. |
| 1527 | sym.push('B' ); |
| 1528 | sym.push(char::from_digit((backref_start - 1) as u32, 36).unwrap()); |
| 1529 | sym.push('_' ); |
| 1530 | |
| 1531 | // Close the `I` at the start. |
| 1532 | sym.push('E' ); |
| 1533 | |
| 1534 | assert_contains!(::demangle(&sym).to_string(), "{recursion limit reached}" ); |
| 1535 | } |
| 1536 | } |
| 1537 | |