oxedyne/fe2o3/fe2o3_net/src/email/header.rs
21.1 KiB, 1 run
created by r1870400018:61442, which is this file's identity for as long as the history lasts, whatever it is later renamed to
download · who wrote it · its history
| 1 | //! Lifted out of `crate::dkim` on 2026-09-23, so the submission server reads a message's |
| 2 | //! header fields exactly as the signer that then signs them does. |
| 3 | //! |
| 4 | //! Made strict on 2026-09-24 (D-06 audit F4-1). The first reader split lines on LF alone, dropped |
| 5 | //! a line with no colon before unfolding, and let a colon clear the address before it, so |
| 6 | //! `From: ceo@x:hello@x;` and `From: ceo@x\r<hello@x>` passed the sender check as `hello@x` |
| 7 | //! while a receiver's parser read `ceo@x`. Whatever cannot be read as RFC 5322 reads it is now an |
| 8 | //! error rather than a guess, because the caller is deciding whose name a message goes out under. |
| 9 | |
| 10 | use oxedyne_fe2o3_core::prelude::*; |
| 11 | |
| 12 | |
| 13 | /// A message split at its first blank line: the header block, without the blank line, and the |
| 14 | /// body, untouched. Only `\r\n\r\n` ends the header section. A message without one is all header |
| 15 | /// block, and [`header_fields`] then refuses the bare line ends it holds, rather than this |
| 16 | /// guessing at a second rule for where the header stops. |
| 17 | pub fn split_headers_body(message: &[u8]) -> (&[u8], &[u8]) { |
| 18 | match find_subseq(message, b"\r\n\r\n") { |
| 19 | Some(i) => (&message[..i], &message[i + 4..]), |
| 20 | None => (message, &[]), |
| 21 | } |
| 22 | } |
| 23 | |
| 24 | fn find_subseq(hay: &[u8], needle: &[u8]) -> Option<usize> { |
| 25 | if needle.is_empty() || needle.len() > hay.len() { return None; } |
| 26 | for i in 0..=hay.len() - needle.len() { |
| 27 | if &hay[i..i + needle.len()] == needle { |
| 28 | return Some(i); |
| 29 | } |
| 30 | } |
| 31 | None |
| 32 | } |
| 33 | |
| 34 | /// Is `c` a line end to some reader somewhere? A control character other than HTAB, or a Unicode |
| 35 | /// line or paragraph separator. |
| 36 | fn breaks_lines(c: char) -> bool { |
| 37 | (c.is_control() && c != '\t') || c == '\u{2028}' || c == '\u{2029}' |
| 38 | } |
| 39 | |
| 40 | /// `(name, unfolded value)` pairs, in the order they appeared, read as RFC 5322 §2.2 reads them. |
| 41 | /// |
| 42 | /// Lines end at CRLF. A CR or an LF standing alone, any other control character, and bytes that |
| 43 | /// are not UTF-8 are refused rather than read one way here and another by the receiver. Folding |
| 44 | /// is undone before any line is judged, so a field's name cannot hide on a line of its own. An |
| 45 | /// empty line, a continuation with nothing before it to continue, and a line with no colon or |
| 46 | /// with a name that is not printable ASCII are all refused. A value keeps its folding WSP, which |
| 47 | /// DKIM's relaxed canonicalisation collapses later; the obsolete WSP before the colon (§4.5) is |
| 48 | /// no part of the name. |
| 49 | pub fn header_fields(headers: &[u8]) -> Outcome<Vec<(String, String)>> { |
| 50 | let text = match std::str::from_utf8(headers) { |
| 51 | Ok(t) => t, |
| 52 | Err(e) => return Err(err!( |
| 53 | "The header section is not UTF-8, from byte {}.", e.valid_up_to(); |
| 54 | Invalid, Input, Decode)), |
| 55 | }; |
| 56 | // Physical lines, each ended by CRLF. A trailing CRLF leaves nothing after it. |
| 57 | let mut lines: Vec<&str> = Vec::new(); |
| 58 | let mut start = 0usize; |
| 59 | let mut chars = text.char_indices(); |
| 60 | while let Some((i, c)) = chars.next() { |
| 61 | match c { |
| 62 | '\r' => match chars.next() { |
| 63 | Some((_, '\n')) => { |
| 64 | lines.push(&text[start..i]); |
| 65 | start = i + 2; |
| 66 | }, |
| 67 | _ => return Err(err!( |
| 68 | "The header section holds a CR not followed by LF, at byte {}.", i; |
| 69 | Invalid, Input)), |
| 70 | }, |
| 71 | '\n' => return Err(err!( |
| 72 | "The header section holds an LF not preceded by CR, at byte {}.", i; |
| 73 | Invalid, Input)), |
| 74 | c if breaks_lines(c) => return Err(err!( |
| 75 | "The header section holds the control character {:?}, at byte {}.", c, i; |
| 76 | Invalid, Input)), |
| 77 | _ => {}, |
| 78 | } |
| 79 | } |
| 80 | if start < text.len() { |
| 81 | lines.push(&text[start..]); |
| 82 | } |
| 83 | // Unfolded fields: a line starting with WSP continues the one before it. |
| 84 | let mut fields: Vec<String> = Vec::new(); |
| 85 | for (n, line) in lines.iter().enumerate() { |
| 86 | if line.is_empty() { |
| 87 | return Err(err!( |
| 88 | "Header line {} is empty, which would end the header section there.", n + 1; |
| 89 | Invalid, Input)); |
| 90 | } |
| 91 | if line.starts_with(' ') || line.starts_with('\t') { |
| 92 | match fields.last_mut() { |
| 93 | Some(f) => f.push_str(line), |
| 94 | None => return Err(err!( |
| 95 | "Header line 1 is a continuation, with no field before it to continue."; |
| 96 | Invalid, Input)), |
| 97 | } |
| 98 | } else { |
| 99 | fields.push(line.to_string()); |
| 100 | } |
| 101 | } |
| 102 | let mut out = Vec::with_capacity(fields.len()); |
| 103 | for (n, field) in fields.into_iter().enumerate() { |
| 104 | let colon = match field.find(':') { |
| 105 | Some(c) => c, |
| 106 | None => return Err(err!( |
| 107 | "Header field {} has no colon, so it is not a field: {:?}.", n + 1, field; |
| 108 | Invalid, Input, Missing)), |
| 109 | }; |
| 110 | let name = field[..colon].trim_end_matches([' ', '\t']); |
| 111 | if name.is_empty() || !name.bytes().all(|b| (33..=126).contains(&b)) { |
| 112 | return Err(err!( |
| 113 | "Header field {} has the name {:?}, which is not printable ASCII.", n + 1, name; |
| 114 | Invalid, Input)); |
| 115 | } |
| 116 | out.push((name.to_string(), field[colon + 1..].to_string())); |
| 117 | } |
| 118 | Ok(out) |
| 119 | } |
| 120 | |
| 121 | /// The addresses in an address-list field such as `From`, `Sender` or `To`, each as its bare |
| 122 | /// addr-spec, in order, with comments and folding white space removed. |
| 123 | /// |
| 124 | /// RFC 5322 §3.4, with the UTF-8 of RFC 6532: `a@b`, `Name <a@b>`, `"Last, First" <a@b>`, |
| 125 | /// comments in parentheses, groups (`Team: a@b, c@d;` and `undisclosed-recipients:;`), and the |
| 126 | /// obsolete empty members and dotted phrases. Anything else is an error rather than a guess -- |
| 127 | /// a member with no address, text after an address, a quote, comment or bracket left open, an |
| 128 | /// obsolete route -- since the caller is deciding whose name a message goes out under. |
| 129 | pub fn addresses(value: &str) -> Outcome<Vec<String>> { |
| 130 | if let Some(c) = value.chars().find(|c| breaks_lines(*c)) { |
| 131 | return Err(err!( |
| 132 | "An address list holds the control character {:?}: {:?}.", c, value; |
| 133 | Invalid, Input)); |
| 134 | } |
| 135 | let toks = res!(tokens(value)); |
| 136 | let mut r = Reader { toks: &toks, pos: 0, value }; |
| 137 | r.address_list() |
| 138 | } |
| 139 | |
| 140 | /// One lexical piece of an address list; comments and white space are gone by now. |
| 141 | #[derive(Clone, Debug, Eq, PartialEq)] |
| 142 | enum Tok { |
| 143 | Atom(String), // a run of atext |
| 144 | Quoted(String), // a quoted-string, quotes and escapes kept as written |
| 145 | Literal(String), // a domain-literal, brackets kept |
| 146 | Special(char), // one of `< > : ; @ , .` |
| 147 | } |
| 148 | |
| 149 | /// Is `c` RFC 5322 atext? Beyond ASCII, anything printable is, per RFC 6532. |
| 150 | fn is_atext(c: char) -> bool { |
| 151 | c.is_ascii_alphanumeric() |
| 152 | || "!#$%&'*+-/=?^_`{|}~".contains(c) |
| 153 | || (!c.is_ascii() && !c.is_whitespace() && !breaks_lines(c)) |
| 154 | } |
| 155 | |
| 156 | fn tokens(value: &str) -> Outcome<Vec<Tok>> { |
| 157 | let mut out = Vec::new(); |
| 158 | let mut chars = value.chars().peekable(); |
| 159 | while let Some(c) = chars.next() { |
| 160 | match c { |
| 161 | ' ' | '\t' => {}, |
| 162 | '(' => { |
| 163 | // A comment, nested comments and quoted pairs included, is white space. |
| 164 | let mut depth = 1usize; |
| 165 | while depth > 0 { |
| 166 | match chars.next() { |
| 167 | Some('\\') => { let _ = chars.next(); }, |
| 168 | Some('(') => depth += 1, |
| 169 | Some(')') => depth -= 1, |
| 170 | Some(_) => {}, |
| 171 | None => return Err(err!( |
| 172 | "An address list leaves a comment open: {:?}.", value; |
| 173 | Invalid, Input)), |
| 174 | } |
| 175 | } |
| 176 | }, |
| 177 | '"' => { |
| 178 | let mut q = String::from('"'); |
| 179 | loop { |
| 180 | match chars.next() { |
| 181 | Some('\\') => match chars.next() { |
| 182 | Some(e) => { |
| 183 | q.push('\\'); |
| 184 | q.push(e); |
| 185 | }, |
| 186 | None => return Err(err!( |
| 187 | "An address list leaves a quoted string open: {:?}.", value; |
| 188 | Invalid, Input)), |
| 189 | }, |
| 190 | Some('"') => { |
| 191 | q.push('"'); |
| 192 | break; |
| 193 | }, |
| 194 | Some(c) => q.push(c), |
| 195 | None => return Err(err!( |
| 196 | "An address list leaves a quoted string open: {:?}.", value; |
| 197 | Invalid, Input)), |
| 198 | } |
| 199 | } |
| 200 | out.push(Tok::Quoted(q)); |
| 201 | }, |
| 202 | '[' => { |
| 203 | let mut l = String::from('['); |
| 204 | loop { |
| 205 | match chars.next() { |
| 206 | Some(']') => { |
| 207 | l.push(']'); |
| 208 | break; |
| 209 | }, |
| 210 | Some(c) if c == '[' || c == '\\' => return Err(err!( |
| 211 | "An address list's domain literal holds {:?}: {:?}.", c, value; |
| 212 | Invalid, Input)), |
| 213 | Some(c) => l.push(c), |
| 214 | None => return Err(err!( |
| 215 | "An address list leaves a domain literal open: {:?}.", value; |
| 216 | Invalid, Input)), |
| 217 | } |
| 218 | } |
| 219 | out.push(Tok::Literal(l)); |
| 220 | }, |
| 221 | '<' | '>' | ':' | ';' | '@' | ',' | '.' => out.push(Tok::Special(c)), |
| 222 | c if is_atext(c) => { |
| 223 | let mut a = String::from(c); |
| 224 | while let Some(&n) = chars.peek() { |
| 225 | if !is_atext(n) { |
| 226 | break; |
| 227 | } |
| 228 | a.push(n); |
| 229 | let _ = chars.next(); |
| 230 | } |
| 231 | out.push(Tok::Atom(a)); |
| 232 | }, |
| 233 | c => return Err(err!( |
| 234 | "An address list holds {:?}, which may not stand there: {:?}.", c, value; |
| 235 | Invalid, Input)), |
| 236 | } |
| 237 | } |
| 238 | Ok(out) |
| 239 | } |
| 240 | |
| 241 | /// A recursive-descent reader of RFC 5322 §3.4 over [`Tok`]s. |
| 242 | struct Reader<'a> { |
| 243 | toks: &'a [Tok], |
| 244 | pos: usize, |
| 245 | value: &'a str, // the field as written, for the errors |
| 246 | } |
| 247 | |
| 248 | impl<'a> Reader<'a> { |
| 249 | fn peek(&self) -> Option<&'a Tok> { |
| 250 | self.toks.get(self.pos) |
| 251 | } |
| 252 | |
| 253 | fn at(&self, c: char) -> bool { |
| 254 | matches!(self.peek(), Some(Tok::Special(s)) if *s == c) |
| 255 | } |
| 256 | |
| 257 | fn expect(&mut self, c: char, phase: &str) -> Outcome<()> { |
| 258 | if self.at(c) { |
| 259 | self.pos += 1; |
| 260 | return Ok(()); |
| 261 | } |
| 262 | Err(err!( |
| 263 | "An address list wants {:?} {}, and has {:?} at token {}: {:?}.", |
| 264 | c, phase, self.peek(), self.pos, self.value; |
| 265 | Invalid, Input)) |
| 266 | } |
| 267 | |
| 268 | /// The first `<`, `:`, `@`, `,` or `;` from here on, which says what the next member is: |
| 269 | /// an angle address, a group, a bare addr-spec, or no address at all. |
| 270 | fn lookahead(&self) -> Option<char> { |
| 271 | self.toks.iter().skip(self.pos).find_map(|t| match t { |
| 272 | Tok::Special(c) if matches!(c, '<' | ':' | '@' | ',' | ';') => Some(*c), |
| 273 | _ => None, |
| 274 | }) |
| 275 | } |
| 276 | |
| 277 | fn address_list(&mut self) -> Outcome<Vec<String>> { |
| 278 | let mut out = Vec::new(); |
| 279 | while self.peek().is_some() { |
| 280 | if self.at(',') { |
| 281 | // The obsolete empty member. |
| 282 | self.pos += 1; |
| 283 | continue; |
| 284 | } |
| 285 | match self.lookahead() { |
| 286 | Some(':') => out.extend(res!(self.group())), |
| 287 | _ => out.push(res!(self.mailbox())), |
| 288 | } |
| 289 | if self.peek().is_some() { |
| 290 | res!(self.expect(',', "after an address")); |
| 291 | } |
| 292 | } |
| 293 | Ok(out) |
| 294 | } |
| 295 | |
| 296 | /// `display-name ":" [group-list] ";"`, whose members are mailboxes and never groups. |
| 297 | fn group(&mut self) -> Outcome<Vec<String>> { |
| 298 | res!(self.phrase(true)); |
| 299 | res!(self.expect(':', "after a group's name")); |
| 300 | let mut out = Vec::new(); |
| 301 | loop { |
| 302 | if self.at(';') { |
| 303 | self.pos += 1; |
| 304 | return Ok(out); |
| 305 | } |
| 306 | if self.at(',') { |
| 307 | self.pos += 1; |
| 308 | continue; |
| 309 | } |
| 310 | if self.peek().is_none() { |
| 311 | return Err(err!( |
| 312 | "An address list leaves a group without its closing ';': {:?}.", self.value; |
| 313 | Invalid, Input, Missing)); |
| 314 | } |
| 315 | out.push(res!(self.mailbox())); |
| 316 | if !self.at(';') { |
| 317 | res!(self.expect(',', "after a group member")); |
| 318 | } |
| 319 | } |
| 320 | } |
| 321 | |
| 322 | /// `name-addr / addr-spec`. |
| 323 | fn mailbox(&mut self) -> Outcome<String> { |
| 324 | match self.lookahead() { |
| 325 | Some('<') => { |
| 326 | res!(self.phrase(false)); |
| 327 | res!(self.expect('<', "before an address")); |
| 328 | if self.at('@') { |
| 329 | return Err(err!( |
| 330 | "An address list holds an obsolete source route, which is not read: \ |
| 331 | {:?}.", self.value; |
| 332 | Invalid, Input, Unimplemented)); |
| 333 | } |
| 334 | let spec = res!(self.addr_spec()); |
| 335 | res!(self.expect('>', "after an address")); |
| 336 | Ok(spec) |
| 337 | }, |
| 338 | Some('@') => self.addr_spec(), |
| 339 | _ => Err(err!( |
| 340 | "An address list member at token {} holds no address: {:?}.", self.pos, self.value; |
| 341 | Invalid, Input, Missing)), |
| 342 | } |
| 343 | } |
| 344 | |
| 345 | /// A display name: words, with the obsolete `.` after the first. A group's is required. |
| 346 | fn phrase(&mut self, required: bool) -> Outcome<()> { |
| 347 | let mut words = 0usize; |
| 348 | loop { |
| 349 | match self.peek() { |
| 350 | Some(Tok::Atom(_)) | Some(Tok::Quoted(_)) => words += 1, |
| 351 | Some(Tok::Special('.')) if words > 0 => {}, |
| 352 | _ => break, |
| 353 | } |
| 354 | self.pos += 1; |
| 355 | } |
| 356 | if required && words == 0 { |
| 357 | return Err(err!( |
| 358 | "An address list holds a group with no name: {:?}.", self.value; |
| 359 | Invalid, Input, Missing)); |
| 360 | } |
| 361 | Ok(()) |
| 362 | } |
| 363 | |
| 364 | /// `local-part "@" domain`, as one string. |
| 365 | fn addr_spec(&mut self) -> Outcome<String> { |
| 366 | let mut spec = res!(self.dotted(true, "local part")); |
| 367 | res!(self.expect('@', "after a local part")); |
| 368 | spec.push('@'); |
| 369 | match self.peek() { |
| 370 | Some(Tok::Literal(l)) => { |
| 371 | spec.push_str(l); |
| 372 | self.pos += 1; |
| 373 | }, |
| 374 | _ => spec.push_str(&res!(self.dotted(false, "domain"))), |
| 375 | } |
| 376 | Ok(spec) |
| 377 | } |
| 378 | |
| 379 | /// `word *("." word)`, where a quoted word may stand in a local part but not in a domain. |
| 380 | fn dotted(&mut self, quoted_ok: bool, what: &str) -> Outcome<String> { |
| 381 | let mut s = String::new(); |
| 382 | loop { |
| 383 | match self.peek() { |
| 384 | Some(Tok::Atom(a)) => s.push_str(a), |
| 385 | Some(Tok::Quoted(q)) if quoted_ok => s.push_str(q), |
| 386 | other => return Err(err!( |
| 387 | "An address's {} wants a word, and has {:?} at token {}: {:?}.", |
| 388 | what, other, self.pos, self.value; |
| 389 | Invalid, Input)), |
| 390 | } |
| 391 | self.pos += 1; |
| 392 | if !self.at('.') { |
| 393 | return Ok(s); |
| 394 | } |
| 395 | self.pos += 1; |
| 396 | s.push('.'); |
| 397 | } |
| 398 | } |
| 399 | } |
| 400 | |
| 401 | |
| 402 | #[cfg(test)] |
| 403 | mod tests { |
| 404 | use super::*; |
| 405 | |
| 406 | fn addrs(v: &str) -> Vec<String> { |
| 407 | match addresses(v) { |
| 408 | Ok(a) => a, |
| 409 | Err(e) => panic!("{:?} would not read: {}", v, e), |
| 410 | } |
| 411 | } |
| 412 | |
| 413 | fn fields(head: &[u8]) -> Vec<(String, String)> { |
| 414 | match header_fields(head) { |
| 415 | Ok(f) => f, |
| 416 | Err(e) => panic!("{:?} would not read: {}", String::from_utf8_lossy(head), e), |
| 417 | } |
| 418 | } |
| 419 | |
| 420 | #[test] |
| 421 | fn the_forms_mail_clients_write_are_read_00() { |
| 422 | assert_eq!(addrs(" hello@oxegen.io"), vec!["hello@oxegen.io"]); |
| 423 | assert_eq!(addrs(" Jason Hoogland <hello@oxegen.io>"), vec!["hello@oxegen.io"]); |
| 424 | assert_eq!(addrs(" \"Hoogland, Jason\" <hello@oxegen.io>"), vec!["hello@oxegen.io"], |
| 425 | "a comma inside a quoted display name is not a separator"); |
| 426 | assert_eq!(addrs(" \"Hoogland <Jason>: CEO@x.test\" <hello@oxegen.io>"), |
| 427 | vec!["hello@oxegen.io"], "nor is an angle bracket, colon or @ inside one"); |
| 428 | assert_eq!(addrs(" hello@oxegen.io (the owner)"), vec!["hello@oxegen.io"]); |
| 429 | assert_eq!(addrs(" =?UTF-8?Q?Caf=C3=A9?= <news@oxegen.io>"), vec!["news@oxegen.io"]); |
| 430 | assert_eq!(addrs(" Café Oxegen <news@oxegen.io>"), vec!["news@oxegen.io"], |
| 431 | "RFC 6532 UTF-8 in a display name"); |
| 432 | assert_eq!(addrs(" a@x.test, \"B\" <b@y.test>"), vec!["a@x.test", "b@y.test"]); |
| 433 | assert_eq!(addrs(" Team: a@x.test, B <b@y.test>;"), vec!["a@x.test", "b@y.test"]); |
| 434 | assert_eq!(addrs(" undisclosed-recipients:;"), Vec::<String>::new()); |
| 435 | assert_eq!(addrs(" \"john doe\"@x.test"), vec!["\"john doe\"@x.test"], |
| 436 | "a quoted local part is part of the address"); |
| 437 | assert_eq!(addrs(" Name (a comment, with a comma) <a@x.test>"), vec!["a@x.test"]); |
| 438 | assert_eq!(addrs(" J. Hoogland <a@x.test>"), vec!["a@x.test"], "the obsolete dotted phrase"); |
| 439 | assert_eq!(addrs(" a@x.test,, b@y.test,"), vec!["a@x.test", "b@y.test"], |
| 440 | "the obsolete empty members"); |
| 441 | assert_eq!(addrs(" a@[192.0.2.1]"), vec!["a@[192.0.2.1]"]); |
| 442 | } |
| 443 | |
| 444 | /// A member with no address, text after an address, or a bracket left open, is refused |
| 445 | /// rather than guessed at. |
| 446 | #[test] |
| 447 | fn what_cannot_be_read_is_refused_00() { |
| 448 | for bad in [ |
| 449 | " Jason Hoogland", |
| 450 | " <>", |
| 451 | " Doe, John <j@x.test>", // an unquoted comma splits the name from its address |
| 452 | " Name <a@x.test", |
| 453 | " \"open quote <a@x.test>", |
| 454 | " (open comment a@x.test", |
| 455 | " <a@x.test> <b@y.test>", |
| 456 | " @x.test", |
| 457 | " a@", |
| 458 | // D-06 audit F4-1: a standard parser reads each of these as ceo@x.test. |
| 459 | " ceo@x.test:hello@x.test;", |
| 460 | " ceo@x.test <hello@x.test>", |
| 461 | " ceo@x.test\r<hello@x.test>", |
| 462 | " ceo@x.test\rX: hello@x.test", |
| 463 | // A colon after an addr-spec does not make it a group's name. |
| 464 | " ceo@x.test: hello@x.test;", |
| 465 | " Team: a@x.test", // a group must be closed |
| 466 | " Outer: Inner: a@x.test;;",// and never nests |
| 467 | " <@route.test:a@x.test>", // the obsolete source route |
| 468 | " hello@x.test;", |
| 469 | " a@x.test\u{2028}", |
| 470 | ] { |
| 471 | assert!(addresses(bad).is_err(), "{:?} was read as {:?}", bad, addresses(bad)); |
| 472 | } |
| 473 | } |
| 474 | |
| 475 | #[test] |
| 476 | fn header_fields_unfold_and_keep_their_order_00() { |
| 477 | let (head, body) = split_headers_body( |
| 478 | b"From: A\r\n <a@x.test>\r\nSubject: hi\r\nFrom: b@y.test\r\n\r\nbody\r\n"); |
| 479 | let f = fields(head); |
| 480 | assert_eq!(f.len(), 3); |
| 481 | assert_eq!(f[0].0, "From"); |
| 482 | assert_eq!(f[0].1, " A <a@x.test>"); |
| 483 | assert_eq!(f[2], (fmt!("From"), fmt!(" b@y.test"))); |
| 484 | assert_eq!(body, b"body\r\n"); |
| 485 | let (head, body) = split_headers_body(b"Subject: no body"); |
| 486 | assert_eq!(head, b"Subject: no body"); |
| 487 | assert!(body.is_empty()); |
| 488 | assert_eq!(fields(b"From : a@x.test"), vec![(fmt!("From"), fmt!(" a@x.test"))], |
| 489 | "the obsolete WSP before a colon is no part of the name"); |
| 490 | } |
| 491 | |
| 492 | /// Lines end at CRLF, and folding is undone before a line is judged, so a name cannot hide |
| 493 | /// on a line of its own (D-06 audit F4-1). |
| 494 | #[test] |
| 495 | fn header_fields_read_lines_as_rfc_5322_does_00() { |
| 496 | // Unfolded first, the field is `From : ceo@x.test`, so there are two `From`s. |
| 497 | let f = fields(b"From\r\n : ceo@x.test\r\nFrom: hello@x.test"); |
| 498 | assert_eq!(f.iter().filter(|(n, _)| n == "From").count(), 2, "{:?}", f); |
| 499 | for bad in [ |
| 500 | &b"From: ceo@x.test\rX: hello@x.test"[..], |
| 501 | b"From: ceo@x.test\r<hello@x.test>", |
| 502 | b"From: hello@x.test\nFrom: ceo@x.test", |
| 503 | b"From: hello@x.test\r\nX: a\x00b", |
| 504 | b"From: hello@x.test\r\nX: a\x0bb", |
| 505 | b"\r\nFrom: hello@x.test", // an empty first line ends the header there |
| 506 | b" From: hello@x.test", // a continuation of nothing |
| 507 | b"From: hello@x.test\r\nNot a field", |
| 508 | b"From: hello@x.test\r\nX Y: z", |
| 509 | b"From: \xff@x.test", |
| 510 | ] { |
| 511 | assert!(header_fields(bad).is_err(), "{:?} was read as {:?}", |
| 512 | String::from_utf8_lossy(bad), header_fields(bad)); |
| 513 | } |
| 514 | // With no CRLF CRLF, the whole message is header, and its bare LFs are refused rather |
| 515 | // than taken for the end of the header. |
| 516 | let (head, body) = split_headers_body(b"From: hello@x.test\n\nX\r\nFrom: ceo@x.test\r\n"); |
| 517 | assert!(body.is_empty()); |
| 518 | assert!(header_fields(head).is_err()); |
| 519 | } |
| 520 | } |