1 | use crate::{ |
2 | custom_flags::Flag, diagnostics::Level, filter::Match, test_result::Errored, Config, Error, |
3 | }; |
4 | use bstr::{ByteSlice, Utf8Error}; |
5 | use color_eyre::eyre::Result; |
6 | use regex::bytes::Regex; |
7 | pub use spanned::*; |
8 | use std::{ |
9 | collections::{BTreeMap, HashMap}, |
10 | num::NonZeroUsize, |
11 | }; |
12 | |
13 | mod spanned; |
14 | #[cfg (test)] |
15 | mod tests; |
16 | |
17 | /// This crate supports various magic comments that get parsed as file-specific |
18 | /// configuration values. This struct parses them all in one go and then they |
19 | /// get processed by their respective use sites. |
20 | #[derive (Debug, Clone)] |
21 | pub struct Comments { |
22 | /// List of revision names to execute. Can only be specified once |
23 | pub revisions: Option<Vec<String>>, |
24 | /// Comments that are only available under specific revisions. |
25 | /// The defaults are in key `vec![]` |
26 | pub revisioned: BTreeMap<Vec<String>, Revisioned>, |
27 | } |
28 | |
29 | impl Default for Comments { |
30 | fn default() -> Self { |
31 | let mut this: Comments = Self { |
32 | revisions: Default::default(), |
33 | revisioned: Default::default(), |
34 | }; |
35 | this.revisioned.insert(key:vec![], value:Revisioned::default()); |
36 | this |
37 | } |
38 | } |
39 | |
40 | impl Comments { |
41 | /// Check that a comment isn't specified twice across multiple differently revisioned statements. |
42 | /// e.g. `//@[foo, bar] error-in-other-file: bop` and `//@[foo, baz] error-in-other-file boop` would end up |
43 | /// specifying two error patterns that are available in revision `foo`. |
44 | pub fn find_one_for_revision<'a, T: 'a>( |
45 | &'a self, |
46 | revision: &'a str, |
47 | kind: &str, |
48 | f: impl Fn(&'a Revisioned) -> OptWithLine<T>, |
49 | ) -> Result<OptWithLine<T>, Errored> { |
50 | let mut result = None; |
51 | let mut errors = vec![]; |
52 | for (k, rev) in &self.revisioned { |
53 | if !k.iter().any(|r| r == revision) { |
54 | continue; |
55 | } |
56 | if let Some(found) = f(rev).into_inner() { |
57 | if result.is_some() { |
58 | errors.push(found.span); |
59 | } else { |
60 | result = found.into(); |
61 | } |
62 | } |
63 | } |
64 | if result.is_none() { |
65 | result = f(&self.revisioned[&[][..]]).into_inner(); |
66 | } |
67 | if errors.is_empty() { |
68 | Ok(result.into()) |
69 | } else { |
70 | Err(Errored { |
71 | command: format!("<finding flags for revision ` {revision}`>" ), |
72 | errors: vec![Error::MultipleRevisionsWithResults { |
73 | kind: kind.to_string(), |
74 | lines: errors, |
75 | }], |
76 | stderr: vec![], |
77 | stdout: vec![], |
78 | }) |
79 | } |
80 | } |
81 | |
82 | /// Returns an iterator over all revisioned comments that match the revision. |
83 | pub fn for_revision<'a>(&'a self, revision: &'a str) -> impl Iterator<Item = &'a Revisioned> { |
84 | [&self.revisioned[&[][..]]].into_iter().chain( |
85 | self.revisioned |
86 | .iter() |
87 | .filter_map(move |(k, v)| k.iter().any(|rev| rev == revision).then_some(v)), |
88 | ) |
89 | } |
90 | |
91 | /// The comments set for all revisions |
92 | pub fn base(&mut self) -> &mut Revisioned { |
93 | self.revisioned.get_mut(&[][..]).unwrap() |
94 | } |
95 | |
96 | /// The comments set for all revisions |
97 | pub fn base_immut(&self) -> &Revisioned { |
98 | self.revisioned.get(&[][..]).unwrap() |
99 | } |
100 | |
101 | pub(crate) fn exit_status(&self, revision: &str) -> Result<Option<Spanned<i32>>, Errored> { |
102 | Ok(self |
103 | .find_one_for_revision(revision, "`exit_status` annotations" , |r| { |
104 | r.exit_status.clone() |
105 | })? |
106 | .into_inner()) |
107 | } |
108 | |
109 | pub(crate) fn require_annotations(&self, revision: &str) -> Option<Spanned<bool>> { |
110 | self.for_revision(revision).fold(None, |acc, elem| { |
111 | elem.require_annotations.as_ref().cloned().or(acc) |
112 | }) |
113 | } |
114 | } |
115 | |
116 | #[derive (Debug, Clone, Default)] |
117 | /// Comments that can be filtered for specific revisions. |
118 | pub struct Revisioned { |
119 | /// The character range in which this revisioned item was first added. |
120 | /// Used for reporting errors on unknown revisions. |
121 | pub span: Span, |
122 | /// Don't run this test if any of these filters apply |
123 | pub ignore: Vec<Condition>, |
124 | /// Only run this test if all of these filters apply |
125 | pub only: Vec<Condition>, |
126 | /// Generate one .stderr file per bit width, by prepending with `.64bit` and similar |
127 | pub stderr_per_bitwidth: bool, |
128 | /// Additional flags to pass to the executable |
129 | pub compile_flags: Vec<String>, |
130 | /// Additional env vars to set for the executable |
131 | pub env_vars: Vec<(String, String)>, |
132 | /// Normalizations to apply to the stderr output before emitting it to disk |
133 | pub normalize_stderr: Vec<(Match, Vec<u8>)>, |
134 | /// Normalizations to apply to the stdout output before emitting it to disk |
135 | pub normalize_stdout: Vec<(Match, Vec<u8>)>, |
136 | /// Arbitrary patterns to look for in the stderr. |
137 | /// The error must be from another file, as errors from the current file must be |
138 | /// checked via `error_matches`. |
139 | pub(crate) error_in_other_files: Vec<Spanned<Pattern>>, |
140 | pub(crate) error_matches: Vec<ErrorMatch>, |
141 | /// Ignore diagnostics below this level. |
142 | /// `None` means pick the lowest level from the `error_pattern`s. |
143 | pub require_annotations_for_level: OptWithLine<Level>, |
144 | /// The exit status that the driver is expected to emit. |
145 | /// If `None`, any exit status is accepted. |
146 | pub exit_status: OptWithLine<i32>, |
147 | /// `Some(true)` means annotations are required |
148 | /// `Some(false)` means annotations are forbidden |
149 | /// `None` means this revision does not change the base annoatation requirement. |
150 | pub require_annotations: OptWithLine<bool>, |
151 | /// Prefix added to all diagnostic code matchers. Note this will make it impossible |
152 | /// match codes which do not contain this prefix. |
153 | pub diagnostic_code_prefix: OptWithLine<String>, |
154 | /// Tester-specific flags. |
155 | /// The keys are just labels for overwriting or retrieving the value later. |
156 | /// They are mostly used by `Config::custom_comments` handlers, |
157 | /// `ui_test` itself only ever looks at the values, not the keys. |
158 | /// |
159 | /// You usually don't modify this directly but use the `add_custom` or `set_custom_once` |
160 | /// helpers. |
161 | pub custom: BTreeMap<&'static str, Spanned<Vec<Box<dyn Flag>>>>, |
162 | } |
163 | |
164 | impl Revisioned { |
165 | /// Append another flag to an existing or new key |
166 | pub fn add_custom(&mut self, key: &'static str, custom: impl Flag + 'static) { |
167 | self.add_custom_spanned(key, custom, Span::default()) |
168 | } |
169 | |
170 | /// Append another flag to an existing or new key |
171 | pub fn add_custom_spanned( |
172 | &mut self, |
173 | key: &'static str, |
174 | custom: impl Flag + 'static, |
175 | span: Span, |
176 | ) { |
177 | self.custom |
178 | .entry(key) |
179 | .or_insert_with(|| Spanned::new(vec![], span)) |
180 | .content |
181 | .push(Box::new(custom)); |
182 | } |
183 | /// Override or set a flag |
184 | pub fn set_custom(&mut self, key: &'static str, custom: impl Flag + 'static) { |
185 | self.custom |
186 | .insert(key, Spanned::dummy(vec![Box::new(custom)])); |
187 | } |
188 | } |
189 | |
190 | /// Main entry point to parsing comments and handling parsing errors. |
191 | #[derive (Debug)] |
192 | pub struct CommentParser<T> { |
193 | /// The comments being built. |
194 | comments: T, |
195 | /// Any errors that ocurred during comment parsing. |
196 | errors: Vec<Error>, |
197 | /// The available commands and their parsing logic |
198 | commands: HashMap<&'static str, CommandParserFunc>, |
199 | /// The symbol(s) that signify the start of a comment. |
200 | comment_start: &'static str, |
201 | } |
202 | |
203 | /// Command parser function type. |
204 | pub type CommandParserFunc = |
205 | fn(&mut CommentParser<&mut Revisioned>, args: Spanned<&str>, span: Span); |
206 | |
207 | impl<T> std::ops::Deref for CommentParser<T> { |
208 | type Target = T; |
209 | |
210 | fn deref(&self) -> &Self::Target { |
211 | &self.comments |
212 | } |
213 | } |
214 | |
215 | impl<T> std::ops::DerefMut for CommentParser<T> { |
216 | fn deref_mut(&mut self) -> &mut Self::Target { |
217 | &mut self.comments |
218 | } |
219 | } |
220 | |
221 | /// The conditions used for "ignore" and "only" filters. |
222 | #[derive (Debug, Clone)] |
223 | pub enum Condition { |
224 | /// One of the given strings must appear in the host triple. |
225 | Host(Vec<TargetSubStr>), |
226 | /// One of the given string must appear in the target triple. |
227 | Target(Vec<TargetSubStr>), |
228 | /// Tests that the bitwidth is one of the given ones. |
229 | Bitwidth(Vec<u8>), |
230 | /// Tests that the target is the host. |
231 | OnHost, |
232 | } |
233 | |
234 | /// A sub string of a target (or a whole target). |
235 | /// |
236 | /// Effectively a `String` that only allows lowercase chars, integers and dashes. |
237 | #[derive (Debug, Clone)] |
238 | pub struct TargetSubStr(String); |
239 | |
240 | impl PartialEq<&str> for TargetSubStr { |
241 | fn eq(&self, other: &&str) -> bool { |
242 | self.0 == *other |
243 | } |
244 | } |
245 | |
246 | impl std::ops::Deref for TargetSubStr { |
247 | type Target = str; |
248 | |
249 | fn deref(&self) -> &Self::Target { |
250 | &self.0 |
251 | } |
252 | } |
253 | |
254 | impl TryFrom<String> for TargetSubStr { |
255 | type Error = String; |
256 | |
257 | fn try_from(value: String) -> std::result::Result<Self, Self::Error> { |
258 | if valueChars<'_> |
259 | .chars() |
260 | .all(|c: char| c.is_ascii_alphanumeric() || c == '-' || c == '_' ) |
261 | { |
262 | Ok(Self(value)) |
263 | } else { |
264 | Err(format!( |
265 | "target strings can only contain integers, basic alphabet characters or dashes" |
266 | )) |
267 | } |
268 | } |
269 | } |
270 | |
271 | /// An error pattern parsed from a `//~` comment. |
272 | #[derive (Debug, Clone)] |
273 | pub enum Pattern { |
274 | /// A substring that must appear in the error message. |
275 | SubString(String), |
276 | /// A regex that must match the error message. |
277 | Regex(Regex), |
278 | } |
279 | |
280 | #[derive (Debug, Clone)] |
281 | pub(crate) enum ErrorMatchKind { |
282 | /// A level and pattern pair parsed from a `//~ LEVEL: Message` comment. |
283 | Pattern { |
284 | pattern: Spanned<Pattern>, |
285 | level: Level, |
286 | }, |
287 | /// An error code parsed from a `//~ error_code` comment. |
288 | Code(Spanned<String>), |
289 | } |
290 | |
291 | #[derive (Debug, Clone)] |
292 | pub(crate) struct ErrorMatch { |
293 | pub(crate) kind: ErrorMatchKind, |
294 | /// The line this pattern is expecting to find a message in. |
295 | pub(crate) line: NonZeroUsize, |
296 | } |
297 | |
298 | impl Condition { |
299 | fn parse(c: &str, args: &str) -> std::result::Result<Self, String> { |
300 | let args: SplitWhitespace<'_> = args.split_whitespace(); |
301 | match c { |
302 | "on-host" => Ok(Condition::OnHost), |
303 | "bitwidth" => { |
304 | let bits: Vec = args.map(|arg: &str| arg.parse::<u8>().map_err(|_err: ParseIntError| { |
305 | format!("invalid ignore/only filter ending in 'bit': {c:?} is not a valid bitwdith" ) |
306 | })).collect::<Result<Vec<_>, _>>()?; |
307 | Ok(Condition::Bitwidth(bits)) |
308 | } |
309 | "target" => Ok(Condition::Target(args.take_while(|&arg: &str| arg != "#" ).map(|arg: &str|TargetSubStr::try_from(arg.to_owned())).collect::<Result<_, _>>()?)), |
310 | "host" => Ok(Condition::Host(args.take_while(|&arg: &str| arg != "#" ).map(|arg: &str|TargetSubStr::try_from(arg.to_owned())).collect::<Result<_, _>>()?)), |
311 | _ => Err(format!("` {c}` is not a valid condition, expected `on-host`, /[0-9]+bit/, /host-.*/, or /target-.*/" )), |
312 | } |
313 | } |
314 | } |
315 | |
316 | enum ParsePatternResult { |
317 | Other, |
318 | ErrorAbove { |
319 | match_line: NonZeroUsize, |
320 | }, |
321 | ErrorBelow { |
322 | span: Span, |
323 | match_line: NonZeroUsize, |
324 | }, |
325 | Fallthrough { |
326 | span: Span, |
327 | idx: usize, |
328 | }, |
329 | } |
330 | |
331 | impl Comments { |
332 | /// Parse comments in `content`. |
333 | /// `path` is only used to emit diagnostics if parsing fails. |
334 | pub(crate) fn parse( |
335 | content: Spanned<&[u8]>, |
336 | config: &Config, |
337 | ) -> std::result::Result<Self, Vec<Error>> { |
338 | CommentParser::new(config).parse(content) |
339 | } |
340 | } |
341 | |
342 | impl CommentParser<Comments> { |
343 | fn new(config: &Config) -> Self { |
344 | let mut this = Self { |
345 | comments: config.comment_defaults.clone(), |
346 | errors: vec![], |
347 | commands: Self::commands(), |
348 | comment_start: config.comment_start, |
349 | }; |
350 | this.commands |
351 | .extend(config.custom_comments.iter().map(|(&k, &v)| (k, v))); |
352 | this |
353 | } |
354 | |
355 | fn parse(mut self, content: Spanned<&[u8]>) -> std::result::Result<Comments, Vec<Error>> { |
356 | // We take out the existing flags so that we can ensure every test only sets them once |
357 | // by checking that they haven't already been set. |
358 | let mut defaults = std::mem::take(self.comments.revisioned.get_mut(&[][..]).unwrap()); |
359 | |
360 | let mut delayed_fallthrough = Vec::new(); |
361 | let mut fallthrough_to = None; // The line that a `|` will refer to. |
362 | let mut last_line = 0; |
363 | for (l, line) in content.lines().enumerate() { |
364 | last_line = l + 1; |
365 | let l = NonZeroUsize::new(l + 1).unwrap(); // enumerate starts at 0, but line numbers start at 1 |
366 | match self.parse_checked_line(fallthrough_to, l, line) { |
367 | Ok(ParsePatternResult::Other) => { |
368 | fallthrough_to = None; |
369 | } |
370 | Ok(ParsePatternResult::ErrorAbove { match_line }) => { |
371 | fallthrough_to = Some(match_line); |
372 | } |
373 | Ok(ParsePatternResult::Fallthrough { span, idx }) => { |
374 | delayed_fallthrough.push((span, l, idx)); |
375 | } |
376 | Ok(ParsePatternResult::ErrorBelow { span, match_line }) => { |
377 | if fallthrough_to.is_some() { |
378 | self.error( |
379 | span, |
380 | "`//~v` comment immediately following a `//~^` comment chain" , |
381 | ); |
382 | } |
383 | |
384 | for (span, line, idx) in delayed_fallthrough.drain(..) { |
385 | if let Some(rev) = self |
386 | .comments |
387 | .revisioned |
388 | .values_mut() |
389 | .find(|rev| rev.error_matches[idx].line == line) |
390 | { |
391 | rev.error_matches[idx].line = match_line; |
392 | } else { |
393 | self.error(span, "`//~|` comment not attached to anchoring matcher" ); |
394 | } |
395 | } |
396 | } |
397 | Err(e) => self.error(e.span, format!("Comment is not utf8: {:?}" , e.content)), |
398 | } |
399 | } |
400 | if let Some(revisions) = &self.comments.revisions { |
401 | for (key, revisioned) in &self.comments.revisioned { |
402 | for rev in key { |
403 | if !revisions.contains(rev) { |
404 | self.errors.push(Error::InvalidComment { |
405 | msg: format!("the revision ` {rev}` is not known" ), |
406 | span: revisioned.span.clone(), |
407 | }) |
408 | } |
409 | } |
410 | } |
411 | } else { |
412 | for (key, revisioned) in &self.comments.revisioned { |
413 | if !key.is_empty() { |
414 | self.errors.push(Error::InvalidComment { |
415 | msg: "there are no revisions in this test" .into(), |
416 | span: revisioned.span.clone(), |
417 | }) |
418 | } |
419 | } |
420 | } |
421 | |
422 | for revisioned in self.comments.revisioned.values() { |
423 | for m in &revisioned.error_matches { |
424 | if m.line.get() > last_line { |
425 | let span = match &m.kind { |
426 | ErrorMatchKind::Pattern { pattern, .. } => pattern.span(), |
427 | ErrorMatchKind::Code(code) => code.span(), |
428 | }; |
429 | self.errors.push(Error::InvalidComment { |
430 | msg: format!( |
431 | "//~v pattern is trying to refer to line {}, but the file only has {} lines" , |
432 | m.line.get(), |
433 | last_line, |
434 | ), |
435 | span, |
436 | }); |
437 | } |
438 | } |
439 | } |
440 | |
441 | for (span, ..) in delayed_fallthrough { |
442 | self.error(span, "`//~|` comment not attached to anchoring matcher" ); |
443 | } |
444 | |
445 | let Revisioned { |
446 | span, |
447 | ignore, |
448 | only, |
449 | stderr_per_bitwidth, |
450 | compile_flags, |
451 | env_vars, |
452 | normalize_stderr, |
453 | normalize_stdout, |
454 | error_in_other_files, |
455 | error_matches, |
456 | require_annotations_for_level, |
457 | exit_status, |
458 | require_annotations, |
459 | diagnostic_code_prefix, |
460 | custom, |
461 | } = &mut defaults; |
462 | |
463 | // We insert into the defaults so that the defaults are first in case of sorted lists |
464 | // like `normalize_stderr`, `compile_flags`, or `env_vars` |
465 | let base = std::mem::take(self.comments.base()); |
466 | if span.is_dummy() { |
467 | *span = base.span; |
468 | } |
469 | ignore.extend(base.ignore); |
470 | only.extend(base.only); |
471 | *stderr_per_bitwidth |= base.stderr_per_bitwidth; |
472 | compile_flags.extend(base.compile_flags); |
473 | env_vars.extend(base.env_vars); |
474 | normalize_stderr.extend(base.normalize_stderr); |
475 | normalize_stdout.extend(base.normalize_stdout); |
476 | error_in_other_files.extend(base.error_in_other_files); |
477 | error_matches.extend(base.error_matches); |
478 | if base.require_annotations_for_level.is_some() { |
479 | *require_annotations_for_level = base.require_annotations_for_level; |
480 | } |
481 | if base.exit_status.is_some() { |
482 | *exit_status = base.exit_status; |
483 | } |
484 | if base.require_annotations.is_some() { |
485 | *require_annotations = base.require_annotations; |
486 | } |
487 | if base.diagnostic_code_prefix.is_some() { |
488 | *diagnostic_code_prefix = base.diagnostic_code_prefix; |
489 | } |
490 | |
491 | for (k, v) in base.custom { |
492 | custom.insert(k, v); |
493 | } |
494 | |
495 | *self.base() = defaults; |
496 | |
497 | if self.errors.is_empty() { |
498 | Ok(self.comments) |
499 | } else { |
500 | Err(self.errors) |
501 | } |
502 | } |
503 | } |
504 | |
505 | impl CommentParser<Comments> { |
506 | fn parse_checked_line( |
507 | &mut self, |
508 | fallthrough_to: Option<NonZeroUsize>, |
509 | current_line: NonZeroUsize, |
510 | line: Spanned<&[u8]>, |
511 | ) -> std::result::Result<ParsePatternResult, Spanned<Utf8Error>> { |
512 | let mut res = ParsePatternResult::Other; |
513 | |
514 | if let Some((_, comment)) = |
515 | line.split_once_str(self.comment_start) |
516 | .filter(|(pre, c)| match &c[..] { |
517 | [b'@' , ..] => pre.is_empty(), |
518 | [b'~' , ..] => true, |
519 | _ => false, |
520 | }) |
521 | { |
522 | if let Some(command) = comment.strip_prefix(b"@" ) { |
523 | self.parse_command(command.to_str()?.trim()) |
524 | } else if let Some(pattern) = comment.strip_prefix(b"~" ) { |
525 | let (revisions, pattern) = self.parse_revisions(pattern.to_str()?); |
526 | self.revisioned(revisions, |this| { |
527 | res = this.parse_pattern(pattern, fallthrough_to, current_line); |
528 | }) |
529 | } else { |
530 | unreachable!() |
531 | } |
532 | } else { |
533 | for pos in line.clone().find_iter(self.comment_start) { |
534 | let (_, rest) = line.clone().to_str()?.split_at(pos + 2); |
535 | for rest in std::iter::once(rest.clone()).chain(rest.strip_prefix(" " )) { |
536 | let c = rest.chars().next(); |
537 | if let Some(Spanned { |
538 | content: '@' | '~' | '[' | ']' | '^' | '|' , |
539 | span, |
540 | }) = c |
541 | { |
542 | self.error( |
543 | span, |
544 | format!( |
545 | "comment looks suspiciously like a test suite command: ` {}` \n\ |
546 | All ` {}@` test suite commands must be at the start of the line. \n\ |
547 | The ` {}` must be directly followed by `@` or `~`." , |
548 | *rest, self.comment_start, self.comment_start, |
549 | ), |
550 | ); |
551 | } else { |
552 | let mut parser = Self { |
553 | errors: vec![], |
554 | comments: Comments::default(), |
555 | commands: std::mem::take(&mut self.commands), |
556 | comment_start: self.comment_start, |
557 | }; |
558 | let span = rest.span(); |
559 | parser.parse_command(rest); |
560 | if parser.errors.is_empty() { |
561 | self.error( |
562 | span, |
563 | format!( |
564 | "a compiletest-rs style comment was detected. \n\ |
565 | Please use text that could not also be interpreted as a command, \n\ |
566 | and prefix all actual commands with ` {}@`" , |
567 | self.comment_start |
568 | ), |
569 | ); |
570 | } |
571 | self.commands = parser.commands; |
572 | } |
573 | } |
574 | } |
575 | } |
576 | Ok(res) |
577 | } |
578 | } |
579 | |
580 | impl<CommentsType> CommentParser<CommentsType> { |
581 | /// Emits an [`InvalidComment`](Error::InvalidComment) error with the given span and message. |
582 | pub fn error(&mut self, span: Span, s: impl Into<String>) { |
583 | self.errors.push(Error::InvalidComment { |
584 | msg: s.into(), |
585 | span, |
586 | }); |
587 | } |
588 | |
589 | /// Checks a condition and emits an error if it is not met. |
590 | pub fn check(&mut self, span: Span, cond: bool, s: impl Into<String>) { |
591 | if !cond { |
592 | self.error(span, s); |
593 | } |
594 | } |
595 | |
596 | /// Checks an option and emits an error if it is `None`. |
597 | pub fn check_some<T>(&mut self, span: Span, opt: Option<T>, s: impl Into<String>) -> Option<T> { |
598 | self.check(span, cond:opt.is_some(), s); |
599 | opt |
600 | } |
601 | } |
602 | |
603 | impl CommentParser<Comments> { |
604 | fn parse_command(&mut self, command: Spanned<&str>) { |
605 | let (revisions, command) = self.parse_revisions(command); |
606 | |
607 | // Commands are letters or dashes, grab everything until the first character that is neither of those. |
608 | let (command, args) = match command |
609 | .char_indices() |
610 | .find_map(|(i, c)| (!c.is_alphanumeric() && c != '-' && c != '_' ).then_some(i)) |
611 | { |
612 | None => { |
613 | let span = command.span().shrink_to_end(); |
614 | (command, Spanned::new("" , span)) |
615 | } |
616 | Some(i) => { |
617 | let (command, args) = command.split_at(i); |
618 | // Commands are separated from their arguments by ':' |
619 | let next = args |
620 | .chars() |
621 | .next() |
622 | .expect("the `position` above guarantees that there is at least one char" ); |
623 | let pos = next.len_utf8(); |
624 | self.check( |
625 | next.span, |
626 | next.content == ':' , |
627 | "test command must be followed by `:` (or end the line)" , |
628 | ); |
629 | (command, args.split_at(pos).1.trim()) |
630 | } |
631 | }; |
632 | |
633 | if *command == "revisions" { |
634 | self.check( |
635 | revisions.span(), |
636 | revisions.is_empty(), |
637 | "revisions cannot be declared under a revision" , |
638 | ); |
639 | self.check( |
640 | revisions.span(), |
641 | self.revisions.is_none(), |
642 | "cannot specify `revisions` twice" , |
643 | ); |
644 | self.revisions = Some(args.split_whitespace().map(|s| s.to_string()).collect()); |
645 | return; |
646 | } |
647 | self.revisioned(revisions, |this| this.parse_command(command, args)); |
648 | } |
649 | |
650 | fn revisioned( |
651 | &mut self, |
652 | revisions: Spanned<Vec<String>>, |
653 | f: impl FnOnce(&mut CommentParser<&mut Revisioned>), |
654 | ) { |
655 | let Spanned { |
656 | content: revisions, |
657 | span, |
658 | } = revisions; |
659 | let mut this = CommentParser { |
660 | comment_start: self.comment_start, |
661 | errors: std::mem::take(&mut self.errors), |
662 | commands: std::mem::take(&mut self.commands), |
663 | comments: self |
664 | .revisioned |
665 | .entry(revisions) |
666 | .or_insert_with(|| Revisioned { |
667 | span, |
668 | ..Default::default() |
669 | }), |
670 | }; |
671 | f(&mut this); |
672 | let CommentParser { |
673 | errors, commands, .. |
674 | } = this; |
675 | self.commands = commands; |
676 | self.errors = errors; |
677 | } |
678 | } |
679 | |
680 | impl CommentParser<&mut Revisioned> { |
681 | fn parse_normalize_test( |
682 | &mut self, |
683 | args: Spanned<&str>, |
684 | mode: &str, |
685 | ) -> Option<(Regex, Vec<u8>)> { |
686 | let (from, rest) = self.parse_str(args); |
687 | |
688 | let to = match rest.strip_prefix("->" ) { |
689 | Some(v) => v, |
690 | None => { |
691 | self.error( |
692 | rest.span(), |
693 | format!( |
694 | "normalize- {mode}-test needs a pattern and replacement separated by `->`" |
695 | ), |
696 | ); |
697 | return None; |
698 | } |
699 | } |
700 | .trim_start(); |
701 | let (to, rest) = self.parse_str(to); |
702 | |
703 | self.check( |
704 | rest.span(), |
705 | rest.is_empty(), |
706 | "trailing text after pattern replacement" , |
707 | ); |
708 | |
709 | let regex = self.parse_regex(from)?.content; |
710 | Some((regex, to.as_bytes().to_owned())) |
711 | } |
712 | |
713 | /// Adds a flag, or errors if it already existed. |
714 | pub fn set_custom_once(&mut self, key: &'static str, custom: impl Flag + 'static, span: Span) { |
715 | let prev = self |
716 | .custom |
717 | .insert(key, Spanned::new(vec![Box::new(custom)], span.clone())); |
718 | self.check( |
719 | span, |
720 | prev.is_none(), |
721 | format!("cannot specify ` {key}` twice" ), |
722 | ); |
723 | } |
724 | } |
725 | |
726 | impl CommentParser<Comments> { |
727 | fn commands() -> HashMap<&'static str, CommandParserFunc> { |
728 | let mut commands = HashMap::<_, CommandParserFunc>::new(); |
729 | macro_rules! commands { |
730 | ($($name:expr => ($this:ident, $args:ident, $span:ident)$block:block)*) => { |
731 | $(commands.insert($name, |$this, $args, $span| { |
732 | $block |
733 | });)* |
734 | }; |
735 | } |
736 | commands! { |
737 | "compile-flags" => (this, args, _span){ |
738 | if let Some(parsed) = comma::parse_command(*args) { |
739 | this.compile_flags.extend(parsed); |
740 | } else { |
741 | this.error(args.span(), format!("` {}` contains an unclosed quotation mark" , *args)); |
742 | } |
743 | } |
744 | "rustc-env" => (this, args, _span){ |
745 | for env in args.split_whitespace() { |
746 | if let Some((k, v)) = this.check_some( |
747 | args.span(), |
748 | env.split_once('=' ), |
749 | "environment variables must be key/value pairs separated by a `=`" , |
750 | ) { |
751 | this.env_vars.push((k.to_string(), v.to_string())); |
752 | } |
753 | } |
754 | } |
755 | "normalize-stderr-test" => (this, args, _span){ |
756 | if let Some((regex, replacement)) = this.parse_normalize_test(args, "stderr" ) { |
757 | this.normalize_stderr.push((regex.into(), replacement)) |
758 | } |
759 | } |
760 | "normalize-stdout-test" => (this, args, _span){ |
761 | if let Some((regex, replacement)) = this.parse_normalize_test(args, "stdout" ) { |
762 | this.normalize_stdout.push((regex.into(), replacement)) |
763 | } |
764 | } |
765 | "error-pattern" => (this, _args, span){ |
766 | this.error(span, "`error-pattern` has been renamed to `error-in-other-file`" ); |
767 | } |
768 | "error-in-other-file" => (this, args, _span){ |
769 | let args = args.trim(); |
770 | let pat = this.parse_error_pattern(args); |
771 | this.error_in_other_files.push(pat); |
772 | } |
773 | "stderr-per-bitwidth" => (this, _args, span){ |
774 | // args are ignored (can be used as comment) |
775 | this.check( |
776 | span, |
777 | !this.stderr_per_bitwidth, |
778 | "cannot specify `stderr-per-bitwidth` twice" , |
779 | ); |
780 | this.stderr_per_bitwidth = true; |
781 | } |
782 | "run-rustfix" => (this, _args, span){ |
783 | this.error(span, "rustfix is now ran by default when applicable suggestions are found" ); |
784 | } |
785 | "check-pass" => (this, _args, span){ |
786 | _ = this.exit_status.set(0, span.clone()); |
787 | this.require_annotations = Spanned::new(false, span.clone()).into(); |
788 | } |
789 | "require-annotations-for-level" => (this, args, span){ |
790 | let args = args.trim(); |
791 | let prev = match args.content.parse() { |
792 | Ok(it) => this.require_annotations_for_level.set(it, args.span()), |
793 | Err(msg) => { |
794 | this.error(args.span(), msg); |
795 | None |
796 | }, |
797 | }; |
798 | |
799 | this.check( |
800 | span, |
801 | prev.is_none(), |
802 | "cannot specify `require-annotations-for-level` twice" , |
803 | ); |
804 | } |
805 | } |
806 | commands |
807 | } |
808 | } |
809 | |
810 | impl CommentParser<&mut Revisioned> { |
811 | fn parse_command(&mut self, command: Spanned<&str>, args: Spanned<&str>) { |
812 | if let Some(command_handler) = self.commands.get(*command) { |
813 | command_handler(self, args, command.span()); |
814 | } else if let Some(rest) = command |
815 | .strip_prefix("ignore-" ) |
816 | .or_else(|| command.strip_prefix("only-" )) |
817 | { |
818 | // args are ignored (can be used as comment) |
819 | match Condition::parse(*rest, *args) { |
820 | Ok(cond) => { |
821 | if command.starts_with("ignore" ) { |
822 | self.ignore.push(cond) |
823 | } else { |
824 | self.only.push(cond) |
825 | } |
826 | } |
827 | Err(msg) => self.error(rest.span(), msg), |
828 | } |
829 | } else { |
830 | let best_match = self |
831 | .commands |
832 | .keys() |
833 | .min_by_key(|key| levenshtein::levenshtein(key, *command)) |
834 | .unwrap(); |
835 | self.error( |
836 | command.span(), |
837 | format!( |
838 | "` {}` is not a command known to `ui_test`, did you mean ` {best_match}`?" , |
839 | *command |
840 | ), |
841 | ); |
842 | } |
843 | } |
844 | } |
845 | |
846 | impl<CommentsType> CommentParser<CommentsType> { |
847 | fn parse_regex(&mut self, regex: Spanned<&str>) -> Option<Spanned<Regex>> { |
848 | match Regex::new(*regex) { |
849 | Ok(r) => Some(regex.map(|_| r)), |
850 | Err(err) => { |
851 | self.error(regex.span(), format!("invalid regex: {err:?}" )); |
852 | None |
853 | } |
854 | } |
855 | } |
856 | |
857 | /// Parses a string literal. `s` has to start with `"`; everything until the next `"` is |
858 | /// returned in the first component. `\` can be used to escape arbitrary character. |
859 | /// Second return component is the rest of the string with leading whitespace removed. |
860 | fn parse_str<'a>(&mut self, s: Spanned<&'a str>) -> (Spanned<&'a str>, Spanned<&'a str>) { |
861 | match s.strip_prefix(" \"" ) { |
862 | Some(s) => { |
863 | let mut escaped = false; |
864 | for (i, c) in s.char_indices() { |
865 | if escaped { |
866 | // Accept any character as literal after a `\`. |
867 | escaped = false; |
868 | } else if c == '"' { |
869 | let (a, b) = s.split_at(i); |
870 | let b = b.split_at(1).1; |
871 | return (a, b.trim_start()); |
872 | } else { |
873 | escaped = c == ' \\' ; |
874 | } |
875 | } |
876 | self.error(s.span(), format!("no closing quotes found for {}" , *s)); |
877 | let span = s.span(); |
878 | (s, Spanned::new("" , span)) |
879 | } |
880 | None => { |
881 | if s.is_empty() { |
882 | self.error(s.span(), "expected quoted string, but found end of line" ) |
883 | } else { |
884 | let c = s.chars().next().unwrap(); |
885 | self.error(c.span, format!("expected ` \"`, got ` {}`" , c.content)) |
886 | } |
887 | let span = s.span(); |
888 | (s, Spanned::new("" , span)) |
889 | } |
890 | } |
891 | } |
892 | |
893 | // parse something like \[[a-z]+(,[a-z]+)*\] |
894 | fn parse_revisions<'a>( |
895 | &mut self, |
896 | pattern: Spanned<&'a str>, |
897 | ) -> (Spanned<Vec<String>>, Spanned<&'a str>) { |
898 | match pattern.strip_prefix("[" ) { |
899 | Some(s) => { |
900 | // revisions |
901 | let end = s.char_indices().find_map(|(i, c)| match c { |
902 | ']' => Some(i), |
903 | _ => None, |
904 | }); |
905 | let Some(end) = end else { |
906 | self.error(s.span(), "`[` without corresponding `]`" ); |
907 | return ( |
908 | Spanned::new(vec![], pattern.span().shrink_to_start()), |
909 | pattern, |
910 | ); |
911 | }; |
912 | let (revision, pattern) = s.split_at(end); |
913 | let revisions = revision.split(',' ).map(|s| s.trim().to_string()).collect(); |
914 | ( |
915 | Spanned::new(revisions, revision.span()), |
916 | // 1.. because `split_at` includes the separator |
917 | pattern.split_at(1).1.trim_start(), |
918 | ) |
919 | } |
920 | _ => ( |
921 | Spanned::new(vec![], pattern.span().shrink_to_start()), |
922 | pattern, |
923 | ), |
924 | } |
925 | } |
926 | } |
927 | |
928 | impl CommentParser<&mut Revisioned> { |
929 | // parse something like: |
930 | // (\[[a-z]+(,[a-z]+)*\])? |
931 | // (?P<offset>\||[\^]+)? * |
932 | // ((?P<level>ERROR|HELP|WARN|NOTE): (?P<text>.*))|(?P<code>[a-z0-9_:]+) |
933 | fn parse_pattern( |
934 | &mut self, |
935 | pattern: Spanned<&str>, |
936 | fallthrough_to: Option<NonZeroUsize>, |
937 | current_line: NonZeroUsize, |
938 | ) -> ParsePatternResult { |
939 | let c = pattern.chars().next(); |
940 | let mut res = ParsePatternResult::Other; |
941 | |
942 | let (match_line, pattern) = match c { |
943 | Some(Spanned { content: '|' , span }) => ( |
944 | match fallthrough_to { |
945 | Some(match_line) => { |
946 | res = ParsePatternResult::ErrorAbove { match_line }; |
947 | match_line |
948 | } |
949 | None => { |
950 | res = ParsePatternResult::Fallthrough { |
951 | span, |
952 | idx: self.error_matches.len(), |
953 | }; |
954 | current_line |
955 | } |
956 | }, |
957 | pattern.split_at(1).1, |
958 | ), |
959 | Some(Spanned { |
960 | content: '^' , |
961 | span: _, |
962 | }) => { |
963 | let offset = pattern.chars().take_while(|c| c.content == '^' ).count(); |
964 | match current_line |
965 | .get() |
966 | .checked_sub(offset) |
967 | .and_then(NonZeroUsize::new) |
968 | { |
969 | // lines are one-indexed, so a target line of 0 is invalid, but also |
970 | // prevented via `NonZeroUsize` |
971 | Some(match_line) => { |
972 | res = ParsePatternResult::ErrorAbove { match_line }; |
973 | (match_line, pattern.split_at(offset).1) |
974 | } |
975 | _ => { |
976 | self.error(pattern.span(), format!( |
977 | " {}~^ pattern is trying to refer to {} lines above, but there are only {} lines above" , |
978 | self.comment_start, |
979 | offset, |
980 | current_line.get() - 1, |
981 | )); |
982 | return ParsePatternResult::ErrorAbove { |
983 | match_line: current_line, |
984 | }; |
985 | } |
986 | } |
987 | } |
988 | Some(Spanned { |
989 | content: 'v' , |
990 | span: _, |
991 | }) => { |
992 | let offset = pattern.chars().take_while(|c| c.content == 'v' ).count(); |
993 | match current_line |
994 | .get() |
995 | .checked_add(offset) |
996 | .and_then(NonZeroUsize::new) |
997 | { |
998 | Some(match_line) => { |
999 | res = ParsePatternResult::ErrorBelow { |
1000 | span: pattern.span(), |
1001 | match_line, |
1002 | }; |
1003 | (match_line, pattern.split_at(offset).1) |
1004 | } |
1005 | _ => { |
1006 | // The line count of the file is not yet known so we can only check |
1007 | // if the resulting line is in the range of a usize. |
1008 | self.error(pattern.span(), format!( |
1009 | " {}~v pattern is trying to refer to {} lines below, which is more than ui_test can count" , |
1010 | self.comment_start, |
1011 | offset, |
1012 | )); |
1013 | return ParsePatternResult::ErrorBelow { |
1014 | span: pattern.span(), |
1015 | match_line: current_line, |
1016 | }; |
1017 | } |
1018 | } |
1019 | } |
1020 | Some(_) => (current_line, pattern), |
1021 | None => { |
1022 | self.error(pattern.span(), "no pattern specified" ); |
1023 | return res; |
1024 | } |
1025 | }; |
1026 | |
1027 | let pattern = pattern.trim_start(); |
1028 | let offset = pattern |
1029 | .bytes() |
1030 | .position(|c| !(c.is_ascii_alphanumeric() || c == b'_' || c == b':' )) |
1031 | .unwrap_or(pattern.len()); |
1032 | |
1033 | let (level_or_code, pattern) = pattern.split_at(offset); |
1034 | if let Some(level) = level_or_code.strip_suffix(":" ) { |
1035 | let level = match (*level).parse() { |
1036 | Ok(level) => level, |
1037 | Err(msg) => { |
1038 | self.error(level.span(), msg); |
1039 | return res; |
1040 | } |
1041 | }; |
1042 | |
1043 | let pattern = pattern.trim(); |
1044 | |
1045 | self.check(pattern.span(), !pattern.is_empty(), "no pattern specified" ); |
1046 | |
1047 | let pattern = self.parse_error_pattern(pattern); |
1048 | |
1049 | self.error_matches.push(ErrorMatch { |
1050 | kind: ErrorMatchKind::Pattern { pattern, level }, |
1051 | line: match_line, |
1052 | }); |
1053 | } else if (*level_or_code).parse::<Level>().is_ok() { |
1054 | // Shouldn't conflict with any real diagnostic code |
1055 | self.error(level_or_code.span(), "no `:` after level found" ); |
1056 | return res; |
1057 | } else if !pattern.trim_start().is_empty() { |
1058 | self.error( |
1059 | pattern.span(), |
1060 | format!("text found after error code ` {}`" , *level_or_code), |
1061 | ); |
1062 | return res; |
1063 | } else { |
1064 | self.error_matches.push(ErrorMatch { |
1065 | kind: ErrorMatchKind::Code(Spanned::new( |
1066 | level_or_code.to_string(), |
1067 | level_or_code.span(), |
1068 | )), |
1069 | line: match_line, |
1070 | }); |
1071 | }; |
1072 | |
1073 | res |
1074 | } |
1075 | } |
1076 | |
1077 | impl Pattern { |
1078 | pub(crate) fn matches(&self, message: &str) -> bool { |
1079 | match self { |
1080 | Pattern::SubString(s: &String) => message.contains(s), |
1081 | Pattern::Regex(r: &Regex) => r.is_match(haystack:message.as_bytes()), |
1082 | } |
1083 | } |
1084 | } |
1085 | |
1086 | impl<CommentsType> CommentParser<CommentsType> { |
1087 | fn parse_error_pattern(&mut self, pattern: Spanned<&str>) -> Spanned<Pattern> { |
1088 | if let Some(regex: Spanned<&str>) = pattern.strip_prefix("/" ) { |
1089 | match regex.strip_suffix("/" ) { |
1090 | Some(regex: Spanned<&str>) => match self.parse_regex(regex) { |
1091 | Some(r: Spanned) => r.map(Pattern::Regex), |
1092 | None => pattern.map(|p: &str| Pattern::SubString(p.to_string())), |
1093 | }, |
1094 | None => { |
1095 | self.error( |
1096 | regex.span(), |
1097 | s:"expected regex pattern due to leading `/`, but found no closing `/`" , |
1098 | ); |
1099 | pattern.map(|p: &str| Pattern::SubString(p.to_string())) |
1100 | } |
1101 | } |
1102 | } else { |
1103 | pattern.map(|p: &str| Pattern::SubString(p.to_string())) |
1104 | } |
1105 | } |
1106 | } |
1107 | |