Show More
@@ -1,1026 +1,1055 b'' | |||
|
1 | 1 | use crate::dirstate::entry::TruncatedTimestamp; |
|
2 | 2 | use crate::dirstate::status::IgnoreFnType; |
|
3 | 3 | use crate::dirstate::status::StatusPath; |
|
4 | 4 | use crate::dirstate_tree::dirstate_map::BorrowedPath; |
|
5 | 5 | use crate::dirstate_tree::dirstate_map::ChildNodesRef; |
|
6 | 6 | use crate::dirstate_tree::dirstate_map::DirstateMap; |
|
7 | 7 | use crate::dirstate_tree::dirstate_map::DirstateVersion; |
|
8 | 8 | use crate::dirstate_tree::dirstate_map::NodeRef; |
|
9 | 9 | use crate::dirstate_tree::on_disk::DirstateV2ParseError; |
|
10 | 10 | use crate::matchers::get_ignore_function; |
|
11 | use crate::matchers::Matcher; | |
|
11 | use crate::matchers::{Matcher, VisitChildrenSet}; | |
|
12 | 12 | use crate::utils::files::get_bytes_from_os_string; |
|
13 | 13 | use crate::utils::files::get_bytes_from_path; |
|
14 | 14 | use crate::utils::files::get_path_from_bytes; |
|
15 | 15 | use crate::utils::hg_path::hg_path_to_path_buf; |
|
16 | 16 | use crate::utils::hg_path::HgPath; |
|
17 | 17 | use crate::BadMatch; |
|
18 | 18 | use crate::BadType; |
|
19 | 19 | use crate::DirstateStatus; |
|
20 | 20 | use crate::HgPathCow; |
|
21 | 21 | use crate::PatternFileWarning; |
|
22 | 22 | use crate::StatusError; |
|
23 | 23 | use crate::StatusOptions; |
|
24 | 24 | use once_cell::sync::OnceCell; |
|
25 | 25 | use rayon::prelude::*; |
|
26 | 26 | use sha1::{Digest, Sha1}; |
|
27 | 27 | use std::borrow::Cow; |
|
28 | 28 | use std::io; |
|
29 | 29 | use std::os::unix::prelude::FileTypeExt; |
|
30 | 30 | use std::path::Path; |
|
31 | 31 | use std::path::PathBuf; |
|
32 | 32 | use std::sync::Mutex; |
|
33 | 33 | use std::time::SystemTime; |
|
34 | 34 | |
|
35 | 35 | /// Returns the status of the working directory compared to its parent |
|
36 | 36 | /// changeset. |
|
37 | 37 | /// |
|
38 | 38 | /// This algorithm is based on traversing the filesystem tree (`fs` in function |
|
39 | 39 | /// and variable names) and dirstate tree at the same time. The core of this |
|
40 | 40 | /// traversal is the recursive `traverse_fs_directory_and_dirstate` function |
|
41 | 41 | /// and its use of `itertools::merge_join_by`. When reaching a path that only |
|
42 | 42 | /// exists in one of the two trees, depending on information requested by |
|
43 | 43 | /// `options` we may need to traverse the remaining subtree. |
|
44 | 44 | #[logging_timer::time("trace")] |
|
45 | 45 | pub fn status<'dirstate>( |
|
46 | 46 | dmap: &'dirstate mut DirstateMap, |
|
47 | 47 | matcher: &(dyn Matcher + Sync), |
|
48 | 48 | root_dir: PathBuf, |
|
49 | 49 | ignore_files: Vec<PathBuf>, |
|
50 | 50 | options: StatusOptions, |
|
51 | 51 | ) -> Result<(DirstateStatus<'dirstate>, Vec<PatternFileWarning>), StatusError> |
|
52 | 52 | { |
|
53 | 53 | // Also cap for a Python caller of this function, but don't complain if |
|
54 | 54 | // the global threadpool has already been set since this code path is also |
|
55 | 55 | // being used by `rhg`, which calls this early. |
|
56 | 56 | let _ = crate::utils::cap_default_rayon_threads(); |
|
57 | 57 | |
|
58 | 58 | let (ignore_fn, warnings, patterns_changed): (IgnoreFnType, _, _) = |
|
59 | 59 | if options.list_ignored || options.list_unknown { |
|
60 | 60 | let (ignore_fn, warnings, changed) = match dmap.dirstate_version { |
|
61 | 61 | DirstateVersion::V1 => { |
|
62 | 62 | let (ignore_fn, warnings) = get_ignore_function( |
|
63 | 63 | ignore_files, |
|
64 | 64 | &root_dir, |
|
65 | 65 | &mut |_source, _pattern_bytes| {}, |
|
66 | 66 | )?; |
|
67 | 67 | (ignore_fn, warnings, None) |
|
68 | 68 | } |
|
69 | 69 | DirstateVersion::V2 => { |
|
70 | 70 | let mut hasher = Sha1::new(); |
|
71 | 71 | let (ignore_fn, warnings) = get_ignore_function( |
|
72 | 72 | ignore_files, |
|
73 | 73 | &root_dir, |
|
74 | 74 | &mut |source, pattern_bytes| { |
|
75 | 75 | // If inside the repo, use the relative version to |
|
76 | 76 | // make it deterministic inside tests. |
|
77 | 77 | // The performance hit should be negligible. |
|
78 | 78 | let source = source |
|
79 | 79 | .strip_prefix(&root_dir) |
|
80 | 80 | .unwrap_or(source); |
|
81 | 81 | let source = get_bytes_from_path(source); |
|
82 | 82 | |
|
83 | 83 | let mut subhasher = Sha1::new(); |
|
84 | 84 | subhasher.update(pattern_bytes); |
|
85 | 85 | let patterns_hash = subhasher.finalize(); |
|
86 | 86 | |
|
87 | 87 | hasher.update(source); |
|
88 | 88 | hasher.update(b" "); |
|
89 | 89 | hasher.update(patterns_hash); |
|
90 | 90 | hasher.update(b"\n"); |
|
91 | 91 | }, |
|
92 | 92 | )?; |
|
93 | 93 | let new_hash = *hasher.finalize().as_ref(); |
|
94 | 94 | let changed = new_hash != dmap.ignore_patterns_hash; |
|
95 | 95 | dmap.ignore_patterns_hash = new_hash; |
|
96 | 96 | (ignore_fn, warnings, Some(changed)) |
|
97 | 97 | } |
|
98 | 98 | }; |
|
99 | 99 | (ignore_fn, warnings, changed) |
|
100 | 100 | } else { |
|
101 | 101 | (Box::new(|&_| true), vec![], None) |
|
102 | 102 | }; |
|
103 | 103 | |
|
104 | 104 | let filesystem_time_at_status_start = |
|
105 | 105 | filesystem_now(&root_dir).ok().map(TruncatedTimestamp::from); |
|
106 | 106 | |
|
107 | 107 | // If the repository is under the current directory, prefer using a |
|
108 | 108 | // relative path, so the kernel needs to traverse fewer directory in every |
|
109 | 109 | // call to `read_dir` or `symlink_metadata`. |
|
110 | 110 | // This is effective in the common case where the current directory is the |
|
111 | 111 | // repository root. |
|
112 | 112 | |
|
113 | 113 | // TODO: Better yet would be to use libc functions like `openat` and |
|
114 | 114 | // `fstatat` to remove such repeated traversals entirely, but the standard |
|
115 | 115 | // library does not provide APIs based on those. |
|
116 | 116 | // Maybe with a crate like https://crates.io/crates/openat instead? |
|
117 | 117 | let root_dir = if let Some(relative) = std::env::current_dir() |
|
118 | 118 | .ok() |
|
119 | 119 | .and_then(|cwd| root_dir.strip_prefix(cwd).ok()) |
|
120 | 120 | { |
|
121 | 121 | relative |
|
122 | 122 | } else { |
|
123 | 123 | &root_dir |
|
124 | 124 | }; |
|
125 | 125 | |
|
126 | 126 | let outcome = DirstateStatus { |
|
127 | 127 | filesystem_time_at_status_start, |
|
128 | 128 | ..Default::default() |
|
129 | 129 | }; |
|
130 | 130 | let common = StatusCommon { |
|
131 | 131 | dmap, |
|
132 | 132 | options, |
|
133 | 133 | matcher, |
|
134 | 134 | ignore_fn, |
|
135 | 135 | outcome: Mutex::new(outcome), |
|
136 | 136 | ignore_patterns_have_changed: patterns_changed, |
|
137 | 137 | new_cacheable_directories: Default::default(), |
|
138 | 138 | outdated_cached_directories: Default::default(), |
|
139 | 139 | filesystem_time_at_status_start, |
|
140 | 140 | }; |
|
141 | 141 | let is_at_repo_root = true; |
|
142 | 142 | let hg_path = &BorrowedPath::OnDisk(HgPath::new("")); |
|
143 | 143 | let has_ignored_ancestor = HasIgnoredAncestor::create(None, hg_path); |
|
144 | 144 | let root_cached_mtime = None; |
|
145 | 145 | // If the path we have for the repository root is a symlink, do follow it. |
|
146 | 146 | // (As opposed to symlinks within the working directory which are not |
|
147 | 147 | // followed, using `std::fs::symlink_metadata`.) |
|
148 | 148 | common.traverse_fs_directory_and_dirstate( |
|
149 | 149 | &has_ignored_ancestor, |
|
150 | 150 | dmap.root.as_ref(), |
|
151 | 151 | hg_path, |
|
152 | 152 | &DirEntry { |
|
153 | 153 | hg_path: Cow::Borrowed(HgPath::new(b"")), |
|
154 | 154 | fs_path: Cow::Borrowed(root_dir), |
|
155 | 155 | symlink_metadata: None, |
|
156 | 156 | file_type: FakeFileType::Directory, |
|
157 | 157 | }, |
|
158 | 158 | root_cached_mtime, |
|
159 | 159 | is_at_repo_root, |
|
160 | 160 | )?; |
|
161 | 161 | if let Some(file_set) = common.matcher.file_set() { |
|
162 | 162 | for file in file_set { |
|
163 | 163 | if !file.is_empty() && !dmap.has_node(file)? { |
|
164 | 164 | let path = hg_path_to_path_buf(file)?; |
|
165 | 165 | if let io::Result::Err(error) = |
|
166 | 166 | root_dir.join(path).symlink_metadata() |
|
167 | 167 | { |
|
168 | 168 | common.io_error(error, file) |
|
169 | 169 | } |
|
170 | 170 | } |
|
171 | 171 | } |
|
172 | 172 | } |
|
173 | 173 | let mut outcome = common.outcome.into_inner().unwrap(); |
|
174 | 174 | let new_cacheable = common.new_cacheable_directories.into_inner().unwrap(); |
|
175 | 175 | let outdated = common.outdated_cached_directories.into_inner().unwrap(); |
|
176 | 176 | |
|
177 | 177 | outcome.dirty = common.ignore_patterns_have_changed == Some(true) |
|
178 | 178 | || !outdated.is_empty() |
|
179 | 179 | || (!new_cacheable.is_empty() |
|
180 | 180 | && dmap.dirstate_version == DirstateVersion::V2); |
|
181 | 181 | |
|
182 | 182 | // Remove outdated mtimes before adding new mtimes, in case a given |
|
183 | 183 | // directory is both |
|
184 | 184 | for path in &outdated { |
|
185 | 185 | dmap.clear_cached_mtime(path)?; |
|
186 | 186 | } |
|
187 | 187 | for (path, mtime) in &new_cacheable { |
|
188 | 188 | dmap.set_cached_mtime(path, *mtime)?; |
|
189 | 189 | } |
|
190 | 190 | |
|
191 | 191 | Ok((outcome, warnings)) |
|
192 | 192 | } |
|
193 | 193 | |
|
194 | 194 | /// Bag of random things needed by various parts of the algorithm. Reduces the |
|
195 | 195 | /// number of parameters passed to functions. |
|
196 | 196 | struct StatusCommon<'a, 'tree, 'on_disk: 'tree> { |
|
197 | 197 | dmap: &'tree DirstateMap<'on_disk>, |
|
198 | 198 | options: StatusOptions, |
|
199 | 199 | matcher: &'a (dyn Matcher + Sync), |
|
200 | 200 | ignore_fn: IgnoreFnType<'a>, |
|
201 | 201 | outcome: Mutex<DirstateStatus<'on_disk>>, |
|
202 | 202 | /// New timestamps of directories to be used for caching their readdirs |
|
203 | 203 | new_cacheable_directories: |
|
204 | 204 | Mutex<Vec<(Cow<'on_disk, HgPath>, TruncatedTimestamp)>>, |
|
205 | 205 | /// Used to invalidate the readdir cache of directories |
|
206 | 206 | outdated_cached_directories: Mutex<Vec<Cow<'on_disk, HgPath>>>, |
|
207 | 207 | |
|
208 | 208 | /// Whether ignore files like `.hgignore` have changed since the previous |
|
209 | 209 | /// time a `status()` call wrote their hash to the dirstate. `None` means |
|
210 | 210 | /// we don’t know as this run doesn’t list either ignored or uknown files |
|
211 | 211 | /// and therefore isn’t reading `.hgignore`. |
|
212 | 212 | ignore_patterns_have_changed: Option<bool>, |
|
213 | 213 | |
|
214 | 214 | /// The current time at the start of the `status()` algorithm, as measured |
|
215 | 215 | /// and possibly truncated by the filesystem. |
|
216 | 216 | filesystem_time_at_status_start: Option<TruncatedTimestamp>, |
|
217 | 217 | } |
|
218 | 218 | |
|
219 | 219 | enum Outcome { |
|
220 | 220 | Modified, |
|
221 | 221 | Added, |
|
222 | 222 | Removed, |
|
223 | 223 | Deleted, |
|
224 | 224 | Clean, |
|
225 | 225 | Ignored, |
|
226 | 226 | Unknown, |
|
227 | 227 | Unsure, |
|
228 | 228 | } |
|
229 | 229 | |
|
230 | 230 | /// Lazy computation of whether a given path has a hgignored |
|
231 | 231 | /// ancestor. |
|
232 | 232 | struct HasIgnoredAncestor<'a> { |
|
233 | 233 | /// `path` and `parent` constitute the inputs to the computation, |
|
234 | 234 | /// `cache` stores the outcome. |
|
235 | 235 | path: &'a HgPath, |
|
236 | 236 | parent: Option<&'a HasIgnoredAncestor<'a>>, |
|
237 | 237 | cache: OnceCell<bool>, |
|
238 | 238 | } |
|
239 | 239 | |
|
240 | 240 | impl<'a> HasIgnoredAncestor<'a> { |
|
241 | 241 | fn create( |
|
242 | 242 | parent: Option<&'a HasIgnoredAncestor<'a>>, |
|
243 | 243 | path: &'a HgPath, |
|
244 | 244 | ) -> HasIgnoredAncestor<'a> { |
|
245 | 245 | Self { |
|
246 | 246 | path, |
|
247 | 247 | parent, |
|
248 | 248 | cache: OnceCell::new(), |
|
249 | 249 | } |
|
250 | 250 | } |
|
251 | 251 | |
|
252 | 252 | fn force<'b>(&self, ignore_fn: &IgnoreFnType<'b>) -> bool { |
|
253 | 253 | match self.parent { |
|
254 | 254 | None => false, |
|
255 | 255 | Some(parent) => { |
|
256 | 256 | *(self.cache.get_or_init(|| { |
|
257 | 257 | parent.force(ignore_fn) || ignore_fn(self.path) |
|
258 | 258 | })) |
|
259 | 259 | } |
|
260 | 260 | } |
|
261 | 261 | } |
|
262 | 262 | } |
|
263 | 263 | |
|
264 | 264 | impl<'a, 'tree, 'on_disk> StatusCommon<'a, 'tree, 'on_disk> { |
|
265 | 265 | fn push_outcome( |
|
266 | 266 | &self, |
|
267 | 267 | which: Outcome, |
|
268 | 268 | dirstate_node: &NodeRef<'tree, 'on_disk>, |
|
269 | 269 | ) -> Result<(), DirstateV2ParseError> { |
|
270 | 270 | let path = dirstate_node |
|
271 | 271 | .full_path_borrowed(self.dmap.on_disk)? |
|
272 | 272 | .detach_from_tree(); |
|
273 | 273 | let copy_source = if self.options.list_copies { |
|
274 | 274 | dirstate_node |
|
275 | 275 | .copy_source_borrowed(self.dmap.on_disk)? |
|
276 | 276 | .map(|source| source.detach_from_tree()) |
|
277 | 277 | } else { |
|
278 | 278 | None |
|
279 | 279 | }; |
|
280 | 280 | self.push_outcome_common(which, path, copy_source); |
|
281 | 281 | Ok(()) |
|
282 | 282 | } |
|
283 | 283 | |
|
284 | 284 | fn push_outcome_without_copy_source( |
|
285 | 285 | &self, |
|
286 | 286 | which: Outcome, |
|
287 | 287 | path: &BorrowedPath<'_, 'on_disk>, |
|
288 | 288 | ) { |
|
289 | 289 | self.push_outcome_common(which, path.detach_from_tree(), None) |
|
290 | 290 | } |
|
291 | 291 | |
|
292 | 292 | fn push_outcome_common( |
|
293 | 293 | &self, |
|
294 | 294 | which: Outcome, |
|
295 | 295 | path: HgPathCow<'on_disk>, |
|
296 | 296 | copy_source: Option<HgPathCow<'on_disk>>, |
|
297 | 297 | ) { |
|
298 | 298 | let mut outcome = self.outcome.lock().unwrap(); |
|
299 | 299 | let vec = match which { |
|
300 | 300 | Outcome::Modified => &mut outcome.modified, |
|
301 | 301 | Outcome::Added => &mut outcome.added, |
|
302 | 302 | Outcome::Removed => &mut outcome.removed, |
|
303 | 303 | Outcome::Deleted => &mut outcome.deleted, |
|
304 | 304 | Outcome::Clean => &mut outcome.clean, |
|
305 | 305 | Outcome::Ignored => &mut outcome.ignored, |
|
306 | 306 | Outcome::Unknown => &mut outcome.unknown, |
|
307 | 307 | Outcome::Unsure => &mut outcome.unsure, |
|
308 | 308 | }; |
|
309 | 309 | vec.push(StatusPath { path, copy_source }); |
|
310 | 310 | } |
|
311 | 311 | |
|
312 | 312 | fn read_dir( |
|
313 | 313 | &self, |
|
314 | 314 | hg_path: &HgPath, |
|
315 | 315 | fs_path: &Path, |
|
316 | 316 | is_at_repo_root: bool, |
|
317 | 317 | ) -> Result<Vec<DirEntry>, ()> { |
|
318 | 318 | DirEntry::read_dir(fs_path, is_at_repo_root) |
|
319 | 319 | .map_err(|error| self.io_error(error, hg_path)) |
|
320 | 320 | } |
|
321 | 321 | |
|
322 | 322 | fn io_error(&self, error: std::io::Error, hg_path: &HgPath) { |
|
323 | 323 | let errno = error.raw_os_error().expect("expected real OS error"); |
|
324 | 324 | self.outcome |
|
325 | 325 | .lock() |
|
326 | 326 | .unwrap() |
|
327 | 327 | .bad |
|
328 | 328 | .push((hg_path.to_owned().into(), BadMatch::OsError(errno))) |
|
329 | 329 | } |
|
330 | 330 | |
|
331 | 331 | fn check_for_outdated_directory_cache( |
|
332 | 332 | &self, |
|
333 | 333 | dirstate_node: &NodeRef<'tree, 'on_disk>, |
|
334 | 334 | ) -> Result<bool, DirstateV2ParseError> { |
|
335 | 335 | if self.ignore_patterns_have_changed == Some(true) |
|
336 | 336 | && dirstate_node.cached_directory_mtime()?.is_some() |
|
337 | 337 | { |
|
338 | 338 | self.outdated_cached_directories.lock().unwrap().push( |
|
339 | 339 | dirstate_node |
|
340 | 340 | .full_path_borrowed(self.dmap.on_disk)? |
|
341 | 341 | .detach_from_tree(), |
|
342 | 342 | ); |
|
343 | 343 | return Ok(true); |
|
344 | 344 | } |
|
345 | 345 | Ok(false) |
|
346 | 346 | } |
|
347 | 347 | |
|
348 | 348 | /// If this returns true, we can get accurate results by only using |
|
349 | 349 | /// `symlink_metadata` for child nodes that exist in the dirstate and don’t |
|
350 | 350 | /// need to call `read_dir`. |
|
351 | 351 | fn can_skip_fs_readdir( |
|
352 | 352 | &self, |
|
353 | 353 | directory_entry: &DirEntry, |
|
354 | 354 | cached_directory_mtime: Option<TruncatedTimestamp>, |
|
355 | 355 | ) -> bool { |
|
356 | 356 | if !self.options.list_unknown && !self.options.list_ignored { |
|
357 | 357 | // All states that we care about listing have corresponding |
|
358 | 358 | // dirstate entries. |
|
359 | 359 | // This happens for example with `hg status -mard`. |
|
360 | 360 | return true; |
|
361 | 361 | } |
|
362 | 362 | if !self.options.list_ignored |
|
363 | 363 | && self.ignore_patterns_have_changed == Some(false) |
|
364 | 364 | { |
|
365 | 365 | if let Some(cached_mtime) = cached_directory_mtime { |
|
366 | 366 | // The dirstate contains a cached mtime for this directory, set |
|
367 | 367 | // by a previous run of the `status` algorithm which found this |
|
368 | 368 | // directory eligible for `read_dir` caching. |
|
369 | 369 | if let Ok(meta) = directory_entry.symlink_metadata() { |
|
370 | 370 | if cached_mtime |
|
371 | 371 | .likely_equal_to_mtime_of(&meta) |
|
372 | 372 | .unwrap_or(false) |
|
373 | 373 | { |
|
374 | 374 | // The mtime of that directory has not changed |
|
375 | 375 | // since then, which means that the results of |
|
376 | 376 | // `read_dir` should also be unchanged. |
|
377 | 377 | return true; |
|
378 | 378 | } |
|
379 | 379 | } |
|
380 | 380 | } |
|
381 | 381 | } |
|
382 | 382 | false |
|
383 | 383 | } |
|
384 | 384 | |
|
385 | fn should_visit(set: &VisitChildrenSet, basename: &HgPath) -> bool { | |
|
386 | match set { | |
|
387 | VisitChildrenSet::This | VisitChildrenSet::Recursive => true, | |
|
388 | VisitChildrenSet::Empty => false, | |
|
389 | VisitChildrenSet::Set(children_to_visit) => { | |
|
390 | children_to_visit.contains(basename) | |
|
391 | } | |
|
392 | } | |
|
393 | } | |
|
394 | ||
|
385 | 395 | /// Returns whether all child entries of the filesystem directory have a |
|
386 | 396 | /// corresponding dirstate node or are ignored. |
|
387 | 397 | fn traverse_fs_directory_and_dirstate<'ancestor>( |
|
388 | 398 | &self, |
|
389 | 399 | has_ignored_ancestor: &'ancestor HasIgnoredAncestor<'ancestor>, |
|
390 | 400 | dirstate_nodes: ChildNodesRef<'tree, 'on_disk>, |
|
391 | 401 | directory_hg_path: &BorrowedPath<'tree, 'on_disk>, |
|
392 | 402 | directory_entry: &DirEntry, |
|
393 | 403 | cached_directory_mtime: Option<TruncatedTimestamp>, |
|
394 | 404 | is_at_repo_root: bool, |
|
395 | 405 | ) -> Result<bool, DirstateV2ParseError> { |
|
406 | let children_set = self.matcher.visit_children_set(directory_hg_path); | |
|
407 | if let VisitChildrenSet::Empty = children_set { | |
|
408 | return Ok(false); | |
|
409 | } | |
|
396 | 410 | if self.can_skip_fs_readdir(directory_entry, cached_directory_mtime) { |
|
397 | 411 | dirstate_nodes |
|
398 | 412 | .par_iter() |
|
399 | 413 | .map(|dirstate_node| { |
|
400 | 414 | let fs_path = &directory_entry.fs_path; |
|
401 | let fs_path = fs_path.join(get_path_from_bytes( | |
|
402 |
dirstate_node.base_name(self.dmap.on_disk)?.as_bytes() |
|
|
403 | )); | |
|
415 | let basename = | |
|
416 | dirstate_node.base_name(self.dmap.on_disk)?.as_bytes(); | |
|
417 | let fs_path = fs_path.join(get_path_from_bytes(basename)); | |
|
418 | if !Self::should_visit( | |
|
419 | &children_set, | |
|
420 | HgPath::new(basename), | |
|
421 | ) { | |
|
422 | return Ok(()); | |
|
423 | } | |
|
404 | 424 | match std::fs::symlink_metadata(&fs_path) { |
|
405 | 425 | Ok(fs_metadata) => { |
|
406 | 426 | let file_type = fs_metadata.file_type().into(); |
|
407 | 427 | let entry = DirEntry { |
|
408 | 428 | hg_path: Cow::Borrowed( |
|
409 | 429 | dirstate_node |
|
410 | 430 | .full_path(self.dmap.on_disk)?, |
|
411 | 431 | ), |
|
412 | 432 | fs_path: Cow::Borrowed(&fs_path), |
|
413 | 433 | symlink_metadata: Some(fs_metadata), |
|
414 | 434 | file_type, |
|
415 | 435 | }; |
|
416 | 436 | self.traverse_fs_and_dirstate( |
|
417 | 437 | &entry, |
|
418 | 438 | dirstate_node, |
|
419 | 439 | has_ignored_ancestor, |
|
420 | 440 | ) |
|
421 | 441 | } |
|
422 | 442 | Err(e) if e.kind() == std::io::ErrorKind::NotFound => { |
|
423 | 443 | self.traverse_dirstate_only(dirstate_node) |
|
424 | 444 | } |
|
425 | 445 | Err(error) => { |
|
426 | 446 | let hg_path = |
|
427 | 447 | dirstate_node.full_path(self.dmap.on_disk)?; |
|
428 | 448 | self.io_error(error, hg_path); |
|
429 | 449 | Ok(()) |
|
430 | 450 | } |
|
431 | 451 | } |
|
432 | 452 | }) |
|
433 | 453 | .collect::<Result<_, _>>()?; |
|
434 | 454 | |
|
435 | 455 | // We don’t know, so conservatively say this isn’t the case |
|
436 | 456 | let children_all_have_dirstate_node_or_are_ignored = false; |
|
437 | 457 | |
|
438 | 458 | return Ok(children_all_have_dirstate_node_or_are_ignored); |
|
439 | 459 | } |
|
440 | 460 | |
|
441 | 461 | let readdir_succeeded; |
|
442 | 462 | let mut fs_entries = if let Ok(entries) = self.read_dir( |
|
443 | 463 | directory_hg_path, |
|
444 | 464 | &directory_entry.fs_path, |
|
445 | 465 | is_at_repo_root, |
|
446 | 466 | ) { |
|
447 | 467 | readdir_succeeded = true; |
|
448 | 468 | entries |
|
449 | 469 | } else { |
|
450 | 470 | // Treat an unreadable directory (typically because of insufficient |
|
451 | 471 | // permissions) like an empty directory. `self.read_dir` has |
|
452 | 472 | // already called `self.io_error` so a warning will be emitted. |
|
453 | 473 | // We still need to remember that there was an error so that we |
|
454 | 474 | // know not to cache this result. |
|
455 | 475 | readdir_succeeded = false; |
|
456 | 476 | Vec::new() |
|
457 | 477 | }; |
|
458 | 478 | |
|
459 | 479 | // `merge_join_by` requires both its input iterators to be sorted: |
|
460 | 480 | |
|
461 | 481 | let dirstate_nodes = dirstate_nodes.sorted(); |
|
462 | 482 | // `sort_unstable_by_key` doesn’t allow keys borrowing from the value: |
|
463 | 483 | // https://github.com/rust-lang/rust/issues/34162 |
|
464 | 484 | fs_entries.sort_unstable_by(|e1, e2| e1.hg_path.cmp(&e2.hg_path)); |
|
465 | 485 | |
|
466 | 486 | // Propagate here any error that would happen inside the comparison |
|
467 | 487 | // callback below |
|
468 | 488 | for dirstate_node in &dirstate_nodes { |
|
469 | 489 | dirstate_node.base_name(self.dmap.on_disk)?; |
|
470 | 490 | } |
|
471 | 491 | itertools::merge_join_by( |
|
472 | 492 | dirstate_nodes, |
|
473 | 493 | &fs_entries, |
|
474 | 494 | |dirstate_node, fs_entry| { |
|
475 | 495 | // This `unwrap` never panics because we already propagated |
|
476 | 496 | // those errors above |
|
477 | 497 | dirstate_node |
|
478 | 498 | .base_name(self.dmap.on_disk) |
|
479 | 499 | .unwrap() |
|
480 | 500 | .cmp(&fs_entry.hg_path) |
|
481 | 501 | }, |
|
482 | 502 | ) |
|
483 | 503 | .par_bridge() |
|
484 | 504 | .map(|pair| { |
|
485 | 505 | use itertools::EitherOrBoth::*; |
|
506 | let basename = match &pair { | |
|
507 | Left(dirstate_node) | Both(dirstate_node, _) => HgPath::new( | |
|
508 | dirstate_node.base_name(self.dmap.on_disk)?.as_bytes(), | |
|
509 | ), | |
|
510 | Right(fs_entry) => &fs_entry.hg_path, | |
|
511 | }; | |
|
512 | if !Self::should_visit(&children_set, basename) { | |
|
513 | return Ok(false); | |
|
514 | } | |
|
486 | 515 | let has_dirstate_node_or_is_ignored = match pair { |
|
487 | 516 | Both(dirstate_node, fs_entry) => { |
|
488 | 517 | self.traverse_fs_and_dirstate( |
|
489 | 518 | fs_entry, |
|
490 | 519 | dirstate_node, |
|
491 | 520 | has_ignored_ancestor, |
|
492 | 521 | )?; |
|
493 | 522 | true |
|
494 | 523 | } |
|
495 | 524 | Left(dirstate_node) => { |
|
496 | 525 | self.traverse_dirstate_only(dirstate_node)?; |
|
497 | 526 | true |
|
498 | 527 | } |
|
499 | 528 | Right(fs_entry) => self.traverse_fs_only( |
|
500 | 529 | has_ignored_ancestor.force(&self.ignore_fn), |
|
501 | 530 | directory_hg_path, |
|
502 | 531 | fs_entry, |
|
503 | 532 | ), |
|
504 | 533 | }; |
|
505 | 534 | Ok(has_dirstate_node_or_is_ignored) |
|
506 | 535 | }) |
|
507 | 536 | .try_reduce(|| true, |a, b| Ok(a && b)) |
|
508 | 537 | .map(|res| res && readdir_succeeded) |
|
509 | 538 | } |
|
510 | 539 | |
|
511 | 540 | fn traverse_fs_and_dirstate<'ancestor>( |
|
512 | 541 | &self, |
|
513 | 542 | fs_entry: &DirEntry, |
|
514 | 543 | dirstate_node: NodeRef<'tree, 'on_disk>, |
|
515 | 544 | has_ignored_ancestor: &'ancestor HasIgnoredAncestor<'ancestor>, |
|
516 | 545 | ) -> Result<(), DirstateV2ParseError> { |
|
517 | 546 | let outdated_dircache = |
|
518 | 547 | self.check_for_outdated_directory_cache(&dirstate_node)?; |
|
519 | 548 | let hg_path = &dirstate_node.full_path_borrowed(self.dmap.on_disk)?; |
|
520 | 549 | let file_or_symlink = fs_entry.is_file() || fs_entry.is_symlink(); |
|
521 | 550 | if !file_or_symlink { |
|
522 | 551 | // If we previously had a file here, it was removed (with |
|
523 | 552 | // `hg rm` or similar) or deleted before it could be |
|
524 | 553 | // replaced by a directory or something else. |
|
525 | 554 | self.mark_removed_or_deleted_if_file(&dirstate_node)?; |
|
526 | 555 | } |
|
527 | 556 | if let Some(bad_type) = fs_entry.is_bad() { |
|
528 | 557 | if self.matcher.exact_match(hg_path) { |
|
529 | 558 | let path = dirstate_node.full_path(self.dmap.on_disk)?; |
|
530 | 559 | self.outcome.lock().unwrap().bad.push(( |
|
531 | 560 | path.to_owned().into(), |
|
532 | 561 | BadMatch::BadType(bad_type), |
|
533 | 562 | )) |
|
534 | 563 | } |
|
535 | 564 | } |
|
536 | 565 | if fs_entry.is_dir() { |
|
537 | 566 | if self.options.collect_traversed_dirs { |
|
538 | 567 | self.outcome |
|
539 | 568 | .lock() |
|
540 | 569 | .unwrap() |
|
541 | 570 | .traversed |
|
542 | 571 | .push(hg_path.detach_from_tree()) |
|
543 | 572 | } |
|
544 | 573 | let is_ignored = HasIgnoredAncestor::create( |
|
545 | 574 | Some(has_ignored_ancestor), |
|
546 | 575 | hg_path, |
|
547 | 576 | ); |
|
548 | 577 | let is_at_repo_root = false; |
|
549 | 578 | let children_all_have_dirstate_node_or_are_ignored = self |
|
550 | 579 | .traverse_fs_directory_and_dirstate( |
|
551 | 580 | &is_ignored, |
|
552 | 581 | dirstate_node.children(self.dmap.on_disk)?, |
|
553 | 582 | hg_path, |
|
554 | 583 | fs_entry, |
|
555 | 584 | dirstate_node.cached_directory_mtime()?, |
|
556 | 585 | is_at_repo_root, |
|
557 | 586 | )?; |
|
558 | 587 | self.maybe_save_directory_mtime( |
|
559 | 588 | children_all_have_dirstate_node_or_are_ignored, |
|
560 | 589 | fs_entry, |
|
561 | 590 | dirstate_node, |
|
562 | 591 | outdated_dircache, |
|
563 | 592 | )? |
|
564 | 593 | } else { |
|
565 | 594 | if file_or_symlink && self.matcher.matches(hg_path) { |
|
566 | 595 | if let Some(entry) = dirstate_node.entry()? { |
|
567 | 596 | if !entry.any_tracked() { |
|
568 | 597 | // Forward-compat if we start tracking unknown/ignored |
|
569 | 598 | // files for caching reasons |
|
570 | 599 | self.mark_unknown_or_ignored( |
|
571 | 600 | has_ignored_ancestor.force(&self.ignore_fn), |
|
572 | 601 | hg_path, |
|
573 | 602 | ); |
|
574 | 603 | } |
|
575 | 604 | if entry.added() { |
|
576 | 605 | self.push_outcome(Outcome::Added, &dirstate_node)?; |
|
577 | 606 | } else if entry.removed() { |
|
578 | 607 | self.push_outcome(Outcome::Removed, &dirstate_node)?; |
|
579 | 608 | } else if entry.modified() { |
|
580 | 609 | self.push_outcome(Outcome::Modified, &dirstate_node)?; |
|
581 | 610 | } else { |
|
582 | 611 | self.handle_normal_file(&dirstate_node, fs_entry)?; |
|
583 | 612 | } |
|
584 | 613 | } else { |
|
585 | 614 | // `node.entry.is_none()` indicates a "directory" |
|
586 | 615 | // node, but the filesystem has a file |
|
587 | 616 | self.mark_unknown_or_ignored( |
|
588 | 617 | has_ignored_ancestor.force(&self.ignore_fn), |
|
589 | 618 | hg_path, |
|
590 | 619 | ); |
|
591 | 620 | } |
|
592 | 621 | } |
|
593 | 622 | |
|
594 | 623 | for child_node in dirstate_node.children(self.dmap.on_disk)?.iter() |
|
595 | 624 | { |
|
596 | 625 | self.traverse_dirstate_only(child_node)? |
|
597 | 626 | } |
|
598 | 627 | } |
|
599 | 628 | Ok(()) |
|
600 | 629 | } |
|
601 | 630 | |
|
602 | 631 | /// Save directory mtime if applicable. |
|
603 | 632 | /// |
|
604 | 633 | /// `outdated_directory_cache` is `true` if we've just invalidated the |
|
605 | 634 | /// cache for this directory in `check_for_outdated_directory_cache`, |
|
606 | 635 | /// which forces the update. |
|
607 | 636 | fn maybe_save_directory_mtime( |
|
608 | 637 | &self, |
|
609 | 638 | children_all_have_dirstate_node_or_are_ignored: bool, |
|
610 | 639 | directory_entry: &DirEntry, |
|
611 | 640 | dirstate_node: NodeRef<'tree, 'on_disk>, |
|
612 | 641 | outdated_directory_cache: bool, |
|
613 | 642 | ) -> Result<(), DirstateV2ParseError> { |
|
614 | 643 | if !children_all_have_dirstate_node_or_are_ignored { |
|
615 | 644 | return Ok(()); |
|
616 | 645 | } |
|
617 | 646 | // All filesystem directory entries from `read_dir` have a |
|
618 | 647 | // corresponding node in the dirstate, so we can reconstitute the |
|
619 | 648 | // names of those entries without calling `read_dir` again. |
|
620 | 649 | |
|
621 | 650 | // TODO: use let-else here and below when available: |
|
622 | 651 | // https://github.com/rust-lang/rust/issues/87335 |
|
623 | 652 | let status_start = if let Some(status_start) = |
|
624 | 653 | &self.filesystem_time_at_status_start |
|
625 | 654 | { |
|
626 | 655 | status_start |
|
627 | 656 | } else { |
|
628 | 657 | return Ok(()); |
|
629 | 658 | }; |
|
630 | 659 | |
|
631 | 660 | // Although the Rust standard library’s `SystemTime` type |
|
632 | 661 | // has nanosecond precision, the times reported for a |
|
633 | 662 | // directory’s (or file’s) modified time may have lower |
|
634 | 663 | // resolution based on the filesystem (for example ext3 |
|
635 | 664 | // only stores integer seconds), kernel (see |
|
636 | 665 | // https://stackoverflow.com/a/14393315/1162888), etc. |
|
637 | 666 | let metadata = match directory_entry.symlink_metadata() { |
|
638 | 667 | Ok(meta) => meta, |
|
639 | 668 | Err(_) => return Ok(()), |
|
640 | 669 | }; |
|
641 | 670 | |
|
642 | 671 | let directory_mtime = match TruncatedTimestamp::for_reliable_mtime_of( |
|
643 | 672 | &metadata, |
|
644 | 673 | status_start, |
|
645 | 674 | ) { |
|
646 | 675 | Ok(Some(directory_mtime)) => directory_mtime, |
|
647 | 676 | Ok(None) => { |
|
648 | 677 | // The directory was modified too recently, |
|
649 | 678 | // don’t cache its `read_dir` results. |
|
650 | 679 | // |
|
651 | 680 | // 1. A change to this directory (direct child was |
|
652 | 681 | // added or removed) cause its mtime to be set |
|
653 | 682 | // (possibly truncated) to `directory_mtime` |
|
654 | 683 | // 2. This `status` algorithm calls `read_dir` |
|
655 | 684 | // 3. An other change is made to the same directory is |
|
656 | 685 | // made so that calling `read_dir` agin would give |
|
657 | 686 | // different results, but soon enough after 1. that |
|
658 | 687 | // the mtime stays the same |
|
659 | 688 | // |
|
660 | 689 | // On a system where the time resolution poor, this |
|
661 | 690 | // scenario is not unlikely if all three steps are caused |
|
662 | 691 | // by the same script. |
|
663 | 692 | return Ok(()); |
|
664 | 693 | } |
|
665 | 694 | Err(_) => { |
|
666 | 695 | // OS/libc does not support mtime? |
|
667 | 696 | return Ok(()); |
|
668 | 697 | } |
|
669 | 698 | }; |
|
670 | 699 | // We’ve observed (through `status_start`) that time has |
|
671 | 700 | // “progressed” since `directory_mtime`, so any further |
|
672 | 701 | // change to this directory is extremely likely to cause a |
|
673 | 702 | // different mtime. |
|
674 | 703 | // |
|
675 | 704 | // Having the same mtime again is not entirely impossible |
|
676 | 705 | // since the system clock is not monotonous. It could jump |
|
677 | 706 | // backward to some point before `directory_mtime`, then a |
|
678 | 707 | // directory change could potentially happen during exactly |
|
679 | 708 | // the wrong tick. |
|
680 | 709 | // |
|
681 | 710 | // We deem this scenario (unlike the previous one) to be |
|
682 | 711 | // unlikely enough in practice. |
|
683 | 712 | |
|
684 | 713 | let is_up_to_date = if let Some(cached) = |
|
685 | 714 | dirstate_node.cached_directory_mtime()? |
|
686 | 715 | { |
|
687 | 716 | !outdated_directory_cache && cached.likely_equal(directory_mtime) |
|
688 | 717 | } else { |
|
689 | 718 | false |
|
690 | 719 | }; |
|
691 | 720 | if !is_up_to_date { |
|
692 | 721 | let hg_path = dirstate_node |
|
693 | 722 | .full_path_borrowed(self.dmap.on_disk)? |
|
694 | 723 | .detach_from_tree(); |
|
695 | 724 | self.new_cacheable_directories |
|
696 | 725 | .lock() |
|
697 | 726 | .unwrap() |
|
698 | 727 | .push((hg_path, directory_mtime)) |
|
699 | 728 | } |
|
700 | 729 | Ok(()) |
|
701 | 730 | } |
|
702 | 731 | |
|
703 | 732 | /// A file that is clean in the dirstate was found in the filesystem |
|
704 | 733 | fn handle_normal_file( |
|
705 | 734 | &self, |
|
706 | 735 | dirstate_node: &NodeRef<'tree, 'on_disk>, |
|
707 | 736 | fs_entry: &DirEntry, |
|
708 | 737 | ) -> Result<(), DirstateV2ParseError> { |
|
709 | 738 | // Keep the low 31 bits |
|
710 | 739 | fn truncate_u64(value: u64) -> i32 { |
|
711 | 740 | (value & 0x7FFF_FFFF) as i32 |
|
712 | 741 | } |
|
713 | 742 | |
|
714 | 743 | let fs_metadata = match fs_entry.symlink_metadata() { |
|
715 | 744 | Ok(meta) => meta, |
|
716 | 745 | Err(_) => return Ok(()), |
|
717 | 746 | }; |
|
718 | 747 | |
|
719 | 748 | let entry = dirstate_node |
|
720 | 749 | .entry()? |
|
721 | 750 | .expect("handle_normal_file called with entry-less node"); |
|
722 | 751 | let mode_changed = |
|
723 | 752 | || self.options.check_exec && entry.mode_changed(&fs_metadata); |
|
724 | 753 | let size = entry.size(); |
|
725 | 754 | let size_changed = size != truncate_u64(fs_metadata.len()); |
|
726 | 755 | if size >= 0 && size_changed && fs_metadata.file_type().is_symlink() { |
|
727 | 756 | // issue6456: Size returned may be longer due to encryption |
|
728 | 757 | // on EXT-4 fscrypt. TODO maybe only do it on EXT4? |
|
729 | 758 | self.push_outcome(Outcome::Unsure, dirstate_node)? |
|
730 | 759 | } else if dirstate_node.has_copy_source() |
|
731 | 760 | || entry.is_from_other_parent() |
|
732 | 761 | || (size >= 0 && (size_changed || mode_changed())) |
|
733 | 762 | { |
|
734 | 763 | self.push_outcome(Outcome::Modified, dirstate_node)? |
|
735 | 764 | } else { |
|
736 | 765 | let mtime_looks_clean = if let Some(dirstate_mtime) = |
|
737 | 766 | entry.truncated_mtime() |
|
738 | 767 | { |
|
739 | 768 | let fs_mtime = TruncatedTimestamp::for_mtime_of(&fs_metadata) |
|
740 | 769 | .expect("OS/libc does not support mtime?"); |
|
741 | 770 | // There might be a change in the future if for example the |
|
742 | 771 | // internal clock become off while process run, but this is a |
|
743 | 772 | // case where the issues the user would face |
|
744 | 773 | // would be a lot worse and there is nothing we |
|
745 | 774 | // can really do. |
|
746 | 775 | fs_mtime.likely_equal(dirstate_mtime) |
|
747 | 776 | } else { |
|
748 | 777 | // No mtime in the dirstate entry |
|
749 | 778 | false |
|
750 | 779 | }; |
|
751 | 780 | if !mtime_looks_clean { |
|
752 | 781 | self.push_outcome(Outcome::Unsure, dirstate_node)? |
|
753 | 782 | } else if self.options.list_clean { |
|
754 | 783 | self.push_outcome(Outcome::Clean, dirstate_node)? |
|
755 | 784 | } |
|
756 | 785 | } |
|
757 | 786 | Ok(()) |
|
758 | 787 | } |
|
759 | 788 | |
|
760 | 789 | /// A node in the dirstate tree has no corresponding filesystem entry |
|
761 | 790 | fn traverse_dirstate_only( |
|
762 | 791 | &self, |
|
763 | 792 | dirstate_node: NodeRef<'tree, 'on_disk>, |
|
764 | 793 | ) -> Result<(), DirstateV2ParseError> { |
|
765 | 794 | self.check_for_outdated_directory_cache(&dirstate_node)?; |
|
766 | 795 | self.mark_removed_or_deleted_if_file(&dirstate_node)?; |
|
767 | 796 | dirstate_node |
|
768 | 797 | .children(self.dmap.on_disk)? |
|
769 | 798 | .par_iter() |
|
770 | 799 | .map(|child_node| self.traverse_dirstate_only(child_node)) |
|
771 | 800 | .collect() |
|
772 | 801 | } |
|
773 | 802 | |
|
774 | 803 | /// A node in the dirstate tree has no corresponding *file* on the |
|
775 | 804 | /// filesystem |
|
776 | 805 | /// |
|
777 | 806 | /// Does nothing on a "directory" node |
|
778 | 807 | fn mark_removed_or_deleted_if_file( |
|
779 | 808 | &self, |
|
780 | 809 | dirstate_node: &NodeRef<'tree, 'on_disk>, |
|
781 | 810 | ) -> Result<(), DirstateV2ParseError> { |
|
782 | 811 | if let Some(entry) = dirstate_node.entry()? { |
|
783 | 812 | if !entry.any_tracked() { |
|
784 | 813 | // Future-compat for when we start storing ignored and unknown |
|
785 | 814 | // files for caching reasons |
|
786 | 815 | return Ok(()); |
|
787 | 816 | } |
|
788 | 817 | let path = dirstate_node.full_path(self.dmap.on_disk)?; |
|
789 | 818 | if self.matcher.matches(path) { |
|
790 | 819 | if entry.removed() { |
|
791 | 820 | self.push_outcome(Outcome::Removed, dirstate_node)? |
|
792 | 821 | } else { |
|
793 | 822 | self.push_outcome(Outcome::Deleted, dirstate_node)? |
|
794 | 823 | } |
|
795 | 824 | } |
|
796 | 825 | } |
|
797 | 826 | Ok(()) |
|
798 | 827 | } |
|
799 | 828 | |
|
800 | 829 | /// Something in the filesystem has no corresponding dirstate node |
|
801 | 830 | /// |
|
802 | 831 | /// Returns whether that path is ignored |
|
803 | 832 | fn traverse_fs_only( |
|
804 | 833 | &self, |
|
805 | 834 | has_ignored_ancestor: bool, |
|
806 | 835 | directory_hg_path: &HgPath, |
|
807 | 836 | fs_entry: &DirEntry, |
|
808 | 837 | ) -> bool { |
|
809 | 838 | let hg_path = directory_hg_path.join(&fs_entry.hg_path); |
|
810 | 839 | let file_or_symlink = fs_entry.is_file() || fs_entry.is_symlink(); |
|
811 | 840 | if fs_entry.is_dir() { |
|
812 | 841 | let is_ignored = |
|
813 | 842 | has_ignored_ancestor || (self.ignore_fn)(&hg_path); |
|
814 | 843 | let traverse_children = if is_ignored { |
|
815 | 844 | // Descendants of an ignored directory are all ignored |
|
816 | 845 | self.options.list_ignored |
|
817 | 846 | } else { |
|
818 | 847 | // Descendants of an unknown directory may be either unknown or |
|
819 | 848 | // ignored |
|
820 | 849 | self.options.list_unknown || self.options.list_ignored |
|
821 | 850 | }; |
|
822 | 851 | if traverse_children { |
|
823 | 852 | let is_at_repo_root = false; |
|
824 | 853 | if let Ok(children_fs_entries) = |
|
825 | 854 | self.read_dir(&hg_path, &fs_entry.fs_path, is_at_repo_root) |
|
826 | 855 | { |
|
827 | 856 | children_fs_entries.par_iter().for_each(|child_fs_entry| { |
|
828 | 857 | self.traverse_fs_only( |
|
829 | 858 | is_ignored, |
|
830 | 859 | &hg_path, |
|
831 | 860 | child_fs_entry, |
|
832 | 861 | ); |
|
833 | 862 | }) |
|
834 | 863 | } |
|
835 | 864 | if self.options.collect_traversed_dirs { |
|
836 | 865 | self.outcome.lock().unwrap().traversed.push(hg_path.into()) |
|
837 | 866 | } |
|
838 | 867 | } |
|
839 | 868 | is_ignored |
|
840 | 869 | } else if file_or_symlink { |
|
841 | 870 | if self.matcher.matches(&hg_path) { |
|
842 | 871 | self.mark_unknown_or_ignored( |
|
843 | 872 | has_ignored_ancestor, |
|
844 | 873 | &BorrowedPath::InMemory(&hg_path), |
|
845 | 874 | ) |
|
846 | 875 | } else { |
|
847 | 876 | // We haven’t computed whether this path is ignored. It |
|
848 | 877 | // might not be, and a future run of status might have a |
|
849 | 878 | // different matcher that matches it. So treat it as not |
|
850 | 879 | // ignored. That is, inhibit readdir caching of the parent |
|
851 | 880 | // directory. |
|
852 | 881 | false |
|
853 | 882 | } |
|
854 | 883 | } else { |
|
855 | 884 | // This is neither a directory, a plain file, or a symlink. |
|
856 | 885 | // Treat it like an ignored file. |
|
857 | 886 | true |
|
858 | 887 | } |
|
859 | 888 | } |
|
860 | 889 | |
|
861 | 890 | /// Returns whether that path is ignored |
|
862 | 891 | fn mark_unknown_or_ignored( |
|
863 | 892 | &self, |
|
864 | 893 | has_ignored_ancestor: bool, |
|
865 | 894 | hg_path: &BorrowedPath<'_, 'on_disk>, |
|
866 | 895 | ) -> bool { |
|
867 | 896 | let is_ignored = has_ignored_ancestor || (self.ignore_fn)(hg_path); |
|
868 | 897 | if is_ignored { |
|
869 | 898 | if self.options.list_ignored { |
|
870 | 899 | self.push_outcome_without_copy_source( |
|
871 | 900 | Outcome::Ignored, |
|
872 | 901 | hg_path, |
|
873 | 902 | ) |
|
874 | 903 | } |
|
875 | 904 | } else if self.options.list_unknown { |
|
876 | 905 | self.push_outcome_without_copy_source(Outcome::Unknown, hg_path) |
|
877 | 906 | } |
|
878 | 907 | is_ignored |
|
879 | 908 | } |
|
880 | 909 | } |
|
881 | 910 | |
|
882 | 911 | /// Since [`std::fs::FileType`] cannot be built directly, we emulate what we |
|
883 | 912 | /// care about. |
|
884 | 913 | #[derive(Copy, Clone, Debug, PartialEq, Eq)] |
|
885 | 914 | enum FakeFileType { |
|
886 | 915 | File, |
|
887 | 916 | Directory, |
|
888 | 917 | Symlink, |
|
889 | 918 | BadType(BadType), |
|
890 | 919 | } |
|
891 | 920 | |
|
892 | 921 | impl From<std::fs::FileType> for FakeFileType { |
|
893 | 922 | fn from(f: std::fs::FileType) -> Self { |
|
894 | 923 | if f.is_dir() { |
|
895 | 924 | Self::Directory |
|
896 | 925 | } else if f.is_file() { |
|
897 | 926 | Self::File |
|
898 | 927 | } else if f.is_symlink() { |
|
899 | 928 | Self::Symlink |
|
900 | 929 | } else if f.is_fifo() { |
|
901 | 930 | Self::BadType(BadType::FIFO) |
|
902 | 931 | } else if f.is_block_device() { |
|
903 | 932 | Self::BadType(BadType::BlockDevice) |
|
904 | 933 | } else if f.is_char_device() { |
|
905 | 934 | Self::BadType(BadType::CharacterDevice) |
|
906 | 935 | } else if f.is_socket() { |
|
907 | 936 | Self::BadType(BadType::Socket) |
|
908 | 937 | } else { |
|
909 | 938 | Self::BadType(BadType::Unknown) |
|
910 | 939 | } |
|
911 | 940 | } |
|
912 | 941 | } |
|
913 | 942 | |
|
914 | 943 | struct DirEntry<'a> { |
|
915 | 944 | /// Path as stored in the dirstate, or just the filename for optimization. |
|
916 | 945 | hg_path: HgPathCow<'a>, |
|
917 | 946 | /// Filesystem path |
|
918 | 947 | fs_path: Cow<'a, Path>, |
|
919 | 948 | /// Lazily computed |
|
920 | 949 | symlink_metadata: Option<std::fs::Metadata>, |
|
921 | 950 | /// Already computed for ergonomics. |
|
922 | 951 | file_type: FakeFileType, |
|
923 | 952 | } |
|
924 | 953 | |
|
925 | 954 | impl<'a> DirEntry<'a> { |
|
926 | 955 | /// Returns **unsorted** entries in the given directory, with name, |
|
927 | 956 | /// metadata and file type. |
|
928 | 957 | /// |
|
929 | 958 | /// If a `.hg` sub-directory is encountered: |
|
930 | 959 | /// |
|
931 | 960 | /// * At the repository root, ignore that sub-directory |
|
932 | 961 | /// * Elsewhere, we’re listing the content of a sub-repo. Return an empty |
|
933 | 962 | /// list instead. |
|
934 | 963 | fn read_dir(path: &Path, is_at_repo_root: bool) -> io::Result<Vec<Self>> { |
|
935 | 964 | // `read_dir` returns a "not found" error for the empty path |
|
936 | 965 | let at_cwd = path == Path::new(""); |
|
937 | 966 | let read_dir_path = if at_cwd { Path::new(".") } else { path }; |
|
938 | 967 | let mut results = Vec::new(); |
|
939 | 968 | for entry in read_dir_path.read_dir()? { |
|
940 | 969 | let entry = entry?; |
|
941 | 970 | let file_type = match entry.file_type() { |
|
942 | 971 | Ok(v) => v, |
|
943 | 972 | Err(e) => { |
|
944 | 973 | // race with file deletion? |
|
945 | 974 | if e.kind() == std::io::ErrorKind::NotFound { |
|
946 | 975 | continue; |
|
947 | 976 | } else { |
|
948 | 977 | return Err(e); |
|
949 | 978 | } |
|
950 | 979 | } |
|
951 | 980 | }; |
|
952 | 981 | let file_name = entry.file_name(); |
|
953 | 982 | // FIXME don't do this when cached |
|
954 | 983 | if file_name == ".hg" { |
|
955 | 984 | if is_at_repo_root { |
|
956 | 985 | // Skip the repo’s own .hg (might be a symlink) |
|
957 | 986 | continue; |
|
958 | 987 | } else if file_type.is_dir() { |
|
959 | 988 | // A .hg sub-directory at another location means a subrepo, |
|
960 | 989 | // skip it entirely. |
|
961 | 990 | return Ok(Vec::new()); |
|
962 | 991 | } |
|
963 | 992 | } |
|
964 | 993 | let full_path = if at_cwd { |
|
965 | 994 | file_name.clone().into() |
|
966 | 995 | } else { |
|
967 | 996 | entry.path() |
|
968 | 997 | }; |
|
969 | 998 | let filename = |
|
970 | 999 | Cow::Owned(get_bytes_from_os_string(file_name).into()); |
|
971 | 1000 | let file_type = FakeFileType::from(file_type); |
|
972 | 1001 | results.push(DirEntry { |
|
973 | 1002 | hg_path: filename, |
|
974 | 1003 | fs_path: Cow::Owned(full_path.to_path_buf()), |
|
975 | 1004 | symlink_metadata: None, |
|
976 | 1005 | file_type, |
|
977 | 1006 | }) |
|
978 | 1007 | } |
|
979 | 1008 | Ok(results) |
|
980 | 1009 | } |
|
981 | 1010 | |
|
982 | 1011 | fn symlink_metadata(&self) -> Result<std::fs::Metadata, std::io::Error> { |
|
983 | 1012 | match &self.symlink_metadata { |
|
984 | 1013 | Some(meta) => Ok(meta.clone()), |
|
985 | 1014 | None => std::fs::symlink_metadata(&self.fs_path), |
|
986 | 1015 | } |
|
987 | 1016 | } |
|
988 | 1017 | |
|
989 | 1018 | fn is_dir(&self) -> bool { |
|
990 | 1019 | self.file_type == FakeFileType::Directory |
|
991 | 1020 | } |
|
992 | 1021 | |
|
993 | 1022 | fn is_file(&self) -> bool { |
|
994 | 1023 | self.file_type == FakeFileType::File |
|
995 | 1024 | } |
|
996 | 1025 | |
|
997 | 1026 | fn is_symlink(&self) -> bool { |
|
998 | 1027 | self.file_type == FakeFileType::Symlink |
|
999 | 1028 | } |
|
1000 | 1029 | |
|
1001 | 1030 | fn is_bad(&self) -> Option<BadType> { |
|
1002 | 1031 | match self.file_type { |
|
1003 | 1032 | FakeFileType::BadType(ty) => Some(ty), |
|
1004 | 1033 | _ => None, |
|
1005 | 1034 | } |
|
1006 | 1035 | } |
|
1007 | 1036 | } |
|
1008 | 1037 | |
|
1009 | 1038 | /// Return the `mtime` of a temporary file newly-created in the `.hg` directory |
|
1010 | 1039 | /// of the give repository. |
|
1011 | 1040 | /// |
|
1012 | 1041 | /// This is similar to `SystemTime::now()`, with the result truncated to the |
|
1013 | 1042 | /// same time resolution as other files’ modification times. Using `.hg` |
|
1014 | 1043 | /// instead of the system’s default temporary directory (such as `/tmp`) makes |
|
1015 | 1044 | /// it more likely the temporary file is in the same disk partition as contents |
|
1016 | 1045 | /// of the working directory, which can matter since different filesystems may |
|
1017 | 1046 | /// store timestamps with different resolutions. |
|
1018 | 1047 | /// |
|
1019 | 1048 | /// This may fail, typically if we lack write permissions. In that case we |
|
1020 | 1049 | /// should continue the `status()` algoritm anyway and consider the current |
|
1021 | 1050 | /// date/time to be unknown. |
|
1022 | 1051 | fn filesystem_now(repo_root: &Path) -> Result<SystemTime, io::Error> { |
|
1023 | 1052 | tempfile::tempfile_in(repo_root.join(".hg"))? |
|
1024 | 1053 | .metadata()? |
|
1025 | 1054 | .modified() |
|
1026 | 1055 | } |
General Comments 0
You need to be logged in to leave comments.
Login now