Show More
@@ -1,1139 +1,1139 b'' | |||
|
1 | 1 | use bytes_cast::BytesCast; |
|
2 | 2 | use micro_timer::timed; |
|
3 | 3 | use std::borrow::Cow; |
|
4 | 4 | use std::path::PathBuf; |
|
5 | 5 | |
|
6 | 6 | use super::on_disk; |
|
7 | 7 | use super::on_disk::DirstateV2ParseError; |
|
8 | 8 | use super::owning::OwningDirstateMap; |
|
9 | 9 | use super::path_with_basename::WithBasename; |
|
10 | 10 | use crate::dirstate::parsers::pack_entry; |
|
11 | 11 | use crate::dirstate::parsers::packed_entry_size; |
|
12 | 12 | use crate::dirstate::parsers::parse_dirstate_entries; |
|
13 | 13 | use crate::dirstate::CopyMapIter; |
|
14 | 14 | use crate::dirstate::StateMapIter; |
|
15 | 15 | use crate::dirstate::TruncatedTimestamp; |
|
16 | 16 | use crate::dirstate::SIZE_FROM_OTHER_PARENT; |
|
17 | 17 | use crate::dirstate::SIZE_NON_NORMAL; |
|
18 | 18 | use crate::matchers::Matcher; |
|
19 | 19 | use crate::utils::hg_path::{HgPath, HgPathBuf}; |
|
20 | 20 | use crate::DirstateEntry; |
|
21 | 21 | use crate::DirstateError; |
|
22 | 22 | use crate::DirstateParents; |
|
23 | 23 | use crate::DirstateStatus; |
|
24 | 24 | use crate::EntryState; |
|
25 | 25 | use crate::FastHashMap; |
|
26 | 26 | use crate::PatternFileWarning; |
|
27 | 27 | use crate::StatusError; |
|
28 | 28 | use crate::StatusOptions; |
|
29 | 29 | |
|
30 | 30 | /// Append to an existing data file if the amount of unreachable data (not used |
|
31 | 31 | /// anymore) is less than this fraction of the total amount of existing data. |
|
32 | 32 | const ACCEPTABLE_UNREACHABLE_BYTES_RATIO: f32 = 0.5; |
|
33 | 33 | |
|
34 | 34 | pub struct DirstateMap<'on_disk> { |
|
35 | 35 | /// Contents of the `.hg/dirstate` file |
|
36 | 36 | pub(super) on_disk: &'on_disk [u8], |
|
37 | 37 | |
|
38 | 38 | pub(super) root: ChildNodes<'on_disk>, |
|
39 | 39 | |
|
40 | 40 | /// Number of nodes anywhere in the tree that have `.entry.is_some()`. |
|
41 | 41 | pub(super) nodes_with_entry_count: u32, |
|
42 | 42 | |
|
43 | 43 | /// Number of nodes anywhere in the tree that have |
|
44 | 44 | /// `.copy_source.is_some()`. |
|
45 | 45 | pub(super) nodes_with_copy_source_count: u32, |
|
46 | 46 | |
|
47 | 47 | /// See on_disk::Header |
|
48 | 48 | pub(super) ignore_patterns_hash: on_disk::IgnorePatternsHash, |
|
49 | 49 | |
|
50 | 50 | /// How many bytes of `on_disk` are not used anymore |
|
51 | 51 | pub(super) unreachable_bytes: u32, |
|
52 | 52 | } |
|
53 | 53 | |
|
54 | 54 | /// Using a plain `HgPathBuf` of the full path from the repository root as a |
|
55 | 55 | /// map key would also work: all paths in a given map have the same parent |
|
56 | 56 | /// path, so comparing full paths gives the same result as comparing base |
|
57 | 57 | /// names. However `HashMap` would waste time always re-hashing the same |
|
58 | 58 | /// string prefix. |
|
59 | 59 | pub(super) type NodeKey<'on_disk> = WithBasename<Cow<'on_disk, HgPath>>; |
|
60 | 60 | |
|
61 | 61 | /// Similar to `&'tree Cow<'on_disk, HgPath>`, but can also be returned |
|
62 | 62 | /// for on-disk nodes that don’t actually have a `Cow` to borrow. |
|
63 | 63 | pub(super) enum BorrowedPath<'tree, 'on_disk> { |
|
64 | 64 | InMemory(&'tree HgPathBuf), |
|
65 | 65 | OnDisk(&'on_disk HgPath), |
|
66 | 66 | } |
|
67 | 67 | |
|
68 | 68 | pub(super) enum ChildNodes<'on_disk> { |
|
69 | 69 | InMemory(FastHashMap<NodeKey<'on_disk>, Node<'on_disk>>), |
|
70 | 70 | OnDisk(&'on_disk [on_disk::Node]), |
|
71 | 71 | } |
|
72 | 72 | |
|
73 | 73 | pub(super) enum ChildNodesRef<'tree, 'on_disk> { |
|
74 | 74 | InMemory(&'tree FastHashMap<NodeKey<'on_disk>, Node<'on_disk>>), |
|
75 | 75 | OnDisk(&'on_disk [on_disk::Node]), |
|
76 | 76 | } |
|
77 | 77 | |
|
78 | 78 | pub(super) enum NodeRef<'tree, 'on_disk> { |
|
79 | 79 | InMemory(&'tree NodeKey<'on_disk>, &'tree Node<'on_disk>), |
|
80 | 80 | OnDisk(&'on_disk on_disk::Node), |
|
81 | 81 | } |
|
82 | 82 | |
|
83 | 83 | impl<'tree, 'on_disk> BorrowedPath<'tree, 'on_disk> { |
|
84 | 84 | pub fn detach_from_tree(&self) -> Cow<'on_disk, HgPath> { |
|
85 | 85 | match *self { |
|
86 | 86 | BorrowedPath::InMemory(in_memory) => Cow::Owned(in_memory.clone()), |
|
87 | 87 | BorrowedPath::OnDisk(on_disk) => Cow::Borrowed(on_disk), |
|
88 | 88 | } |
|
89 | 89 | } |
|
90 | 90 | } |
|
91 | 91 | |
|
92 | 92 | impl<'tree, 'on_disk> std::ops::Deref for BorrowedPath<'tree, 'on_disk> { |
|
93 | 93 | type Target = HgPath; |
|
94 | 94 | |
|
95 | 95 | fn deref(&self) -> &HgPath { |
|
96 | 96 | match *self { |
|
97 | 97 | BorrowedPath::InMemory(in_memory) => in_memory, |
|
98 | 98 | BorrowedPath::OnDisk(on_disk) => on_disk, |
|
99 | 99 | } |
|
100 | 100 | } |
|
101 | 101 | } |
|
102 | 102 | |
|
103 | 103 | impl Default for ChildNodes<'_> { |
|
104 | 104 | fn default() -> Self { |
|
105 | 105 | ChildNodes::InMemory(Default::default()) |
|
106 | 106 | } |
|
107 | 107 | } |
|
108 | 108 | |
|
109 | 109 | impl<'on_disk> ChildNodes<'on_disk> { |
|
110 | 110 | pub(super) fn as_ref<'tree>( |
|
111 | 111 | &'tree self, |
|
112 | 112 | ) -> ChildNodesRef<'tree, 'on_disk> { |
|
113 | 113 | match self { |
|
114 | 114 | ChildNodes::InMemory(nodes) => ChildNodesRef::InMemory(nodes), |
|
115 | 115 | ChildNodes::OnDisk(nodes) => ChildNodesRef::OnDisk(nodes), |
|
116 | 116 | } |
|
117 | 117 | } |
|
118 | 118 | |
|
119 | 119 | pub(super) fn is_empty(&self) -> bool { |
|
120 | 120 | match self { |
|
121 | 121 | ChildNodes::InMemory(nodes) => nodes.is_empty(), |
|
122 | 122 | ChildNodes::OnDisk(nodes) => nodes.is_empty(), |
|
123 | 123 | } |
|
124 | 124 | } |
|
125 | 125 | |
|
126 | 126 | fn make_mut( |
|
127 | 127 | &mut self, |
|
128 | 128 | on_disk: &'on_disk [u8], |
|
129 | 129 | unreachable_bytes: &mut u32, |
|
130 | 130 | ) -> Result< |
|
131 | 131 | &mut FastHashMap<NodeKey<'on_disk>, Node<'on_disk>>, |
|
132 | 132 | DirstateV2ParseError, |
|
133 | 133 | > { |
|
134 | 134 | match self { |
|
135 | 135 | ChildNodes::InMemory(nodes) => Ok(nodes), |
|
136 | 136 | ChildNodes::OnDisk(nodes) => { |
|
137 | 137 | *unreachable_bytes += |
|
138 | 138 | std::mem::size_of_val::<[on_disk::Node]>(nodes) as u32; |
|
139 | 139 | let nodes = nodes |
|
140 | 140 | .iter() |
|
141 | 141 | .map(|node| { |
|
142 | 142 | Ok(( |
|
143 | 143 | node.path(on_disk)?, |
|
144 | 144 | node.to_in_memory_node(on_disk)?, |
|
145 | 145 | )) |
|
146 | 146 | }) |
|
147 | 147 | .collect::<Result<_, _>>()?; |
|
148 | 148 | *self = ChildNodes::InMemory(nodes); |
|
149 | 149 | match self { |
|
150 | 150 | ChildNodes::InMemory(nodes) => Ok(nodes), |
|
151 | 151 | ChildNodes::OnDisk(_) => unreachable!(), |
|
152 | 152 | } |
|
153 | 153 | } |
|
154 | 154 | } |
|
155 | 155 | } |
|
156 | 156 | } |
|
157 | 157 | |
|
158 | 158 | impl<'tree, 'on_disk> ChildNodesRef<'tree, 'on_disk> { |
|
159 | 159 | pub(super) fn get( |
|
160 | 160 | &self, |
|
161 | 161 | base_name: &HgPath, |
|
162 | 162 | on_disk: &'on_disk [u8], |
|
163 | 163 | ) -> Result<Option<NodeRef<'tree, 'on_disk>>, DirstateV2ParseError> { |
|
164 | 164 | match self { |
|
165 | 165 | ChildNodesRef::InMemory(nodes) => Ok(nodes |
|
166 | 166 | .get_key_value(base_name) |
|
167 | 167 | .map(|(k, v)| NodeRef::InMemory(k, v))), |
|
168 | 168 | ChildNodesRef::OnDisk(nodes) => { |
|
169 | 169 | let mut parse_result = Ok(()); |
|
170 | 170 | let search_result = nodes.binary_search_by(|node| { |
|
171 | 171 | match node.base_name(on_disk) { |
|
172 | 172 | Ok(node_base_name) => node_base_name.cmp(base_name), |
|
173 | 173 | Err(e) => { |
|
174 | 174 | parse_result = Err(e); |
|
175 | 175 | // Dummy comparison result, `search_result` won’t |
|
176 | 176 | // be used since `parse_result` is an error |
|
177 | 177 | std::cmp::Ordering::Equal |
|
178 | 178 | } |
|
179 | 179 | } |
|
180 | 180 | }); |
|
181 | 181 | parse_result.map(|()| { |
|
182 | 182 | search_result.ok().map(|i| NodeRef::OnDisk(&nodes[i])) |
|
183 | 183 | }) |
|
184 | 184 | } |
|
185 | 185 | } |
|
186 | 186 | } |
|
187 | 187 | |
|
188 | 188 | /// Iterate in undefined order |
|
189 | 189 | pub(super) fn iter( |
|
190 | 190 | &self, |
|
191 | 191 | ) -> impl Iterator<Item = NodeRef<'tree, 'on_disk>> { |
|
192 | 192 | match self { |
|
193 | 193 | ChildNodesRef::InMemory(nodes) => itertools::Either::Left( |
|
194 | 194 | nodes.iter().map(|(k, v)| NodeRef::InMemory(k, v)), |
|
195 | 195 | ), |
|
196 | 196 | ChildNodesRef::OnDisk(nodes) => { |
|
197 | 197 | itertools::Either::Right(nodes.iter().map(NodeRef::OnDisk)) |
|
198 | 198 | } |
|
199 | 199 | } |
|
200 | 200 | } |
|
201 | 201 | |
|
202 | 202 | /// Iterate in parallel in undefined order |
|
203 | 203 | pub(super) fn par_iter( |
|
204 | 204 | &self, |
|
205 | 205 | ) -> impl rayon::iter::ParallelIterator<Item = NodeRef<'tree, 'on_disk>> |
|
206 | 206 | { |
|
207 | 207 | use rayon::prelude::*; |
|
208 | 208 | match self { |
|
209 | 209 | ChildNodesRef::InMemory(nodes) => rayon::iter::Either::Left( |
|
210 | 210 | nodes.par_iter().map(|(k, v)| NodeRef::InMemory(k, v)), |
|
211 | 211 | ), |
|
212 | 212 | ChildNodesRef::OnDisk(nodes) => rayon::iter::Either::Right( |
|
213 | 213 | nodes.par_iter().map(NodeRef::OnDisk), |
|
214 | 214 | ), |
|
215 | 215 | } |
|
216 | 216 | } |
|
217 | 217 | |
|
218 | 218 | pub(super) fn sorted(&self) -> Vec<NodeRef<'tree, 'on_disk>> { |
|
219 | 219 | match self { |
|
220 | 220 | ChildNodesRef::InMemory(nodes) => { |
|
221 | 221 | let mut vec: Vec<_> = nodes |
|
222 | 222 | .iter() |
|
223 | 223 | .map(|(k, v)| NodeRef::InMemory(k, v)) |
|
224 | 224 | .collect(); |
|
225 | 225 | fn sort_key<'a>(node: &'a NodeRef) -> &'a HgPath { |
|
226 | 226 | match node { |
|
227 | 227 | NodeRef::InMemory(path, _node) => path.base_name(), |
|
228 | 228 | NodeRef::OnDisk(_) => unreachable!(), |
|
229 | 229 | } |
|
230 | 230 | } |
|
231 | 231 | // `sort_unstable_by_key` doesn’t allow keys borrowing from the |
|
232 | 232 | // value: https://github.com/rust-lang/rust/issues/34162 |
|
233 | 233 | vec.sort_unstable_by(|a, b| sort_key(a).cmp(sort_key(b))); |
|
234 | 234 | vec |
|
235 | 235 | } |
|
236 | 236 | ChildNodesRef::OnDisk(nodes) => { |
|
237 | 237 | // Nodes on disk are already sorted |
|
238 | 238 | nodes.iter().map(NodeRef::OnDisk).collect() |
|
239 | 239 | } |
|
240 | 240 | } |
|
241 | 241 | } |
|
242 | 242 | } |
|
243 | 243 | |
|
244 | 244 | impl<'tree, 'on_disk> NodeRef<'tree, 'on_disk> { |
|
245 | 245 | pub(super) fn full_path( |
|
246 | 246 | &self, |
|
247 | 247 | on_disk: &'on_disk [u8], |
|
248 | 248 | ) -> Result<&'tree HgPath, DirstateV2ParseError> { |
|
249 | 249 | match self { |
|
250 | 250 | NodeRef::InMemory(path, _node) => Ok(path.full_path()), |
|
251 | 251 | NodeRef::OnDisk(node) => node.full_path(on_disk), |
|
252 | 252 | } |
|
253 | 253 | } |
|
254 | 254 | |
|
255 | 255 | /// Returns a `BorrowedPath`, which can be turned into a `Cow<'on_disk, |
|
256 | 256 | /// HgPath>` detached from `'tree` |
|
257 | 257 | pub(super) fn full_path_borrowed( |
|
258 | 258 | &self, |
|
259 | 259 | on_disk: &'on_disk [u8], |
|
260 | 260 | ) -> Result<BorrowedPath<'tree, 'on_disk>, DirstateV2ParseError> { |
|
261 | 261 | match self { |
|
262 | 262 | NodeRef::InMemory(path, _node) => match path.full_path() { |
|
263 | 263 | Cow::Borrowed(on_disk) => Ok(BorrowedPath::OnDisk(on_disk)), |
|
264 | 264 | Cow::Owned(in_memory) => Ok(BorrowedPath::InMemory(in_memory)), |
|
265 | 265 | }, |
|
266 | 266 | NodeRef::OnDisk(node) => { |
|
267 | 267 | Ok(BorrowedPath::OnDisk(node.full_path(on_disk)?)) |
|
268 | 268 | } |
|
269 | 269 | } |
|
270 | 270 | } |
|
271 | 271 | |
|
272 | 272 | pub(super) fn base_name( |
|
273 | 273 | &self, |
|
274 | 274 | on_disk: &'on_disk [u8], |
|
275 | 275 | ) -> Result<&'tree HgPath, DirstateV2ParseError> { |
|
276 | 276 | match self { |
|
277 | 277 | NodeRef::InMemory(path, _node) => Ok(path.base_name()), |
|
278 | 278 | NodeRef::OnDisk(node) => node.base_name(on_disk), |
|
279 | 279 | } |
|
280 | 280 | } |
|
281 | 281 | |
|
282 | 282 | pub(super) fn children( |
|
283 | 283 | &self, |
|
284 | 284 | on_disk: &'on_disk [u8], |
|
285 | 285 | ) -> Result<ChildNodesRef<'tree, 'on_disk>, DirstateV2ParseError> { |
|
286 | 286 | match self { |
|
287 | 287 | NodeRef::InMemory(_path, node) => Ok(node.children.as_ref()), |
|
288 | 288 | NodeRef::OnDisk(node) => { |
|
289 | 289 | Ok(ChildNodesRef::OnDisk(node.children(on_disk)?)) |
|
290 | 290 | } |
|
291 | 291 | } |
|
292 | 292 | } |
|
293 | 293 | |
|
294 | 294 | pub(super) fn has_copy_source(&self) -> bool { |
|
295 | 295 | match self { |
|
296 | 296 | NodeRef::InMemory(_path, node) => node.copy_source.is_some(), |
|
297 | 297 | NodeRef::OnDisk(node) => node.has_copy_source(), |
|
298 | 298 | } |
|
299 | 299 | } |
|
300 | 300 | |
|
301 | 301 | pub(super) fn copy_source( |
|
302 | 302 | &self, |
|
303 | 303 | on_disk: &'on_disk [u8], |
|
304 | 304 | ) -> Result<Option<&'tree HgPath>, DirstateV2ParseError> { |
|
305 | 305 | match self { |
|
306 | 306 | NodeRef::InMemory(_path, node) => { |
|
307 | 307 | Ok(node.copy_source.as_ref().map(|s| &**s)) |
|
308 | 308 | } |
|
309 | 309 | NodeRef::OnDisk(node) => node.copy_source(on_disk), |
|
310 | 310 | } |
|
311 | 311 | } |
|
312 | 312 | |
|
313 | 313 | pub(super) fn entry( |
|
314 | 314 | &self, |
|
315 | 315 | ) -> Result<Option<DirstateEntry>, DirstateV2ParseError> { |
|
316 | 316 | match self { |
|
317 | 317 | NodeRef::InMemory(_path, node) => { |
|
318 | 318 | Ok(node.data.as_entry().copied()) |
|
319 | 319 | } |
|
320 | 320 | NodeRef::OnDisk(node) => node.entry(), |
|
321 | 321 | } |
|
322 | 322 | } |
|
323 | 323 | |
|
324 | 324 | pub(super) fn state( |
|
325 | 325 | &self, |
|
326 | 326 | ) -> Result<Option<EntryState>, DirstateV2ParseError> { |
|
327 | 327 | Ok(self.entry()?.map(|e| e.state())) |
|
328 | 328 | } |
|
329 | 329 | |
|
330 | 330 | pub(super) fn cached_directory_mtime( |
|
331 | 331 | &self, |
|
332 | 332 | ) -> Result<Option<TruncatedTimestamp>, DirstateV2ParseError> { |
|
333 | 333 | match self { |
|
334 | 334 | NodeRef::InMemory(_path, node) => Ok(match node.data { |
|
335 | 335 | NodeData::CachedDirectory { mtime } => Some(mtime), |
|
336 | 336 | _ => None, |
|
337 | 337 | }), |
|
338 | 338 | NodeRef::OnDisk(node) => node.cached_directory_mtime(), |
|
339 | 339 | } |
|
340 | 340 | } |
|
341 | 341 | |
|
342 | 342 | pub(super) fn descendants_with_entry_count(&self) -> u32 { |
|
343 | 343 | match self { |
|
344 | 344 | NodeRef::InMemory(_path, node) => { |
|
345 | 345 | node.descendants_with_entry_count |
|
346 | 346 | } |
|
347 | 347 | NodeRef::OnDisk(node) => node.descendants_with_entry_count.get(), |
|
348 | 348 | } |
|
349 | 349 | } |
|
350 | 350 | |
|
351 | 351 | pub(super) fn tracked_descendants_count(&self) -> u32 { |
|
352 | 352 | match self { |
|
353 | 353 | NodeRef::InMemory(_path, node) => node.tracked_descendants_count, |
|
354 | 354 | NodeRef::OnDisk(node) => node.tracked_descendants_count.get(), |
|
355 | 355 | } |
|
356 | 356 | } |
|
357 | 357 | } |
|
358 | 358 | |
|
359 | 359 | /// Represents a file or a directory |
|
360 | 360 | #[derive(Default)] |
|
361 | 361 | pub(super) struct Node<'on_disk> { |
|
362 | 362 | pub(super) data: NodeData, |
|
363 | 363 | |
|
364 | 364 | pub(super) copy_source: Option<Cow<'on_disk, HgPath>>, |
|
365 | 365 | |
|
366 | 366 | pub(super) children: ChildNodes<'on_disk>, |
|
367 | 367 | |
|
368 | 368 | /// How many (non-inclusive) descendants of this node have an entry. |
|
369 | 369 | pub(super) descendants_with_entry_count: u32, |
|
370 | 370 | |
|
371 | 371 | /// How many (non-inclusive) descendants of this node have an entry whose |
|
372 | 372 | /// state is "tracked". |
|
373 | 373 | pub(super) tracked_descendants_count: u32, |
|
374 | 374 | } |
|
375 | 375 | |
|
376 | 376 | pub(super) enum NodeData { |
|
377 | 377 | Entry(DirstateEntry), |
|
378 | 378 | CachedDirectory { mtime: TruncatedTimestamp }, |
|
379 | 379 | None, |
|
380 | 380 | } |
|
381 | 381 | |
|
382 | 382 | impl Default for NodeData { |
|
383 | 383 | fn default() -> Self { |
|
384 | 384 | NodeData::None |
|
385 | 385 | } |
|
386 | 386 | } |
|
387 | 387 | |
|
388 | 388 | impl NodeData { |
|
389 | 389 | fn has_entry(&self) -> bool { |
|
390 | 390 | match self { |
|
391 | 391 | NodeData::Entry(_) => true, |
|
392 | 392 | _ => false, |
|
393 | 393 | } |
|
394 | 394 | } |
|
395 | 395 | |
|
396 | 396 | fn as_entry(&self) -> Option<&DirstateEntry> { |
|
397 | 397 | match self { |
|
398 | 398 | NodeData::Entry(entry) => Some(entry), |
|
399 | 399 | _ => None, |
|
400 | 400 | } |
|
401 | 401 | } |
|
402 | 402 | } |
|
403 | 403 | |
|
404 | 404 | impl<'on_disk> DirstateMap<'on_disk> { |
|
405 | 405 | pub(super) fn empty(on_disk: &'on_disk [u8]) -> Self { |
|
406 | 406 | Self { |
|
407 | 407 | on_disk, |
|
408 | 408 | root: ChildNodes::default(), |
|
409 | 409 | nodes_with_entry_count: 0, |
|
410 | 410 | nodes_with_copy_source_count: 0, |
|
411 | 411 | ignore_patterns_hash: [0; on_disk::IGNORE_PATTERNS_HASH_LEN], |
|
412 | 412 | unreachable_bytes: 0, |
|
413 | 413 | } |
|
414 | 414 | } |
|
415 | 415 | |
|
416 | 416 | #[timed] |
|
417 | 417 | pub fn new_v2( |
|
418 | 418 | on_disk: &'on_disk [u8], |
|
419 | 419 | data_size: usize, |
|
420 | 420 | metadata: &[u8], |
|
421 | 421 | ) -> Result<Self, DirstateError> { |
|
422 | 422 | if let Some(data) = on_disk.get(..data_size) { |
|
423 | 423 | Ok(on_disk::read(data, metadata)?) |
|
424 | 424 | } else { |
|
425 | 425 | Err(DirstateV2ParseError.into()) |
|
426 | 426 | } |
|
427 | 427 | } |
|
428 | 428 | |
|
429 | 429 | #[timed] |
|
430 | 430 | pub fn new_v1( |
|
431 | 431 | on_disk: &'on_disk [u8], |
|
432 | 432 | ) -> Result<(Self, Option<DirstateParents>), DirstateError> { |
|
433 | 433 | let mut map = Self::empty(on_disk); |
|
434 | 434 | if map.on_disk.is_empty() { |
|
435 | 435 | return Ok((map, None)); |
|
436 | 436 | } |
|
437 | 437 | |
|
438 | 438 | let parents = parse_dirstate_entries( |
|
439 | 439 | map.on_disk, |
|
440 | 440 | |path, entry, copy_source| { |
|
441 | 441 | let tracked = entry.state().is_tracked(); |
|
442 | 442 | let node = Self::get_or_insert_node( |
|
443 | 443 | map.on_disk, |
|
444 | 444 | &mut map.unreachable_bytes, |
|
445 | 445 | &mut map.root, |
|
446 | 446 | path, |
|
447 | 447 | WithBasename::to_cow_borrowed, |
|
448 | 448 | |ancestor| { |
|
449 | 449 | if tracked { |
|
450 | 450 | ancestor.tracked_descendants_count += 1 |
|
451 | 451 | } |
|
452 | 452 | ancestor.descendants_with_entry_count += 1 |
|
453 | 453 | }, |
|
454 | 454 | )?; |
|
455 | 455 | assert!( |
|
456 | 456 | !node.data.has_entry(), |
|
457 | 457 | "duplicate dirstate entry in read" |
|
458 | 458 | ); |
|
459 | 459 | assert!( |
|
460 | 460 | node.copy_source.is_none(), |
|
461 | 461 | "duplicate dirstate entry in read" |
|
462 | 462 | ); |
|
463 | 463 | node.data = NodeData::Entry(*entry); |
|
464 | 464 | node.copy_source = copy_source.map(Cow::Borrowed); |
|
465 | 465 | map.nodes_with_entry_count += 1; |
|
466 | 466 | if copy_source.is_some() { |
|
467 | 467 | map.nodes_with_copy_source_count += 1 |
|
468 | 468 | } |
|
469 | 469 | Ok(()) |
|
470 | 470 | }, |
|
471 | 471 | )?; |
|
472 | 472 | let parents = Some(parents.clone()); |
|
473 | 473 | |
|
474 | 474 | Ok((map, parents)) |
|
475 | 475 | } |
|
476 | 476 | |
|
477 | 477 | /// Assuming dirstate-v2 format, returns whether the next write should |
|
478 | 478 | /// append to the existing data file that contains `self.on_disk` (true), |
|
479 | 479 | /// or create a new data file from scratch (false). |
|
480 | 480 | pub(super) fn write_should_append(&self) -> bool { |
|
481 | 481 | let ratio = self.unreachable_bytes as f32 / self.on_disk.len() as f32; |
|
482 | 482 | ratio < ACCEPTABLE_UNREACHABLE_BYTES_RATIO |
|
483 | 483 | } |
|
484 | 484 | |
|
485 | 485 | fn get_node<'tree>( |
|
486 | 486 | &'tree self, |
|
487 | 487 | path: &HgPath, |
|
488 | 488 | ) -> Result<Option<NodeRef<'tree, 'on_disk>>, DirstateV2ParseError> { |
|
489 | 489 | let mut children = self.root.as_ref(); |
|
490 | 490 | let mut components = path.components(); |
|
491 | 491 | let mut component = |
|
492 | 492 | components.next().expect("expected at least one components"); |
|
493 | 493 | loop { |
|
494 | 494 | if let Some(child) = children.get(component, self.on_disk)? { |
|
495 | 495 | if let Some(next_component) = components.next() { |
|
496 | 496 | component = next_component; |
|
497 | 497 | children = child.children(self.on_disk)?; |
|
498 | 498 | } else { |
|
499 | 499 | return Ok(Some(child)); |
|
500 | 500 | } |
|
501 | 501 | } else { |
|
502 | 502 | return Ok(None); |
|
503 | 503 | } |
|
504 | 504 | } |
|
505 | 505 | } |
|
506 | 506 | |
|
507 | 507 | /// Returns a mutable reference to the node at `path` if it exists |
|
508 | 508 | /// |
|
509 | 509 | /// This takes `root` instead of `&mut self` so that callers can mutate |
|
510 | 510 | /// other fields while the returned borrow is still valid |
|
511 | 511 | fn get_node_mut<'tree>( |
|
512 | 512 | on_disk: &'on_disk [u8], |
|
513 | 513 | unreachable_bytes: &mut u32, |
|
514 | 514 | root: &'tree mut ChildNodes<'on_disk>, |
|
515 | 515 | path: &HgPath, |
|
516 | 516 | ) -> Result<Option<&'tree mut Node<'on_disk>>, DirstateV2ParseError> { |
|
517 | 517 | let mut children = root; |
|
518 | 518 | let mut components = path.components(); |
|
519 | 519 | let mut component = |
|
520 | 520 | components.next().expect("expected at least one components"); |
|
521 | 521 | loop { |
|
522 | 522 | if let Some(child) = children |
|
523 | 523 | .make_mut(on_disk, unreachable_bytes)? |
|
524 | 524 | .get_mut(component) |
|
525 | 525 | { |
|
526 | 526 | if let Some(next_component) = components.next() { |
|
527 | 527 | component = next_component; |
|
528 | 528 | children = &mut child.children; |
|
529 | 529 | } else { |
|
530 | 530 | return Ok(Some(child)); |
|
531 | 531 | } |
|
532 | 532 | } else { |
|
533 | 533 | return Ok(None); |
|
534 | 534 | } |
|
535 | 535 | } |
|
536 | 536 | } |
|
537 | 537 | |
|
538 | 538 | pub(super) fn get_or_insert<'tree, 'path>( |
|
539 | 539 | &'tree mut self, |
|
540 | 540 | path: &HgPath, |
|
541 | 541 | ) -> Result<&'tree mut Node<'on_disk>, DirstateV2ParseError> { |
|
542 | 542 | Self::get_or_insert_node( |
|
543 | 543 | self.on_disk, |
|
544 | 544 | &mut self.unreachable_bytes, |
|
545 | 545 | &mut self.root, |
|
546 | 546 | path, |
|
547 | 547 | WithBasename::to_cow_owned, |
|
548 | 548 | |_| {}, |
|
549 | 549 | ) |
|
550 | 550 | } |
|
551 | 551 | |
|
552 | 552 | fn get_or_insert_node<'tree, 'path>( |
|
553 | 553 | on_disk: &'on_disk [u8], |
|
554 | 554 | unreachable_bytes: &mut u32, |
|
555 | 555 | root: &'tree mut ChildNodes<'on_disk>, |
|
556 | 556 | path: &'path HgPath, |
|
557 | 557 | to_cow: impl Fn( |
|
558 | 558 | WithBasename<&'path HgPath>, |
|
559 | 559 | ) -> WithBasename<Cow<'on_disk, HgPath>>, |
|
560 | 560 | mut each_ancestor: impl FnMut(&mut Node), |
|
561 | 561 | ) -> Result<&'tree mut Node<'on_disk>, DirstateV2ParseError> { |
|
562 | 562 | let mut child_nodes = root; |
|
563 | 563 | let mut inclusive_ancestor_paths = |
|
564 | 564 | WithBasename::inclusive_ancestors_of(path); |
|
565 | 565 | let mut ancestor_path = inclusive_ancestor_paths |
|
566 | 566 | .next() |
|
567 | 567 | .expect("expected at least one inclusive ancestor"); |
|
568 | 568 | loop { |
|
569 | 569 | // TODO: can we avoid allocating an owned key in cases where the |
|
570 | 570 | // map already contains that key, without introducing double |
|
571 | 571 | // lookup? |
|
572 | 572 | let child_node = child_nodes |
|
573 | 573 | .make_mut(on_disk, unreachable_bytes)? |
|
574 | 574 | .entry(to_cow(ancestor_path)) |
|
575 | 575 | .or_default(); |
|
576 | 576 | if let Some(next) = inclusive_ancestor_paths.next() { |
|
577 | 577 | each_ancestor(child_node); |
|
578 | 578 | ancestor_path = next; |
|
579 | 579 | child_nodes = &mut child_node.children; |
|
580 | 580 | } else { |
|
581 | 581 | return Ok(child_node); |
|
582 | 582 | } |
|
583 | 583 | } |
|
584 | 584 | } |
|
585 | 585 | |
|
586 | 586 | fn add_or_remove_file( |
|
587 | 587 | &mut self, |
|
588 | 588 | path: &HgPath, |
|
589 | 589 | old_state: Option<EntryState>, |
|
590 | 590 | new_entry: DirstateEntry, |
|
591 | 591 | ) -> Result<(), DirstateV2ParseError> { |
|
592 | 592 | let had_entry = old_state.is_some(); |
|
593 | 593 | let was_tracked = old_state.map_or(false, |s| s.is_tracked()); |
|
594 | 594 | let tracked_count_increment = |
|
595 | 595 | match (was_tracked, new_entry.state().is_tracked()) { |
|
596 | 596 | (false, true) => 1, |
|
597 | 597 | (true, false) => -1, |
|
598 | 598 | _ => 0, |
|
599 | 599 | }; |
|
600 | 600 | |
|
601 | 601 | let node = Self::get_or_insert_node( |
|
602 | 602 | self.on_disk, |
|
603 | 603 | &mut self.unreachable_bytes, |
|
604 | 604 | &mut self.root, |
|
605 | 605 | path, |
|
606 | 606 | WithBasename::to_cow_owned, |
|
607 | 607 | |ancestor| { |
|
608 | 608 | if !had_entry { |
|
609 | 609 | ancestor.descendants_with_entry_count += 1; |
|
610 | 610 | } |
|
611 | 611 | |
|
612 | 612 | // We can’t use `+= increment` because the counter is unsigned, |
|
613 | 613 | // and we want debug builds to detect accidental underflow |
|
614 | 614 | // through zero |
|
615 | 615 | match tracked_count_increment { |
|
616 | 616 | 1 => ancestor.tracked_descendants_count += 1, |
|
617 | 617 | -1 => ancestor.tracked_descendants_count -= 1, |
|
618 | 618 | _ => {} |
|
619 | 619 | } |
|
620 | 620 | }, |
|
621 | 621 | )?; |
|
622 | 622 | if !had_entry { |
|
623 | 623 | self.nodes_with_entry_count += 1 |
|
624 | 624 | } |
|
625 | 625 | node.data = NodeData::Entry(new_entry); |
|
626 | 626 | Ok(()) |
|
627 | 627 | } |
|
628 | 628 | |
|
629 | 629 | fn iter_nodes<'tree>( |
|
630 | 630 | &'tree self, |
|
631 | 631 | ) -> impl Iterator< |
|
632 | 632 | Item = Result<NodeRef<'tree, 'on_disk>, DirstateV2ParseError>, |
|
633 | 633 | > + 'tree { |
|
634 | 634 | // Depth first tree traversal. |
|
635 | 635 | // |
|
636 | 636 | // If we could afford internal iteration and recursion, |
|
637 | 637 | // this would look like: |
|
638 | 638 | // |
|
639 | 639 | // ``` |
|
640 | 640 | // fn traverse_children( |
|
641 | 641 | // children: &ChildNodes, |
|
642 | 642 | // each: &mut impl FnMut(&Node), |
|
643 | 643 | // ) { |
|
644 | 644 | // for child in children.values() { |
|
645 | 645 | // traverse_children(&child.children, each); |
|
646 | 646 | // each(child); |
|
647 | 647 | // } |
|
648 | 648 | // } |
|
649 | 649 | // ``` |
|
650 | 650 | // |
|
651 | 651 | // However we want an external iterator and therefore can’t use the |
|
652 | 652 | // call stack. Use an explicit stack instead: |
|
653 | 653 | let mut stack = Vec::new(); |
|
654 | 654 | let mut iter = self.root.as_ref().iter(); |
|
655 | 655 | std::iter::from_fn(move || { |
|
656 | 656 | while let Some(child_node) = iter.next() { |
|
657 | 657 | let children = match child_node.children(self.on_disk) { |
|
658 | 658 | Ok(children) => children, |
|
659 | 659 | Err(error) => return Some(Err(error)), |
|
660 | 660 | }; |
|
661 | 661 | // Pseudo-recursion |
|
662 | 662 | let new_iter = children.iter(); |
|
663 | 663 | let old_iter = std::mem::replace(&mut iter, new_iter); |
|
664 | 664 | stack.push((child_node, old_iter)); |
|
665 | 665 | } |
|
666 | 666 | // Found the end of a `children.iter()` iterator. |
|
667 | 667 | if let Some((child_node, next_iter)) = stack.pop() { |
|
668 | 668 | // "Return" from pseudo-recursion by restoring state from the |
|
669 | 669 | // explicit stack |
|
670 | 670 | iter = next_iter; |
|
671 | 671 | |
|
672 | 672 | Some(Ok(child_node)) |
|
673 | 673 | } else { |
|
674 | 674 | // Reached the bottom of the stack, we’re done |
|
675 | 675 | None |
|
676 | 676 | } |
|
677 | 677 | }) |
|
678 | 678 | } |
|
679 | 679 | |
|
680 | 680 | fn count_dropped_path(unreachable_bytes: &mut u32, path: &Cow<HgPath>) { |
|
681 | 681 | if let Cow::Borrowed(path) = path { |
|
682 | 682 | *unreachable_bytes += path.len() as u32 |
|
683 | 683 | } |
|
684 | 684 | } |
|
685 | 685 | } |
|
686 | 686 | |
|
687 | 687 | /// Like `Iterator::filter_map`, but over a fallible iterator of `Result`s. |
|
688 | 688 | /// |
|
689 | 689 | /// The callback is only called for incoming `Ok` values. Errors are passed |
|
690 | 690 | /// through as-is. In order to let it use the `?` operator the callback is |
|
691 | 691 | /// expected to return a `Result` of `Option`, instead of an `Option` of |
|
692 | 692 | /// `Result`. |
|
693 | 693 | fn filter_map_results<'a, I, F, A, B, E>( |
|
694 | 694 | iter: I, |
|
695 | 695 | f: F, |
|
696 | 696 | ) -> impl Iterator<Item = Result<B, E>> + 'a |
|
697 | 697 | where |
|
698 | 698 | I: Iterator<Item = Result<A, E>> + 'a, |
|
699 | 699 | F: Fn(A) -> Result<Option<B>, E> + 'a, |
|
700 | 700 | { |
|
701 | 701 | iter.filter_map(move |result| match result { |
|
702 | 702 | Ok(node) => f(node).transpose(), |
|
703 | 703 | Err(e) => Some(Err(e)), |
|
704 | 704 | }) |
|
705 | 705 | } |
|
706 | 706 | |
|
707 | 707 | impl OwningDirstateMap { |
|
708 | 708 | pub fn clear(&mut self) { |
|
709 | 709 | let map = self.get_map_mut(); |
|
710 | 710 | map.root = Default::default(); |
|
711 | 711 | map.nodes_with_entry_count = 0; |
|
712 | 712 | map.nodes_with_copy_source_count = 0; |
|
713 | 713 | } |
|
714 | 714 | |
|
715 | 715 | pub fn set_entry( |
|
716 | 716 | &mut self, |
|
717 | 717 | filename: &HgPath, |
|
718 | 718 | entry: DirstateEntry, |
|
719 | 719 | ) -> Result<(), DirstateV2ParseError> { |
|
720 | 720 | let map = self.get_map_mut(); |
|
721 | 721 | map.get_or_insert(&filename)?.data = NodeData::Entry(entry); |
|
722 | 722 | Ok(()) |
|
723 | 723 | } |
|
724 | 724 | |
|
725 | 725 | pub fn add_file( |
|
726 | 726 | &mut self, |
|
727 | 727 | filename: &HgPath, |
|
728 | 728 | entry: DirstateEntry, |
|
729 | 729 | ) -> Result<(), DirstateError> { |
|
730 | 730 | let old_state = self.get(filename)?.map(|e| e.state()); |
|
731 | 731 | let map = self.get_map_mut(); |
|
732 | 732 | Ok(map.add_or_remove_file(filename, old_state, entry)?) |
|
733 | 733 | } |
|
734 | 734 | |
|
735 | 735 | pub fn remove_file( |
|
736 | 736 | &mut self, |
|
737 | 737 | filename: &HgPath, |
|
738 | 738 | in_merge: bool, |
|
739 | 739 | ) -> Result<(), DirstateError> { |
|
740 | 740 | let old_entry_opt = self.get(filename)?; |
|
741 | 741 | let old_state = old_entry_opt.map(|e| e.state()); |
|
742 | 742 | let mut size = 0; |
|
743 | 743 | if in_merge { |
|
744 | 744 | // XXX we should not be able to have 'm' state and 'FROM_P2' if not |
|
745 | 745 | // during a merge. So I (marmoute) am not sure we need the |
|
746 | 746 | // conditionnal at all. Adding double checking this with assert |
|
747 | 747 | // would be nice. |
|
748 | 748 | if let Some(old_entry) = old_entry_opt { |
|
749 | 749 | // backup the previous state |
|
750 | 750 | if old_entry.state() == EntryState::Merged { |
|
751 | 751 | size = SIZE_NON_NORMAL; |
|
752 | 752 | } else if old_entry.state() == EntryState::Normal |
|
753 | 753 | && old_entry.size() == SIZE_FROM_OTHER_PARENT |
|
754 | 754 | { |
|
755 | 755 | // other parent |
|
756 | 756 | size = SIZE_FROM_OTHER_PARENT; |
|
757 | 757 | } |
|
758 | 758 | } |
|
759 | 759 | } |
|
760 | 760 | if size == 0 { |
|
761 | 761 | self.copy_map_remove(filename)?; |
|
762 | 762 | } |
|
763 | 763 | let map = self.get_map_mut(); |
|
764 | 764 | let entry = DirstateEntry::new_removed(size); |
|
765 | 765 | Ok(map.add_or_remove_file(filename, old_state, entry)?) |
|
766 | 766 | } |
|
767 | 767 | |
|
768 | 768 | pub fn drop_entry_and_copy_source( |
|
769 | 769 | &mut self, |
|
770 | 770 | filename: &HgPath, |
|
771 | 771 | ) -> Result<(), DirstateError> { |
|
772 | 772 | let was_tracked = self |
|
773 | 773 | .get(filename)? |
|
774 | 774 | .map_or(false, |e| e.state().is_tracked()); |
|
775 | 775 | let map = self.get_map_mut(); |
|
776 | 776 | struct Dropped { |
|
777 | 777 | was_tracked: bool, |
|
778 | 778 | had_entry: bool, |
|
779 | 779 | had_copy_source: bool, |
|
780 | 780 | } |
|
781 | 781 | |
|
782 | 782 | /// If this returns `Ok(Some((dropped, removed)))`, then |
|
783 | 783 | /// |
|
784 | 784 | /// * `dropped` is about the leaf node that was at `filename` |
|
785 | 785 | /// * `removed` is whether this particular level of recursion just |
|
786 | 786 | /// removed a node in `nodes`. |
|
787 | 787 | fn recur<'on_disk>( |
|
788 | 788 | on_disk: &'on_disk [u8], |
|
789 | 789 | unreachable_bytes: &mut u32, |
|
790 | 790 | nodes: &mut ChildNodes<'on_disk>, |
|
791 | 791 | path: &HgPath, |
|
792 | 792 | ) -> Result<Option<(Dropped, bool)>, DirstateV2ParseError> { |
|
793 | 793 | let (first_path_component, rest_of_path) = |
|
794 | 794 | path.split_first_component(); |
|
795 | 795 | let nodes = nodes.make_mut(on_disk, unreachable_bytes)?; |
|
796 | 796 | let node = if let Some(node) = nodes.get_mut(first_path_component) |
|
797 | 797 | { |
|
798 | 798 | node |
|
799 | 799 | } else { |
|
800 | 800 | return Ok(None); |
|
801 | 801 | }; |
|
802 | 802 | let dropped; |
|
803 | 803 | if let Some(rest) = rest_of_path { |
|
804 | 804 | if let Some((d, removed)) = recur( |
|
805 | 805 | on_disk, |
|
806 | 806 | unreachable_bytes, |
|
807 | 807 | &mut node.children, |
|
808 | 808 | rest, |
|
809 | 809 | )? { |
|
810 | 810 | dropped = d; |
|
811 | 811 | if dropped.had_entry { |
|
812 | 812 | node.descendants_with_entry_count -= 1; |
|
813 | 813 | } |
|
814 | 814 | if dropped.was_tracked { |
|
815 | 815 | node.tracked_descendants_count -= 1; |
|
816 | 816 | } |
|
817 | 817 | |
|
818 | 818 | // Directory caches must be invalidated when removing a |
|
819 | 819 | // child node |
|
820 | 820 | if removed { |
|
821 | 821 | if let NodeData::CachedDirectory { .. } = &node.data { |
|
822 | 822 | node.data = NodeData::None |
|
823 | 823 | } |
|
824 | 824 | } |
|
825 | 825 | } else { |
|
826 | 826 | return Ok(None); |
|
827 | 827 | } |
|
828 | 828 | } else { |
|
829 | 829 | let had_entry = node.data.has_entry(); |
|
830 | 830 | if had_entry { |
|
831 | 831 | node.data = NodeData::None |
|
832 | 832 | } |
|
833 | 833 | if let Some(source) = &node.copy_source { |
|
834 | 834 | DirstateMap::count_dropped_path(unreachable_bytes, source); |
|
835 | 835 | node.copy_source = None |
|
836 | 836 | } |
|
837 | 837 | dropped = Dropped { |
|
838 | 838 | was_tracked: node |
|
839 | 839 | .data |
|
840 | 840 | .as_entry() |
|
841 | 841 | .map_or(false, |entry| entry.state().is_tracked()), |
|
842 | 842 | had_entry, |
|
843 | 843 | had_copy_source: node.copy_source.take().is_some(), |
|
844 | 844 | }; |
|
845 | 845 | } |
|
846 | 846 | // After recursion, for both leaf (rest_of_path is None) nodes and |
|
847 | 847 | // parent nodes, remove a node if it just became empty. |
|
848 | 848 | let remove = !node.data.has_entry() |
|
849 | 849 | && node.copy_source.is_none() |
|
850 | 850 | && node.children.is_empty(); |
|
851 | 851 | if remove { |
|
852 | 852 | let (key, _) = |
|
853 | 853 | nodes.remove_entry(first_path_component).unwrap(); |
|
854 | 854 | DirstateMap::count_dropped_path( |
|
855 | 855 | unreachable_bytes, |
|
856 | 856 | key.full_path(), |
|
857 | 857 | ) |
|
858 | 858 | } |
|
859 | 859 | Ok(Some((dropped, remove))) |
|
860 | 860 | } |
|
861 | 861 | |
|
862 | 862 | if let Some((dropped, _removed)) = recur( |
|
863 | 863 | map.on_disk, |
|
864 | 864 | &mut map.unreachable_bytes, |
|
865 | 865 | &mut map.root, |
|
866 | 866 | filename, |
|
867 | 867 | )? { |
|
868 | 868 | if dropped.had_entry { |
|
869 | 869 | map.nodes_with_entry_count -= 1 |
|
870 | 870 | } |
|
871 | 871 | if dropped.had_copy_source { |
|
872 | 872 | map.nodes_with_copy_source_count -= 1 |
|
873 | 873 | } |
|
874 | 874 | } else { |
|
875 | 875 | debug_assert!(!was_tracked); |
|
876 | 876 | } |
|
877 | 877 | Ok(()) |
|
878 | 878 | } |
|
879 | 879 | |
|
880 | 880 | pub fn has_tracked_dir( |
|
881 | 881 | &mut self, |
|
882 | 882 | directory: &HgPath, |
|
883 | 883 | ) -> Result<bool, DirstateError> { |
|
884 | 884 | let map = self.get_map_mut(); |
|
885 | 885 | if let Some(node) = map.get_node(directory)? { |
|
886 | 886 | // A node without a `DirstateEntry` was created to hold child |
|
887 | 887 | // nodes, and is therefore a directory. |
|
888 | 888 | let state = node.state()?; |
|
889 | 889 | Ok(state.is_none() && node.tracked_descendants_count() > 0) |
|
890 | 890 | } else { |
|
891 | 891 | Ok(false) |
|
892 | 892 | } |
|
893 | 893 | } |
|
894 | 894 | |
|
895 | 895 | pub fn has_dir( |
|
896 | 896 | &mut self, |
|
897 | 897 | directory: &HgPath, |
|
898 | 898 | ) -> Result<bool, DirstateError> { |
|
899 | 899 | let map = self.get_map_mut(); |
|
900 | 900 | if let Some(node) = map.get_node(directory)? { |
|
901 | 901 | // A node without a `DirstateEntry` was created to hold child |
|
902 | 902 | // nodes, and is therefore a directory. |
|
903 | 903 | let state = node.state()?; |
|
904 | 904 | Ok(state.is_none() && node.descendants_with_entry_count() > 0) |
|
905 | 905 | } else { |
|
906 | 906 | Ok(false) |
|
907 | 907 | } |
|
908 | 908 | } |
|
909 | 909 | |
|
910 | 910 | #[timed] |
|
911 | 911 | pub fn pack_v1( |
|
912 | 912 | &self, |
|
913 | 913 | parents: DirstateParents, |
|
914 | 914 | ) -> Result<Vec<u8>, DirstateError> { |
|
915 | 915 | let map = self.get_map(); |
|
916 | 916 | // Optizimation (to be measured?): pre-compute size to avoid `Vec` |
|
917 | 917 | // reallocations |
|
918 | 918 | let mut size = parents.as_bytes().len(); |
|
919 | 919 | for node in map.iter_nodes() { |
|
920 | 920 | let node = node?; |
|
921 | 921 | if node.entry()?.is_some() { |
|
922 | 922 | size += packed_entry_size( |
|
923 | 923 | node.full_path(map.on_disk)?, |
|
924 | 924 | node.copy_source(map.on_disk)?, |
|
925 | 925 | ); |
|
926 | 926 | } |
|
927 | 927 | } |
|
928 | 928 | |
|
929 | 929 | let mut packed = Vec::with_capacity(size); |
|
930 | 930 | packed.extend(parents.as_bytes()); |
|
931 | 931 | |
|
932 | 932 | for node in map.iter_nodes() { |
|
933 | 933 | let node = node?; |
|
934 | 934 | if let Some(entry) = node.entry()? { |
|
935 | 935 | pack_entry( |
|
936 | 936 | node.full_path(map.on_disk)?, |
|
937 | 937 | &entry, |
|
938 | 938 | node.copy_source(map.on_disk)?, |
|
939 | 939 | &mut packed, |
|
940 | 940 | ); |
|
941 | 941 | } |
|
942 | 942 | } |
|
943 | 943 | Ok(packed) |
|
944 | 944 | } |
|
945 | 945 | |
|
946 | 946 | /// Returns new data and metadata together with whether that data should be |
|
947 | 947 | /// appended to the existing data file whose content is at |
|
948 | 948 | /// `map.on_disk` (true), instead of written to a new data file |
|
949 | 949 | /// (false). |
|
950 | 950 | #[timed] |
|
951 | 951 | pub fn pack_v2( |
|
952 | 952 | &self, |
|
953 | 953 | can_append: bool, |
|
954 |
) -> Result<(Vec<u8>, |
|
|
954 | ) -> Result<(Vec<u8>, on_disk::TreeMetadata, bool), DirstateError> { | |
|
955 | 955 | let map = self.get_map(); |
|
956 | 956 | on_disk::write(map, can_append) |
|
957 | 957 | } |
|
958 | 958 | |
|
959 | 959 | pub fn status<'a>( |
|
960 | 960 | &'a mut self, |
|
961 | 961 | matcher: &'a (dyn Matcher + Sync), |
|
962 | 962 | root_dir: PathBuf, |
|
963 | 963 | ignore_files: Vec<PathBuf>, |
|
964 | 964 | options: StatusOptions, |
|
965 | 965 | ) -> Result<(DirstateStatus<'a>, Vec<PatternFileWarning>), StatusError> |
|
966 | 966 | { |
|
967 | 967 | let map = self.get_map_mut(); |
|
968 | 968 | super::status::status(map, matcher, root_dir, ignore_files, options) |
|
969 | 969 | } |
|
970 | 970 | |
|
971 | 971 | pub fn copy_map_len(&self) -> usize { |
|
972 | 972 | let map = self.get_map(); |
|
973 | 973 | map.nodes_with_copy_source_count as usize |
|
974 | 974 | } |
|
975 | 975 | |
|
976 | 976 | pub fn copy_map_iter(&self) -> CopyMapIter<'_> { |
|
977 | 977 | let map = self.get_map(); |
|
978 | 978 | Box::new(filter_map_results(map.iter_nodes(), move |node| { |
|
979 | 979 | Ok(if let Some(source) = node.copy_source(map.on_disk)? { |
|
980 | 980 | Some((node.full_path(map.on_disk)?, source)) |
|
981 | 981 | } else { |
|
982 | 982 | None |
|
983 | 983 | }) |
|
984 | 984 | })) |
|
985 | 985 | } |
|
986 | 986 | |
|
987 | 987 | pub fn copy_map_contains_key( |
|
988 | 988 | &self, |
|
989 | 989 | key: &HgPath, |
|
990 | 990 | ) -> Result<bool, DirstateV2ParseError> { |
|
991 | 991 | let map = self.get_map(); |
|
992 | 992 | Ok(if let Some(node) = map.get_node(key)? { |
|
993 | 993 | node.has_copy_source() |
|
994 | 994 | } else { |
|
995 | 995 | false |
|
996 | 996 | }) |
|
997 | 997 | } |
|
998 | 998 | |
|
999 | 999 | pub fn copy_map_get( |
|
1000 | 1000 | &self, |
|
1001 | 1001 | key: &HgPath, |
|
1002 | 1002 | ) -> Result<Option<&HgPath>, DirstateV2ParseError> { |
|
1003 | 1003 | let map = self.get_map(); |
|
1004 | 1004 | if let Some(node) = map.get_node(key)? { |
|
1005 | 1005 | if let Some(source) = node.copy_source(map.on_disk)? { |
|
1006 | 1006 | return Ok(Some(source)); |
|
1007 | 1007 | } |
|
1008 | 1008 | } |
|
1009 | 1009 | Ok(None) |
|
1010 | 1010 | } |
|
1011 | 1011 | |
|
1012 | 1012 | pub fn copy_map_remove( |
|
1013 | 1013 | &mut self, |
|
1014 | 1014 | key: &HgPath, |
|
1015 | 1015 | ) -> Result<Option<HgPathBuf>, DirstateV2ParseError> { |
|
1016 | 1016 | let map = self.get_map_mut(); |
|
1017 | 1017 | let count = &mut map.nodes_with_copy_source_count; |
|
1018 | 1018 | let unreachable_bytes = &mut map.unreachable_bytes; |
|
1019 | 1019 | Ok(DirstateMap::get_node_mut( |
|
1020 | 1020 | map.on_disk, |
|
1021 | 1021 | unreachable_bytes, |
|
1022 | 1022 | &mut map.root, |
|
1023 | 1023 | key, |
|
1024 | 1024 | )? |
|
1025 | 1025 | .and_then(|node| { |
|
1026 | 1026 | if let Some(source) = &node.copy_source { |
|
1027 | 1027 | *count -= 1; |
|
1028 | 1028 | DirstateMap::count_dropped_path(unreachable_bytes, source); |
|
1029 | 1029 | } |
|
1030 | 1030 | node.copy_source.take().map(Cow::into_owned) |
|
1031 | 1031 | })) |
|
1032 | 1032 | } |
|
1033 | 1033 | |
|
1034 | 1034 | pub fn copy_map_insert( |
|
1035 | 1035 | &mut self, |
|
1036 | 1036 | key: HgPathBuf, |
|
1037 | 1037 | value: HgPathBuf, |
|
1038 | 1038 | ) -> Result<Option<HgPathBuf>, DirstateV2ParseError> { |
|
1039 | 1039 | let map = self.get_map_mut(); |
|
1040 | 1040 | let node = DirstateMap::get_or_insert_node( |
|
1041 | 1041 | map.on_disk, |
|
1042 | 1042 | &mut map.unreachable_bytes, |
|
1043 | 1043 | &mut map.root, |
|
1044 | 1044 | &key, |
|
1045 | 1045 | WithBasename::to_cow_owned, |
|
1046 | 1046 | |_ancestor| {}, |
|
1047 | 1047 | )?; |
|
1048 | 1048 | if node.copy_source.is_none() { |
|
1049 | 1049 | map.nodes_with_copy_source_count += 1 |
|
1050 | 1050 | } |
|
1051 | 1051 | Ok(node.copy_source.replace(value.into()).map(Cow::into_owned)) |
|
1052 | 1052 | } |
|
1053 | 1053 | |
|
1054 | 1054 | pub fn len(&self) -> usize { |
|
1055 | 1055 | let map = self.get_map(); |
|
1056 | 1056 | map.nodes_with_entry_count as usize |
|
1057 | 1057 | } |
|
1058 | 1058 | |
|
1059 | 1059 | pub fn contains_key( |
|
1060 | 1060 | &self, |
|
1061 | 1061 | key: &HgPath, |
|
1062 | 1062 | ) -> Result<bool, DirstateV2ParseError> { |
|
1063 | 1063 | Ok(self.get(key)?.is_some()) |
|
1064 | 1064 | } |
|
1065 | 1065 | |
|
1066 | 1066 | pub fn get( |
|
1067 | 1067 | &self, |
|
1068 | 1068 | key: &HgPath, |
|
1069 | 1069 | ) -> Result<Option<DirstateEntry>, DirstateV2ParseError> { |
|
1070 | 1070 | let map = self.get_map(); |
|
1071 | 1071 | Ok(if let Some(node) = map.get_node(key)? { |
|
1072 | 1072 | node.entry()? |
|
1073 | 1073 | } else { |
|
1074 | 1074 | None |
|
1075 | 1075 | }) |
|
1076 | 1076 | } |
|
1077 | 1077 | |
|
1078 | 1078 | pub fn iter(&self) -> StateMapIter<'_> { |
|
1079 | 1079 | let map = self.get_map(); |
|
1080 | 1080 | Box::new(filter_map_results(map.iter_nodes(), move |node| { |
|
1081 | 1081 | Ok(if let Some(entry) = node.entry()? { |
|
1082 | 1082 | Some((node.full_path(map.on_disk)?, entry)) |
|
1083 | 1083 | } else { |
|
1084 | 1084 | None |
|
1085 | 1085 | }) |
|
1086 | 1086 | })) |
|
1087 | 1087 | } |
|
1088 | 1088 | |
|
1089 | 1089 | pub fn iter_tracked_dirs( |
|
1090 | 1090 | &mut self, |
|
1091 | 1091 | ) -> Result< |
|
1092 | 1092 | Box< |
|
1093 | 1093 | dyn Iterator<Item = Result<&HgPath, DirstateV2ParseError>> |
|
1094 | 1094 | + Send |
|
1095 | 1095 | + '_, |
|
1096 | 1096 | >, |
|
1097 | 1097 | DirstateError, |
|
1098 | 1098 | > { |
|
1099 | 1099 | let map = self.get_map_mut(); |
|
1100 | 1100 | let on_disk = map.on_disk; |
|
1101 | 1101 | Ok(Box::new(filter_map_results( |
|
1102 | 1102 | map.iter_nodes(), |
|
1103 | 1103 | move |node| { |
|
1104 | 1104 | Ok(if node.tracked_descendants_count() > 0 { |
|
1105 | 1105 | Some(node.full_path(on_disk)?) |
|
1106 | 1106 | } else { |
|
1107 | 1107 | None |
|
1108 | 1108 | }) |
|
1109 | 1109 | }, |
|
1110 | 1110 | ))) |
|
1111 | 1111 | } |
|
1112 | 1112 | |
|
1113 | 1113 | pub fn debug_iter( |
|
1114 | 1114 | &self, |
|
1115 | 1115 | all: bool, |
|
1116 | 1116 | ) -> Box< |
|
1117 | 1117 | dyn Iterator< |
|
1118 | 1118 | Item = Result< |
|
1119 | 1119 | (&HgPath, (u8, i32, i32, i32)), |
|
1120 | 1120 | DirstateV2ParseError, |
|
1121 | 1121 | >, |
|
1122 | 1122 | > + Send |
|
1123 | 1123 | + '_, |
|
1124 | 1124 | > { |
|
1125 | 1125 | let map = self.get_map(); |
|
1126 | 1126 | Box::new(filter_map_results(map.iter_nodes(), move |node| { |
|
1127 | 1127 | let debug_tuple = if let Some(entry) = node.entry()? { |
|
1128 | 1128 | entry.debug_tuple() |
|
1129 | 1129 | } else if !all { |
|
1130 | 1130 | return Ok(None); |
|
1131 | 1131 | } else if let Some(mtime) = node.cached_directory_mtime()? { |
|
1132 | 1132 | (b' ', 0, -1, mtime.truncated_seconds() as i32) |
|
1133 | 1133 | } else { |
|
1134 | 1134 | (b' ', 0, -1, -1) |
|
1135 | 1135 | }; |
|
1136 | 1136 | Ok(Some((node.full_path(map.on_disk)?, debug_tuple))) |
|
1137 | 1137 | })) |
|
1138 | 1138 | } |
|
1139 | 1139 | } |
@@ -1,791 +1,837 b'' | |||
|
1 | 1 | //! The "version 2" disk representation of the dirstate |
|
2 | 2 | //! |
|
3 | 3 | //! See `mercurial/helptext/internals/dirstate-v2.txt` |
|
4 | 4 | |
|
5 | 5 | use crate::dirstate::TruncatedTimestamp; |
|
6 | 6 | use crate::dirstate_tree::dirstate_map::{self, DirstateMap, NodeRef}; |
|
7 | 7 | use crate::dirstate_tree::path_with_basename::WithBasename; |
|
8 | 8 | use crate::errors::HgError; |
|
9 | 9 | use crate::utils::hg_path::HgPath; |
|
10 | 10 | use crate::DirstateEntry; |
|
11 | 11 | use crate::DirstateError; |
|
12 | 12 | use crate::DirstateParents; |
|
13 | 13 | use bitflags::bitflags; |
|
14 | 14 | use bytes_cast::unaligned::{U16Be, U32Be}; |
|
15 | 15 | use bytes_cast::BytesCast; |
|
16 | 16 | use format_bytes::format_bytes; |
|
17 | use rand::Rng; | |
|
17 | 18 | use std::borrow::Cow; |
|
18 | 19 | use std::convert::{TryFrom, TryInto}; |
|
20 | use std::fmt::Write; | |
|
19 | 21 | |
|
20 | 22 | /// Added at the start of `.hg/dirstate` when the "v2" format is used. |
|
21 | 23 | /// This a redundant sanity check more than an actual "magic number" since |
|
22 | 24 | /// `.hg/requires` already governs which format should be used. |
|
23 | 25 | pub const V2_FORMAT_MARKER: &[u8; 12] = b"dirstate-v2\n"; |
|
24 | 26 | |
|
25 | 27 | /// Keep space for 256-bit hashes |
|
26 | 28 | const STORED_NODE_ID_BYTES: usize = 32; |
|
27 | 29 | |
|
28 | 30 | /// … even though only 160 bits are used for now, with SHA-1 |
|
29 | 31 | const USED_NODE_ID_BYTES: usize = 20; |
|
30 | 32 | |
|
31 | 33 | pub(super) const IGNORE_PATTERNS_HASH_LEN: usize = 20; |
|
32 | 34 | pub(super) type IgnorePatternsHash = [u8; IGNORE_PATTERNS_HASH_LEN]; |
|
33 | 35 | |
|
34 | 36 | /// Must match constants of the same names in `mercurial/dirstateutils/v2.py` |
|
35 | 37 | const TREE_METADATA_SIZE: usize = 44; |
|
36 | 38 | const NODE_SIZE: usize = 44; |
|
37 | 39 | |
|
38 | 40 | /// Make sure that size-affecting changes are made knowingly |
|
39 | 41 | #[allow(unused)] |
|
40 | 42 | fn static_assert_size_of() { |
|
41 | 43 | let _ = std::mem::transmute::<TreeMetadata, [u8; TREE_METADATA_SIZE]>; |
|
42 | 44 | let _ = std::mem::transmute::<DocketHeader, [u8; TREE_METADATA_SIZE + 81]>; |
|
43 | 45 | let _ = std::mem::transmute::<Node, [u8; NODE_SIZE]>; |
|
44 | 46 | } |
|
45 | 47 | |
|
46 | 48 | // Must match `HEADER` in `mercurial/dirstateutils/docket.py` |
|
47 | 49 | #[derive(BytesCast)] |
|
48 | 50 | #[repr(C)] |
|
49 | 51 | struct DocketHeader { |
|
50 | 52 | marker: [u8; V2_FORMAT_MARKER.len()], |
|
51 | 53 | parent_1: [u8; STORED_NODE_ID_BYTES], |
|
52 | 54 | parent_2: [u8; STORED_NODE_ID_BYTES], |
|
53 | 55 | |
|
54 | 56 | metadata: TreeMetadata, |
|
55 | 57 | |
|
56 | 58 | /// Counted in bytes |
|
57 | 59 | data_size: Size, |
|
58 | 60 | |
|
59 | 61 | uuid_size: u8, |
|
60 | 62 | } |
|
61 | 63 | |
|
62 | 64 | pub struct Docket<'on_disk> { |
|
63 | 65 | header: &'on_disk DocketHeader, |
|
64 | 66 | pub uuid: &'on_disk [u8], |
|
65 | 67 | } |
|
66 | 68 | |
|
67 | 69 | /// Fields are documented in the *Tree metadata in the docket file* |
|
68 | 70 | /// section of `mercurial/helptext/internals/dirstate-v2.txt` |
|
69 | 71 | #[derive(BytesCast)] |
|
70 | 72 | #[repr(C)] |
|
71 | struct TreeMetadata { | |
|
73 | pub struct TreeMetadata { | |
|
72 | 74 | root_nodes: ChildNodes, |
|
73 | 75 | nodes_with_entry_count: Size, |
|
74 | 76 | nodes_with_copy_source_count: Size, |
|
75 | 77 | unreachable_bytes: Size, |
|
76 | 78 | unused: [u8; 4], |
|
77 | 79 | |
|
78 | 80 | /// See *Optional hash of ignore patterns* section of |
|
79 | 81 | /// `mercurial/helptext/internals/dirstate-v2.txt` |
|
80 | 82 | ignore_patterns_hash: IgnorePatternsHash, |
|
81 | 83 | } |
|
82 | 84 | |
|
83 | 85 | /// Fields are documented in the *The data file format* |
|
84 | 86 | /// section of `mercurial/helptext/internals/dirstate-v2.txt` |
|
85 | 87 | #[derive(BytesCast)] |
|
86 | 88 | #[repr(C)] |
|
87 | 89 | pub(super) struct Node { |
|
88 | 90 | full_path: PathSlice, |
|
89 | 91 | |
|
90 | 92 | /// In bytes from `self.full_path.start` |
|
91 | 93 | base_name_start: PathSize, |
|
92 | 94 | |
|
93 | 95 | copy_source: OptPathSlice, |
|
94 | 96 | children: ChildNodes, |
|
95 | 97 | pub(super) descendants_with_entry_count: Size, |
|
96 | 98 | pub(super) tracked_descendants_count: Size, |
|
97 | 99 | flags: U16Be, |
|
98 | 100 | size: U32Be, |
|
99 | 101 | mtime: PackedTruncatedTimestamp, |
|
100 | 102 | } |
|
101 | 103 | |
|
102 | 104 | bitflags! { |
|
103 | 105 | #[repr(C)] |
|
104 | 106 | struct Flags: u16 { |
|
105 | 107 | const WDIR_TRACKED = 1 << 0; |
|
106 | 108 | const P1_TRACKED = 1 << 1; |
|
107 | 109 | const P2_INFO = 1 << 2; |
|
108 | 110 | const MODE_EXEC_PERM = 1 << 3; |
|
109 | 111 | const MODE_IS_SYMLINK = 1 << 4; |
|
110 | 112 | const HAS_FALLBACK_EXEC = 1 << 5; |
|
111 | 113 | const FALLBACK_EXEC = 1 << 6; |
|
112 | 114 | const HAS_FALLBACK_SYMLINK = 1 << 7; |
|
113 | 115 | const FALLBACK_SYMLINK = 1 << 8; |
|
114 | 116 | const EXPECTED_STATE_IS_MODIFIED = 1 << 9; |
|
115 | 117 | const HAS_MODE_AND_SIZE = 1 <<10; |
|
116 | 118 | const HAS_MTIME = 1 <<11; |
|
117 | 119 | const MTIME_SECOND_AMBIGUOUS = 1 << 12; |
|
118 | 120 | const DIRECTORY = 1 <<13; |
|
119 | 121 | const ALL_UNKNOWN_RECORDED = 1 <<14; |
|
120 | 122 | const ALL_IGNORED_RECORDED = 1 <<15; |
|
121 | 123 | } |
|
122 | 124 | } |
|
123 | 125 | |
|
124 | 126 | /// Duration since the Unix epoch |
|
125 | 127 | #[derive(BytesCast, Copy, Clone)] |
|
126 | 128 | #[repr(C)] |
|
127 | 129 | struct PackedTruncatedTimestamp { |
|
128 | 130 | truncated_seconds: U32Be, |
|
129 | 131 | nanoseconds: U32Be, |
|
130 | 132 | } |
|
131 | 133 | |
|
132 | 134 | /// Counted in bytes from the start of the file |
|
133 | 135 | /// |
|
134 | 136 | /// NOTE: not supporting `.hg/dirstate` files larger than 4 GiB. |
|
135 | 137 | type Offset = U32Be; |
|
136 | 138 | |
|
137 | 139 | /// Counted in number of items |
|
138 | 140 | /// |
|
139 | 141 | /// NOTE: we choose not to support counting more than 4 billion nodes anywhere. |
|
140 | 142 | type Size = U32Be; |
|
141 | 143 | |
|
142 | 144 | /// Counted in bytes |
|
143 | 145 | /// |
|
144 | 146 | /// NOTE: we choose not to support file names/paths longer than 64 KiB. |
|
145 | 147 | type PathSize = U16Be; |
|
146 | 148 | |
|
147 | 149 | /// A contiguous sequence of `len` times `Node`, representing the child nodes |
|
148 | 150 | /// of either some other node or of the repository root. |
|
149 | 151 | /// |
|
150 | 152 | /// Always sorted by ascending `full_path`, to allow binary search. |
|
151 | 153 | /// Since nodes with the same parent nodes also have the same parent path, |
|
152 | 154 | /// only the `base_name`s need to be compared during binary search. |
|
153 | 155 | #[derive(BytesCast, Copy, Clone)] |
|
154 | 156 | #[repr(C)] |
|
155 | 157 | struct ChildNodes { |
|
156 | 158 | start: Offset, |
|
157 | 159 | len: Size, |
|
158 | 160 | } |
|
159 | 161 | |
|
160 | 162 | /// A `HgPath` of `len` bytes |
|
161 | 163 | #[derive(BytesCast, Copy, Clone)] |
|
162 | 164 | #[repr(C)] |
|
163 | 165 | struct PathSlice { |
|
164 | 166 | start: Offset, |
|
165 | 167 | len: PathSize, |
|
166 | 168 | } |
|
167 | 169 | |
|
168 | 170 | /// Either nothing if `start == 0`, or a `HgPath` of `len` bytes |
|
169 | 171 | type OptPathSlice = PathSlice; |
|
170 | 172 | |
|
171 | 173 | /// Unexpected file format found in `.hg/dirstate` with the "v2" format. |
|
172 | 174 | /// |
|
173 | 175 | /// This should only happen if Mercurial is buggy or a repository is corrupted. |
|
174 | 176 | #[derive(Debug)] |
|
175 | 177 | pub struct DirstateV2ParseError; |
|
176 | 178 | |
|
177 | 179 | impl From<DirstateV2ParseError> for HgError { |
|
178 | 180 | fn from(_: DirstateV2ParseError) -> Self { |
|
179 | 181 | HgError::corrupted("dirstate-v2 parse error") |
|
180 | 182 | } |
|
181 | 183 | } |
|
182 | 184 | |
|
183 | 185 | impl From<DirstateV2ParseError> for crate::DirstateError { |
|
184 | 186 | fn from(error: DirstateV2ParseError) -> Self { |
|
185 | 187 | HgError::from(error).into() |
|
186 | 188 | } |
|
187 | 189 | } |
|
188 | 190 | |
|
191 | impl TreeMetadata { | |
|
192 | pub fn as_bytes(&self) -> &[u8] { | |
|
193 | BytesCast::as_bytes(self) | |
|
194 | } | |
|
195 | } | |
|
196 | ||
|
189 | 197 | impl<'on_disk> Docket<'on_disk> { |
|
198 | /// Generate the identifier for a new data file | |
|
199 | /// | |
|
200 | /// TODO: support the `HGTEST_UUIDFILE` environment variable. | |
|
201 | /// See `mercurial/revlogutils/docket.py` | |
|
202 | pub fn new_uid() -> String { | |
|
203 | const ID_LENGTH: usize = 8; | |
|
204 | let mut id = String::with_capacity(ID_LENGTH); | |
|
205 | let mut rng = rand::thread_rng(); | |
|
206 | for _ in 0..ID_LENGTH { | |
|
207 | // One random hexadecimal digit. | |
|
208 | // `unwrap` never panics because `impl Write for String` | |
|
209 | // never returns an error. | |
|
210 | write!(&mut id, "{:x}", rng.gen_range(0, 16)).unwrap(); | |
|
211 | } | |
|
212 | id | |
|
213 | } | |
|
214 | ||
|
215 | pub fn serialize( | |
|
216 | parents: DirstateParents, | |
|
217 | tree_metadata: TreeMetadata, | |
|
218 | data_size: u64, | |
|
219 | uuid: &[u8], | |
|
220 | ) -> Result<Vec<u8>, std::num::TryFromIntError> { | |
|
221 | let header = DocketHeader { | |
|
222 | marker: *V2_FORMAT_MARKER, | |
|
223 | parent_1: parents.p1.pad_to_256_bits(), | |
|
224 | parent_2: parents.p2.pad_to_256_bits(), | |
|
225 | metadata: tree_metadata, | |
|
226 | data_size: u32::try_from(data_size)?.into(), | |
|
227 | uuid_size: uuid.len().try_into()?, | |
|
228 | }; | |
|
229 | let header = header.as_bytes(); | |
|
230 | let mut docket = Vec::with_capacity(header.len() + uuid.len()); | |
|
231 | docket.extend_from_slice(header); | |
|
232 | docket.extend_from_slice(uuid); | |
|
233 | Ok(docket) | |
|
234 | } | |
|
235 | ||
|
190 | 236 | pub fn parents(&self) -> DirstateParents { |
|
191 | 237 | use crate::Node; |
|
192 | 238 | let p1 = Node::try_from(&self.header.parent_1[..USED_NODE_ID_BYTES]) |
|
193 | 239 | .unwrap() |
|
194 | 240 | .clone(); |
|
195 | 241 | let p2 = Node::try_from(&self.header.parent_2[..USED_NODE_ID_BYTES]) |
|
196 | 242 | .unwrap() |
|
197 | 243 | .clone(); |
|
198 | 244 | DirstateParents { p1, p2 } |
|
199 | 245 | } |
|
200 | 246 | |
|
201 | 247 | pub fn tree_metadata(&self) -> &[u8] { |
|
202 | 248 | self.header.metadata.as_bytes() |
|
203 | 249 | } |
|
204 | 250 | |
|
205 | 251 | pub fn data_size(&self) -> usize { |
|
206 | 252 | // This `unwrap` could only panic on a 16-bit CPU |
|
207 | 253 | self.header.data_size.get().try_into().unwrap() |
|
208 | 254 | } |
|
209 | 255 | |
|
210 | 256 | pub fn data_filename(&self) -> String { |
|
211 | 257 | String::from_utf8(format_bytes!(b"dirstate.{}", self.uuid)).unwrap() |
|
212 | 258 | } |
|
213 | 259 | } |
|
214 | 260 | |
|
215 | 261 | pub fn read_docket( |
|
216 | 262 | on_disk: &[u8], |
|
217 | 263 | ) -> Result<Docket<'_>, DirstateV2ParseError> { |
|
218 | 264 | let (header, uuid) = |
|
219 | 265 | DocketHeader::from_bytes(on_disk).map_err(|_| DirstateV2ParseError)?; |
|
220 | 266 | let uuid_size = header.uuid_size as usize; |
|
221 | 267 | if header.marker == *V2_FORMAT_MARKER && uuid.len() == uuid_size { |
|
222 | 268 | Ok(Docket { header, uuid }) |
|
223 | 269 | } else { |
|
224 | 270 | Err(DirstateV2ParseError) |
|
225 | 271 | } |
|
226 | 272 | } |
|
227 | 273 | |
|
228 | 274 | pub(super) fn read<'on_disk>( |
|
229 | 275 | on_disk: &'on_disk [u8], |
|
230 | 276 | metadata: &[u8], |
|
231 | 277 | ) -> Result<DirstateMap<'on_disk>, DirstateV2ParseError> { |
|
232 | 278 | if on_disk.is_empty() { |
|
233 | 279 | return Ok(DirstateMap::empty(on_disk)); |
|
234 | 280 | } |
|
235 | 281 | let (meta, _) = TreeMetadata::from_bytes(metadata) |
|
236 | 282 | .map_err(|_| DirstateV2ParseError)?; |
|
237 | 283 | let dirstate_map = DirstateMap { |
|
238 | 284 | on_disk, |
|
239 | 285 | root: dirstate_map::ChildNodes::OnDisk(read_nodes( |
|
240 | 286 | on_disk, |
|
241 | 287 | meta.root_nodes, |
|
242 | 288 | )?), |
|
243 | 289 | nodes_with_entry_count: meta.nodes_with_entry_count.get(), |
|
244 | 290 | nodes_with_copy_source_count: meta.nodes_with_copy_source_count.get(), |
|
245 | 291 | ignore_patterns_hash: meta.ignore_patterns_hash, |
|
246 | 292 | unreachable_bytes: meta.unreachable_bytes.get(), |
|
247 | 293 | }; |
|
248 | 294 | Ok(dirstate_map) |
|
249 | 295 | } |
|
250 | 296 | |
|
251 | 297 | impl Node { |
|
252 | 298 | pub(super) fn full_path<'on_disk>( |
|
253 | 299 | &self, |
|
254 | 300 | on_disk: &'on_disk [u8], |
|
255 | 301 | ) -> Result<&'on_disk HgPath, DirstateV2ParseError> { |
|
256 | 302 | read_hg_path(on_disk, self.full_path) |
|
257 | 303 | } |
|
258 | 304 | |
|
259 | 305 | pub(super) fn base_name_start<'on_disk>( |
|
260 | 306 | &self, |
|
261 | 307 | ) -> Result<usize, DirstateV2ParseError> { |
|
262 | 308 | let start = self.base_name_start.get(); |
|
263 | 309 | if start < self.full_path.len.get() { |
|
264 | 310 | let start = usize::try_from(start) |
|
265 | 311 | // u32 -> usize, could only panic on a 16-bit CPU |
|
266 | 312 | .expect("dirstate-v2 base_name_start out of bounds"); |
|
267 | 313 | Ok(start) |
|
268 | 314 | } else { |
|
269 | 315 | Err(DirstateV2ParseError) |
|
270 | 316 | } |
|
271 | 317 | } |
|
272 | 318 | |
|
273 | 319 | pub(super) fn base_name<'on_disk>( |
|
274 | 320 | &self, |
|
275 | 321 | on_disk: &'on_disk [u8], |
|
276 | 322 | ) -> Result<&'on_disk HgPath, DirstateV2ParseError> { |
|
277 | 323 | let full_path = self.full_path(on_disk)?; |
|
278 | 324 | let base_name_start = self.base_name_start()?; |
|
279 | 325 | Ok(HgPath::new(&full_path.as_bytes()[base_name_start..])) |
|
280 | 326 | } |
|
281 | 327 | |
|
282 | 328 | pub(super) fn path<'on_disk>( |
|
283 | 329 | &self, |
|
284 | 330 | on_disk: &'on_disk [u8], |
|
285 | 331 | ) -> Result<dirstate_map::NodeKey<'on_disk>, DirstateV2ParseError> { |
|
286 | 332 | Ok(WithBasename::from_raw_parts( |
|
287 | 333 | Cow::Borrowed(self.full_path(on_disk)?), |
|
288 | 334 | self.base_name_start()?, |
|
289 | 335 | )) |
|
290 | 336 | } |
|
291 | 337 | |
|
292 | 338 | pub(super) fn has_copy_source<'on_disk>(&self) -> bool { |
|
293 | 339 | self.copy_source.start.get() != 0 |
|
294 | 340 | } |
|
295 | 341 | |
|
296 | 342 | pub(super) fn copy_source<'on_disk>( |
|
297 | 343 | &self, |
|
298 | 344 | on_disk: &'on_disk [u8], |
|
299 | 345 | ) -> Result<Option<&'on_disk HgPath>, DirstateV2ParseError> { |
|
300 | 346 | Ok(if self.has_copy_source() { |
|
301 | 347 | Some(read_hg_path(on_disk, self.copy_source)?) |
|
302 | 348 | } else { |
|
303 | 349 | None |
|
304 | 350 | }) |
|
305 | 351 | } |
|
306 | 352 | |
|
307 | 353 | fn flags(&self) -> Flags { |
|
308 | 354 | Flags::from_bits_truncate(self.flags.get()) |
|
309 | 355 | } |
|
310 | 356 | |
|
311 | 357 | fn has_entry(&self) -> bool { |
|
312 | 358 | self.flags().intersects( |
|
313 | 359 | Flags::WDIR_TRACKED | Flags::P1_TRACKED | Flags::P2_INFO, |
|
314 | 360 | ) |
|
315 | 361 | } |
|
316 | 362 | |
|
317 | 363 | pub(super) fn node_data( |
|
318 | 364 | &self, |
|
319 | 365 | ) -> Result<dirstate_map::NodeData, DirstateV2ParseError> { |
|
320 | 366 | if self.has_entry() { |
|
321 | 367 | Ok(dirstate_map::NodeData::Entry(self.assume_entry()?)) |
|
322 | 368 | } else if let Some(mtime) = self.cached_directory_mtime()? { |
|
323 | 369 | Ok(dirstate_map::NodeData::CachedDirectory { mtime }) |
|
324 | 370 | } else { |
|
325 | 371 | Ok(dirstate_map::NodeData::None) |
|
326 | 372 | } |
|
327 | 373 | } |
|
328 | 374 | |
|
329 | 375 | pub(super) fn cached_directory_mtime( |
|
330 | 376 | &self, |
|
331 | 377 | ) -> Result<Option<TruncatedTimestamp>, DirstateV2ParseError> { |
|
332 | 378 | // For now we do not have code to handle the absence of |
|
333 | 379 | // ALL_UNKNOWN_RECORDED, so we ignore the mtime if the flag is |
|
334 | 380 | // unset. |
|
335 | 381 | if self.flags().contains(Flags::DIRECTORY) |
|
336 | 382 | && self.flags().contains(Flags::HAS_MTIME) |
|
337 | 383 | && self.flags().contains(Flags::ALL_UNKNOWN_RECORDED) |
|
338 | 384 | { |
|
339 | 385 | Ok(Some(self.mtime.try_into()?)) |
|
340 | 386 | } else { |
|
341 | 387 | Ok(None) |
|
342 | 388 | } |
|
343 | 389 | } |
|
344 | 390 | |
|
345 | 391 | fn synthesize_unix_mode(&self) -> u32 { |
|
346 | 392 | let file_type = if self.flags().contains(Flags::MODE_IS_SYMLINK) { |
|
347 | 393 | libc::S_IFLNK |
|
348 | 394 | } else { |
|
349 | 395 | libc::S_IFREG |
|
350 | 396 | }; |
|
351 | 397 | let permisions = if self.flags().contains(Flags::MODE_EXEC_PERM) { |
|
352 | 398 | 0o755 |
|
353 | 399 | } else { |
|
354 | 400 | 0o644 |
|
355 | 401 | }; |
|
356 | 402 | file_type | permisions |
|
357 | 403 | } |
|
358 | 404 | |
|
359 | 405 | fn assume_entry(&self) -> Result<DirstateEntry, DirstateV2ParseError> { |
|
360 | 406 | // TODO: convert through raw bits instead? |
|
361 | 407 | let wdir_tracked = self.flags().contains(Flags::WDIR_TRACKED); |
|
362 | 408 | let p1_tracked = self.flags().contains(Flags::P1_TRACKED); |
|
363 | 409 | let p2_info = self.flags().contains(Flags::P2_INFO); |
|
364 | 410 | let mode_size = if self.flags().contains(Flags::HAS_MODE_AND_SIZE) |
|
365 | 411 | && !self.flags().contains(Flags::EXPECTED_STATE_IS_MODIFIED) |
|
366 | 412 | { |
|
367 | 413 | Some((self.synthesize_unix_mode(), self.size.into())) |
|
368 | 414 | } else { |
|
369 | 415 | None |
|
370 | 416 | }; |
|
371 | 417 | let mtime = if self.flags().contains(Flags::HAS_MTIME) |
|
372 | 418 | && !self.flags().contains(Flags::DIRECTORY) |
|
373 | 419 | && !self.flags().contains(Flags::EXPECTED_STATE_IS_MODIFIED) |
|
374 | 420 | { |
|
375 | 421 | let mut m: TruncatedTimestamp = self.mtime.try_into()?; |
|
376 | 422 | if self.flags().contains(Flags::MTIME_SECOND_AMBIGUOUS) { |
|
377 | 423 | m.second_ambiguous = true; |
|
378 | 424 | } |
|
379 | 425 | Some(m) |
|
380 | 426 | } else { |
|
381 | 427 | None |
|
382 | 428 | }; |
|
383 | 429 | let fallback_exec = if self.flags().contains(Flags::HAS_FALLBACK_EXEC) |
|
384 | 430 | { |
|
385 | 431 | Some(self.flags().contains(Flags::FALLBACK_EXEC)) |
|
386 | 432 | } else { |
|
387 | 433 | None |
|
388 | 434 | }; |
|
389 | 435 | let fallback_symlink = |
|
390 | 436 | if self.flags().contains(Flags::HAS_FALLBACK_SYMLINK) { |
|
391 | 437 | Some(self.flags().contains(Flags::FALLBACK_SYMLINK)) |
|
392 | 438 | } else { |
|
393 | 439 | None |
|
394 | 440 | }; |
|
395 | 441 | Ok(DirstateEntry::from_v2_data( |
|
396 | 442 | wdir_tracked, |
|
397 | 443 | p1_tracked, |
|
398 | 444 | p2_info, |
|
399 | 445 | mode_size, |
|
400 | 446 | mtime, |
|
401 | 447 | fallback_exec, |
|
402 | 448 | fallback_symlink, |
|
403 | 449 | )) |
|
404 | 450 | } |
|
405 | 451 | |
|
406 | 452 | pub(super) fn entry( |
|
407 | 453 | &self, |
|
408 | 454 | ) -> Result<Option<DirstateEntry>, DirstateV2ParseError> { |
|
409 | 455 | if self.has_entry() { |
|
410 | 456 | Ok(Some(self.assume_entry()?)) |
|
411 | 457 | } else { |
|
412 | 458 | Ok(None) |
|
413 | 459 | } |
|
414 | 460 | } |
|
415 | 461 | |
|
416 | 462 | pub(super) fn children<'on_disk>( |
|
417 | 463 | &self, |
|
418 | 464 | on_disk: &'on_disk [u8], |
|
419 | 465 | ) -> Result<&'on_disk [Node], DirstateV2ParseError> { |
|
420 | 466 | read_nodes(on_disk, self.children) |
|
421 | 467 | } |
|
422 | 468 | |
|
423 | 469 | pub(super) fn to_in_memory_node<'on_disk>( |
|
424 | 470 | &self, |
|
425 | 471 | on_disk: &'on_disk [u8], |
|
426 | 472 | ) -> Result<dirstate_map::Node<'on_disk>, DirstateV2ParseError> { |
|
427 | 473 | Ok(dirstate_map::Node { |
|
428 | 474 | children: dirstate_map::ChildNodes::OnDisk( |
|
429 | 475 | self.children(on_disk)?, |
|
430 | 476 | ), |
|
431 | 477 | copy_source: self.copy_source(on_disk)?.map(Cow::Borrowed), |
|
432 | 478 | data: self.node_data()?, |
|
433 | 479 | descendants_with_entry_count: self |
|
434 | 480 | .descendants_with_entry_count |
|
435 | 481 | .get(), |
|
436 | 482 | tracked_descendants_count: self.tracked_descendants_count.get(), |
|
437 | 483 | }) |
|
438 | 484 | } |
|
439 | 485 | |
|
440 | 486 | fn from_dirstate_entry( |
|
441 | 487 | entry: &DirstateEntry, |
|
442 | 488 | ) -> (Flags, U32Be, PackedTruncatedTimestamp) { |
|
443 | 489 | let ( |
|
444 | 490 | wdir_tracked, |
|
445 | 491 | p1_tracked, |
|
446 | 492 | p2_info, |
|
447 | 493 | mode_size_opt, |
|
448 | 494 | mtime_opt, |
|
449 | 495 | fallback_exec, |
|
450 | 496 | fallback_symlink, |
|
451 | 497 | ) = entry.v2_data(); |
|
452 | 498 | // TODO: convert throug raw flag bits instead? |
|
453 | 499 | let mut flags = Flags::empty(); |
|
454 | 500 | flags.set(Flags::WDIR_TRACKED, wdir_tracked); |
|
455 | 501 | flags.set(Flags::P1_TRACKED, p1_tracked); |
|
456 | 502 | flags.set(Flags::P2_INFO, p2_info); |
|
457 | 503 | let size = if let Some((m, s)) = mode_size_opt { |
|
458 | 504 | let exec_perm = m & libc::S_IXUSR != 0; |
|
459 | 505 | let is_symlink = m & libc::S_IFMT == libc::S_IFLNK; |
|
460 | 506 | flags.set(Flags::MODE_EXEC_PERM, exec_perm); |
|
461 | 507 | flags.set(Flags::MODE_IS_SYMLINK, is_symlink); |
|
462 | 508 | flags.insert(Flags::HAS_MODE_AND_SIZE); |
|
463 | 509 | s.into() |
|
464 | 510 | } else { |
|
465 | 511 | 0.into() |
|
466 | 512 | }; |
|
467 | 513 | let mtime = if let Some(m) = mtime_opt { |
|
468 | 514 | flags.insert(Flags::HAS_MTIME); |
|
469 | 515 | if m.second_ambiguous { |
|
470 | 516 | flags.insert(Flags::MTIME_SECOND_AMBIGUOUS); |
|
471 | 517 | }; |
|
472 | 518 | m.into() |
|
473 | 519 | } else { |
|
474 | 520 | PackedTruncatedTimestamp::null() |
|
475 | 521 | }; |
|
476 | 522 | if let Some(f_exec) = fallback_exec { |
|
477 | 523 | flags.insert(Flags::HAS_FALLBACK_EXEC); |
|
478 | 524 | if f_exec { |
|
479 | 525 | flags.insert(Flags::FALLBACK_EXEC); |
|
480 | 526 | } |
|
481 | 527 | } |
|
482 | 528 | if let Some(f_symlink) = fallback_symlink { |
|
483 | 529 | flags.insert(Flags::HAS_FALLBACK_SYMLINK); |
|
484 | 530 | if f_symlink { |
|
485 | 531 | flags.insert(Flags::FALLBACK_SYMLINK); |
|
486 | 532 | } |
|
487 | 533 | } |
|
488 | 534 | (flags, size, mtime) |
|
489 | 535 | } |
|
490 | 536 | } |
|
491 | 537 | |
|
492 | 538 | fn read_hg_path( |
|
493 | 539 | on_disk: &[u8], |
|
494 | 540 | slice: PathSlice, |
|
495 | 541 | ) -> Result<&HgPath, DirstateV2ParseError> { |
|
496 | 542 | read_slice(on_disk, slice.start, slice.len.get()).map(HgPath::new) |
|
497 | 543 | } |
|
498 | 544 | |
|
499 | 545 | fn read_nodes( |
|
500 | 546 | on_disk: &[u8], |
|
501 | 547 | slice: ChildNodes, |
|
502 | 548 | ) -> Result<&[Node], DirstateV2ParseError> { |
|
503 | 549 | read_slice(on_disk, slice.start, slice.len.get()) |
|
504 | 550 | } |
|
505 | 551 | |
|
506 | 552 | fn read_slice<T, Len>( |
|
507 | 553 | on_disk: &[u8], |
|
508 | 554 | start: Offset, |
|
509 | 555 | len: Len, |
|
510 | 556 | ) -> Result<&[T], DirstateV2ParseError> |
|
511 | 557 | where |
|
512 | 558 | T: BytesCast, |
|
513 | 559 | Len: TryInto<usize>, |
|
514 | 560 | { |
|
515 | 561 | // Either `usize::MAX` would result in "out of bounds" error since a single |
|
516 | 562 | // `&[u8]` cannot occupy the entire addess space. |
|
517 | 563 | let start = start.get().try_into().unwrap_or(std::usize::MAX); |
|
518 | 564 | let len = len.try_into().unwrap_or(std::usize::MAX); |
|
519 | 565 | on_disk |
|
520 | 566 | .get(start..) |
|
521 | 567 | .and_then(|bytes| T::slice_from_bytes(bytes, len).ok()) |
|
522 | 568 | .map(|(slice, _rest)| slice) |
|
523 | 569 | .ok_or_else(|| DirstateV2ParseError) |
|
524 | 570 | } |
|
525 | 571 | |
|
526 | 572 | pub(crate) fn for_each_tracked_path<'on_disk>( |
|
527 | 573 | on_disk: &'on_disk [u8], |
|
528 | 574 | metadata: &[u8], |
|
529 | 575 | mut f: impl FnMut(&'on_disk HgPath), |
|
530 | 576 | ) -> Result<(), DirstateV2ParseError> { |
|
531 | 577 | let (meta, _) = TreeMetadata::from_bytes(metadata) |
|
532 | 578 | .map_err(|_| DirstateV2ParseError)?; |
|
533 | 579 | fn recur<'on_disk>( |
|
534 | 580 | on_disk: &'on_disk [u8], |
|
535 | 581 | nodes: ChildNodes, |
|
536 | 582 | f: &mut impl FnMut(&'on_disk HgPath), |
|
537 | 583 | ) -> Result<(), DirstateV2ParseError> { |
|
538 | 584 | for node in read_nodes(on_disk, nodes)? { |
|
539 | 585 | if let Some(entry) = node.entry()? { |
|
540 | 586 | if entry.state().is_tracked() { |
|
541 | 587 | f(node.full_path(on_disk)?) |
|
542 | 588 | } |
|
543 | 589 | } |
|
544 | 590 | recur(on_disk, node.children, f)? |
|
545 | 591 | } |
|
546 | 592 | Ok(()) |
|
547 | 593 | } |
|
548 | 594 | recur(on_disk, meta.root_nodes, &mut f) |
|
549 | 595 | } |
|
550 | 596 | |
|
551 | 597 | /// Returns new data and metadata, together with whether that data should be |
|
552 | 598 | /// appended to the existing data file whose content is at |
|
553 | 599 | /// `dirstate_map.on_disk` (true), instead of written to a new data file |
|
554 | 600 | /// (false). |
|
555 | 601 | pub(super) fn write( |
|
556 | 602 | dirstate_map: &DirstateMap, |
|
557 | 603 | can_append: bool, |
|
558 |
) -> Result<(Vec<u8>, |
|
|
604 | ) -> Result<(Vec<u8>, TreeMetadata, bool), DirstateError> { | |
|
559 | 605 | let append = can_append && dirstate_map.write_should_append(); |
|
560 | 606 | |
|
561 | 607 | // This ignores the space for paths, and for nodes without an entry. |
|
562 | 608 | // TODO: better estimate? Skip the `Vec` and write to a file directly? |
|
563 | 609 | let size_guess = std::mem::size_of::<Node>() |
|
564 | 610 | * dirstate_map.nodes_with_entry_count as usize; |
|
565 | 611 | |
|
566 | 612 | let mut writer = Writer { |
|
567 | 613 | dirstate_map, |
|
568 | 614 | append, |
|
569 | 615 | out: Vec::with_capacity(size_guess), |
|
570 | 616 | }; |
|
571 | 617 | |
|
572 | 618 | let root_nodes = writer.write_nodes(dirstate_map.root.as_ref())?; |
|
573 | 619 | |
|
574 | 620 | let meta = TreeMetadata { |
|
575 | 621 | root_nodes, |
|
576 | 622 | nodes_with_entry_count: dirstate_map.nodes_with_entry_count.into(), |
|
577 | 623 | nodes_with_copy_source_count: dirstate_map |
|
578 | 624 | .nodes_with_copy_source_count |
|
579 | 625 | .into(), |
|
580 | 626 | unreachable_bytes: dirstate_map.unreachable_bytes.into(), |
|
581 | 627 | unused: [0; 4], |
|
582 | 628 | ignore_patterns_hash: dirstate_map.ignore_patterns_hash, |
|
583 | 629 | }; |
|
584 |
Ok((writer.out, meta |
|
|
630 | Ok((writer.out, meta, append)) | |
|
585 | 631 | } |
|
586 | 632 | |
|
587 | 633 | struct Writer<'dmap, 'on_disk> { |
|
588 | 634 | dirstate_map: &'dmap DirstateMap<'on_disk>, |
|
589 | 635 | append: bool, |
|
590 | 636 | out: Vec<u8>, |
|
591 | 637 | } |
|
592 | 638 | |
|
593 | 639 | impl Writer<'_, '_> { |
|
594 | 640 | fn write_nodes( |
|
595 | 641 | &mut self, |
|
596 | 642 | nodes: dirstate_map::ChildNodesRef, |
|
597 | 643 | ) -> Result<ChildNodes, DirstateError> { |
|
598 | 644 | // Reuse already-written nodes if possible |
|
599 | 645 | if self.append { |
|
600 | 646 | if let dirstate_map::ChildNodesRef::OnDisk(nodes_slice) = nodes { |
|
601 | 647 | let start = self.on_disk_offset_of(nodes_slice).expect( |
|
602 | 648 | "dirstate-v2 OnDisk nodes not found within on_disk", |
|
603 | 649 | ); |
|
604 | 650 | let len = child_nodes_len_from_usize(nodes_slice.len()); |
|
605 | 651 | return Ok(ChildNodes { start, len }); |
|
606 | 652 | } |
|
607 | 653 | } |
|
608 | 654 | |
|
609 | 655 | // `dirstate_map::ChildNodes::InMemory` contains a `HashMap` which has |
|
610 | 656 | // undefined iteration order. Sort to enable binary search in the |
|
611 | 657 | // written file. |
|
612 | 658 | let nodes = nodes.sorted(); |
|
613 | 659 | let nodes_len = nodes.len(); |
|
614 | 660 | |
|
615 | 661 | // First accumulate serialized nodes in a `Vec` |
|
616 | 662 | let mut on_disk_nodes = Vec::with_capacity(nodes_len); |
|
617 | 663 | for node in nodes { |
|
618 | 664 | let children = |
|
619 | 665 | self.write_nodes(node.children(self.dirstate_map.on_disk)?)?; |
|
620 | 666 | let full_path = node.full_path(self.dirstate_map.on_disk)?; |
|
621 | 667 | let full_path = self.write_path(full_path.as_bytes()); |
|
622 | 668 | let copy_source = if let Some(source) = |
|
623 | 669 | node.copy_source(self.dirstate_map.on_disk)? |
|
624 | 670 | { |
|
625 | 671 | self.write_path(source.as_bytes()) |
|
626 | 672 | } else { |
|
627 | 673 | PathSlice { |
|
628 | 674 | start: 0.into(), |
|
629 | 675 | len: 0.into(), |
|
630 | 676 | } |
|
631 | 677 | }; |
|
632 | 678 | on_disk_nodes.push(match node { |
|
633 | 679 | NodeRef::InMemory(path, node) => { |
|
634 | 680 | let (flags, size, mtime) = match &node.data { |
|
635 | 681 | dirstate_map::NodeData::Entry(entry) => { |
|
636 | 682 | Node::from_dirstate_entry(entry) |
|
637 | 683 | } |
|
638 | 684 | dirstate_map::NodeData::CachedDirectory { mtime } => ( |
|
639 | 685 | // we currently never set a mtime if unknown file |
|
640 | 686 | // are present. |
|
641 | 687 | // So if we have a mtime for a directory, we know |
|
642 | 688 | // they are no unknown |
|
643 | 689 | // files and we |
|
644 | 690 | // blindly set ALL_UNKNOWN_RECORDED. |
|
645 | 691 | // |
|
646 | 692 | // We never set ALL_IGNORED_RECORDED since we |
|
647 | 693 | // don't track that case |
|
648 | 694 | // currently. |
|
649 | 695 | Flags::DIRECTORY |
|
650 | 696 | | Flags::HAS_MTIME |
|
651 | 697 | | Flags::ALL_UNKNOWN_RECORDED, |
|
652 | 698 | 0.into(), |
|
653 | 699 | (*mtime).into(), |
|
654 | 700 | ), |
|
655 | 701 | dirstate_map::NodeData::None => ( |
|
656 | 702 | Flags::DIRECTORY, |
|
657 | 703 | 0.into(), |
|
658 | 704 | PackedTruncatedTimestamp::null(), |
|
659 | 705 | ), |
|
660 | 706 | }; |
|
661 | 707 | Node { |
|
662 | 708 | children, |
|
663 | 709 | copy_source, |
|
664 | 710 | full_path, |
|
665 | 711 | base_name_start: u16::try_from(path.base_name_start()) |
|
666 | 712 | // Could only panic for paths over 64 KiB |
|
667 | 713 | .expect("dirstate-v2 path length overflow") |
|
668 | 714 | .into(), |
|
669 | 715 | descendants_with_entry_count: node |
|
670 | 716 | .descendants_with_entry_count |
|
671 | 717 | .into(), |
|
672 | 718 | tracked_descendants_count: node |
|
673 | 719 | .tracked_descendants_count |
|
674 | 720 | .into(), |
|
675 | 721 | flags: flags.bits().into(), |
|
676 | 722 | size, |
|
677 | 723 | mtime, |
|
678 | 724 | } |
|
679 | 725 | } |
|
680 | 726 | NodeRef::OnDisk(node) => Node { |
|
681 | 727 | children, |
|
682 | 728 | copy_source, |
|
683 | 729 | full_path, |
|
684 | 730 | ..*node |
|
685 | 731 | }, |
|
686 | 732 | }) |
|
687 | 733 | } |
|
688 | 734 | // … so we can write them contiguously, after writing everything else |
|
689 | 735 | // they refer to. |
|
690 | 736 | let start = self.current_offset(); |
|
691 | 737 | let len = child_nodes_len_from_usize(nodes_len); |
|
692 | 738 | self.out.extend(on_disk_nodes.as_bytes()); |
|
693 | 739 | Ok(ChildNodes { start, len }) |
|
694 | 740 | } |
|
695 | 741 | |
|
696 | 742 | /// If the given slice of items is within `on_disk`, returns its offset |
|
697 | 743 | /// from the start of `on_disk`. |
|
698 | 744 | fn on_disk_offset_of<T>(&self, slice: &[T]) -> Option<Offset> |
|
699 | 745 | where |
|
700 | 746 | T: BytesCast, |
|
701 | 747 | { |
|
702 | 748 | fn address_range(slice: &[u8]) -> std::ops::RangeInclusive<usize> { |
|
703 | 749 | let start = slice.as_ptr() as usize; |
|
704 | 750 | let end = start + slice.len(); |
|
705 | 751 | start..=end |
|
706 | 752 | } |
|
707 | 753 | let slice_addresses = address_range(slice.as_bytes()); |
|
708 | 754 | let on_disk_addresses = address_range(self.dirstate_map.on_disk); |
|
709 | 755 | if on_disk_addresses.contains(slice_addresses.start()) |
|
710 | 756 | && on_disk_addresses.contains(slice_addresses.end()) |
|
711 | 757 | { |
|
712 | 758 | let offset = slice_addresses.start() - on_disk_addresses.start(); |
|
713 | 759 | Some(offset_from_usize(offset)) |
|
714 | 760 | } else { |
|
715 | 761 | None |
|
716 | 762 | } |
|
717 | 763 | } |
|
718 | 764 | |
|
719 | 765 | fn current_offset(&mut self) -> Offset { |
|
720 | 766 | let mut offset = self.out.len(); |
|
721 | 767 | if self.append { |
|
722 | 768 | offset += self.dirstate_map.on_disk.len() |
|
723 | 769 | } |
|
724 | 770 | offset_from_usize(offset) |
|
725 | 771 | } |
|
726 | 772 | |
|
727 | 773 | fn write_path(&mut self, slice: &[u8]) -> PathSlice { |
|
728 | 774 | let len = path_len_from_usize(slice.len()); |
|
729 | 775 | // Reuse an already-written path if possible |
|
730 | 776 | if self.append { |
|
731 | 777 | if let Some(start) = self.on_disk_offset_of(slice) { |
|
732 | 778 | return PathSlice { start, len }; |
|
733 | 779 | } |
|
734 | 780 | } |
|
735 | 781 | let start = self.current_offset(); |
|
736 | 782 | self.out.extend(slice.as_bytes()); |
|
737 | 783 | PathSlice { start, len } |
|
738 | 784 | } |
|
739 | 785 | } |
|
740 | 786 | |
|
741 | 787 | fn offset_from_usize(x: usize) -> Offset { |
|
742 | 788 | u32::try_from(x) |
|
743 | 789 | // Could only panic for a dirstate file larger than 4 GiB |
|
744 | 790 | .expect("dirstate-v2 offset overflow") |
|
745 | 791 | .into() |
|
746 | 792 | } |
|
747 | 793 | |
|
748 | 794 | fn child_nodes_len_from_usize(x: usize) -> Size { |
|
749 | 795 | u32::try_from(x) |
|
750 | 796 | // Could only panic with over 4 billion nodes |
|
751 | 797 | .expect("dirstate-v2 slice length overflow") |
|
752 | 798 | .into() |
|
753 | 799 | } |
|
754 | 800 | |
|
755 | 801 | fn path_len_from_usize(x: usize) -> PathSize { |
|
756 | 802 | u16::try_from(x) |
|
757 | 803 | // Could only panic for paths over 64 KiB |
|
758 | 804 | .expect("dirstate-v2 path length overflow") |
|
759 | 805 | .into() |
|
760 | 806 | } |
|
761 | 807 | |
|
762 | 808 | impl From<TruncatedTimestamp> for PackedTruncatedTimestamp { |
|
763 | 809 | fn from(timestamp: TruncatedTimestamp) -> Self { |
|
764 | 810 | Self { |
|
765 | 811 | truncated_seconds: timestamp.truncated_seconds().into(), |
|
766 | 812 | nanoseconds: timestamp.nanoseconds().into(), |
|
767 | 813 | } |
|
768 | 814 | } |
|
769 | 815 | } |
|
770 | 816 | |
|
771 | 817 | impl TryFrom<PackedTruncatedTimestamp> for TruncatedTimestamp { |
|
772 | 818 | type Error = DirstateV2ParseError; |
|
773 | 819 | |
|
774 | 820 | fn try_from( |
|
775 | 821 | timestamp: PackedTruncatedTimestamp, |
|
776 | 822 | ) -> Result<Self, Self::Error> { |
|
777 | 823 | Self::from_already_truncated( |
|
778 | 824 | timestamp.truncated_seconds.get(), |
|
779 | 825 | timestamp.nanoseconds.get(), |
|
780 | 826 | false, |
|
781 | 827 | ) |
|
782 | 828 | } |
|
783 | 829 | } |
|
784 | 830 | impl PackedTruncatedTimestamp { |
|
785 | 831 | fn null() -> Self { |
|
786 | 832 | Self { |
|
787 | 833 | truncated_seconds: 0.into(), |
|
788 | 834 | nanoseconds: 0.into(), |
|
789 | 835 | } |
|
790 | 836 | } |
|
791 | 837 | } |
@@ -1,464 +1,532 b'' | |||
|
1 | 1 | use crate::changelog::Changelog; |
|
2 | 2 | use crate::config::{Config, ConfigError, ConfigParseError}; |
|
3 | 3 | use crate::dirstate::DirstateParents; |
|
4 | 4 | use crate::dirstate_tree::dirstate_map::DirstateMap; |
|
5 | use crate::dirstate_tree::on_disk::Docket as DirstateDocket; | |
|
5 | 6 | use crate::dirstate_tree::owning::OwningDirstateMap; |
|
6 | use crate::errors::HgError; | |
|
7 | 7 | use crate::errors::HgResultExt; |
|
8 | use crate::errors::{HgError, IoResultExt}; | |
|
8 | 9 | use crate::exit_codes; |
|
9 | 10 | use crate::lock::{try_with_lock_no_wait, LockError}; |
|
10 | 11 | use crate::manifest::{Manifest, Manifestlog}; |
|
11 | 12 | use crate::revlog::filelog::Filelog; |
|
12 | 13 | use crate::revlog::revlog::RevlogError; |
|
13 | 14 | use crate::utils::files::get_path_from_bytes; |
|
14 | 15 | use crate::utils::hg_path::HgPath; |
|
15 | 16 | use crate::utils::SliceExt; |
|
16 | 17 | use crate::vfs::{is_dir, is_file, Vfs}; |
|
17 | 18 | use crate::{requirements, NodePrefix}; |
|
18 | 19 | use crate::{DirstateError, Revision}; |
|
19 | 20 | use std::cell::{Ref, RefCell, RefMut}; |
|
20 | 21 | use std::collections::HashSet; |
|
22 | use std::io::Seek; | |
|
23 | use std::io::SeekFrom; | |
|
24 | use std::io::Write as IoWrite; | |
|
21 | 25 | use std::path::{Path, PathBuf}; |
|
22 | 26 | |
|
23 | 27 | /// A repository on disk |
|
24 | 28 | pub struct Repo { |
|
25 | 29 | working_directory: PathBuf, |
|
26 | 30 | dot_hg: PathBuf, |
|
27 | 31 | store: PathBuf, |
|
28 | 32 | requirements: HashSet<String>, |
|
29 | 33 | config: Config, |
|
30 | 34 | dirstate_parents: LazyCell<DirstateParents, HgError>, |
|
31 | 35 | dirstate_data_file_uuid: LazyCell<Option<Vec<u8>>, HgError>, |
|
32 | 36 | dirstate_map: LazyCell<OwningDirstateMap, DirstateError>, |
|
33 | 37 | changelog: LazyCell<Changelog, HgError>, |
|
34 | 38 | manifestlog: LazyCell<Manifestlog, HgError>, |
|
35 | 39 | } |
|
36 | 40 | |
|
37 | 41 | #[derive(Debug, derive_more::From)] |
|
38 | 42 | pub enum RepoError { |
|
39 | 43 | NotFound { |
|
40 | 44 | at: PathBuf, |
|
41 | 45 | }, |
|
42 | 46 | #[from] |
|
43 | 47 | ConfigParseError(ConfigParseError), |
|
44 | 48 | #[from] |
|
45 | 49 | Other(HgError), |
|
46 | 50 | } |
|
47 | 51 | |
|
48 | 52 | impl From<ConfigError> for RepoError { |
|
49 | 53 | fn from(error: ConfigError) -> Self { |
|
50 | 54 | match error { |
|
51 | 55 | ConfigError::Parse(error) => error.into(), |
|
52 | 56 | ConfigError::Other(error) => error.into(), |
|
53 | 57 | } |
|
54 | 58 | } |
|
55 | 59 | } |
|
56 | 60 | |
|
57 | 61 | impl Repo { |
|
58 | 62 | /// tries to find nearest repository root in current working directory or |
|
59 | 63 | /// its ancestors |
|
60 | 64 | pub fn find_repo_root() -> Result<PathBuf, RepoError> { |
|
61 | 65 | let current_directory = crate::utils::current_dir()?; |
|
62 | 66 | // ancestors() is inclusive: it first yields `current_directory` |
|
63 | 67 | // as-is. |
|
64 | 68 | for ancestor in current_directory.ancestors() { |
|
65 | 69 | if is_dir(ancestor.join(".hg"))? { |
|
66 | 70 | return Ok(ancestor.to_path_buf()); |
|
67 | 71 | } |
|
68 | 72 | } |
|
69 | 73 | return Err(RepoError::NotFound { |
|
70 | 74 | at: current_directory, |
|
71 | 75 | }); |
|
72 | 76 | } |
|
73 | 77 | |
|
74 | 78 | /// Find a repository, either at the given path (which must contain a `.hg` |
|
75 | 79 | /// sub-directory) or by searching the current directory and its |
|
76 | 80 | /// ancestors. |
|
77 | 81 | /// |
|
78 | 82 | /// A method with two very different "modes" like this usually a code smell |
|
79 | 83 | /// to make two methods instead, but in this case an `Option` is what rhg |
|
80 | 84 | /// sub-commands get from Clap for the `-R` / `--repository` CLI argument. |
|
81 | 85 | /// Having two methods would just move that `if` to almost all callers. |
|
82 | 86 | pub fn find( |
|
83 | 87 | config: &Config, |
|
84 | 88 | explicit_path: Option<PathBuf>, |
|
85 | 89 | ) -> Result<Self, RepoError> { |
|
86 | 90 | if let Some(root) = explicit_path { |
|
87 | 91 | if is_dir(root.join(".hg"))? { |
|
88 | 92 | Self::new_at_path(root.to_owned(), config) |
|
89 | 93 | } else if is_file(&root)? { |
|
90 | 94 | Err(HgError::unsupported("bundle repository").into()) |
|
91 | 95 | } else { |
|
92 | 96 | Err(RepoError::NotFound { |
|
93 | 97 | at: root.to_owned(), |
|
94 | 98 | }) |
|
95 | 99 | } |
|
96 | 100 | } else { |
|
97 | 101 | let root = Self::find_repo_root()?; |
|
98 | 102 | Self::new_at_path(root, config) |
|
99 | 103 | } |
|
100 | 104 | } |
|
101 | 105 | |
|
102 | 106 | /// To be called after checking that `.hg` is a sub-directory |
|
103 | 107 | fn new_at_path( |
|
104 | 108 | working_directory: PathBuf, |
|
105 | 109 | config: &Config, |
|
106 | 110 | ) -> Result<Self, RepoError> { |
|
107 | 111 | let dot_hg = working_directory.join(".hg"); |
|
108 | 112 | |
|
109 | 113 | let mut repo_config_files = Vec::new(); |
|
110 | 114 | repo_config_files.push(dot_hg.join("hgrc")); |
|
111 | 115 | repo_config_files.push(dot_hg.join("hgrc-not-shared")); |
|
112 | 116 | |
|
113 | 117 | let hg_vfs = Vfs { base: &dot_hg }; |
|
114 | 118 | let mut reqs = requirements::load_if_exists(hg_vfs)?; |
|
115 | 119 | let relative = |
|
116 | 120 | reqs.contains(requirements::RELATIVE_SHARED_REQUIREMENT); |
|
117 | 121 | let shared = |
|
118 | 122 | reqs.contains(requirements::SHARED_REQUIREMENT) || relative; |
|
119 | 123 | |
|
120 | 124 | // From `mercurial/localrepo.py`: |
|
121 | 125 | // |
|
122 | 126 | // if .hg/requires contains the sharesafe requirement, it means |
|
123 | 127 | // there exists a `.hg/store/requires` too and we should read it |
|
124 | 128 | // NOTE: presence of SHARESAFE_REQUIREMENT imply that store requirement |
|
125 | 129 | // is present. We never write SHARESAFE_REQUIREMENT for a repo if store |
|
126 | 130 | // is not present, refer checkrequirementscompat() for that |
|
127 | 131 | // |
|
128 | 132 | // However, if SHARESAFE_REQUIREMENT is not present, it means that the |
|
129 | 133 | // repository was shared the old way. We check the share source |
|
130 | 134 | // .hg/requires for SHARESAFE_REQUIREMENT to detect whether the |
|
131 | 135 | // current repository needs to be reshared |
|
132 | 136 | let share_safe = reqs.contains(requirements::SHARESAFE_REQUIREMENT); |
|
133 | 137 | |
|
134 | 138 | let store_path; |
|
135 | 139 | if !shared { |
|
136 | 140 | store_path = dot_hg.join("store"); |
|
137 | 141 | } else { |
|
138 | 142 | let bytes = hg_vfs.read("sharedpath")?; |
|
139 | 143 | let mut shared_path = |
|
140 | 144 | get_path_from_bytes(bytes.trim_end_matches(|b| b == b'\n')) |
|
141 | 145 | .to_owned(); |
|
142 | 146 | if relative { |
|
143 | 147 | shared_path = dot_hg.join(shared_path) |
|
144 | 148 | } |
|
145 | 149 | if !is_dir(&shared_path)? { |
|
146 | 150 | return Err(HgError::corrupted(format!( |
|
147 | 151 | ".hg/sharedpath points to nonexistent directory {}", |
|
148 | 152 | shared_path.display() |
|
149 | 153 | )) |
|
150 | 154 | .into()); |
|
151 | 155 | } |
|
152 | 156 | |
|
153 | 157 | store_path = shared_path.join("store"); |
|
154 | 158 | |
|
155 | 159 | let source_is_share_safe = |
|
156 | 160 | requirements::load(Vfs { base: &shared_path })? |
|
157 | 161 | .contains(requirements::SHARESAFE_REQUIREMENT); |
|
158 | 162 | |
|
159 | 163 | if share_safe && !source_is_share_safe { |
|
160 | 164 | return Err(match config |
|
161 | 165 | .get(b"share", b"safe-mismatch.source-not-safe") |
|
162 | 166 | { |
|
163 | 167 | Some(b"abort") | None => HgError::abort( |
|
164 | 168 | "abort: share source does not support share-safe requirement\n\ |
|
165 | 169 | (see `hg help config.format.use-share-safe` for more information)", |
|
166 | 170 | exit_codes::ABORT, |
|
167 | 171 | ), |
|
168 | 172 | _ => HgError::unsupported("share-safe downgrade"), |
|
169 | 173 | } |
|
170 | 174 | .into()); |
|
171 | 175 | } else if source_is_share_safe && !share_safe { |
|
172 | 176 | return Err( |
|
173 | 177 | match config.get(b"share", b"safe-mismatch.source-safe") { |
|
174 | 178 | Some(b"abort") | None => HgError::abort( |
|
175 | 179 | "abort: version mismatch: source uses share-safe \ |
|
176 | 180 | functionality while the current share does not\n\ |
|
177 | 181 | (see `hg help config.format.use-share-safe` for more information)", |
|
178 | 182 | exit_codes::ABORT, |
|
179 | 183 | ), |
|
180 | 184 | _ => HgError::unsupported("share-safe upgrade"), |
|
181 | 185 | } |
|
182 | 186 | .into(), |
|
183 | 187 | ); |
|
184 | 188 | } |
|
185 | 189 | |
|
186 | 190 | if share_safe { |
|
187 | 191 | repo_config_files.insert(0, shared_path.join("hgrc")) |
|
188 | 192 | } |
|
189 | 193 | } |
|
190 | 194 | if share_safe { |
|
191 | 195 | reqs.extend(requirements::load(Vfs { base: &store_path })?); |
|
192 | 196 | } |
|
193 | 197 | |
|
194 | 198 | let repo_config = if std::env::var_os("HGRCSKIPREPO").is_none() { |
|
195 | 199 | config.combine_with_repo(&repo_config_files)? |
|
196 | 200 | } else { |
|
197 | 201 | config.clone() |
|
198 | 202 | }; |
|
199 | 203 | |
|
200 | 204 | let repo = Self { |
|
201 | 205 | requirements: reqs, |
|
202 | 206 | working_directory, |
|
203 | 207 | store: store_path, |
|
204 | 208 | dot_hg, |
|
205 | 209 | config: repo_config, |
|
206 | 210 | dirstate_parents: LazyCell::new(Self::read_dirstate_parents), |
|
207 | 211 | dirstate_data_file_uuid: LazyCell::new( |
|
208 | 212 | Self::read_dirstate_data_file_uuid, |
|
209 | 213 | ), |
|
210 | 214 | dirstate_map: LazyCell::new(Self::new_dirstate_map), |
|
211 | 215 | changelog: LazyCell::new(Changelog::open), |
|
212 | 216 | manifestlog: LazyCell::new(Manifestlog::open), |
|
213 | 217 | }; |
|
214 | 218 | |
|
215 | 219 | requirements::check(&repo)?; |
|
216 | 220 | |
|
217 | 221 | Ok(repo) |
|
218 | 222 | } |
|
219 | 223 | |
|
220 | 224 | pub fn working_directory_path(&self) -> &Path { |
|
221 | 225 | &self.working_directory |
|
222 | 226 | } |
|
223 | 227 | |
|
224 | 228 | pub fn requirements(&self) -> &HashSet<String> { |
|
225 | 229 | &self.requirements |
|
226 | 230 | } |
|
227 | 231 | |
|
228 | 232 | pub fn config(&self) -> &Config { |
|
229 | 233 | &self.config |
|
230 | 234 | } |
|
231 | 235 | |
|
232 | 236 | /// For accessing repository files (in `.hg`), except for the store |
|
233 | 237 | /// (`.hg/store`). |
|
234 | 238 | pub fn hg_vfs(&self) -> Vfs<'_> { |
|
235 | 239 | Vfs { base: &self.dot_hg } |
|
236 | 240 | } |
|
237 | 241 | |
|
238 | 242 | /// For accessing repository store files (in `.hg/store`) |
|
239 | 243 | pub fn store_vfs(&self) -> Vfs<'_> { |
|
240 | 244 | Vfs { base: &self.store } |
|
241 | 245 | } |
|
242 | 246 | |
|
243 | 247 | /// For accessing the working copy |
|
244 | 248 | pub fn working_directory_vfs(&self) -> Vfs<'_> { |
|
245 | 249 | Vfs { |
|
246 | 250 | base: &self.working_directory, |
|
247 | 251 | } |
|
248 | 252 | } |
|
249 | 253 | |
|
250 | 254 | pub fn try_with_wlock_no_wait<R>( |
|
251 | 255 | &self, |
|
252 | 256 | f: impl FnOnce() -> R, |
|
253 | 257 | ) -> Result<R, LockError> { |
|
254 | 258 | try_with_lock_no_wait(self.hg_vfs(), "wlock", f) |
|
255 | 259 | } |
|
256 | 260 | |
|
257 | 261 | pub fn has_dirstate_v2(&self) -> bool { |
|
258 | 262 | self.requirements |
|
259 | 263 | .contains(requirements::DIRSTATE_V2_REQUIREMENT) |
|
260 | 264 | } |
|
261 | 265 | |
|
262 | 266 | pub fn has_sparse(&self) -> bool { |
|
263 | 267 | self.requirements.contains(requirements::SPARSE_REQUIREMENT) |
|
264 | 268 | } |
|
265 | 269 | |
|
266 | 270 | pub fn has_narrow(&self) -> bool { |
|
267 | 271 | self.requirements.contains(requirements::NARROW_REQUIREMENT) |
|
268 | 272 | } |
|
269 | 273 | |
|
270 | 274 | fn dirstate_file_contents(&self) -> Result<Vec<u8>, HgError> { |
|
271 | 275 | Ok(self |
|
272 | 276 | .hg_vfs() |
|
273 | 277 | .read("dirstate") |
|
274 | 278 | .io_not_found_as_none()? |
|
275 | 279 | .unwrap_or(Vec::new())) |
|
276 | 280 | } |
|
277 | 281 | |
|
278 | 282 | pub fn dirstate_parents(&self) -> Result<DirstateParents, HgError> { |
|
279 | 283 | Ok(*self.dirstate_parents.get_or_init(self)?) |
|
280 | 284 | } |
|
281 | 285 | |
|
282 | 286 | fn read_dirstate_parents(&self) -> Result<DirstateParents, HgError> { |
|
283 | 287 | let dirstate = self.dirstate_file_contents()?; |
|
284 | 288 | let parents = if dirstate.is_empty() { |
|
285 | 289 | if self.has_dirstate_v2() { |
|
286 | 290 | self.dirstate_data_file_uuid.set(None); |
|
287 | 291 | } |
|
288 | 292 | DirstateParents::NULL |
|
289 | 293 | } else if self.has_dirstate_v2() { |
|
290 | 294 | let docket = |
|
291 | 295 | crate::dirstate_tree::on_disk::read_docket(&dirstate)?; |
|
292 | 296 | self.dirstate_data_file_uuid |
|
293 | 297 | .set(Some(docket.uuid.to_owned())); |
|
294 | 298 | docket.parents() |
|
295 | 299 | } else { |
|
296 | 300 | crate::dirstate::parsers::parse_dirstate_parents(&dirstate)? |
|
297 | 301 | .clone() |
|
298 | 302 | }; |
|
299 | 303 | self.dirstate_parents.set(parents); |
|
300 | 304 | Ok(parents) |
|
301 | 305 | } |
|
302 | 306 | |
|
303 | 307 | fn read_dirstate_data_file_uuid( |
|
304 | 308 | &self, |
|
305 | 309 | ) -> Result<Option<Vec<u8>>, HgError> { |
|
306 | 310 | assert!( |
|
307 | 311 | self.has_dirstate_v2(), |
|
308 | 312 | "accessing dirstate data file ID without dirstate-v2" |
|
309 | 313 | ); |
|
310 | 314 | let dirstate = self.dirstate_file_contents()?; |
|
311 | 315 | if dirstate.is_empty() { |
|
312 | 316 | self.dirstate_parents.set(DirstateParents::NULL); |
|
313 | 317 | Ok(None) |
|
314 | 318 | } else { |
|
315 | 319 | let docket = |
|
316 | 320 | crate::dirstate_tree::on_disk::read_docket(&dirstate)?; |
|
317 | 321 | self.dirstate_parents.set(docket.parents()); |
|
318 | 322 | Ok(Some(docket.uuid.to_owned())) |
|
319 | 323 | } |
|
320 | 324 | } |
|
321 | 325 | |
|
322 | 326 | fn new_dirstate_map(&self) -> Result<OwningDirstateMap, DirstateError> { |
|
323 | 327 | let dirstate_file_contents = self.dirstate_file_contents()?; |
|
324 | 328 | if dirstate_file_contents.is_empty() { |
|
325 | 329 | self.dirstate_parents.set(DirstateParents::NULL); |
|
326 | 330 | if self.has_dirstate_v2() { |
|
327 | 331 | self.dirstate_data_file_uuid.set(None); |
|
328 | 332 | } |
|
329 | 333 | Ok(OwningDirstateMap::new_empty(Vec::new())) |
|
330 | 334 | } else if self.has_dirstate_v2() { |
|
331 | 335 | let docket = crate::dirstate_tree::on_disk::read_docket( |
|
332 | 336 | &dirstate_file_contents, |
|
333 | 337 | )?; |
|
334 | 338 | self.dirstate_parents.set(docket.parents()); |
|
335 | 339 | self.dirstate_data_file_uuid |
|
336 | 340 | .set(Some(docket.uuid.to_owned())); |
|
337 | 341 | let data_size = docket.data_size(); |
|
338 | 342 | let metadata = docket.tree_metadata(); |
|
339 | 343 | let mut map = if let Some(data_mmap) = self |
|
340 | 344 | .hg_vfs() |
|
341 | 345 | .mmap_open(docket.data_filename()) |
|
342 | 346 | .io_not_found_as_none()? |
|
343 | 347 | { |
|
344 | 348 | OwningDirstateMap::new_empty(data_mmap) |
|
345 | 349 | } else { |
|
346 | 350 | OwningDirstateMap::new_empty(Vec::new()) |
|
347 | 351 | }; |
|
348 | 352 | let (on_disk, placeholder) = map.get_pair_mut(); |
|
349 | 353 | *placeholder = DirstateMap::new_v2(on_disk, data_size, metadata)?; |
|
350 | 354 | Ok(map) |
|
351 | 355 | } else { |
|
352 | 356 | let mut map = OwningDirstateMap::new_empty(dirstate_file_contents); |
|
353 | 357 | let (on_disk, placeholder) = map.get_pair_mut(); |
|
354 | 358 | let (inner, parents) = DirstateMap::new_v1(on_disk)?; |
|
355 | 359 | self.dirstate_parents |
|
356 | 360 | .set(parents.unwrap_or(DirstateParents::NULL)); |
|
357 | 361 | *placeholder = inner; |
|
358 | 362 | Ok(map) |
|
359 | 363 | } |
|
360 | 364 | } |
|
361 | 365 | |
|
362 | 366 | pub fn dirstate_map( |
|
363 | 367 | &self, |
|
364 | 368 | ) -> Result<Ref<OwningDirstateMap>, DirstateError> { |
|
365 | 369 | self.dirstate_map.get_or_init(self) |
|
366 | 370 | } |
|
367 | 371 | |
|
368 | 372 | pub fn dirstate_map_mut( |
|
369 | 373 | &self, |
|
370 | 374 | ) -> Result<RefMut<OwningDirstateMap>, DirstateError> { |
|
371 | 375 | self.dirstate_map.get_mut_or_init(self) |
|
372 | 376 | } |
|
373 | 377 | |
|
374 | 378 | pub fn changelog(&self) -> Result<Ref<Changelog>, HgError> { |
|
375 | 379 | self.changelog.get_or_init(self) |
|
376 | 380 | } |
|
377 | 381 | |
|
378 | 382 | pub fn changelog_mut(&self) -> Result<RefMut<Changelog>, HgError> { |
|
379 | 383 | self.changelog.get_mut_or_init(self) |
|
380 | 384 | } |
|
381 | 385 | |
|
382 | 386 | pub fn manifestlog(&self) -> Result<Ref<Manifestlog>, HgError> { |
|
383 | 387 | self.manifestlog.get_or_init(self) |
|
384 | 388 | } |
|
385 | 389 | |
|
386 | 390 | pub fn manifestlog_mut(&self) -> Result<RefMut<Manifestlog>, HgError> { |
|
387 | 391 | self.manifestlog.get_mut_or_init(self) |
|
388 | 392 | } |
|
389 | 393 | |
|
390 | 394 | /// Returns the manifest of the *changeset* with the given node ID |
|
391 | 395 | pub fn manifest_for_node( |
|
392 | 396 | &self, |
|
393 | 397 | node: impl Into<NodePrefix>, |
|
394 | 398 | ) -> Result<Manifest, RevlogError> { |
|
395 | 399 | self.manifestlog()?.data_for_node( |
|
396 | 400 | self.changelog()? |
|
397 | 401 | .data_for_node(node.into())? |
|
398 | 402 | .manifest_node()? |
|
399 | 403 | .into(), |
|
400 | 404 | ) |
|
401 | 405 | } |
|
402 | 406 | |
|
403 | 407 | /// Returns the manifest of the *changeset* with the given revision number |
|
404 | 408 | pub fn manifest_for_rev( |
|
405 | 409 | &self, |
|
406 | 410 | revision: Revision, |
|
407 | 411 | ) -> Result<Manifest, RevlogError> { |
|
408 | 412 | self.manifestlog()?.data_for_node( |
|
409 | 413 | self.changelog()? |
|
410 | 414 | .data_for_rev(revision)? |
|
411 | 415 | .manifest_node()? |
|
412 | 416 | .into(), |
|
413 | 417 | ) |
|
414 | 418 | } |
|
415 | 419 | |
|
416 | 420 | pub fn filelog(&self, path: &HgPath) -> Result<Filelog, HgError> { |
|
417 | 421 | Filelog::open(self, path) |
|
418 | 422 | } |
|
423 | ||
|
424 | /// Write to disk any updates that were made through `dirstate_map_mut`. | |
|
425 | /// | |
|
426 | /// The "wlock" must be held while calling this. | |
|
427 | /// See for example `try_with_wlock_no_wait`. | |
|
428 | /// | |
|
429 | /// TODO: have a `WritableRepo` type only accessible while holding the | |
|
430 | /// lock? | |
|
431 | pub fn write_dirstate(&self) -> Result<(), DirstateError> { | |
|
432 | let map = self.dirstate_map()?; | |
|
433 | // TODO: Maintain a `DirstateMap::dirty` flag, and return early here if | |
|
434 | // it’s unset | |
|
435 | let parents = self.dirstate_parents()?; | |
|
436 | let packed_dirstate = if self.has_dirstate_v2() { | |
|
437 | let uuid = self.dirstate_data_file_uuid.get_or_init(self)?; | |
|
438 | let mut uuid = uuid.as_ref(); | |
|
439 | let can_append = uuid.is_some(); | |
|
440 | let (data, tree_metadata, append) = map.pack_v2(can_append)?; | |
|
441 | if !append { | |
|
442 | uuid = None | |
|
443 | } | |
|
444 | let uuid = if let Some(uuid) = uuid { | |
|
445 | std::str::from_utf8(uuid) | |
|
446 | .map_err(|_| { | |
|
447 | HgError::corrupted("non-UTF-8 dirstate data file ID") | |
|
448 | })? | |
|
449 | .to_owned() | |
|
450 | } else { | |
|
451 | DirstateDocket::new_uid() | |
|
452 | }; | |
|
453 | let data_filename = format!("dirstate.{}", uuid); | |
|
454 | let data_filename = self.hg_vfs().join(data_filename); | |
|
455 | let mut options = std::fs::OpenOptions::new(); | |
|
456 | if append { | |
|
457 | options.append(true); | |
|
458 | } else { | |
|
459 | options.write(true).create_new(true); | |
|
460 | } | |
|
461 | let data_size = (|| { | |
|
462 | // TODO: loop and try another random ID if !append and this | |
|
463 | // returns `ErrorKind::AlreadyExists`? Collision chance of two | |
|
464 | // random IDs is one in 2**32 | |
|
465 | let mut file = options.open(&data_filename)?; | |
|
466 | file.write_all(&data)?; | |
|
467 | file.flush()?; | |
|
468 | // TODO: use https://doc.rust-lang.org/std/io/trait.Seek.html#method.stream_position when we require Rust 1.51+ | |
|
469 | file.seek(SeekFrom::Current(0)) | |
|
470 | })() | |
|
471 | .when_writing_file(&data_filename)?; | |
|
472 | DirstateDocket::serialize( | |
|
473 | parents, | |
|
474 | tree_metadata, | |
|
475 | data_size, | |
|
476 | uuid.as_bytes(), | |
|
477 | ) | |
|
478 | .map_err(|_: std::num::TryFromIntError| { | |
|
479 | HgError::corrupted("overflow in dirstate docket serialization") | |
|
480 | })? | |
|
481 | } else { | |
|
482 | map.pack_v1(parents)? | |
|
483 | }; | |
|
484 | self.hg_vfs().atomic_write("dirstate", &packed_dirstate)?; | |
|
485 | Ok(()) | |
|
486 | } | |
|
419 | 487 | } |
|
420 | 488 | |
|
421 | 489 | /// Lazily-initialized component of `Repo` with interior mutability |
|
422 | 490 | /// |
|
423 | 491 | /// This differs from `OnceCell` in that the value can still be "deinitialized" |
|
424 | 492 | /// later by setting its inner `Option` to `None`. |
|
425 | 493 | struct LazyCell<T, E> { |
|
426 | 494 | value: RefCell<Option<T>>, |
|
427 | 495 | // `Fn`s that don’t capture environment are zero-size, so this box does |
|
428 | 496 | // not allocate: |
|
429 | 497 | init: Box<dyn Fn(&Repo) -> Result<T, E>>, |
|
430 | 498 | } |
|
431 | 499 | |
|
432 | 500 | impl<T, E> LazyCell<T, E> { |
|
433 | 501 | fn new(init: impl Fn(&Repo) -> Result<T, E> + 'static) -> Self { |
|
434 | 502 | Self { |
|
435 | 503 | value: RefCell::new(None), |
|
436 | 504 | init: Box::new(init), |
|
437 | 505 | } |
|
438 | 506 | } |
|
439 | 507 | |
|
440 | 508 | fn set(&self, value: T) { |
|
441 | 509 | *self.value.borrow_mut() = Some(value) |
|
442 | 510 | } |
|
443 | 511 | |
|
444 | 512 | fn get_or_init(&self, repo: &Repo) -> Result<Ref<T>, E> { |
|
445 | 513 | let mut borrowed = self.value.borrow(); |
|
446 | 514 | if borrowed.is_none() { |
|
447 | 515 | drop(borrowed); |
|
448 | 516 | // Only use `borrow_mut` if it is really needed to avoid panic in |
|
449 | 517 | // case there is another outstanding borrow but mutation is not |
|
450 | 518 | // needed. |
|
451 | 519 | *self.value.borrow_mut() = Some((self.init)(repo)?); |
|
452 | 520 | borrowed = self.value.borrow() |
|
453 | 521 | } |
|
454 | 522 | Ok(Ref::map(borrowed, |option| option.as_ref().unwrap())) |
|
455 | 523 | } |
|
456 | 524 | |
|
457 | 525 | fn get_mut_or_init(&self, repo: &Repo) -> Result<RefMut<T>, E> { |
|
458 | 526 | let mut borrowed = self.value.borrow_mut(); |
|
459 | 527 | if borrowed.is_none() { |
|
460 | 528 | *borrowed = Some((self.init)(repo)?); |
|
461 | 529 | } |
|
462 | 530 | Ok(RefMut::map(borrowed, |option| option.as_mut().unwrap())) |
|
463 | 531 | } |
|
464 | 532 | } |
@@ -1,415 +1,421 b'' | |||
|
1 | 1 | // Copyright 2019-2020 Georges Racinet <georges.racinet@octobus.net> |
|
2 | 2 | // |
|
3 | 3 | // This software may be used and distributed according to the terms of the |
|
4 | 4 | // GNU General Public License version 2 or any later version. |
|
5 | 5 | |
|
6 | 6 | //! Definitions and utilities for Revision nodes |
|
7 | 7 | //! |
|
8 | 8 | //! In Mercurial code base, it is customary to call "a node" the binary SHA |
|
9 | 9 | //! of a revision. |
|
10 | 10 | |
|
11 | 11 | use crate::errors::HgError; |
|
12 | 12 | use bytes_cast::BytesCast; |
|
13 | 13 | use std::convert::{TryFrom, TryInto}; |
|
14 | 14 | use std::fmt; |
|
15 | 15 | |
|
16 | 16 | /// The length in bytes of a `Node` |
|
17 | 17 | /// |
|
18 | 18 | /// This constant is meant to ease refactors of this module, and |
|
19 | 19 | /// are private so that calling code does not expect all nodes have |
|
20 | 20 | /// the same size, should we support several formats concurrently in |
|
21 | 21 | /// the future. |
|
22 | 22 | pub const NODE_BYTES_LENGTH: usize = 20; |
|
23 | 23 | |
|
24 | 24 | /// Id of the null node. |
|
25 | 25 | /// |
|
26 | 26 | /// Used to indicate the absence of node. |
|
27 | 27 | pub const NULL_NODE_ID: [u8; NODE_BYTES_LENGTH] = [0u8; NODE_BYTES_LENGTH]; |
|
28 | 28 | |
|
29 | 29 | /// The length in bytes of a `Node` |
|
30 | 30 | /// |
|
31 | 31 | /// see also `NODES_BYTES_LENGTH` about it being private. |
|
32 | 32 | const NODE_NYBBLES_LENGTH: usize = 2 * NODE_BYTES_LENGTH; |
|
33 | 33 | |
|
34 | 34 | /// Default for UI presentation |
|
35 | 35 | const SHORT_PREFIX_DEFAULT_NYBBLES_LENGTH: u8 = 12; |
|
36 | 36 | |
|
37 | 37 | /// Private alias for readability and to ease future change |
|
38 | 38 | type NodeData = [u8; NODE_BYTES_LENGTH]; |
|
39 | 39 | |
|
40 | 40 | /// Binary revision SHA |
|
41 | 41 | /// |
|
42 | 42 | /// ## Future changes of hash size |
|
43 | 43 | /// |
|
44 | 44 | /// To accomodate future changes of hash size, Rust callers |
|
45 | 45 | /// should use the conversion methods at the boundaries (FFI, actual |
|
46 | 46 | /// computation of hashes and I/O) only, and only if required. |
|
47 | 47 | /// |
|
48 | 48 | /// All other callers outside of unit tests should just handle `Node` values |
|
49 | 49 | /// and never make any assumption on the actual length, using [`nybbles_len`] |
|
50 | 50 | /// if they need a loop boundary. |
|
51 | 51 | /// |
|
52 | 52 | /// All methods that create a `Node` either take a type that enforces |
|
53 | 53 | /// the size or return an error at runtime. |
|
54 | 54 | /// |
|
55 | 55 | /// [`nybbles_len`]: #method.nybbles_len |
|
56 | 56 | #[derive(Copy, Clone, Debug, PartialEq, BytesCast, derive_more::From)] |
|
57 | 57 | #[repr(transparent)] |
|
58 | 58 | pub struct Node { |
|
59 | 59 | data: NodeData, |
|
60 | 60 | } |
|
61 | 61 | |
|
62 | 62 | /// The node value for NULL_REVISION |
|
63 | 63 | pub const NULL_NODE: Node = Node { |
|
64 | 64 | data: [0; NODE_BYTES_LENGTH], |
|
65 | 65 | }; |
|
66 | 66 | |
|
67 | 67 | /// Return an error if the slice has an unexpected length |
|
68 | 68 | impl<'a> TryFrom<&'a [u8]> for &'a Node { |
|
69 | 69 | type Error = (); |
|
70 | 70 | |
|
71 | 71 | #[inline] |
|
72 | 72 | fn try_from(bytes: &'a [u8]) -> Result<Self, Self::Error> { |
|
73 | 73 | match Node::from_bytes(bytes) { |
|
74 | 74 | Ok((node, rest)) if rest.is_empty() => Ok(node), |
|
75 | 75 | _ => Err(()), |
|
76 | 76 | } |
|
77 | 77 | } |
|
78 | 78 | } |
|
79 | 79 | |
|
80 | 80 | /// Return an error if the slice has an unexpected length |
|
81 | 81 | impl TryFrom<&'_ [u8]> for Node { |
|
82 | 82 | type Error = std::array::TryFromSliceError; |
|
83 | 83 | |
|
84 | 84 | #[inline] |
|
85 | 85 | fn try_from(bytes: &'_ [u8]) -> Result<Self, Self::Error> { |
|
86 | 86 | let data = bytes.try_into()?; |
|
87 | 87 | Ok(Self { data }) |
|
88 | 88 | } |
|
89 | 89 | } |
|
90 | 90 | |
|
91 | 91 | impl From<&'_ NodeData> for Node { |
|
92 | 92 | #[inline] |
|
93 | 93 | fn from(data: &'_ NodeData) -> Self { |
|
94 | 94 | Self { data: *data } |
|
95 | 95 | } |
|
96 | 96 | } |
|
97 | 97 | |
|
98 | 98 | impl fmt::LowerHex for Node { |
|
99 | 99 | fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { |
|
100 | 100 | for &byte in &self.data { |
|
101 | 101 | write!(f, "{:02x}", byte)? |
|
102 | 102 | } |
|
103 | 103 | Ok(()) |
|
104 | 104 | } |
|
105 | 105 | } |
|
106 | 106 | |
|
107 | 107 | #[derive(Debug)] |
|
108 | 108 | pub struct FromHexError; |
|
109 | 109 | |
|
110 | 110 | /// Low level utility function, also for prefixes |
|
111 | 111 | fn get_nybble(s: &[u8], i: usize) -> u8 { |
|
112 | 112 | if i % 2 == 0 { |
|
113 | 113 | s[i / 2] >> 4 |
|
114 | 114 | } else { |
|
115 | 115 | s[i / 2] & 0x0f |
|
116 | 116 | } |
|
117 | 117 | } |
|
118 | 118 | |
|
119 | 119 | impl Node { |
|
120 | 120 | /// Retrieve the `i`th half-byte of the binary data. |
|
121 | 121 | /// |
|
122 | 122 | /// This is also the `i`th hexadecimal digit in numeric form, |
|
123 | 123 | /// also called a [nybble](https://en.wikipedia.org/wiki/Nibble). |
|
124 | 124 | pub fn get_nybble(&self, i: usize) -> u8 { |
|
125 | 125 | get_nybble(&self.data, i) |
|
126 | 126 | } |
|
127 | 127 | |
|
128 | 128 | /// Length of the data, in nybbles |
|
129 | 129 | pub fn nybbles_len(&self) -> usize { |
|
130 | 130 | // public exposure as an instance method only, so that we can |
|
131 | 131 | // easily support several sizes of hashes if needed in the future. |
|
132 | 132 | NODE_NYBBLES_LENGTH |
|
133 | 133 | } |
|
134 | 134 | |
|
135 | 135 | /// Convert from hexadecimal string representation |
|
136 | 136 | /// |
|
137 | 137 | /// Exact length is required. |
|
138 | 138 | /// |
|
139 | 139 | /// To be used in FFI and I/O only, in order to facilitate future |
|
140 | 140 | /// changes of hash format. |
|
141 | 141 | pub fn from_hex(hex: impl AsRef<[u8]>) -> Result<Node, FromHexError> { |
|
142 | 142 | let prefix = NodePrefix::from_hex(hex)?; |
|
143 | 143 | if prefix.nybbles_len() == NODE_NYBBLES_LENGTH { |
|
144 | 144 | Ok(Self { data: prefix.data }) |
|
145 | 145 | } else { |
|
146 | 146 | Err(FromHexError) |
|
147 | 147 | } |
|
148 | 148 | } |
|
149 | 149 | |
|
150 | 150 | /// `from_hex`, but for input from an internal file of the repository such |
|
151 | 151 | /// as a changelog or manifest entry. |
|
152 | 152 | /// |
|
153 | 153 | /// An error is treated as repository corruption. |
|
154 | 154 | pub fn from_hex_for_repo(hex: impl AsRef<[u8]>) -> Result<Node, HgError> { |
|
155 | 155 | Self::from_hex(hex.as_ref()).map_err(|FromHexError| { |
|
156 | 156 | HgError::CorruptedRepository(format!( |
|
157 | 157 | "Expected a full hexadecimal node ID, found {}", |
|
158 | 158 | String::from_utf8_lossy(hex.as_ref()) |
|
159 | 159 | )) |
|
160 | 160 | }) |
|
161 | 161 | } |
|
162 | 162 | |
|
163 | 163 | /// Provide access to binary data |
|
164 | 164 | /// |
|
165 | 165 | /// This is needed by FFI layers, for instance to return expected |
|
166 | 166 | /// binary values to Python. |
|
167 | 167 | pub fn as_bytes(&self) -> &[u8] { |
|
168 | 168 | &self.data |
|
169 | 169 | } |
|
170 | 170 | |
|
171 | 171 | pub fn short(&self) -> NodePrefix { |
|
172 | 172 | NodePrefix { |
|
173 | 173 | nybbles_len: SHORT_PREFIX_DEFAULT_NYBBLES_LENGTH, |
|
174 | 174 | data: self.data, |
|
175 | 175 | } |
|
176 | 176 | } |
|
177 | ||
|
178 | pub fn pad_to_256_bits(&self) -> [u8; 32] { | |
|
179 | let mut bits = [0; 32]; | |
|
180 | bits[..NODE_BYTES_LENGTH].copy_from_slice(&self.data); | |
|
181 | bits | |
|
182 | } | |
|
177 | 183 | } |
|
178 | 184 | |
|
179 | 185 | /// The beginning of a binary revision SHA. |
|
180 | 186 | /// |
|
181 | 187 | /// Since it can potentially come from an hexadecimal representation with |
|
182 | 188 | /// odd length, it needs to carry around whether the last 4 bits are relevant |
|
183 | 189 | /// or not. |
|
184 | 190 | #[derive(Debug, PartialEq, Copy, Clone)] |
|
185 | 191 | pub struct NodePrefix { |
|
186 | 192 | /// In `1..=NODE_NYBBLES_LENGTH` |
|
187 | 193 | nybbles_len: u8, |
|
188 | 194 | /// The first `4 * length_in_nybbles` bits are used (considering bits |
|
189 | 195 | /// within a bytes in big-endian: most significant first), the rest |
|
190 | 196 | /// are zero. |
|
191 | 197 | data: NodeData, |
|
192 | 198 | } |
|
193 | 199 | |
|
194 | 200 | impl NodePrefix { |
|
195 | 201 | /// Convert from hexadecimal string representation |
|
196 | 202 | /// |
|
197 | 203 | /// Similarly to `hex::decode`, can be used with Unicode string types |
|
198 | 204 | /// (`String`, `&str`) as well as bytes. |
|
199 | 205 | /// |
|
200 | 206 | /// To be used in FFI and I/O only, in order to facilitate future |
|
201 | 207 | /// changes of hash format. |
|
202 | 208 | pub fn from_hex(hex: impl AsRef<[u8]>) -> Result<Self, FromHexError> { |
|
203 | 209 | let hex = hex.as_ref(); |
|
204 | 210 | let len = hex.len(); |
|
205 | 211 | if len > NODE_NYBBLES_LENGTH || len == 0 { |
|
206 | 212 | return Err(FromHexError); |
|
207 | 213 | } |
|
208 | 214 | |
|
209 | 215 | let mut data = [0; NODE_BYTES_LENGTH]; |
|
210 | 216 | let mut nybbles_len = 0; |
|
211 | 217 | for &ascii_byte in hex { |
|
212 | 218 | let nybble = match char::from(ascii_byte).to_digit(16) { |
|
213 | 219 | Some(digit) => digit as u8, |
|
214 | 220 | None => return Err(FromHexError), |
|
215 | 221 | }; |
|
216 | 222 | // Fill in the upper half of a byte first, then the lower half. |
|
217 | 223 | let shift = if nybbles_len % 2 == 0 { 4 } else { 0 }; |
|
218 | 224 | data[nybbles_len as usize / 2] |= nybble << shift; |
|
219 | 225 | nybbles_len += 1; |
|
220 | 226 | } |
|
221 | 227 | Ok(Self { data, nybbles_len }) |
|
222 | 228 | } |
|
223 | 229 | |
|
224 | 230 | pub fn nybbles_len(&self) -> usize { |
|
225 | 231 | self.nybbles_len as _ |
|
226 | 232 | } |
|
227 | 233 | |
|
228 | 234 | pub fn is_prefix_of(&self, node: &Node) -> bool { |
|
229 | 235 | let full_bytes = self.nybbles_len() / 2; |
|
230 | 236 | if self.data[..full_bytes] != node.data[..full_bytes] { |
|
231 | 237 | return false; |
|
232 | 238 | } |
|
233 | 239 | if self.nybbles_len() % 2 == 0 { |
|
234 | 240 | return true; |
|
235 | 241 | } |
|
236 | 242 | let last = self.nybbles_len() - 1; |
|
237 | 243 | self.get_nybble(last) == node.get_nybble(last) |
|
238 | 244 | } |
|
239 | 245 | |
|
240 | 246 | /// Retrieve the `i`th half-byte from the prefix. |
|
241 | 247 | /// |
|
242 | 248 | /// This is also the `i`th hexadecimal digit in numeric form, |
|
243 | 249 | /// also called a [nybble](https://en.wikipedia.org/wiki/Nibble). |
|
244 | 250 | pub fn get_nybble(&self, i: usize) -> u8 { |
|
245 | 251 | assert!(i < self.nybbles_len()); |
|
246 | 252 | get_nybble(&self.data, i) |
|
247 | 253 | } |
|
248 | 254 | |
|
249 | 255 | fn iter_nybbles(&self) -> impl Iterator<Item = u8> + '_ { |
|
250 | 256 | (0..self.nybbles_len()).map(move |i| get_nybble(&self.data, i)) |
|
251 | 257 | } |
|
252 | 258 | |
|
253 | 259 | /// Return the index first nybble that's different from `node` |
|
254 | 260 | /// |
|
255 | 261 | /// If the return value is `None` that means that `self` is |
|
256 | 262 | /// a prefix of `node`, but the current method is a bit slower |
|
257 | 263 | /// than `is_prefix_of`. |
|
258 | 264 | /// |
|
259 | 265 | /// Returned index is as in `get_nybble`, i.e., starting at 0. |
|
260 | 266 | pub fn first_different_nybble(&self, node: &Node) -> Option<usize> { |
|
261 | 267 | self.iter_nybbles() |
|
262 | 268 | .zip(NodePrefix::from(*node).iter_nybbles()) |
|
263 | 269 | .position(|(a, b)| a != b) |
|
264 | 270 | } |
|
265 | 271 | } |
|
266 | 272 | |
|
267 | 273 | impl fmt::LowerHex for NodePrefix { |
|
268 | 274 | fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { |
|
269 | 275 | let full_bytes = self.nybbles_len() / 2; |
|
270 | 276 | for &byte in &self.data[..full_bytes] { |
|
271 | 277 | write!(f, "{:02x}", byte)? |
|
272 | 278 | } |
|
273 | 279 | if self.nybbles_len() % 2 == 1 { |
|
274 | 280 | let last = self.nybbles_len() - 1; |
|
275 | 281 | write!(f, "{:x}", self.get_nybble(last))? |
|
276 | 282 | } |
|
277 | 283 | Ok(()) |
|
278 | 284 | } |
|
279 | 285 | } |
|
280 | 286 | |
|
281 | 287 | /// A shortcut for full `Node` references |
|
282 | 288 | impl From<&'_ Node> for NodePrefix { |
|
283 | 289 | fn from(node: &'_ Node) -> Self { |
|
284 | 290 | NodePrefix { |
|
285 | 291 | nybbles_len: node.nybbles_len() as _, |
|
286 | 292 | data: node.data, |
|
287 | 293 | } |
|
288 | 294 | } |
|
289 | 295 | } |
|
290 | 296 | |
|
291 | 297 | /// A shortcut for full `Node` references |
|
292 | 298 | impl From<Node> for NodePrefix { |
|
293 | 299 | fn from(node: Node) -> Self { |
|
294 | 300 | NodePrefix { |
|
295 | 301 | nybbles_len: node.nybbles_len() as _, |
|
296 | 302 | data: node.data, |
|
297 | 303 | } |
|
298 | 304 | } |
|
299 | 305 | } |
|
300 | 306 | |
|
301 | 307 | impl PartialEq<Node> for NodePrefix { |
|
302 | 308 | fn eq(&self, other: &Node) -> bool { |
|
303 | 309 | Self::from(*other) == *self |
|
304 | 310 | } |
|
305 | 311 | } |
|
306 | 312 | |
|
307 | 313 | #[cfg(test)] |
|
308 | 314 | mod tests { |
|
309 | 315 | use super::*; |
|
310 | 316 | |
|
311 | 317 | const SAMPLE_NODE_HEX: &str = "0123456789abcdeffedcba9876543210deadbeef"; |
|
312 | 318 | const SAMPLE_NODE: Node = Node { |
|
313 | 319 | data: [ |
|
314 | 320 | 0x01, 0x23, 0x45, 0x67, 0x89, 0xab, 0xcd, 0xef, 0xfe, 0xdc, 0xba, |
|
315 | 321 | 0x98, 0x76, 0x54, 0x32, 0x10, 0xde, 0xad, 0xbe, 0xef, |
|
316 | 322 | ], |
|
317 | 323 | }; |
|
318 | 324 | |
|
319 | 325 | /// Pad an hexadecimal string to reach `NODE_NYBBLES_LENGTH` |
|
320 | 326 | /// The padding is made with zeros. |
|
321 | 327 | pub fn hex_pad_right(hex: &str) -> String { |
|
322 | 328 | let mut res = hex.to_string(); |
|
323 | 329 | while res.len() < NODE_NYBBLES_LENGTH { |
|
324 | 330 | res.push('0'); |
|
325 | 331 | } |
|
326 | 332 | res |
|
327 | 333 | } |
|
328 | 334 | |
|
329 | 335 | #[test] |
|
330 | 336 | fn test_node_from_hex() { |
|
331 | 337 | let not_hex = "012... oops"; |
|
332 | 338 | let too_short = "0123"; |
|
333 | 339 | let too_long = format!("{}0", SAMPLE_NODE_HEX); |
|
334 | 340 | assert_eq!(Node::from_hex(SAMPLE_NODE_HEX).unwrap(), SAMPLE_NODE); |
|
335 | 341 | assert!(Node::from_hex(not_hex).is_err()); |
|
336 | 342 | assert!(Node::from_hex(too_short).is_err()); |
|
337 | 343 | assert!(Node::from_hex(&too_long).is_err()); |
|
338 | 344 | } |
|
339 | 345 | |
|
340 | 346 | #[test] |
|
341 | 347 | fn test_node_encode_hex() { |
|
342 | 348 | assert_eq!(format!("{:x}", SAMPLE_NODE), SAMPLE_NODE_HEX); |
|
343 | 349 | } |
|
344 | 350 | |
|
345 | 351 | #[test] |
|
346 | 352 | fn test_prefix_from_to_hex() -> Result<(), FromHexError> { |
|
347 | 353 | assert_eq!(format!("{:x}", NodePrefix::from_hex("0e1")?), "0e1"); |
|
348 | 354 | assert_eq!(format!("{:x}", NodePrefix::from_hex("0e1a")?), "0e1a"); |
|
349 | 355 | assert_eq!( |
|
350 | 356 | format!("{:x}", NodePrefix::from_hex(SAMPLE_NODE_HEX)?), |
|
351 | 357 | SAMPLE_NODE_HEX |
|
352 | 358 | ); |
|
353 | 359 | Ok(()) |
|
354 | 360 | } |
|
355 | 361 | |
|
356 | 362 | #[test] |
|
357 | 363 | fn test_prefix_from_hex_errors() { |
|
358 | 364 | assert!(NodePrefix::from_hex("testgr").is_err()); |
|
359 | 365 | let mut long = format!("{:x}", NULL_NODE); |
|
360 | 366 | long.push('c'); |
|
361 | 367 | assert!(NodePrefix::from_hex(&long).is_err()) |
|
362 | 368 | } |
|
363 | 369 | |
|
364 | 370 | #[test] |
|
365 | 371 | fn test_is_prefix_of() -> Result<(), FromHexError> { |
|
366 | 372 | let mut node_data = [0; NODE_BYTES_LENGTH]; |
|
367 | 373 | node_data[0] = 0x12; |
|
368 | 374 | node_data[1] = 0xca; |
|
369 | 375 | let node = Node::from(node_data); |
|
370 | 376 | assert!(NodePrefix::from_hex("12")?.is_prefix_of(&node)); |
|
371 | 377 | assert!(!NodePrefix::from_hex("1a")?.is_prefix_of(&node)); |
|
372 | 378 | assert!(NodePrefix::from_hex("12c")?.is_prefix_of(&node)); |
|
373 | 379 | assert!(!NodePrefix::from_hex("12d")?.is_prefix_of(&node)); |
|
374 | 380 | Ok(()) |
|
375 | 381 | } |
|
376 | 382 | |
|
377 | 383 | #[test] |
|
378 | 384 | fn test_get_nybble() -> Result<(), FromHexError> { |
|
379 | 385 | let prefix = NodePrefix::from_hex("dead6789cafe")?; |
|
380 | 386 | assert_eq!(prefix.get_nybble(0), 13); |
|
381 | 387 | assert_eq!(prefix.get_nybble(7), 9); |
|
382 | 388 | Ok(()) |
|
383 | 389 | } |
|
384 | 390 | |
|
385 | 391 | #[test] |
|
386 | 392 | fn test_first_different_nybble_even_prefix() { |
|
387 | 393 | let prefix = NodePrefix::from_hex("12ca").unwrap(); |
|
388 | 394 | let mut node = Node::from([0; NODE_BYTES_LENGTH]); |
|
389 | 395 | assert_eq!(prefix.first_different_nybble(&node), Some(0)); |
|
390 | 396 | node.data[0] = 0x13; |
|
391 | 397 | assert_eq!(prefix.first_different_nybble(&node), Some(1)); |
|
392 | 398 | node.data[0] = 0x12; |
|
393 | 399 | assert_eq!(prefix.first_different_nybble(&node), Some(2)); |
|
394 | 400 | node.data[1] = 0xca; |
|
395 | 401 | // now it is a prefix |
|
396 | 402 | assert_eq!(prefix.first_different_nybble(&node), None); |
|
397 | 403 | } |
|
398 | 404 | |
|
399 | 405 | #[test] |
|
400 | 406 | fn test_first_different_nybble_odd_prefix() { |
|
401 | 407 | let prefix = NodePrefix::from_hex("12c").unwrap(); |
|
402 | 408 | let mut node = Node::from([0; NODE_BYTES_LENGTH]); |
|
403 | 409 | assert_eq!(prefix.first_different_nybble(&node), Some(0)); |
|
404 | 410 | node.data[0] = 0x13; |
|
405 | 411 | assert_eq!(prefix.first_different_nybble(&node), Some(1)); |
|
406 | 412 | node.data[0] = 0x12; |
|
407 | 413 | assert_eq!(prefix.first_different_nybble(&node), Some(2)); |
|
408 | 414 | node.data[1] = 0xca; |
|
409 | 415 | // now it is a prefix |
|
410 | 416 | assert_eq!(prefix.first_different_nybble(&node), None); |
|
411 | 417 | } |
|
412 | 418 | } |
|
413 | 419 | |
|
414 | 420 | #[cfg(test)] |
|
415 | 421 | pub use tests::hex_pad_right; |
@@ -1,499 +1,499 b'' | |||
|
1 | 1 | // dirstate_map.rs |
|
2 | 2 | // |
|
3 | 3 | // Copyright 2019 Raphaël Gomès <rgomes@octobus.net> |
|
4 | 4 | // |
|
5 | 5 | // This software may be used and distributed according to the terms of the |
|
6 | 6 | // GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | //! Bindings for the `hg::dirstate::dirstate_map` file provided by the |
|
9 | 9 | //! `hg-core` package. |
|
10 | 10 | |
|
11 | 11 | use std::cell::{RefCell, RefMut}; |
|
12 | 12 | use std::convert::TryInto; |
|
13 | 13 | |
|
14 | 14 | use cpython::{ |
|
15 | 15 | exc, PyBool, PyBytes, PyClone, PyDict, PyErr, PyList, PyNone, PyObject, |
|
16 | 16 | PyResult, Python, PythonObject, ToPyObject, UnsafePyLeaked, |
|
17 | 17 | }; |
|
18 | 18 | |
|
19 | 19 | use crate::{ |
|
20 | 20 | dirstate::copymap::{CopyMap, CopyMapItemsIterator, CopyMapKeysIterator}, |
|
21 | 21 | dirstate::item::DirstateItem, |
|
22 | 22 | pybytes_deref::PyBytesDeref, |
|
23 | 23 | }; |
|
24 | 24 | use hg::{ |
|
25 | 25 | dirstate::StateMapIter, |
|
26 | 26 | dirstate_tree::dirstate_map::DirstateMap as TreeDirstateMap, |
|
27 | 27 | dirstate_tree::on_disk::DirstateV2ParseError, |
|
28 | 28 | dirstate_tree::owning::OwningDirstateMap, |
|
29 | 29 | revlog::Node, |
|
30 | 30 | utils::files::normalize_case, |
|
31 | 31 | utils::hg_path::{HgPath, HgPathBuf}, |
|
32 | 32 | DirstateEntry, DirstateError, DirstateParents, EntryState, |
|
33 | 33 | }; |
|
34 | 34 | |
|
35 | 35 | // TODO |
|
36 | 36 | // This object needs to share references to multiple members of its Rust |
|
37 | 37 | // inner struct, namely `copy_map`, `dirs` and `all_dirs`. |
|
38 | 38 | // Right now `CopyMap` is done, but it needs to have an explicit reference |
|
39 | 39 | // to `RustDirstateMap` which itself needs to have an encapsulation for |
|
40 | 40 | // every method in `CopyMap` (copymapcopy, etc.). |
|
41 | 41 | // This is ugly and hard to maintain. |
|
42 | 42 | // The same logic applies to `dirs` and `all_dirs`, however the `Dirs` |
|
43 | 43 | // `py_class!` is already implemented and does not mention |
|
44 | 44 | // `RustDirstateMap`, rightfully so. |
|
45 | 45 | // All attributes also have to have a separate refcount data attribute for |
|
46 | 46 | // leaks, with all methods that go along for reference sharing. |
|
47 | 47 | py_class!(pub class DirstateMap |py| { |
|
48 | 48 | @shared data inner: OwningDirstateMap; |
|
49 | 49 | |
|
50 | 50 | /// Returns a `(dirstate_map, parents)` tuple |
|
51 | 51 | @staticmethod |
|
52 | 52 | def new_v1( |
|
53 | 53 | on_disk: PyBytes, |
|
54 | 54 | ) -> PyResult<PyObject> { |
|
55 | 55 | let on_disk = PyBytesDeref::new(py, on_disk); |
|
56 | 56 | let mut map = OwningDirstateMap::new_empty(on_disk); |
|
57 | 57 | let (on_disk, map_placeholder) = map.get_pair_mut(); |
|
58 | 58 | |
|
59 | 59 | let (actual_map, parents) = TreeDirstateMap::new_v1(on_disk) |
|
60 | 60 | .map_err(|e| dirstate_error(py, e))?; |
|
61 | 61 | *map_placeholder = actual_map; |
|
62 | 62 | let map = Self::create_instance(py, map)?; |
|
63 | 63 | let parents = parents.map(|p| { |
|
64 | 64 | let p1 = PyBytes::new(py, p.p1.as_bytes()); |
|
65 | 65 | let p2 = PyBytes::new(py, p.p2.as_bytes()); |
|
66 | 66 | (p1, p2) |
|
67 | 67 | }); |
|
68 | 68 | Ok((map, parents).to_py_object(py).into_object()) |
|
69 | 69 | } |
|
70 | 70 | |
|
71 | 71 | /// Returns a DirstateMap |
|
72 | 72 | @staticmethod |
|
73 | 73 | def new_v2( |
|
74 | 74 | on_disk: PyBytes, |
|
75 | 75 | data_size: usize, |
|
76 | 76 | tree_metadata: PyBytes, |
|
77 | 77 | ) -> PyResult<PyObject> { |
|
78 | 78 | let dirstate_error = |e: DirstateError| { |
|
79 | 79 | PyErr::new::<exc::OSError, _>(py, format!("Dirstate error: {:?}", e)) |
|
80 | 80 | }; |
|
81 | 81 | let on_disk = PyBytesDeref::new(py, on_disk); |
|
82 | 82 | let mut map = OwningDirstateMap::new_empty(on_disk); |
|
83 | 83 | let (on_disk, map_placeholder) = map.get_pair_mut(); |
|
84 | 84 | *map_placeholder = TreeDirstateMap::new_v2( |
|
85 | 85 | on_disk, data_size, tree_metadata.data(py), |
|
86 | 86 | ).map_err(dirstate_error)?; |
|
87 | 87 | let map = Self::create_instance(py, map)?; |
|
88 | 88 | Ok(map.into_object()) |
|
89 | 89 | } |
|
90 | 90 | |
|
91 | 91 | def clear(&self) -> PyResult<PyObject> { |
|
92 | 92 | self.inner(py).borrow_mut().clear(); |
|
93 | 93 | Ok(py.None()) |
|
94 | 94 | } |
|
95 | 95 | |
|
96 | 96 | def get( |
|
97 | 97 | &self, |
|
98 | 98 | key: PyObject, |
|
99 | 99 | default: Option<PyObject> = None |
|
100 | 100 | ) -> PyResult<Option<PyObject>> { |
|
101 | 101 | let key = key.extract::<PyBytes>(py)?; |
|
102 | 102 | match self |
|
103 | 103 | .inner(py) |
|
104 | 104 | .borrow() |
|
105 | 105 | .get(HgPath::new(key.data(py))) |
|
106 | 106 | .map_err(|e| v2_error(py, e))? |
|
107 | 107 | { |
|
108 | 108 | Some(entry) => { |
|
109 | 109 | Ok(Some(DirstateItem::new_as_pyobject(py, entry)?)) |
|
110 | 110 | }, |
|
111 | 111 | None => Ok(default) |
|
112 | 112 | } |
|
113 | 113 | } |
|
114 | 114 | |
|
115 | 115 | def set_dirstate_item( |
|
116 | 116 | &self, |
|
117 | 117 | path: PyObject, |
|
118 | 118 | item: DirstateItem |
|
119 | 119 | ) -> PyResult<PyObject> { |
|
120 | 120 | let f = path.extract::<PyBytes>(py)?; |
|
121 | 121 | let filename = HgPath::new(f.data(py)); |
|
122 | 122 | self.inner(py) |
|
123 | 123 | .borrow_mut() |
|
124 | 124 | .set_entry(filename, item.get_entry(py)) |
|
125 | 125 | .map_err(|e| v2_error(py, e))?; |
|
126 | 126 | Ok(py.None()) |
|
127 | 127 | } |
|
128 | 128 | |
|
129 | 129 | def addfile( |
|
130 | 130 | &self, |
|
131 | 131 | f: PyBytes, |
|
132 | 132 | item: DirstateItem, |
|
133 | 133 | ) -> PyResult<PyNone> { |
|
134 | 134 | let filename = HgPath::new(f.data(py)); |
|
135 | 135 | let entry = item.get_entry(py); |
|
136 | 136 | self.inner(py) |
|
137 | 137 | .borrow_mut() |
|
138 | 138 | .add_file(filename, entry) |
|
139 | 139 | .map_err(|e |dirstate_error(py, e))?; |
|
140 | 140 | Ok(PyNone) |
|
141 | 141 | } |
|
142 | 142 | |
|
143 | 143 | def removefile( |
|
144 | 144 | &self, |
|
145 | 145 | f: PyObject, |
|
146 | 146 | in_merge: PyObject |
|
147 | 147 | ) -> PyResult<PyObject> { |
|
148 | 148 | self.inner(py).borrow_mut() |
|
149 | 149 | .remove_file( |
|
150 | 150 | HgPath::new(f.extract::<PyBytes>(py)?.data(py)), |
|
151 | 151 | in_merge.extract::<PyBool>(py)?.is_true(), |
|
152 | 152 | ) |
|
153 | 153 | .or_else(|_| { |
|
154 | 154 | Err(PyErr::new::<exc::OSError, _>( |
|
155 | 155 | py, |
|
156 | 156 | "Dirstate error".to_string(), |
|
157 | 157 | )) |
|
158 | 158 | })?; |
|
159 | 159 | Ok(py.None()) |
|
160 | 160 | } |
|
161 | 161 | |
|
162 | 162 | def drop_item_and_copy_source( |
|
163 | 163 | &self, |
|
164 | 164 | f: PyBytes, |
|
165 | 165 | ) -> PyResult<PyNone> { |
|
166 | 166 | self.inner(py) |
|
167 | 167 | .borrow_mut() |
|
168 | 168 | .drop_entry_and_copy_source(HgPath::new(f.data(py))) |
|
169 | 169 | .map_err(|e |dirstate_error(py, e))?; |
|
170 | 170 | Ok(PyNone) |
|
171 | 171 | } |
|
172 | 172 | |
|
173 | 173 | def hastrackeddir(&self, d: PyObject) -> PyResult<PyBool> { |
|
174 | 174 | let d = d.extract::<PyBytes>(py)?; |
|
175 | 175 | Ok(self.inner(py).borrow_mut() |
|
176 | 176 | .has_tracked_dir(HgPath::new(d.data(py))) |
|
177 | 177 | .map_err(|e| { |
|
178 | 178 | PyErr::new::<exc::ValueError, _>(py, e.to_string()) |
|
179 | 179 | })? |
|
180 | 180 | .to_py_object(py)) |
|
181 | 181 | } |
|
182 | 182 | |
|
183 | 183 | def hasdir(&self, d: PyObject) -> PyResult<PyBool> { |
|
184 | 184 | let d = d.extract::<PyBytes>(py)?; |
|
185 | 185 | Ok(self.inner(py).borrow_mut() |
|
186 | 186 | .has_dir(HgPath::new(d.data(py))) |
|
187 | 187 | .map_err(|e| { |
|
188 | 188 | PyErr::new::<exc::ValueError, _>(py, e.to_string()) |
|
189 | 189 | })? |
|
190 | 190 | .to_py_object(py)) |
|
191 | 191 | } |
|
192 | 192 | |
|
193 | 193 | def write_v1( |
|
194 | 194 | &self, |
|
195 | 195 | p1: PyObject, |
|
196 | 196 | p2: PyObject, |
|
197 | 197 | ) -> PyResult<PyBytes> { |
|
198 | 198 | let inner = self.inner(py).borrow(); |
|
199 | 199 | let parents = DirstateParents { |
|
200 | 200 | p1: extract_node_id(py, &p1)?, |
|
201 | 201 | p2: extract_node_id(py, &p2)?, |
|
202 | 202 | }; |
|
203 | 203 | let result = inner.pack_v1(parents); |
|
204 | 204 | match result { |
|
205 | 205 | Ok(packed) => Ok(PyBytes::new(py, &packed)), |
|
206 | 206 | Err(_) => Err(PyErr::new::<exc::OSError, _>( |
|
207 | 207 | py, |
|
208 | 208 | "Dirstate error".to_string(), |
|
209 | 209 | )), |
|
210 | 210 | } |
|
211 | 211 | } |
|
212 | 212 | |
|
213 | 213 | /// Returns new data together with whether that data should be appended to |
|
214 | 214 | /// the existing data file whose content is at `self.on_disk` (True), |
|
215 | 215 | /// instead of written to a new data file (False). |
|
216 | 216 | def write_v2( |
|
217 | 217 | &self, |
|
218 | 218 | can_append: bool, |
|
219 | 219 | ) -> PyResult<PyObject> { |
|
220 | 220 | let inner = self.inner(py).borrow(); |
|
221 | 221 | let result = inner.pack_v2(can_append); |
|
222 | 222 | match result { |
|
223 | 223 | Ok((packed, tree_metadata, append)) => { |
|
224 | 224 | let packed = PyBytes::new(py, &packed); |
|
225 |
let tree_metadata = PyBytes::new(py, |
|
|
225 | let tree_metadata = PyBytes::new(py, tree_metadata.as_bytes()); | |
|
226 | 226 | let tuple = (packed, tree_metadata, append); |
|
227 | 227 | Ok(tuple.to_py_object(py).into_object()) |
|
228 | 228 | }, |
|
229 | 229 | Err(_) => Err(PyErr::new::<exc::OSError, _>( |
|
230 | 230 | py, |
|
231 | 231 | "Dirstate error".to_string(), |
|
232 | 232 | )), |
|
233 | 233 | } |
|
234 | 234 | } |
|
235 | 235 | |
|
236 | 236 | def filefoldmapasdict(&self) -> PyResult<PyDict> { |
|
237 | 237 | let dict = PyDict::new(py); |
|
238 | 238 | for item in self.inner(py).borrow_mut().iter() { |
|
239 | 239 | let (path, entry) = item.map_err(|e| v2_error(py, e))?; |
|
240 | 240 | if entry.state() != EntryState::Removed { |
|
241 | 241 | let key = normalize_case(path); |
|
242 | 242 | let value = path; |
|
243 | 243 | dict.set_item( |
|
244 | 244 | py, |
|
245 | 245 | PyBytes::new(py, key.as_bytes()).into_object(), |
|
246 | 246 | PyBytes::new(py, value.as_bytes()).into_object(), |
|
247 | 247 | )?; |
|
248 | 248 | } |
|
249 | 249 | } |
|
250 | 250 | Ok(dict) |
|
251 | 251 | } |
|
252 | 252 | |
|
253 | 253 | def __len__(&self) -> PyResult<usize> { |
|
254 | 254 | Ok(self.inner(py).borrow().len()) |
|
255 | 255 | } |
|
256 | 256 | |
|
257 | 257 | def __contains__(&self, key: PyObject) -> PyResult<bool> { |
|
258 | 258 | let key = key.extract::<PyBytes>(py)?; |
|
259 | 259 | self.inner(py) |
|
260 | 260 | .borrow() |
|
261 | 261 | .contains_key(HgPath::new(key.data(py))) |
|
262 | 262 | .map_err(|e| v2_error(py, e)) |
|
263 | 263 | } |
|
264 | 264 | |
|
265 | 265 | def __getitem__(&self, key: PyObject) -> PyResult<PyObject> { |
|
266 | 266 | let key = key.extract::<PyBytes>(py)?; |
|
267 | 267 | let key = HgPath::new(key.data(py)); |
|
268 | 268 | match self |
|
269 | 269 | .inner(py) |
|
270 | 270 | .borrow() |
|
271 | 271 | .get(key) |
|
272 | 272 | .map_err(|e| v2_error(py, e))? |
|
273 | 273 | { |
|
274 | 274 | Some(entry) => { |
|
275 | 275 | Ok(DirstateItem::new_as_pyobject(py, entry)?) |
|
276 | 276 | }, |
|
277 | 277 | None => Err(PyErr::new::<exc::KeyError, _>( |
|
278 | 278 | py, |
|
279 | 279 | String::from_utf8_lossy(key.as_bytes()), |
|
280 | 280 | )), |
|
281 | 281 | } |
|
282 | 282 | } |
|
283 | 283 | |
|
284 | 284 | def keys(&self) -> PyResult<DirstateMapKeysIterator> { |
|
285 | 285 | let leaked_ref = self.inner(py).leak_immutable(); |
|
286 | 286 | DirstateMapKeysIterator::from_inner( |
|
287 | 287 | py, |
|
288 | 288 | unsafe { leaked_ref.map(py, |o| o.iter()) }, |
|
289 | 289 | ) |
|
290 | 290 | } |
|
291 | 291 | |
|
292 | 292 | def items(&self) -> PyResult<DirstateMapItemsIterator> { |
|
293 | 293 | let leaked_ref = self.inner(py).leak_immutable(); |
|
294 | 294 | DirstateMapItemsIterator::from_inner( |
|
295 | 295 | py, |
|
296 | 296 | unsafe { leaked_ref.map(py, |o| o.iter()) }, |
|
297 | 297 | ) |
|
298 | 298 | } |
|
299 | 299 | |
|
300 | 300 | def __iter__(&self) -> PyResult<DirstateMapKeysIterator> { |
|
301 | 301 | let leaked_ref = self.inner(py).leak_immutable(); |
|
302 | 302 | DirstateMapKeysIterator::from_inner( |
|
303 | 303 | py, |
|
304 | 304 | unsafe { leaked_ref.map(py, |o| o.iter()) }, |
|
305 | 305 | ) |
|
306 | 306 | } |
|
307 | 307 | |
|
308 | 308 | // TODO all copymap* methods, see docstring above |
|
309 | 309 | def copymapcopy(&self) -> PyResult<PyDict> { |
|
310 | 310 | let dict = PyDict::new(py); |
|
311 | 311 | for item in self.inner(py).borrow().copy_map_iter() { |
|
312 | 312 | let (key, value) = item.map_err(|e| v2_error(py, e))?; |
|
313 | 313 | dict.set_item( |
|
314 | 314 | py, |
|
315 | 315 | PyBytes::new(py, key.as_bytes()), |
|
316 | 316 | PyBytes::new(py, value.as_bytes()), |
|
317 | 317 | )?; |
|
318 | 318 | } |
|
319 | 319 | Ok(dict) |
|
320 | 320 | } |
|
321 | 321 | |
|
322 | 322 | def copymapgetitem(&self, key: PyObject) -> PyResult<PyBytes> { |
|
323 | 323 | let key = key.extract::<PyBytes>(py)?; |
|
324 | 324 | match self |
|
325 | 325 | .inner(py) |
|
326 | 326 | .borrow() |
|
327 | 327 | .copy_map_get(HgPath::new(key.data(py))) |
|
328 | 328 | .map_err(|e| v2_error(py, e))? |
|
329 | 329 | { |
|
330 | 330 | Some(copy) => Ok(PyBytes::new(py, copy.as_bytes())), |
|
331 | 331 | None => Err(PyErr::new::<exc::KeyError, _>( |
|
332 | 332 | py, |
|
333 | 333 | String::from_utf8_lossy(key.data(py)), |
|
334 | 334 | )), |
|
335 | 335 | } |
|
336 | 336 | } |
|
337 | 337 | def copymap(&self) -> PyResult<CopyMap> { |
|
338 | 338 | CopyMap::from_inner(py, self.clone_ref(py)) |
|
339 | 339 | } |
|
340 | 340 | |
|
341 | 341 | def copymaplen(&self) -> PyResult<usize> { |
|
342 | 342 | Ok(self.inner(py).borrow().copy_map_len()) |
|
343 | 343 | } |
|
344 | 344 | def copymapcontains(&self, key: PyObject) -> PyResult<bool> { |
|
345 | 345 | let key = key.extract::<PyBytes>(py)?; |
|
346 | 346 | self.inner(py) |
|
347 | 347 | .borrow() |
|
348 | 348 | .copy_map_contains_key(HgPath::new(key.data(py))) |
|
349 | 349 | .map_err(|e| v2_error(py, e)) |
|
350 | 350 | } |
|
351 | 351 | def copymapget( |
|
352 | 352 | &self, |
|
353 | 353 | key: PyObject, |
|
354 | 354 | default: Option<PyObject> |
|
355 | 355 | ) -> PyResult<Option<PyObject>> { |
|
356 | 356 | let key = key.extract::<PyBytes>(py)?; |
|
357 | 357 | match self |
|
358 | 358 | .inner(py) |
|
359 | 359 | .borrow() |
|
360 | 360 | .copy_map_get(HgPath::new(key.data(py))) |
|
361 | 361 | .map_err(|e| v2_error(py, e))? |
|
362 | 362 | { |
|
363 | 363 | Some(copy) => Ok(Some( |
|
364 | 364 | PyBytes::new(py, copy.as_bytes()).into_object(), |
|
365 | 365 | )), |
|
366 | 366 | None => Ok(default), |
|
367 | 367 | } |
|
368 | 368 | } |
|
369 | 369 | def copymapsetitem( |
|
370 | 370 | &self, |
|
371 | 371 | key: PyObject, |
|
372 | 372 | value: PyObject |
|
373 | 373 | ) -> PyResult<PyObject> { |
|
374 | 374 | let key = key.extract::<PyBytes>(py)?; |
|
375 | 375 | let value = value.extract::<PyBytes>(py)?; |
|
376 | 376 | self.inner(py) |
|
377 | 377 | .borrow_mut() |
|
378 | 378 | .copy_map_insert( |
|
379 | 379 | HgPathBuf::from_bytes(key.data(py)), |
|
380 | 380 | HgPathBuf::from_bytes(value.data(py)), |
|
381 | 381 | ) |
|
382 | 382 | .map_err(|e| v2_error(py, e))?; |
|
383 | 383 | Ok(py.None()) |
|
384 | 384 | } |
|
385 | 385 | def copymappop( |
|
386 | 386 | &self, |
|
387 | 387 | key: PyObject, |
|
388 | 388 | default: Option<PyObject> |
|
389 | 389 | ) -> PyResult<Option<PyObject>> { |
|
390 | 390 | let key = key.extract::<PyBytes>(py)?; |
|
391 | 391 | match self |
|
392 | 392 | .inner(py) |
|
393 | 393 | .borrow_mut() |
|
394 | 394 | .copy_map_remove(HgPath::new(key.data(py))) |
|
395 | 395 | .map_err(|e| v2_error(py, e))? |
|
396 | 396 | { |
|
397 | 397 | Some(copy) => Ok(Some( |
|
398 | 398 | PyBytes::new(py, copy.as_bytes()).into_object(), |
|
399 | 399 | )), |
|
400 | 400 | None => Ok(default), |
|
401 | 401 | } |
|
402 | 402 | } |
|
403 | 403 | |
|
404 | 404 | def copymapiter(&self) -> PyResult<CopyMapKeysIterator> { |
|
405 | 405 | let leaked_ref = self.inner(py).leak_immutable(); |
|
406 | 406 | CopyMapKeysIterator::from_inner( |
|
407 | 407 | py, |
|
408 | 408 | unsafe { leaked_ref.map(py, |o| o.copy_map_iter()) }, |
|
409 | 409 | ) |
|
410 | 410 | } |
|
411 | 411 | |
|
412 | 412 | def copymapitemsiter(&self) -> PyResult<CopyMapItemsIterator> { |
|
413 | 413 | let leaked_ref = self.inner(py).leak_immutable(); |
|
414 | 414 | CopyMapItemsIterator::from_inner( |
|
415 | 415 | py, |
|
416 | 416 | unsafe { leaked_ref.map(py, |o| o.copy_map_iter()) }, |
|
417 | 417 | ) |
|
418 | 418 | } |
|
419 | 419 | |
|
420 | 420 | def tracked_dirs(&self) -> PyResult<PyList> { |
|
421 | 421 | let dirs = PyList::new(py, &[]); |
|
422 | 422 | for path in self.inner(py).borrow_mut().iter_tracked_dirs() |
|
423 | 423 | .map_err(|e |dirstate_error(py, e))? |
|
424 | 424 | { |
|
425 | 425 | let path = path.map_err(|e| v2_error(py, e))?; |
|
426 | 426 | let path = PyBytes::new(py, path.as_bytes()); |
|
427 | 427 | dirs.append(py, path.into_object()) |
|
428 | 428 | } |
|
429 | 429 | Ok(dirs) |
|
430 | 430 | } |
|
431 | 431 | |
|
432 | 432 | def debug_iter(&self, all: bool) -> PyResult<PyList> { |
|
433 | 433 | let dirs = PyList::new(py, &[]); |
|
434 | 434 | for item in self.inner(py).borrow().debug_iter(all) { |
|
435 | 435 | let (path, (state, mode, size, mtime)) = |
|
436 | 436 | item.map_err(|e| v2_error(py, e))?; |
|
437 | 437 | let path = PyBytes::new(py, path.as_bytes()); |
|
438 | 438 | let item = (path, state, mode, size, mtime); |
|
439 | 439 | dirs.append(py, item.to_py_object(py).into_object()) |
|
440 | 440 | } |
|
441 | 441 | Ok(dirs) |
|
442 | 442 | } |
|
443 | 443 | }); |
|
444 | 444 | |
|
445 | 445 | impl DirstateMap { |
|
446 | 446 | pub fn get_inner_mut<'a>( |
|
447 | 447 | &'a self, |
|
448 | 448 | py: Python<'a>, |
|
449 | 449 | ) -> RefMut<'a, OwningDirstateMap> { |
|
450 | 450 | self.inner(py).borrow_mut() |
|
451 | 451 | } |
|
452 | 452 | fn translate_key( |
|
453 | 453 | py: Python, |
|
454 | 454 | res: Result<(&HgPath, DirstateEntry), DirstateV2ParseError>, |
|
455 | 455 | ) -> PyResult<Option<PyBytes>> { |
|
456 | 456 | let (f, _entry) = res.map_err(|e| v2_error(py, e))?; |
|
457 | 457 | Ok(Some(PyBytes::new(py, f.as_bytes()))) |
|
458 | 458 | } |
|
459 | 459 | fn translate_key_value( |
|
460 | 460 | py: Python, |
|
461 | 461 | res: Result<(&HgPath, DirstateEntry), DirstateV2ParseError>, |
|
462 | 462 | ) -> PyResult<Option<(PyBytes, PyObject)>> { |
|
463 | 463 | let (f, entry) = res.map_err(|e| v2_error(py, e))?; |
|
464 | 464 | Ok(Some(( |
|
465 | 465 | PyBytes::new(py, f.as_bytes()), |
|
466 | 466 | DirstateItem::new_as_pyobject(py, entry)?, |
|
467 | 467 | ))) |
|
468 | 468 | } |
|
469 | 469 | } |
|
470 | 470 | |
|
471 | 471 | py_shared_iterator!( |
|
472 | 472 | DirstateMapKeysIterator, |
|
473 | 473 | UnsafePyLeaked<StateMapIter<'static>>, |
|
474 | 474 | DirstateMap::translate_key, |
|
475 | 475 | Option<PyBytes> |
|
476 | 476 | ); |
|
477 | 477 | |
|
478 | 478 | py_shared_iterator!( |
|
479 | 479 | DirstateMapItemsIterator, |
|
480 | 480 | UnsafePyLeaked<StateMapIter<'static>>, |
|
481 | 481 | DirstateMap::translate_key_value, |
|
482 | 482 | Option<(PyBytes, PyObject)> |
|
483 | 483 | ); |
|
484 | 484 | |
|
485 | 485 | fn extract_node_id(py: Python, obj: &PyObject) -> PyResult<Node> { |
|
486 | 486 | let bytes = obj.extract::<PyBytes>(py)?; |
|
487 | 487 | match bytes.data(py).try_into() { |
|
488 | 488 | Ok(s) => Ok(s), |
|
489 | 489 | Err(e) => Err(PyErr::new::<exc::ValueError, _>(py, e.to_string())), |
|
490 | 490 | } |
|
491 | 491 | } |
|
492 | 492 | |
|
493 | 493 | pub(super) fn v2_error(py: Python<'_>, _: DirstateV2ParseError) -> PyErr { |
|
494 | 494 | PyErr::new::<exc::ValueError, _>(py, "corrupted dirstate-v2") |
|
495 | 495 | } |
|
496 | 496 | |
|
497 | 497 | fn dirstate_error(py: Python<'_>, e: DirstateError) -> PyErr { |
|
498 | 498 | PyErr::new::<exc::OSError, _>(py, format!("Dirstate error: {:?}", e)) |
|
499 | 499 | } |
General Comments 0
You need to be logged in to leave comments.
Login now