Show More
@@ -1,376 +1,484 b'' | |||
|
1 | 1 | The *dirstate* is what Mercurial uses internally to track |
|
2 | 2 | the state of files in the working directory, |
|
3 | 3 | such as set by commands like `hg add` and `hg rm`. |
|
4 | 4 | It also contains some cached data that help make `hg status` faster. |
|
5 | 5 | The name refers both to `.hg/dirstate` on the filesystem |
|
6 | 6 | and the corresponding data structure in memory while a Mercurial process |
|
7 | 7 | is running. |
|
8 | 8 | |
|
9 | 9 | The original file format, retroactively dubbed `dirstate-v1`, |
|
10 | 10 | is described at https://www.mercurial-scm.org/wiki/DirState. |
|
11 | 11 | It is made of a flat sequence of unordered variable-size entries, |
|
12 | 12 | so accessing any information in it requires parsing all of it. |
|
13 | 13 | Similarly, saving changes requires rewriting the entire file. |
|
14 | 14 | |
|
15 | 15 | The newer `dirsate-v2` file format is designed to fix these limitations |
|
16 | 16 | and make `hg status` faster. |
|
17 | 17 | |
|
18 | 18 | User guide |
|
19 | 19 | ========== |
|
20 | 20 | |
|
21 | 21 | Compatibility |
|
22 | 22 | ------------- |
|
23 | 23 | |
|
24 | 24 | The file format is experimental and may still change. |
|
25 | 25 | Different versions of Mercurial may not be compatible with each other |
|
26 | 26 | when working on a local repository that uses this format. |
|
27 | 27 | When using an incompatible version with the experimental format, |
|
28 | 28 | anything can happen including data corruption. |
|
29 | 29 | |
|
30 | 30 | Since the dirstate is entirely local and not relevant to the wire protocol, |
|
31 | 31 | `dirstate-v2` does not affect compatibility with remote Mercurial versions. |
|
32 | 32 | |
|
33 | 33 | When `share-safe` is enabled, different repositories sharing the same store |
|
34 | 34 | can use different dirstate formats. |
|
35 | 35 | |
|
36 | 36 | Enabling `dirsate-v2` for new local repositories |
|
37 | 37 | ------------------------------------------------ |
|
38 | 38 | |
|
39 | 39 | When creating a new local repository such as with `hg init` or `hg clone`, |
|
40 | 40 | the `exp-dirstate-v2` boolean in the `format` configuration section |
|
41 | 41 | controls whether to use this file format. |
|
42 | 42 | This is disabled by default as of this writing. |
|
43 | 43 | To enable it for a single repository, run for example:: |
|
44 | 44 | |
|
45 | 45 | $ hg init my-project --config format.exp-dirstate-v2=1 |
|
46 | 46 | |
|
47 | 47 | Checking the format of an existing local repsitory |
|
48 | 48 | -------------------------------------------------- |
|
49 | 49 | |
|
50 | 50 | The `debugformat` commands prints information about |
|
51 | 51 | which of multiple optional formats are used in the current repository, |
|
52 | 52 | including `dirstate-v2`:: |
|
53 | 53 | |
|
54 | 54 | $ hg debugformat |
|
55 | 55 | format-variant repo |
|
56 | 56 | fncache: yes |
|
57 | 57 | dirstate-v2: yes |
|
58 | 58 | [β¦] |
|
59 | 59 | |
|
60 | 60 | Upgrading or downgrading an existing local repository |
|
61 | 61 | ----------------------------------------------------- |
|
62 | 62 | |
|
63 | 63 | The `debugupgrade` command does various upgrades or downgrades |
|
64 | 64 | on a local repository |
|
65 | 65 | based on the current Mercurial version and on configuration. |
|
66 | 66 | The same `format.exp-dirstate-v2` configuration is used again. |
|
67 | 67 | |
|
68 | 68 | Example to upgrade:: |
|
69 | 69 | |
|
70 | 70 | $ hg debugupgrade --config format.exp-dirstate-v2=1 |
|
71 | 71 | |
|
72 | 72 | Example to downgrade to `dirstate-v1`:: |
|
73 | 73 | |
|
74 | 74 | $ hg debugupgrade --config format.exp-dirstate-v2=0 |
|
75 | 75 | |
|
76 | 76 | Both of this commands do nothing but print a list of proposed changes, |
|
77 | 77 | which may include changes unrelated to the dirstate. |
|
78 | 78 | Those other changes are controlled by their own configuration keys. |
|
79 | 79 | Add `--run` to a command to actually apply the proposed changes. |
|
80 | 80 | |
|
81 | 81 | Backups of `.hg/requires` and `.hg/dirstate` are created |
|
82 | 82 | in a `.hg/upgradebackup.*` directory. |
|
83 | 83 | If something goes wrong, restoring those files should undo the change. |
|
84 | 84 | |
|
85 | 85 | Note that upgrading affects compatibility with older versions of Mercurial |
|
86 | 86 | as noted above. |
|
87 | 87 | This can be relevant when a repositoryβs files are on a USB drive |
|
88 | 88 | or some other removable media, or shared over the network, etc. |
|
89 | 89 | |
|
90 | 90 | Internal filesystem representation |
|
91 | 91 | ================================== |
|
92 | 92 | |
|
93 | 93 | Requirements file |
|
94 | 94 | ----------------- |
|
95 | 95 | |
|
96 | 96 | The `.hg/requires` file indicates which of various optional file formats |
|
97 | 97 | are used by a given repository. |
|
98 | 98 | Mercurial aborts when seeing a requirement it does not know about, |
|
99 | 99 | which avoids older version accidentally messing up a respository |
|
100 | 100 | that uses a format that was introduced later. |
|
101 | 101 | For versions that do support a format, the presence or absence of |
|
102 | 102 | the corresponding requirement indicates whether to use that format. |
|
103 | 103 | |
|
104 | 104 | When the file contains a `exp-dirstate-v2` line, |
|
105 | 105 | the `dirstate-v2` format is used. |
|
106 | 106 | With no such line `dirstate-v1` is used. |
|
107 | 107 | |
|
108 | 108 | High level description |
|
109 | 109 | ---------------------- |
|
110 | 110 | |
|
111 | 111 | Whereas `dirstate-v1` uses a single `.hg/disrtate` file, |
|
112 | 112 | in `dirstate-v2` that file is a "docket" file |
|
113 | 113 | that only contains some metadata |
|
114 | 114 | and points to separate data file named `.hg/dirstate.{ID}`, |
|
115 | 115 | where `{ID}` is a random identifier. |
|
116 | 116 | |
|
117 | 117 | This separation allows making data files append-only |
|
118 | 118 | and therefore safer to memory-map. |
|
119 | 119 | Creating a new data file (occasionally to clean up unused data) |
|
120 | 120 | can be done with a different ID |
|
121 | 121 | without disrupting another Mercurial process |
|
122 | 122 | that could still be using the previous data file. |
|
123 | 123 | |
|
124 | 124 | Both files have a format designed to reduce the need for parsing, |
|
125 | 125 | by using fixed-size binary components as much as possible. |
|
126 | 126 | For data that is not fixed-size, |
|
127 | 127 | references to other parts of a file can be made by storing "pseudo-pointers": |
|
128 | 128 | integers counted in bytes from the start of a file. |
|
129 | 129 | For read-only access no data structure is needed, |
|
130 | 130 | only a bytes buffer (possibly memory-mapped directly from the filesystem) |
|
131 | 131 | with specific parts read on demand. |
|
132 | 132 | |
|
133 | 133 | The data file contains "nodes" organized in a tree. |
|
134 | 134 | Each node represents a file or directory inside the working directory |
|
135 | 135 | or its parent changeset. |
|
136 | 136 | This tree has the same structure as the filesystem, |
|
137 | 137 | so a node representing a directory has child nodes representing |
|
138 | 138 | the files and subdirectories contained directly in that directory. |
|
139 | 139 | |
|
140 | 140 | The docket file format |
|
141 | 141 | ---------------------- |
|
142 | 142 | |
|
143 | 143 | This is implemented in `rust/hg-core/src/dirstate_tree/on_disk.rs` |
|
144 | 144 | and `mercurial/dirstateutils/docket.py`. |
|
145 | 145 | |
|
146 | 146 | Components of the docket file are found at fixed offsets, |
|
147 | 147 | counted in bytes from the start of the file: |
|
148 | 148 | |
|
149 | 149 | * Offset 0: |
|
150 | 150 | The 12-bytes marker string "dirstate-v2\n" ending with a newline character. |
|
151 | 151 | This makes it easier to tell a dirstate-v2 file from a dirstate-v1 file, |
|
152 | 152 | although it is not strictly necessary |
|
153 | 153 | since `.hg/requires` determines which format to use. |
|
154 | 154 | |
|
155 | 155 | * Offset 12: |
|
156 | 156 | The changeset node ID on the first parent of the working directory, |
|
157 | 157 | as up to 32 binary bytes. |
|
158 | 158 | If a node ID is shorter (20 bytes for SHA-1), |
|
159 | 159 | it is start-aligned and the rest of the bytes are set to zero. |
|
160 | 160 | |
|
161 | 161 | * Offset 44: |
|
162 | 162 | The changeset node ID on the second parent of the working directory, |
|
163 | 163 | or all zeros if there isnβt one. |
|
164 | 164 | Also 32 binary bytes. |
|
165 | 165 | |
|
166 | 166 | * Offset 76: |
|
167 | 167 | Tree metadata on 44 bytes, described below. |
|
168 | 168 | Its separation in this documentation from the rest of the docket |
|
169 | 169 | reflects a detail of the current implementation. |
|
170 | 170 | Since tree metadata is also made of fields at fixed offsets, those could |
|
171 | 171 | be inlined here by adding 76 bytes to each offset. |
|
172 | 172 | |
|
173 | 173 | * Offset 120: |
|
174 | 174 | The used size of the data file, as a 32-bit big-endian integer. |
|
175 | 175 | The actual size of the data file may be larger |
|
176 | 176 | (if another Mercurial processis in appending to it |
|
177 | 177 | but has not updated the docket yet). |
|
178 | 178 | That extra data must be ignored. |
|
179 | 179 | |
|
180 | 180 | * Offset 124: |
|
181 | 181 | The length of the data file identifier, as a 8-bit integer. |
|
182 | 182 | |
|
183 | 183 | * Offset 125: |
|
184 | 184 | The data file identifier. |
|
185 | 185 | |
|
186 | 186 | * Any additional data is current ignored, and dropped when updating the file. |
|
187 | 187 | |
|
188 | 188 | Tree metadata in the docket file |
|
189 | 189 | -------------------------------- |
|
190 | 190 | |
|
191 | 191 | Tree metadata is similarly made of components at fixed offsets. |
|
192 | 192 | These offsets are counted in bytes from the start of tree metadata, |
|
193 | 193 | which is 76 bytes after the start of the docket file. |
|
194 | 194 | |
|
195 | 195 | This metadata can be thought of as the singular root of the tree |
|
196 | 196 | formed by nodes in the data file. |
|
197 | 197 | |
|
198 | 198 | * Offset 0: |
|
199 | 199 | Pseudo-pointer to the start of root nodes, |
|
200 | 200 | counted in bytes from the start of the data file, |
|
201 | 201 | as a 32-bit big-endian integer. |
|
202 | 202 | These nodes describe files and directories found directly |
|
203 | 203 | at the root of the working directory. |
|
204 | 204 | |
|
205 | 205 | * Offset 4: |
|
206 | 206 | Number of root nodes, as a 32-bit big-endian integer. |
|
207 | 207 | |
|
208 | 208 | * Offset 8: |
|
209 | 209 | Total number of nodes in the entire tree that "have a dirstate entry", |
|
210 | 210 | as a 32-bit big-endian integer. |
|
211 | 211 | Those nodes represent files that would be present at all in `dirstate-v1`. |
|
212 | 212 | This is typically less than the total number of nodes. |
|
213 | 213 | This counter is used to implement `len(dirstatemap)`. |
|
214 | 214 | |
|
215 | 215 | * Offset 12: |
|
216 | 216 | Number of nodes in the entire tree that have a copy source, |
|
217 | 217 | as a 32-bit big-endian integer. |
|
218 | 218 | At the next commit, these files are recorded |
|
219 | 219 | as having been copied or moved/renamed from that source. |
|
220 | 220 | (A move is recorded as a copy and separate removal of the source.) |
|
221 | 221 | This counter is used to implement `len(dirstatemap.copymap)`. |
|
222 | 222 | |
|
223 | 223 | * Offset 16: |
|
224 | 224 | An estimation of how many bytes of the data file |
|
225 | 225 | (within its used size) are unused, as a 32-bit big-endian integer. |
|
226 | 226 | When appending to an existing data file, |
|
227 | 227 | some existing nodes or paths can be unreachable from the new root |
|
228 | 228 | but they still take up space. |
|
229 | 229 | This counter is used to decide when to write a new data file from scratch |
|
230 | 230 | instead of appending to an existing one, |
|
231 | 231 | in order to get rid of that unreachable data |
|
232 | 232 | and avoid unbounded file size growth. |
|
233 | 233 | |
|
234 | 234 | * Offset 20: |
|
235 | 235 | These four bytes are currently ignored |
|
236 | 236 | and reset to zero when updating a docket file. |
|
237 | 237 | This is an attempt at forward compatibility: |
|
238 | 238 | future Mercurial versions could use this as a bit field |
|
239 | 239 | to indicate that a dirstate has additional data or constraints. |
|
240 | 240 | Finding a dirstate file with the relevant bit unset indicates that |
|
241 | 241 | it was written by a then-older version |
|
242 | 242 | which is not aware of that future change. |
|
243 | 243 | |
|
244 | 244 | * Offset 24: |
|
245 | 245 | Either 20 zero bytes, or a SHA-1 hash as 20 binary bytes. |
|
246 | 246 | When present, the hash is of ignore patterns |
|
247 | 247 | that were used for some previous run of the `status` algorithm. |
|
248 | 248 | |
|
249 | 249 | * (Offset 44: end of tree metadata) |
|
250 | 250 | |
|
251 | 251 | Optional hash of ignore patterns |
|
252 | 252 | -------------------------------- |
|
253 | 253 | |
|
254 | 254 | The implementation of `status` at `rust/hg-core/src/dirstate_tree/status.rs` |
|
255 | 255 | has been optimized such that its run time is dominated by calls |
|
256 | 256 | to `stat` for reading the filesystem metadata of a file or directory, |
|
257 | 257 | and to `readdir` for listing the contents of a directory. |
|
258 | 258 | In some cases the algorithm can skip calls to `readdir` |
|
259 | 259 | (saving significant time) |
|
260 | 260 | because the dirstate already contains enough of the relevant information |
|
261 | 261 | to build the correct `status` results. |
|
262 | 262 | |
|
263 | 263 | The default configuration of `hg status` is to list unknown files |
|
264 | 264 | but not ignored files. |
|
265 | 265 | In this case, it matters for the `readdir`-skipping optimization |
|
266 | 266 | if a given file used to be ignored but became unknown |
|
267 | 267 | because `.hgignore` changed. |
|
268 | 268 | To detect the possibility of such a change, |
|
269 | 269 | the tree metadata contains an optional hash of all ignore patterns. |
|
270 | 270 | |
|
271 | 271 | We define: |
|
272 | 272 | |
|
273 | 273 | * "Root" ignore files as: |
|
274 | 274 | |
|
275 | 275 | - `.hgignore` at the root of the repository if it exists |
|
276 | 276 | - And all files from `ui.ignore.*` config. |
|
277 | 277 | |
|
278 | 278 | This set of files is sorted by the string representation of their path. |
|
279 | 279 | |
|
280 | 280 | * The "expanded contents" of an ignore files is the byte string made |
|
281 | 281 | by the concatenation of its contents followed by the "expanded contents" |
|
282 | 282 | of other files included with `include:` or `subinclude:` directives, |
|
283 | 283 | in inclusion order. This definition is recursive, as included files can |
|
284 | 284 | themselves include more files. |
|
285 | 285 | |
|
286 | 286 | This hash is defined as the SHA-1 of the concatenation (in sorted |
|
287 | 287 | order) of the "expanded contents" of each "root" ignore file. |
|
288 | 288 | (Note that computing this does not require actually concatenating |
|
289 | 289 | into a single contiguous byte sequence. |
|
290 | 290 | Instead a SHA-1 hasher object can be created |
|
291 | 291 | and fed separate chunks one by one.) |
|
292 | 292 | |
|
293 | 293 | The data file format |
|
294 | 294 | -------------------- |
|
295 | 295 | |
|
296 | 296 | This is implemented in `rust/hg-core/src/dirstate_tree/on_disk.rs` |
|
297 | 297 | and `mercurial/dirstateutils/v2.py`. |
|
298 | 298 | |
|
299 | 299 | The data file contains two types of data: paths and nodes. |
|
300 | 300 | |
|
301 | 301 | Paths and nodes can be organized in any order in the file, except that sibling |
|
302 | 302 | nodes must be next to each other and sorted by their path. |
|
303 | 303 | Contiguity lets the parent refer to them all |
|
304 | 304 | by their count and a single pseudo-pointer, |
|
305 | 305 | instead of storing one pseudo-pointer per child node. |
|
306 | 306 | Sorting allows using binary seach to find a child node with a given name |
|
307 | 307 | in `O(log(n))` byte sequence comparisons. |
|
308 | 308 | |
|
309 | 309 | The current implemention writes paths and child node before a given node |
|
310 | 310 | for ease of figuring out the value of pseudo-pointers by the time the are to be |
|
311 | 311 | written, but this is not an obligation and readers must not rely on it. |
|
312 | 312 | |
|
313 | 313 | A path is stored as a byte string anywhere in the file, without delimiter. |
|
314 | 314 | It is refered to by one or more node by a pseudo-pointer to its start, and its |
|
315 | 315 | length in bytes. Since there is no delimiter, |
|
316 | 316 | when a path is a substring of another the same bytes could be reused, |
|
317 | 317 | although the implementation does not exploit this as of this writing. |
|
318 | 318 | |
|
319 | 319 | A node is stored on 43 bytes with components at fixed offsets. Paths and |
|
320 | 320 | child nodes relevant to a node are stored externally and referenced though |
|
321 | 321 | pseudo-pointers. |
|
322 | 322 | |
|
323 | 323 | All integers are stored in big-endian. All pseudo-pointers are 32-bit integers |
|
324 | 324 | counting bytes from the start of the data file. Path lengths and positions |
|
325 | 325 | are 16-bit integers, also counted in bytes. |
|
326 | 326 | |
|
327 | 327 | Node components are: |
|
328 | 328 | |
|
329 | 329 | * Offset 0: |
|
330 | 330 | Pseudo-pointer to the full path of this node, |
|
331 | 331 | from the working directory root. |
|
332 | 332 | |
|
333 | 333 | * Offset 4: |
|
334 | 334 | Length of the full path. |
|
335 | 335 | |
|
336 | 336 | * Offset 6: |
|
337 | 337 | Position of the last `/` path separator within the full path, |
|
338 | 338 | in bytes from the start of the full path, |
|
339 | 339 | or zero if there isnβt one. |
|
340 | 340 | The part of the full path after this position is the "base name". |
|
341 | 341 | Since sibling nodes have the same parent, only their base name vary |
|
342 | 342 | and needs to be considered when doing binary search to find a given path. |
|
343 | 343 | |
|
344 | 344 | * Offset 8: |
|
345 | 345 | Pseudo-pointer to the "copy source" path for this node, |
|
346 | 346 | or zero if there is no copy source. |
|
347 | 347 | |
|
348 | 348 | * Offset 12: |
|
349 | 349 | Length of the copy source path, or zero if there isnβt one. |
|
350 | 350 | |
|
351 | 351 | * Offset 14: |
|
352 | 352 | Pseudo-pointer to the start of child nodes. |
|
353 | 353 | |
|
354 | 354 | * Offset 18: |
|
355 | 355 | Number of child nodes, as a 32-bit integer. |
|
356 | 356 | They occupy 43 times this number of bytes |
|
357 | 357 | (not counting space for paths, and further descendants). |
|
358 | 358 | |
|
359 | 359 | * Offset 22: |
|
360 | 360 | Number as a 32-bit integer of descendant nodes in this subtree, |
|
361 | 361 | not including this node itself, |
|
362 | 362 | that "have a dirstate entry". |
|
363 | 363 | Those nodes represent files that would be present at all in `dirstate-v1`. |
|
364 | 364 | This is typically less than the total number of descendants. |
|
365 | 365 | This counter is used to implement `has_dir`. |
|
366 | 366 | |
|
367 | 367 | * Offset 26: |
|
368 | 368 | Number as a 32-bit integer of descendant nodes in this subtree, |
|
369 | 369 | not including this node itself, |
|
370 | 370 | that represent files tracked in the working directory. |
|
371 | 371 | (For example, `hg rm` makes a file untracked.) |
|
372 | 372 | This counter is used to implement `has_tracked_dir`. |
|
373 | 373 | |
|
374 |
* Offset 30 |
|
|
375 | **TODO:** docs not written yet | |
|
376 | as this part of the format might be changing soon. | |
|
374 | * Offset 30: | |
|
375 | Some boolean values packed as bits of a single byte. | |
|
376 | Starting from least-significant, bit masks are:: | |
|
377 | ||
|
378 | WDIR_TRACKED = 1 << 0 | |
|
379 | P1_TRACKED = 1 << 1 | |
|
380 | P2_INFO = 1 << 2 | |
|
381 | HAS_MODE_AND_SIZE = 1 << 3 | |
|
382 | HAS_MTIME = 1 << 4 | |
|
383 | ||
|
384 | Other bits are unset. The meaning of these bits are: | |
|
385 | ||
|
386 | `WDIR_TRACKED` | |
|
387 | Set if the working directory contains a tracked file at this nodeβs path. | |
|
388 | This is typically set and unset by `hg add` and `hg rm`. | |
|
389 | ||
|
390 | `P1_TRACKED` | |
|
391 | set if the working directoryβs first parent changeset | |
|
392 | (whose node identifier is found in tree metadata) | |
|
393 | contains a tracked file at this nodeβs path. | |
|
394 | This is a cache to reduce manifest lookups. | |
|
395 | ||
|
396 | `P2_INFO` | |
|
397 | Set if the file has been involved in some merge operation. | |
|
398 | Either because it was actually merged, | |
|
399 | or because the version in the second parent p2 version was ahead, | |
|
400 | or because some rename moved it there. | |
|
401 | In either case `hg status` will want it displayed as modified. | |
|
402 | ||
|
403 | Files that would be mentioned at all in the `dirstate-v1` file format | |
|
404 | have a node with at least one of the above three bits set in `dirstate-v2`. | |
|
405 | Letβs call these files "tracked anywhere", | |
|
406 | and "untracked" the nodes with all three of these bits unset. | |
|
407 | Untracked nodes are typically for directories: | |
|
408 | they hold child nodes and form the tree structure. | |
|
409 | Additional untracked nodes may also exist. | |
|
410 | Although implementations should strive to clean up nodes | |
|
411 | that are entirely unused, other untracked nodes may also exist. | |
|
412 | For example, a future version of Mercurial might in some cases | |
|
413 | add nodes for untracked files or/and ignored files in the working directory | |
|
414 | in order to optimize `hg status` | |
|
415 | by enabling it to skip `readdir` in more cases. | |
|
416 | ||
|
417 | When a node is for a file tracked anywhere, | |
|
418 | the rest of the node data is three fields: | |
|
419 | ||
|
420 | * Offset 31: | |
|
421 | If `HAS_MODE_AND_SIZE` is unset, four zero bytes. | |
|
422 | Otherwise, a 32-bit integer for the Unix mode (as in `stat_result.st_mode`) | |
|
423 | expected for this file to be considered clean. | |
|
424 | Only the `S_IXUSR` bit (owner has execute permission) is considered. | |
|
425 | ||
|
426 | * Offset 35: | |
|
427 | If `HAS_MTIME` is unset, four zero bytes. | |
|
428 | Otherwise, a 32-bit integer for expected modified time of the file | |
|
429 | (as in `stat_result.st_mtime`), | |
|
430 | truncated to its 31 least-significant bits. | |
|
431 | Unlike in dirstate-v1, negative values are not used. | |
|
432 | ||
|
433 | * Offset 39: | |
|
434 | If `HAS_MODE_AND_SIZE` is unset, four zero bytes. | |
|
435 | Otherwise, a 32-bit integer for expected size of the file | |
|
436 | truncated to its 31 least-significant bits. | |
|
437 | Unlike in dirstate-v1, negative values are not used. | |
|
438 | ||
|
439 | If an untracked node `HAS_MTIME` *unset*, this space is unused: | |
|
440 | ||
|
441 | * Offset 31: | |
|
442 | 12 bytes set to zero | |
|
443 | ||
|
444 | If an untracked node `HAS_MTIME` *set*, | |
|
445 | what follows is the modification time of a directory | |
|
446 | represented with separated second and sub-second components | |
|
447 | since the Unix epoch: | |
|
448 | ||
|
449 | * Offset 31: | |
|
450 | The number of seconds as a signed (twoβs complement) 64-bit integer. | |
|
451 | ||
|
452 | * Offset 39: | |
|
453 | The number of nanoseconds as 32-bit integer. | |
|
454 | Always greater than or equal to zero, and strictly less than a billion. | |
|
455 | Increasing this component makes the modification time | |
|
456 | go forward or backward in time dependening | |
|
457 | on the sign of the integral seconds components. | |
|
458 | (Note: this is buggy because there is no negative zero integer, | |
|
459 | but will be changed soon.) | |
|
460 | ||
|
461 | The presence of a directory modification time means that at some point, | |
|
462 | this path in the working directory was observed: | |
|
463 | ||
|
464 | - To be a directory | |
|
465 | - With the given modification time | |
|
466 | - That time was already strictly in the past when observed, | |
|
467 | meaning that later changes cannot happen in the same clock tick | |
|
468 | and must cause a different modification time | |
|
469 | (unless the system clock jumps back and we get unlucky, | |
|
470 | which is not impossible but deemed unlikely enough). | |
|
471 | - All direct children of this directory | |
|
472 | (as returned by `std::fs::read_dir`) | |
|
473 | either have a corresponding dirstate node, | |
|
474 | or are ignored by ignore patterns whose hash is in tree metadata. | |
|
475 | ||
|
476 | This means that if `std::fs::symlink_metadata` later reports | |
|
477 | the same modification time | |
|
478 | and ignored patterns havenβt changed, | |
|
479 | a run of status that is not listing ignored files | |
|
480 | can skip calling `std::fs::read_dir` again for this directory, | |
|
481 | and iterate child dirstate nodes instead. | |
|
482 | ||
|
483 | ||
|
484 | * (Offset 43: end of this node) |
@@ -1,736 +1,736 b'' | |||
|
1 | 1 | # parsers.py - Python implementation of parsers.c |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2009 Olivia Mackall <olivia@selenic.com> and others |
|
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 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import struct |
|
11 | 11 | import zlib |
|
12 | 12 | |
|
13 | 13 | from ..node import ( |
|
14 | 14 | nullrev, |
|
15 | 15 | sha1nodeconstants, |
|
16 | 16 | ) |
|
17 | 17 | from ..thirdparty import attr |
|
18 | 18 | from .. import ( |
|
19 | 19 | error, |
|
20 | 20 | pycompat, |
|
21 | 21 | revlogutils, |
|
22 | 22 | util, |
|
23 | 23 | ) |
|
24 | 24 | |
|
25 | 25 | from ..revlogutils import nodemap as nodemaputil |
|
26 | 26 | from ..revlogutils import constants as revlog_constants |
|
27 | 27 | |
|
28 | 28 | stringio = pycompat.bytesio |
|
29 | 29 | |
|
30 | 30 | |
|
31 | 31 | _pack = struct.pack |
|
32 | 32 | _unpack = struct.unpack |
|
33 | 33 | _compress = zlib.compress |
|
34 | 34 | _decompress = zlib.decompress |
|
35 | 35 | |
|
36 | 36 | |
|
37 | 37 | # a special value used internally for `size` if the file come from the other parent |
|
38 | 38 | FROM_P2 = -2 |
|
39 | 39 | |
|
40 | 40 | # a special value used internally for `size` if the file is modified/merged/added |
|
41 | 41 | NONNORMAL = -1 |
|
42 | 42 | |
|
43 | 43 | # a special value used internally for `time` if the time is ambigeous |
|
44 | 44 | AMBIGUOUS_TIME = -1 |
|
45 | 45 | |
|
46 | 46 | |
|
47 | 47 | @attr.s(slots=True, init=False) |
|
48 | 48 | class DirstateItem(object): |
|
49 | 49 | """represent a dirstate entry |
|
50 | 50 | |
|
51 | 51 | It hold multiple attributes |
|
52 | 52 | |
|
53 | 53 | # about file tracking |
|
54 | 54 | - wc_tracked: is the file tracked by the working copy |
|
55 | 55 | - p1_tracked: is the file tracked in working copy first parent |
|
56 | 56 | - p2_info: the file has been involved in some merge operation. Either |
|
57 | 57 | because it was actually merged, or because the p2 version was |
|
58 |
ahead, or because some rename |
|
|
58 | ahead, or because some rename moved it there. In either case | |
|
59 | 59 | `hg status` will want it displayed as modified. |
|
60 | 60 | |
|
61 | 61 | # about the file state expected from p1 manifest: |
|
62 | 62 | - mode: the file mode in p1 |
|
63 | 63 | - size: the file size in p1 |
|
64 | 64 | |
|
65 | 65 | These value can be set to None, which mean we don't have a meaningful value |
|
66 | 66 | to compare with. Either because we don't really care about them as there |
|
67 | 67 | `status` is known without having to look at the disk or because we don't |
|
68 | 68 | know these right now and a full comparison will be needed to find out if |
|
69 | 69 | the file is clean. |
|
70 | 70 | |
|
71 | 71 | # about the file state on disk last time we saw it: |
|
72 | 72 | - mtime: the last known clean mtime for the file. |
|
73 | 73 | |
|
74 | 74 | This value can be set to None if no cachable state exist. Either because we |
|
75 | 75 | do not care (see previous section) or because we could not cache something |
|
76 | 76 | yet. |
|
77 | 77 | """ |
|
78 | 78 | |
|
79 | 79 | _wc_tracked = attr.ib() |
|
80 | 80 | _p1_tracked = attr.ib() |
|
81 | 81 | _p2_info = attr.ib() |
|
82 | 82 | _mode = attr.ib() |
|
83 | 83 | _size = attr.ib() |
|
84 | 84 | _mtime = attr.ib() |
|
85 | 85 | |
|
86 | 86 | def __init__( |
|
87 | 87 | self, |
|
88 | 88 | wc_tracked=False, |
|
89 | 89 | p1_tracked=False, |
|
90 | 90 | p2_info=False, |
|
91 | 91 | has_meaningful_data=True, |
|
92 | 92 | has_meaningful_mtime=True, |
|
93 | 93 | parentfiledata=None, |
|
94 | 94 | ): |
|
95 | 95 | self._wc_tracked = wc_tracked |
|
96 | 96 | self._p1_tracked = p1_tracked |
|
97 | 97 | self._p2_info = p2_info |
|
98 | 98 | |
|
99 | 99 | self._mode = None |
|
100 | 100 | self._size = None |
|
101 | 101 | self._mtime = None |
|
102 | 102 | if parentfiledata is None: |
|
103 | 103 | has_meaningful_mtime = False |
|
104 | 104 | has_meaningful_data = False |
|
105 | 105 | if has_meaningful_data: |
|
106 | 106 | self._mode = parentfiledata[0] |
|
107 | 107 | self._size = parentfiledata[1] |
|
108 | 108 | if has_meaningful_mtime: |
|
109 | 109 | self._mtime = parentfiledata[2] |
|
110 | 110 | |
|
111 | 111 | @classmethod |
|
112 | 112 | def from_v1_data(cls, state, mode, size, mtime): |
|
113 | 113 | """Build a new DirstateItem object from V1 data |
|
114 | 114 | |
|
115 | 115 | Since the dirstate-v1 format is frozen, the signature of this function |
|
116 | 116 | is not expected to change, unlike the __init__ one. |
|
117 | 117 | """ |
|
118 | 118 | if state == b'm': |
|
119 | 119 | return cls(wc_tracked=True, p1_tracked=True, p2_info=True) |
|
120 | 120 | elif state == b'a': |
|
121 | 121 | return cls(wc_tracked=True) |
|
122 | 122 | elif state == b'r': |
|
123 | 123 | if size == NONNORMAL: |
|
124 | 124 | p1_tracked = True |
|
125 | 125 | p2_info = True |
|
126 | 126 | elif size == FROM_P2: |
|
127 | 127 | p1_tracked = False |
|
128 | 128 | p2_info = True |
|
129 | 129 | else: |
|
130 | 130 | p1_tracked = True |
|
131 | 131 | p2_info = False |
|
132 | 132 | return cls(p1_tracked=p1_tracked, p2_info=p2_info) |
|
133 | 133 | elif state == b'n': |
|
134 | 134 | if size == FROM_P2: |
|
135 | 135 | return cls(wc_tracked=True, p2_info=True) |
|
136 | 136 | elif size == NONNORMAL: |
|
137 | 137 | return cls(wc_tracked=True, p1_tracked=True) |
|
138 | 138 | elif mtime == AMBIGUOUS_TIME: |
|
139 | 139 | return cls( |
|
140 | 140 | wc_tracked=True, |
|
141 | 141 | p1_tracked=True, |
|
142 | 142 | has_meaningful_mtime=False, |
|
143 | 143 | parentfiledata=(mode, size, 42), |
|
144 | 144 | ) |
|
145 | 145 | else: |
|
146 | 146 | return cls( |
|
147 | 147 | wc_tracked=True, |
|
148 | 148 | p1_tracked=True, |
|
149 | 149 | parentfiledata=(mode, size, mtime), |
|
150 | 150 | ) |
|
151 | 151 | else: |
|
152 | 152 | raise RuntimeError(b'unknown state: %s' % state) |
|
153 | 153 | |
|
154 | 154 | def set_possibly_dirty(self): |
|
155 | 155 | """Mark a file as "possibly dirty" |
|
156 | 156 | |
|
157 | 157 | This means the next status call will have to actually check its content |
|
158 | 158 | to make sure it is correct. |
|
159 | 159 | """ |
|
160 | 160 | self._mtime = None |
|
161 | 161 | |
|
162 | 162 | def set_clean(self, mode, size, mtime): |
|
163 | 163 | """mark a file as "clean" cancelling potential "possibly dirty call" |
|
164 | 164 | |
|
165 | 165 | Note: this function is a descendant of `dirstate.normal` and is |
|
166 | 166 | currently expected to be call on "normal" entry only. There are not |
|
167 | 167 | reason for this to not change in the future as long as the ccode is |
|
168 | 168 | updated to preserve the proper state of the non-normal files. |
|
169 | 169 | """ |
|
170 | 170 | self._wc_tracked = True |
|
171 | 171 | self._p1_tracked = True |
|
172 | 172 | self._mode = mode |
|
173 | 173 | self._size = size |
|
174 | 174 | self._mtime = mtime |
|
175 | 175 | |
|
176 | 176 | def set_tracked(self): |
|
177 | 177 | """mark a file as tracked in the working copy |
|
178 | 178 | |
|
179 | 179 | This will ultimately be called by command like `hg add`. |
|
180 | 180 | """ |
|
181 | 181 | self._wc_tracked = True |
|
182 | 182 | # `set_tracked` is replacing various `normallookup` call. So we mark |
|
183 | 183 | # the files as needing lookup |
|
184 | 184 | # |
|
185 | 185 | # Consider dropping this in the future in favor of something less broad. |
|
186 | 186 | self._mtime = None |
|
187 | 187 | |
|
188 | 188 | def set_untracked(self): |
|
189 | 189 | """mark a file as untracked in the working copy |
|
190 | 190 | |
|
191 | 191 | This will ultimately be called by command like `hg remove`. |
|
192 | 192 | """ |
|
193 | 193 | self._wc_tracked = False |
|
194 | 194 | self._mode = None |
|
195 | 195 | self._size = None |
|
196 | 196 | self._mtime = None |
|
197 | 197 | |
|
198 | 198 | def drop_merge_data(self): |
|
199 | 199 | """remove all "merge-only" from a DirstateItem |
|
200 | 200 | |
|
201 | 201 | This is to be call by the dirstatemap code when the second parent is dropped |
|
202 | 202 | """ |
|
203 | 203 | if self._p2_info: |
|
204 | 204 | self._p2_info = False |
|
205 | 205 | self._mode = None |
|
206 | 206 | self._size = None |
|
207 | 207 | self._mtime = None |
|
208 | 208 | |
|
209 | 209 | @property |
|
210 | 210 | def mode(self): |
|
211 | 211 | return self.v1_mode() |
|
212 | 212 | |
|
213 | 213 | @property |
|
214 | 214 | def size(self): |
|
215 | 215 | return self.v1_size() |
|
216 | 216 | |
|
217 | 217 | @property |
|
218 | 218 | def mtime(self): |
|
219 | 219 | return self.v1_mtime() |
|
220 | 220 | |
|
221 | 221 | @property |
|
222 | 222 | def state(self): |
|
223 | 223 | """ |
|
224 | 224 | States are: |
|
225 | 225 | n normal |
|
226 | 226 | m needs merging |
|
227 | 227 | r marked for removal |
|
228 | 228 | a marked for addition |
|
229 | 229 | |
|
230 | 230 | XXX This "state" is a bit obscure and mostly a direct expression of the |
|
231 | 231 | dirstatev1 format. It would make sense to ultimately deprecate it in |
|
232 | 232 | favor of the more "semantic" attributes. |
|
233 | 233 | """ |
|
234 | 234 | if not self.any_tracked: |
|
235 | 235 | return b'?' |
|
236 | 236 | return self.v1_state() |
|
237 | 237 | |
|
238 | 238 | @property |
|
239 | 239 | def tracked(self): |
|
240 | 240 | """True is the file is tracked in the working copy""" |
|
241 | 241 | return self._wc_tracked |
|
242 | 242 | |
|
243 | 243 | @property |
|
244 | 244 | def any_tracked(self): |
|
245 | 245 | """True is the file is tracked anywhere (wc or parents)""" |
|
246 | 246 | return self._wc_tracked or self._p1_tracked or self._p2_info |
|
247 | 247 | |
|
248 | 248 | @property |
|
249 | 249 | def added(self): |
|
250 | 250 | """True if the file has been added""" |
|
251 | 251 | return self._wc_tracked and not (self._p1_tracked or self._p2_info) |
|
252 | 252 | |
|
253 | 253 | @property |
|
254 | 254 | def maybe_clean(self): |
|
255 | 255 | """True if the file has a chance to be in the "clean" state""" |
|
256 | 256 | if not self._wc_tracked: |
|
257 | 257 | return False |
|
258 | 258 | elif not self._p1_tracked: |
|
259 | 259 | return False |
|
260 | 260 | elif self._p2_info: |
|
261 | 261 | return False |
|
262 | 262 | return True |
|
263 | 263 | |
|
264 | 264 | @property |
|
265 | 265 | def p1_tracked(self): |
|
266 | 266 | """True if the file is tracked in the first parent manifest""" |
|
267 | 267 | return self._p1_tracked |
|
268 | 268 | |
|
269 | 269 | @property |
|
270 | 270 | def p2_info(self): |
|
271 | 271 | """True if the file needed to merge or apply any input from p2 |
|
272 | 272 | |
|
273 | 273 | See the class documentation for details. |
|
274 | 274 | """ |
|
275 | 275 | return self._wc_tracked and self._p2_info |
|
276 | 276 | |
|
277 | 277 | @property |
|
278 | 278 | def removed(self): |
|
279 | 279 | """True if the file has been removed""" |
|
280 | 280 | return not self._wc_tracked and (self._p1_tracked or self._p2_info) |
|
281 | 281 | |
|
282 | 282 | def v1_state(self): |
|
283 | 283 | """return a "state" suitable for v1 serialization""" |
|
284 | 284 | if not self.any_tracked: |
|
285 | 285 | # the object has no state to record, this is -currently- |
|
286 | 286 | # unsupported |
|
287 | 287 | raise RuntimeError('untracked item') |
|
288 | 288 | elif self.removed: |
|
289 | 289 | return b'r' |
|
290 | 290 | elif self._p1_tracked and self._p2_info: |
|
291 | 291 | return b'm' |
|
292 | 292 | elif self.added: |
|
293 | 293 | return b'a' |
|
294 | 294 | else: |
|
295 | 295 | return b'n' |
|
296 | 296 | |
|
297 | 297 | def v1_mode(self): |
|
298 | 298 | """return a "mode" suitable for v1 serialization""" |
|
299 | 299 | return self._mode if self._mode is not None else 0 |
|
300 | 300 | |
|
301 | 301 | def v1_size(self): |
|
302 | 302 | """return a "size" suitable for v1 serialization""" |
|
303 | 303 | if not self.any_tracked: |
|
304 | 304 | # the object has no state to record, this is -currently- |
|
305 | 305 | # unsupported |
|
306 | 306 | raise RuntimeError('untracked item') |
|
307 | 307 | elif self.removed and self._p1_tracked and self._p2_info: |
|
308 | 308 | return NONNORMAL |
|
309 | 309 | elif self._p2_info: |
|
310 | 310 | return FROM_P2 |
|
311 | 311 | elif self.removed: |
|
312 | 312 | return 0 |
|
313 | 313 | elif self.added: |
|
314 | 314 | return NONNORMAL |
|
315 | 315 | elif self._size is None: |
|
316 | 316 | return NONNORMAL |
|
317 | 317 | else: |
|
318 | 318 | return self._size |
|
319 | 319 | |
|
320 | 320 | def v1_mtime(self): |
|
321 | 321 | """return a "mtime" suitable for v1 serialization""" |
|
322 | 322 | if not self.any_tracked: |
|
323 | 323 | # the object has no state to record, this is -currently- |
|
324 | 324 | # unsupported |
|
325 | 325 | raise RuntimeError('untracked item') |
|
326 | 326 | elif self.removed: |
|
327 | 327 | return 0 |
|
328 | 328 | elif self._mtime is None: |
|
329 | 329 | return AMBIGUOUS_TIME |
|
330 | 330 | elif self._p2_info: |
|
331 | 331 | return AMBIGUOUS_TIME |
|
332 | 332 | elif not self._p1_tracked: |
|
333 | 333 | return AMBIGUOUS_TIME |
|
334 | 334 | else: |
|
335 | 335 | return self._mtime |
|
336 | 336 | |
|
337 | 337 | def need_delay(self, now): |
|
338 | 338 | """True if the stored mtime would be ambiguous with the current time""" |
|
339 | 339 | return self.v1_state() == b'n' and self.v1_mtime() == now |
|
340 | 340 | |
|
341 | 341 | |
|
342 | 342 | def gettype(q): |
|
343 | 343 | return int(q & 0xFFFF) |
|
344 | 344 | |
|
345 | 345 | |
|
346 | 346 | class BaseIndexObject(object): |
|
347 | 347 | # Can I be passed to an algorithme implemented in Rust ? |
|
348 | 348 | rust_ext_compat = 0 |
|
349 | 349 | # Format of an index entry according to Python's `struct` language |
|
350 | 350 | index_format = revlog_constants.INDEX_ENTRY_V1 |
|
351 | 351 | # Size of a C unsigned long long int, platform independent |
|
352 | 352 | big_int_size = struct.calcsize(b'>Q') |
|
353 | 353 | # Size of a C long int, platform independent |
|
354 | 354 | int_size = struct.calcsize(b'>i') |
|
355 | 355 | # An empty index entry, used as a default value to be overridden, or nullrev |
|
356 | 356 | null_item = ( |
|
357 | 357 | 0, |
|
358 | 358 | 0, |
|
359 | 359 | 0, |
|
360 | 360 | -1, |
|
361 | 361 | -1, |
|
362 | 362 | -1, |
|
363 | 363 | -1, |
|
364 | 364 | sha1nodeconstants.nullid, |
|
365 | 365 | 0, |
|
366 | 366 | 0, |
|
367 | 367 | revlog_constants.COMP_MODE_INLINE, |
|
368 | 368 | revlog_constants.COMP_MODE_INLINE, |
|
369 | 369 | ) |
|
370 | 370 | |
|
371 | 371 | @util.propertycache |
|
372 | 372 | def entry_size(self): |
|
373 | 373 | return self.index_format.size |
|
374 | 374 | |
|
375 | 375 | @property |
|
376 | 376 | def nodemap(self): |
|
377 | 377 | msg = b"index.nodemap is deprecated, use index.[has_node|rev|get_rev]" |
|
378 | 378 | util.nouideprecwarn(msg, b'5.3', stacklevel=2) |
|
379 | 379 | return self._nodemap |
|
380 | 380 | |
|
381 | 381 | @util.propertycache |
|
382 | 382 | def _nodemap(self): |
|
383 | 383 | nodemap = nodemaputil.NodeMap({sha1nodeconstants.nullid: nullrev}) |
|
384 | 384 | for r in range(0, len(self)): |
|
385 | 385 | n = self[r][7] |
|
386 | 386 | nodemap[n] = r |
|
387 | 387 | return nodemap |
|
388 | 388 | |
|
389 | 389 | def has_node(self, node): |
|
390 | 390 | """return True if the node exist in the index""" |
|
391 | 391 | return node in self._nodemap |
|
392 | 392 | |
|
393 | 393 | def rev(self, node): |
|
394 | 394 | """return a revision for a node |
|
395 | 395 | |
|
396 | 396 | If the node is unknown, raise a RevlogError""" |
|
397 | 397 | return self._nodemap[node] |
|
398 | 398 | |
|
399 | 399 | def get_rev(self, node): |
|
400 | 400 | """return a revision for a node |
|
401 | 401 | |
|
402 | 402 | If the node is unknown, return None""" |
|
403 | 403 | return self._nodemap.get(node) |
|
404 | 404 | |
|
405 | 405 | def _stripnodes(self, start): |
|
406 | 406 | if '_nodemap' in vars(self): |
|
407 | 407 | for r in range(start, len(self)): |
|
408 | 408 | n = self[r][7] |
|
409 | 409 | del self._nodemap[n] |
|
410 | 410 | |
|
411 | 411 | def clearcaches(self): |
|
412 | 412 | self.__dict__.pop('_nodemap', None) |
|
413 | 413 | |
|
414 | 414 | def __len__(self): |
|
415 | 415 | return self._lgt + len(self._extra) |
|
416 | 416 | |
|
417 | 417 | def append(self, tup): |
|
418 | 418 | if '_nodemap' in vars(self): |
|
419 | 419 | self._nodemap[tup[7]] = len(self) |
|
420 | 420 | data = self._pack_entry(len(self), tup) |
|
421 | 421 | self._extra.append(data) |
|
422 | 422 | |
|
423 | 423 | def _pack_entry(self, rev, entry): |
|
424 | 424 | assert entry[8] == 0 |
|
425 | 425 | assert entry[9] == 0 |
|
426 | 426 | return self.index_format.pack(*entry[:8]) |
|
427 | 427 | |
|
428 | 428 | def _check_index(self, i): |
|
429 | 429 | if not isinstance(i, int): |
|
430 | 430 | raise TypeError(b"expecting int indexes") |
|
431 | 431 | if i < 0 or i >= len(self): |
|
432 | 432 | raise IndexError |
|
433 | 433 | |
|
434 | 434 | def __getitem__(self, i): |
|
435 | 435 | if i == -1: |
|
436 | 436 | return self.null_item |
|
437 | 437 | self._check_index(i) |
|
438 | 438 | if i >= self._lgt: |
|
439 | 439 | data = self._extra[i - self._lgt] |
|
440 | 440 | else: |
|
441 | 441 | index = self._calculate_index(i) |
|
442 | 442 | data = self._data[index : index + self.entry_size] |
|
443 | 443 | r = self._unpack_entry(i, data) |
|
444 | 444 | if self._lgt and i == 0: |
|
445 | 445 | offset = revlogutils.offset_type(0, gettype(r[0])) |
|
446 | 446 | r = (offset,) + r[1:] |
|
447 | 447 | return r |
|
448 | 448 | |
|
449 | 449 | def _unpack_entry(self, rev, data): |
|
450 | 450 | r = self.index_format.unpack(data) |
|
451 | 451 | r = r + ( |
|
452 | 452 | 0, |
|
453 | 453 | 0, |
|
454 | 454 | revlog_constants.COMP_MODE_INLINE, |
|
455 | 455 | revlog_constants.COMP_MODE_INLINE, |
|
456 | 456 | ) |
|
457 | 457 | return r |
|
458 | 458 | |
|
459 | 459 | def pack_header(self, header): |
|
460 | 460 | """pack header information as binary""" |
|
461 | 461 | v_fmt = revlog_constants.INDEX_HEADER |
|
462 | 462 | return v_fmt.pack(header) |
|
463 | 463 | |
|
464 | 464 | def entry_binary(self, rev): |
|
465 | 465 | """return the raw binary string representing a revision""" |
|
466 | 466 | entry = self[rev] |
|
467 | 467 | p = revlog_constants.INDEX_ENTRY_V1.pack(*entry[:8]) |
|
468 | 468 | if rev == 0: |
|
469 | 469 | p = p[revlog_constants.INDEX_HEADER.size :] |
|
470 | 470 | return p |
|
471 | 471 | |
|
472 | 472 | |
|
473 | 473 | class IndexObject(BaseIndexObject): |
|
474 | 474 | def __init__(self, data): |
|
475 | 475 | assert len(data) % self.entry_size == 0, ( |
|
476 | 476 | len(data), |
|
477 | 477 | self.entry_size, |
|
478 | 478 | len(data) % self.entry_size, |
|
479 | 479 | ) |
|
480 | 480 | self._data = data |
|
481 | 481 | self._lgt = len(data) // self.entry_size |
|
482 | 482 | self._extra = [] |
|
483 | 483 | |
|
484 | 484 | def _calculate_index(self, i): |
|
485 | 485 | return i * self.entry_size |
|
486 | 486 | |
|
487 | 487 | def __delitem__(self, i): |
|
488 | 488 | if not isinstance(i, slice) or not i.stop == -1 or i.step is not None: |
|
489 | 489 | raise ValueError(b"deleting slices only supports a:-1 with step 1") |
|
490 | 490 | i = i.start |
|
491 | 491 | self._check_index(i) |
|
492 | 492 | self._stripnodes(i) |
|
493 | 493 | if i < self._lgt: |
|
494 | 494 | self._data = self._data[: i * self.entry_size] |
|
495 | 495 | self._lgt = i |
|
496 | 496 | self._extra = [] |
|
497 | 497 | else: |
|
498 | 498 | self._extra = self._extra[: i - self._lgt] |
|
499 | 499 | |
|
500 | 500 | |
|
501 | 501 | class PersistentNodeMapIndexObject(IndexObject): |
|
502 | 502 | """a Debug oriented class to test persistent nodemap |
|
503 | 503 | |
|
504 | 504 | We need a simple python object to test API and higher level behavior. See |
|
505 | 505 | the Rust implementation for more serious usage. This should be used only |
|
506 | 506 | through the dedicated `devel.persistent-nodemap` config. |
|
507 | 507 | """ |
|
508 | 508 | |
|
509 | 509 | def nodemap_data_all(self): |
|
510 | 510 | """Return bytes containing a full serialization of a nodemap |
|
511 | 511 | |
|
512 | 512 | The nodemap should be valid for the full set of revisions in the |
|
513 | 513 | index.""" |
|
514 | 514 | return nodemaputil.persistent_data(self) |
|
515 | 515 | |
|
516 | 516 | def nodemap_data_incremental(self): |
|
517 | 517 | """Return bytes containing a incremental update to persistent nodemap |
|
518 | 518 | |
|
519 | 519 | This containst the data for an append-only update of the data provided |
|
520 | 520 | in the last call to `update_nodemap_data`. |
|
521 | 521 | """ |
|
522 | 522 | if self._nm_root is None: |
|
523 | 523 | return None |
|
524 | 524 | docket = self._nm_docket |
|
525 | 525 | changed, data = nodemaputil.update_persistent_data( |
|
526 | 526 | self, self._nm_root, self._nm_max_idx, self._nm_docket.tip_rev |
|
527 | 527 | ) |
|
528 | 528 | |
|
529 | 529 | self._nm_root = self._nm_max_idx = self._nm_docket = None |
|
530 | 530 | return docket, changed, data |
|
531 | 531 | |
|
532 | 532 | def update_nodemap_data(self, docket, nm_data): |
|
533 | 533 | """provide full block of persisted binary data for a nodemap |
|
534 | 534 | |
|
535 | 535 | The data are expected to come from disk. See `nodemap_data_all` for a |
|
536 | 536 | produceur of such data.""" |
|
537 | 537 | if nm_data is not None: |
|
538 | 538 | self._nm_root, self._nm_max_idx = nodemaputil.parse_data(nm_data) |
|
539 | 539 | if self._nm_root: |
|
540 | 540 | self._nm_docket = docket |
|
541 | 541 | else: |
|
542 | 542 | self._nm_root = self._nm_max_idx = self._nm_docket = None |
|
543 | 543 | |
|
544 | 544 | |
|
545 | 545 | class InlinedIndexObject(BaseIndexObject): |
|
546 | 546 | def __init__(self, data, inline=0): |
|
547 | 547 | self._data = data |
|
548 | 548 | self._lgt = self._inline_scan(None) |
|
549 | 549 | self._inline_scan(self._lgt) |
|
550 | 550 | self._extra = [] |
|
551 | 551 | |
|
552 | 552 | def _inline_scan(self, lgt): |
|
553 | 553 | off = 0 |
|
554 | 554 | if lgt is not None: |
|
555 | 555 | self._offsets = [0] * lgt |
|
556 | 556 | count = 0 |
|
557 | 557 | while off <= len(self._data) - self.entry_size: |
|
558 | 558 | start = off + self.big_int_size |
|
559 | 559 | (s,) = struct.unpack( |
|
560 | 560 | b'>i', |
|
561 | 561 | self._data[start : start + self.int_size], |
|
562 | 562 | ) |
|
563 | 563 | if lgt is not None: |
|
564 | 564 | self._offsets[count] = off |
|
565 | 565 | count += 1 |
|
566 | 566 | off += self.entry_size + s |
|
567 | 567 | if off != len(self._data): |
|
568 | 568 | raise ValueError(b"corrupted data") |
|
569 | 569 | return count |
|
570 | 570 | |
|
571 | 571 | def __delitem__(self, i): |
|
572 | 572 | if not isinstance(i, slice) or not i.stop == -1 or i.step is not None: |
|
573 | 573 | raise ValueError(b"deleting slices only supports a:-1 with step 1") |
|
574 | 574 | i = i.start |
|
575 | 575 | self._check_index(i) |
|
576 | 576 | self._stripnodes(i) |
|
577 | 577 | if i < self._lgt: |
|
578 | 578 | self._offsets = self._offsets[:i] |
|
579 | 579 | self._lgt = i |
|
580 | 580 | self._extra = [] |
|
581 | 581 | else: |
|
582 | 582 | self._extra = self._extra[: i - self._lgt] |
|
583 | 583 | |
|
584 | 584 | def _calculate_index(self, i): |
|
585 | 585 | return self._offsets[i] |
|
586 | 586 | |
|
587 | 587 | |
|
588 | 588 | def parse_index2(data, inline, revlogv2=False): |
|
589 | 589 | if not inline: |
|
590 | 590 | cls = IndexObject2 if revlogv2 else IndexObject |
|
591 | 591 | return cls(data), None |
|
592 | 592 | cls = InlinedIndexObject |
|
593 | 593 | return cls(data, inline), (0, data) |
|
594 | 594 | |
|
595 | 595 | |
|
596 | 596 | def parse_index_cl_v2(data): |
|
597 | 597 | return IndexChangelogV2(data), None |
|
598 | 598 | |
|
599 | 599 | |
|
600 | 600 | class IndexObject2(IndexObject): |
|
601 | 601 | index_format = revlog_constants.INDEX_ENTRY_V2 |
|
602 | 602 | |
|
603 | 603 | def replace_sidedata_info( |
|
604 | 604 | self, |
|
605 | 605 | rev, |
|
606 | 606 | sidedata_offset, |
|
607 | 607 | sidedata_length, |
|
608 | 608 | offset_flags, |
|
609 | 609 | compression_mode, |
|
610 | 610 | ): |
|
611 | 611 | """ |
|
612 | 612 | Replace an existing index entry's sidedata offset and length with new |
|
613 | 613 | ones. |
|
614 | 614 | This cannot be used outside of the context of sidedata rewriting, |
|
615 | 615 | inside the transaction that creates the revision `rev`. |
|
616 | 616 | """ |
|
617 | 617 | if rev < 0: |
|
618 | 618 | raise KeyError |
|
619 | 619 | self._check_index(rev) |
|
620 | 620 | if rev < self._lgt: |
|
621 | 621 | msg = b"cannot rewrite entries outside of this transaction" |
|
622 | 622 | raise KeyError(msg) |
|
623 | 623 | else: |
|
624 | 624 | entry = list(self[rev]) |
|
625 | 625 | entry[0] = offset_flags |
|
626 | 626 | entry[8] = sidedata_offset |
|
627 | 627 | entry[9] = sidedata_length |
|
628 | 628 | entry[11] = compression_mode |
|
629 | 629 | entry = tuple(entry) |
|
630 | 630 | new = self._pack_entry(rev, entry) |
|
631 | 631 | self._extra[rev - self._lgt] = new |
|
632 | 632 | |
|
633 | 633 | def _unpack_entry(self, rev, data): |
|
634 | 634 | data = self.index_format.unpack(data) |
|
635 | 635 | entry = data[:10] |
|
636 | 636 | data_comp = data[10] & 3 |
|
637 | 637 | sidedata_comp = (data[10] & (3 << 2)) >> 2 |
|
638 | 638 | return entry + (data_comp, sidedata_comp) |
|
639 | 639 | |
|
640 | 640 | def _pack_entry(self, rev, entry): |
|
641 | 641 | data = entry[:10] |
|
642 | 642 | data_comp = entry[10] & 3 |
|
643 | 643 | sidedata_comp = (entry[11] & 3) << 2 |
|
644 | 644 | data += (data_comp | sidedata_comp,) |
|
645 | 645 | |
|
646 | 646 | return self.index_format.pack(*data) |
|
647 | 647 | |
|
648 | 648 | def entry_binary(self, rev): |
|
649 | 649 | """return the raw binary string representing a revision""" |
|
650 | 650 | entry = self[rev] |
|
651 | 651 | return self._pack_entry(rev, entry) |
|
652 | 652 | |
|
653 | 653 | def pack_header(self, header): |
|
654 | 654 | """pack header information as binary""" |
|
655 | 655 | msg = 'version header should go in the docket, not the index: %d' |
|
656 | 656 | msg %= header |
|
657 | 657 | raise error.ProgrammingError(msg) |
|
658 | 658 | |
|
659 | 659 | |
|
660 | 660 | class IndexChangelogV2(IndexObject2): |
|
661 | 661 | index_format = revlog_constants.INDEX_ENTRY_CL_V2 |
|
662 | 662 | |
|
663 | 663 | def _unpack_entry(self, rev, data, r=True): |
|
664 | 664 | items = self.index_format.unpack(data) |
|
665 | 665 | entry = items[:3] + (rev, rev) + items[3:8] |
|
666 | 666 | data_comp = items[8] & 3 |
|
667 | 667 | sidedata_comp = (items[8] >> 2) & 3 |
|
668 | 668 | return entry + (data_comp, sidedata_comp) |
|
669 | 669 | |
|
670 | 670 | def _pack_entry(self, rev, entry): |
|
671 | 671 | assert entry[3] == rev, entry[3] |
|
672 | 672 | assert entry[4] == rev, entry[4] |
|
673 | 673 | data = entry[:3] + entry[5:10] |
|
674 | 674 | data_comp = entry[10] & 3 |
|
675 | 675 | sidedata_comp = (entry[11] & 3) << 2 |
|
676 | 676 | data += (data_comp | sidedata_comp,) |
|
677 | 677 | return self.index_format.pack(*data) |
|
678 | 678 | |
|
679 | 679 | |
|
680 | 680 | def parse_index_devel_nodemap(data, inline): |
|
681 | 681 | """like parse_index2, but alway return a PersistentNodeMapIndexObject""" |
|
682 | 682 | return PersistentNodeMapIndexObject(data), None |
|
683 | 683 | |
|
684 | 684 | |
|
685 | 685 | def parse_dirstate(dmap, copymap, st): |
|
686 | 686 | parents = [st[:20], st[20:40]] |
|
687 | 687 | # dereference fields so they will be local in loop |
|
688 | 688 | format = b">cllll" |
|
689 | 689 | e_size = struct.calcsize(format) |
|
690 | 690 | pos1 = 40 |
|
691 | 691 | l = len(st) |
|
692 | 692 | |
|
693 | 693 | # the inner loop |
|
694 | 694 | while pos1 < l: |
|
695 | 695 | pos2 = pos1 + e_size |
|
696 | 696 | e = _unpack(b">cllll", st[pos1:pos2]) # a literal here is faster |
|
697 | 697 | pos1 = pos2 + e[4] |
|
698 | 698 | f = st[pos2:pos1] |
|
699 | 699 | if b'\0' in f: |
|
700 | 700 | f, c = f.split(b'\0') |
|
701 | 701 | copymap[f] = c |
|
702 | 702 | dmap[f] = DirstateItem.from_v1_data(*e[:4]) |
|
703 | 703 | return parents |
|
704 | 704 | |
|
705 | 705 | |
|
706 | 706 | def pack_dirstate(dmap, copymap, pl, now): |
|
707 | 707 | now = int(now) |
|
708 | 708 | cs = stringio() |
|
709 | 709 | write = cs.write |
|
710 | 710 | write(b"".join(pl)) |
|
711 | 711 | for f, e in pycompat.iteritems(dmap): |
|
712 | 712 | if e.need_delay(now): |
|
713 | 713 | # The file was last modified "simultaneously" with the current |
|
714 | 714 | # write to dirstate (i.e. within the same second for file- |
|
715 | 715 | # systems with a granularity of 1 sec). This commonly happens |
|
716 | 716 | # for at least a couple of files on 'update'. |
|
717 | 717 | # The user could change the file without changing its size |
|
718 | 718 | # within the same second. Invalidate the file's mtime in |
|
719 | 719 | # dirstate, forcing future 'status' calls to compare the |
|
720 | 720 | # contents of the file if the size is the same. This prevents |
|
721 | 721 | # mistakenly treating such files as clean. |
|
722 | 722 | e.set_possibly_dirty() |
|
723 | 723 | |
|
724 | 724 | if f in copymap: |
|
725 | 725 | f = b"%s\0%s" % (f, copymap[f]) |
|
726 | 726 | e = _pack( |
|
727 | 727 | b">cllll", |
|
728 | 728 | e.v1_state(), |
|
729 | 729 | e.v1_mode(), |
|
730 | 730 | e.v1_size(), |
|
731 | 731 | e.v1_mtime(), |
|
732 | 732 | len(f), |
|
733 | 733 | ) |
|
734 | 734 | write(e) |
|
735 | 735 | write(f) |
|
736 | 736 | return cs.getvalue() |
@@ -1,792 +1,733 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_tree::dirstate_map::{self, DirstateMap, NodeRef}; |
|
6 | 6 | use crate::dirstate_tree::path_with_basename::WithBasename; |
|
7 | 7 | use crate::errors::HgError; |
|
8 | 8 | use crate::utils::hg_path::HgPath; |
|
9 | 9 | use crate::DirstateEntry; |
|
10 | 10 | use crate::DirstateError; |
|
11 | 11 | use crate::DirstateParents; |
|
12 | 12 | use bitflags::bitflags; |
|
13 | 13 | use bytes_cast::unaligned::{I32Be, I64Be, U16Be, U32Be}; |
|
14 | 14 | use bytes_cast::BytesCast; |
|
15 | 15 | use format_bytes::format_bytes; |
|
16 | 16 | use std::borrow::Cow; |
|
17 | 17 | use std::convert::{TryFrom, TryInto}; |
|
18 | 18 | use std::time::{Duration, SystemTime, UNIX_EPOCH}; |
|
19 | 19 | |
|
20 | 20 | /// Added at the start of `.hg/dirstate` when the "v2" format is used. |
|
21 | 21 | /// This a redundant sanity check more than an actual "magic number" since |
|
22 | 22 | /// `.hg/requires` already governs which format should be used. |
|
23 | 23 | pub const V2_FORMAT_MARKER: &[u8; 12] = b"dirstate-v2\n"; |
|
24 | 24 | |
|
25 | 25 | /// Keep space for 256-bit hashes |
|
26 | 26 | const STORED_NODE_ID_BYTES: usize = 32; |
|
27 | 27 | |
|
28 | 28 | /// β¦ even though only 160 bits are used for now, with SHA-1 |
|
29 | 29 | const USED_NODE_ID_BYTES: usize = 20; |
|
30 | 30 | |
|
31 | 31 | pub(super) const IGNORE_PATTERNS_HASH_LEN: usize = 20; |
|
32 | 32 | pub(super) type IgnorePatternsHash = [u8; IGNORE_PATTERNS_HASH_LEN]; |
|
33 | 33 | |
|
34 | 34 | /// Must match the constant of the same name in |
|
35 | 35 | /// `mercurial/dirstateutils/docket.py` |
|
36 | 36 | const TREE_METADATA_SIZE: usize = 44; |
|
37 | 37 | |
|
38 | 38 | /// Make sure that size-affecting changes are made knowingly |
|
39 | 39 | #[allow(unused)] |
|
40 | 40 | fn static_assert_size_of() { |
|
41 | 41 | let _ = std::mem::transmute::<TreeMetadata, [u8; TREE_METADATA_SIZE]>; |
|
42 | 42 | let _ = std::mem::transmute::<DocketHeader, [u8; TREE_METADATA_SIZE + 81]>; |
|
43 | 43 | let _ = std::mem::transmute::<Node, [u8; 43]>; |
|
44 | 44 | } |
|
45 | 45 | |
|
46 | 46 | // Must match `HEADER` in `mercurial/dirstateutils/docket.py` |
|
47 | 47 | #[derive(BytesCast)] |
|
48 | 48 | #[repr(C)] |
|
49 | 49 | struct DocketHeader { |
|
50 | 50 | marker: [u8; V2_FORMAT_MARKER.len()], |
|
51 | 51 | parent_1: [u8; STORED_NODE_ID_BYTES], |
|
52 | 52 | parent_2: [u8; STORED_NODE_ID_BYTES], |
|
53 | 53 | |
|
54 | 54 | metadata: TreeMetadata, |
|
55 | 55 | |
|
56 | 56 | /// Counted in bytes |
|
57 | 57 | data_size: Size, |
|
58 | 58 | |
|
59 | 59 | uuid_size: u8, |
|
60 | 60 | } |
|
61 | 61 | |
|
62 | 62 | pub struct Docket<'on_disk> { |
|
63 | 63 | header: &'on_disk DocketHeader, |
|
64 | 64 | uuid: &'on_disk [u8], |
|
65 | 65 | } |
|
66 | 66 | |
|
67 | /// Fields are documented in the *Tree metadata in the docket file* | |
|
68 | /// section of `mercurial/helptext/internals/dirstate-v2.txt` | |
|
67 | 69 | #[derive(BytesCast)] |
|
68 | 70 | #[repr(C)] |
|
69 | 71 | struct TreeMetadata { |
|
70 | 72 | root_nodes: ChildNodes, |
|
71 | 73 | nodes_with_entry_count: Size, |
|
72 | 74 | nodes_with_copy_source_count: Size, |
|
73 | ||
|
74 | /// How many bytes of this data file are not used anymore | |
|
75 | 75 | unreachable_bytes: Size, |
|
76 | ||
|
77 | /// Current version always sets these bytes to zero when creating or | |
|
78 | /// updating a dirstate. Future versions could assign some bits to signal | |
|
79 | /// for example "the version that last wrote/updated this dirstate did so | |
|
80 | /// in such and such way that can be relied on by versions that know to." | |
|
81 | 76 | unused: [u8; 4], |
|
82 | 77 | |
|
83 | /// If non-zero, a hash of ignore files that were used for some previous | |
|
84 | /// run of the `status` algorithm. | |
|
85 | /// | |
|
86 | /// We define: | |
|
87 | /// | |
|
88 | /// * "Root" ignore files are `.hgignore` at the root of the repository if | |
|
89 | /// it exists, and files from `ui.ignore.*` config. This set of files is | |
|
90 | /// then sorted by the string representation of their path. | |
|
91 | /// * The "expanded contents" of an ignore files is the byte string made | |
|
92 | /// by concatenating its contents with the "expanded contents" of other | |
|
93 | /// files included with `include:` or `subinclude:` files, in inclusion | |
|
94 | /// order. This definition is recursive, as included files can | |
|
95 | /// themselves include more files. | |
|
96 | /// | |
|
97 | /// This hash is defined as the SHA-1 of the concatenation (in sorted | |
|
98 | /// order) of the "expanded contents" of each "root" ignore file. | |
|
99 | /// (Note that computing this does not require actually concatenating byte | |
|
100 | /// strings into contiguous memory, instead SHA-1 hashing can be done | |
|
101 | /// incrementally.) | |
|
78 | /// See *Optional hash of ignore patterns* section of | |
|
79 | /// `mercurial/helptext/internals/dirstate-v2.txt` | |
|
102 | 80 | ignore_patterns_hash: IgnorePatternsHash, |
|
103 | 81 | } |
|
104 | 82 | |
|
83 | /// Fields are documented in the *The data file format* | |
|
84 | /// section of `mercurial/helptext/internals/dirstate-v2.txt` | |
|
105 | 85 | #[derive(BytesCast)] |
|
106 | 86 | #[repr(C)] |
|
107 | 87 | pub(super) struct Node { |
|
108 | 88 | full_path: PathSlice, |
|
109 | 89 | |
|
110 | 90 | /// In bytes from `self.full_path.start` |
|
111 | 91 | base_name_start: PathSize, |
|
112 | 92 | |
|
113 | 93 | copy_source: OptPathSlice, |
|
114 | 94 | children: ChildNodes, |
|
115 | 95 | pub(super) descendants_with_entry_count: Size, |
|
116 | 96 | pub(super) tracked_descendants_count: Size, |
|
117 | ||
|
118 | /// Depending on the bits in `flags`: | |
|
119 | /// | |
|
120 | /// * If any of `WDIR_TRACKED`, `P1_TRACKED`, or `P2_INFO` are set, the | |
|
121 | /// node has an entry. | |
|
122 | /// | |
|
123 | /// - If `HAS_MODE_AND_SIZE` is set, `data.mode` and `data.size` are | |
|
124 | /// meaningful. Otherwise they are set to zero | |
|
125 | /// - If `HAS_MTIME` is set, `data.mtime` is meaningful. Otherwise it is | |
|
126 | /// set to zero. | |
|
127 | /// | |
|
128 | /// * If none of `WDIR_TRACKED`, `P1_TRACKED`, `P2_INFO`, or `HAS_MTIME` | |
|
129 | /// are set, the node does not have an entry and `data` is set to all | |
|
130 | /// zeros. | |
|
131 | /// | |
|
132 | /// * If none of `WDIR_TRACKED`, `P1_TRACKED`, `P2_INFO` are set, but | |
|
133 | /// `HAS_MTIME` is set, the bytes of `data` should instead be | |
|
134 | /// interpreted as the `Timestamp` for the mtime of a cached directory. | |
|
135 | /// | |
|
136 | /// The presence of this combination of flags means that at some point, | |
|
137 | /// this path in the working directory was observed: | |
|
138 | /// | |
|
139 | /// - To be a directory | |
|
140 | /// - With the modification time as given by `Timestamp` | |
|
141 | /// - That timestamp was already strictly in the past when observed, | |
|
142 | /// meaning that later changes cannot happen in the same clock tick | |
|
143 | /// and must cause a different modification time (unless the system | |
|
144 | /// clock jumps back and we get unlucky, which is not impossible but | |
|
145 | /// but deemed unlikely enough). | |
|
146 | /// - All direct children of this directory (as returned by | |
|
147 | /// `std::fs::read_dir`) either have a corresponding dirstate node, or | |
|
148 | /// are ignored by ignore patterns whose hash is in | |
|
149 | /// `TreeMetadata::ignore_patterns_hash`. | |
|
150 | /// | |
|
151 | /// This means that if `std::fs::symlink_metadata` later reports the | |
|
152 | /// same modification time and ignored patterns havenβt changed, a run | |
|
153 | /// of status that is not listing ignored files can skip calling | |
|
154 | /// `std::fs::read_dir` again for this directory, iterate child | |
|
155 | /// dirstate nodes instead. | |
|
156 | 97 | flags: Flags, |
|
157 | 98 | data: Entry, |
|
158 | 99 | } |
|
159 | 100 | |
|
160 | 101 | bitflags! { |
|
161 | 102 | #[derive(BytesCast)] |
|
162 | 103 | #[repr(C)] |
|
163 | 104 | struct Flags: u8 { |
|
164 | 105 | const WDIR_TRACKED = 1 << 0; |
|
165 | 106 | const P1_TRACKED = 1 << 1; |
|
166 | 107 | const P2_INFO = 1 << 2; |
|
167 | 108 | const HAS_MODE_AND_SIZE = 1 << 3; |
|
168 | 109 | const HAS_MTIME = 1 << 4; |
|
169 | 110 | } |
|
170 | 111 | } |
|
171 | 112 | |
|
172 | 113 | #[derive(BytesCast, Copy, Clone, Debug)] |
|
173 | 114 | #[repr(C)] |
|
174 | 115 | struct Entry { |
|
175 | 116 | mode: I32Be, |
|
176 | 117 | mtime: I32Be, |
|
177 | 118 | size: I32Be, |
|
178 | 119 | } |
|
179 | 120 | |
|
180 | 121 | /// Duration since the Unix epoch |
|
181 | 122 | #[derive(BytesCast, Copy, Clone, PartialEq)] |
|
182 | 123 | #[repr(C)] |
|
183 | 124 | pub(super) struct Timestamp { |
|
184 | 125 | seconds: I64Be, |
|
185 | 126 | |
|
186 | 127 | /// In `0 .. 1_000_000_000`. |
|
187 | 128 | /// |
|
188 | 129 | /// This timestamp is later or earlier than `(seconds, 0)` by this many |
|
189 | 130 | /// nanoseconds, if `seconds` is non-negative or negative, respectively. |
|
190 | 131 | nanoseconds: U32Be, |
|
191 | 132 | } |
|
192 | 133 | |
|
193 | 134 | /// Counted in bytes from the start of the file |
|
194 | 135 | /// |
|
195 | 136 | /// NOTE: not supporting `.hg/dirstate` files larger than 4 GiB. |
|
196 | 137 | type Offset = U32Be; |
|
197 | 138 | |
|
198 | 139 | /// Counted in number of items |
|
199 | 140 | /// |
|
200 | 141 | /// NOTE: we choose not to support counting more than 4 billion nodes anywhere. |
|
201 | 142 | type Size = U32Be; |
|
202 | 143 | |
|
203 | 144 | /// Counted in bytes |
|
204 | 145 | /// |
|
205 | 146 | /// NOTE: we choose not to support file names/paths longer than 64 KiB. |
|
206 | 147 | type PathSize = U16Be; |
|
207 | 148 | |
|
208 | 149 | /// A contiguous sequence of `len` times `Node`, representing the child nodes |
|
209 | 150 | /// of either some other node or of the repository root. |
|
210 | 151 | /// |
|
211 | 152 | /// Always sorted by ascending `full_path`, to allow binary search. |
|
212 | 153 | /// Since nodes with the same parent nodes also have the same parent path, |
|
213 | 154 | /// only the `base_name`s need to be compared during binary search. |
|
214 | 155 | #[derive(BytesCast, Copy, Clone)] |
|
215 | 156 | #[repr(C)] |
|
216 | 157 | struct ChildNodes { |
|
217 | 158 | start: Offset, |
|
218 | 159 | len: Size, |
|
219 | 160 | } |
|
220 | 161 | |
|
221 | 162 | /// A `HgPath` of `len` bytes |
|
222 | 163 | #[derive(BytesCast, Copy, Clone)] |
|
223 | 164 | #[repr(C)] |
|
224 | 165 | struct PathSlice { |
|
225 | 166 | start: Offset, |
|
226 | 167 | len: PathSize, |
|
227 | 168 | } |
|
228 | 169 | |
|
229 | 170 | /// Either nothing if `start == 0`, or a `HgPath` of `len` bytes |
|
230 | 171 | type OptPathSlice = PathSlice; |
|
231 | 172 | |
|
232 | 173 | /// Unexpected file format found in `.hg/dirstate` with the "v2" format. |
|
233 | 174 | /// |
|
234 | 175 | /// This should only happen if Mercurial is buggy or a repository is corrupted. |
|
235 | 176 | #[derive(Debug)] |
|
236 | 177 | pub struct DirstateV2ParseError; |
|
237 | 178 | |
|
238 | 179 | impl From<DirstateV2ParseError> for HgError { |
|
239 | 180 | fn from(_: DirstateV2ParseError) -> Self { |
|
240 | 181 | HgError::corrupted("dirstate-v2 parse error") |
|
241 | 182 | } |
|
242 | 183 | } |
|
243 | 184 | |
|
244 | 185 | impl From<DirstateV2ParseError> for crate::DirstateError { |
|
245 | 186 | fn from(error: DirstateV2ParseError) -> Self { |
|
246 | 187 | HgError::from(error).into() |
|
247 | 188 | } |
|
248 | 189 | } |
|
249 | 190 | |
|
250 | 191 | impl<'on_disk> Docket<'on_disk> { |
|
251 | 192 | pub fn parents(&self) -> DirstateParents { |
|
252 | 193 | use crate::Node; |
|
253 | 194 | let p1 = Node::try_from(&self.header.parent_1[..USED_NODE_ID_BYTES]) |
|
254 | 195 | .unwrap() |
|
255 | 196 | .clone(); |
|
256 | 197 | let p2 = Node::try_from(&self.header.parent_2[..USED_NODE_ID_BYTES]) |
|
257 | 198 | .unwrap() |
|
258 | 199 | .clone(); |
|
259 | 200 | DirstateParents { p1, p2 } |
|
260 | 201 | } |
|
261 | 202 | |
|
262 | 203 | pub fn tree_metadata(&self) -> &[u8] { |
|
263 | 204 | self.header.metadata.as_bytes() |
|
264 | 205 | } |
|
265 | 206 | |
|
266 | 207 | pub fn data_size(&self) -> usize { |
|
267 | 208 | // This `unwrap` could only panic on a 16-bit CPU |
|
268 | 209 | self.header.data_size.get().try_into().unwrap() |
|
269 | 210 | } |
|
270 | 211 | |
|
271 | 212 | pub fn data_filename(&self) -> String { |
|
272 | 213 | String::from_utf8(format_bytes!(b"dirstate.{}", self.uuid)).unwrap() |
|
273 | 214 | } |
|
274 | 215 | } |
|
275 | 216 | |
|
276 | 217 | pub fn read_docket( |
|
277 | 218 | on_disk: &[u8], |
|
278 | 219 | ) -> Result<Docket<'_>, DirstateV2ParseError> { |
|
279 | 220 | let (header, uuid) = |
|
280 | 221 | DocketHeader::from_bytes(on_disk).map_err(|_| DirstateV2ParseError)?; |
|
281 | 222 | let uuid_size = header.uuid_size as usize; |
|
282 | 223 | if header.marker == *V2_FORMAT_MARKER && uuid.len() == uuid_size { |
|
283 | 224 | Ok(Docket { header, uuid }) |
|
284 | 225 | } else { |
|
285 | 226 | Err(DirstateV2ParseError) |
|
286 | 227 | } |
|
287 | 228 | } |
|
288 | 229 | |
|
289 | 230 | pub(super) fn read<'on_disk>( |
|
290 | 231 | on_disk: &'on_disk [u8], |
|
291 | 232 | metadata: &[u8], |
|
292 | 233 | ) -> Result<DirstateMap<'on_disk>, DirstateV2ParseError> { |
|
293 | 234 | if on_disk.is_empty() { |
|
294 | 235 | return Ok(DirstateMap::empty(on_disk)); |
|
295 | 236 | } |
|
296 | 237 | let (meta, _) = TreeMetadata::from_bytes(metadata) |
|
297 | 238 | .map_err(|_| DirstateV2ParseError)?; |
|
298 | 239 | let dirstate_map = DirstateMap { |
|
299 | 240 | on_disk, |
|
300 | 241 | root: dirstate_map::ChildNodes::OnDisk(read_nodes( |
|
301 | 242 | on_disk, |
|
302 | 243 | meta.root_nodes, |
|
303 | 244 | )?), |
|
304 | 245 | nodes_with_entry_count: meta.nodes_with_entry_count.get(), |
|
305 | 246 | nodes_with_copy_source_count: meta.nodes_with_copy_source_count.get(), |
|
306 | 247 | ignore_patterns_hash: meta.ignore_patterns_hash, |
|
307 | 248 | unreachable_bytes: meta.unreachable_bytes.get(), |
|
308 | 249 | }; |
|
309 | 250 | Ok(dirstate_map) |
|
310 | 251 | } |
|
311 | 252 | |
|
312 | 253 | impl Node { |
|
313 | 254 | pub(super) fn full_path<'on_disk>( |
|
314 | 255 | &self, |
|
315 | 256 | on_disk: &'on_disk [u8], |
|
316 | 257 | ) -> Result<&'on_disk HgPath, DirstateV2ParseError> { |
|
317 | 258 | read_hg_path(on_disk, self.full_path) |
|
318 | 259 | } |
|
319 | 260 | |
|
320 | 261 | pub(super) fn base_name_start<'on_disk>( |
|
321 | 262 | &self, |
|
322 | 263 | ) -> Result<usize, DirstateV2ParseError> { |
|
323 | 264 | let start = self.base_name_start.get(); |
|
324 | 265 | if start < self.full_path.len.get() { |
|
325 | 266 | let start = usize::try_from(start) |
|
326 | 267 | // u32 -> usize, could only panic on a 16-bit CPU |
|
327 | 268 | .expect("dirstate-v2 base_name_start out of bounds"); |
|
328 | 269 | Ok(start) |
|
329 | 270 | } else { |
|
330 | 271 | Err(DirstateV2ParseError) |
|
331 | 272 | } |
|
332 | 273 | } |
|
333 | 274 | |
|
334 | 275 | pub(super) fn base_name<'on_disk>( |
|
335 | 276 | &self, |
|
336 | 277 | on_disk: &'on_disk [u8], |
|
337 | 278 | ) -> Result<&'on_disk HgPath, DirstateV2ParseError> { |
|
338 | 279 | let full_path = self.full_path(on_disk)?; |
|
339 | 280 | let base_name_start = self.base_name_start()?; |
|
340 | 281 | Ok(HgPath::new(&full_path.as_bytes()[base_name_start..])) |
|
341 | 282 | } |
|
342 | 283 | |
|
343 | 284 | pub(super) fn path<'on_disk>( |
|
344 | 285 | &self, |
|
345 | 286 | on_disk: &'on_disk [u8], |
|
346 | 287 | ) -> Result<dirstate_map::NodeKey<'on_disk>, DirstateV2ParseError> { |
|
347 | 288 | Ok(WithBasename::from_raw_parts( |
|
348 | 289 | Cow::Borrowed(self.full_path(on_disk)?), |
|
349 | 290 | self.base_name_start()?, |
|
350 | 291 | )) |
|
351 | 292 | } |
|
352 | 293 | |
|
353 | 294 | pub(super) fn has_copy_source<'on_disk>(&self) -> bool { |
|
354 | 295 | self.copy_source.start.get() != 0 |
|
355 | 296 | } |
|
356 | 297 | |
|
357 | 298 | pub(super) fn copy_source<'on_disk>( |
|
358 | 299 | &self, |
|
359 | 300 | on_disk: &'on_disk [u8], |
|
360 | 301 | ) -> Result<Option<&'on_disk HgPath>, DirstateV2ParseError> { |
|
361 | 302 | Ok(if self.has_copy_source() { |
|
362 | 303 | Some(read_hg_path(on_disk, self.copy_source)?) |
|
363 | 304 | } else { |
|
364 | 305 | None |
|
365 | 306 | }) |
|
366 | 307 | } |
|
367 | 308 | |
|
368 | 309 | fn has_entry(&self) -> bool { |
|
369 | 310 | self.flags.intersects( |
|
370 | 311 | Flags::WDIR_TRACKED | Flags::P1_TRACKED | Flags::P2_INFO, |
|
371 | 312 | ) |
|
372 | 313 | } |
|
373 | 314 | |
|
374 | 315 | pub(super) fn node_data( |
|
375 | 316 | &self, |
|
376 | 317 | ) -> Result<dirstate_map::NodeData, DirstateV2ParseError> { |
|
377 | 318 | if self.has_entry() { |
|
378 | 319 | Ok(dirstate_map::NodeData::Entry(self.assume_entry())) |
|
379 | 320 | } else if let Some(&mtime) = self.cached_directory_mtime() { |
|
380 | 321 | Ok(dirstate_map::NodeData::CachedDirectory { mtime }) |
|
381 | 322 | } else { |
|
382 | 323 | Ok(dirstate_map::NodeData::None) |
|
383 | 324 | } |
|
384 | 325 | } |
|
385 | 326 | |
|
386 | 327 | pub(super) fn cached_directory_mtime(&self) -> Option<&Timestamp> { |
|
387 | 328 | if self.flags.contains(Flags::HAS_MTIME) && !self.has_entry() { |
|
388 | 329 | Some(self.data.as_timestamp()) |
|
389 | 330 | } else { |
|
390 | 331 | None |
|
391 | 332 | } |
|
392 | 333 | } |
|
393 | 334 | |
|
394 | 335 | fn assume_entry(&self) -> DirstateEntry { |
|
395 | 336 | // TODO: convert through raw bits instead? |
|
396 | 337 | let wdir_tracked = self.flags.contains(Flags::WDIR_TRACKED); |
|
397 | 338 | let p1_tracked = self.flags.contains(Flags::P1_TRACKED); |
|
398 | 339 | let p2_info = self.flags.contains(Flags::P2_INFO); |
|
399 | 340 | let mode_size = if self.flags.contains(Flags::HAS_MODE_AND_SIZE) { |
|
400 | 341 | Some((self.data.mode.into(), self.data.size.into())) |
|
401 | 342 | } else { |
|
402 | 343 | None |
|
403 | 344 | }; |
|
404 | 345 | let mtime = if self.flags.contains(Flags::HAS_MTIME) { |
|
405 | 346 | Some(self.data.mtime.into()) |
|
406 | 347 | } else { |
|
407 | 348 | None |
|
408 | 349 | }; |
|
409 | 350 | DirstateEntry::from_v2_data( |
|
410 | 351 | wdir_tracked, |
|
411 | 352 | p1_tracked, |
|
412 | 353 | p2_info, |
|
413 | 354 | mode_size, |
|
414 | 355 | mtime, |
|
415 | 356 | ) |
|
416 | 357 | } |
|
417 | 358 | |
|
418 | 359 | pub(super) fn entry( |
|
419 | 360 | &self, |
|
420 | 361 | ) -> Result<Option<DirstateEntry>, DirstateV2ParseError> { |
|
421 | 362 | if self.has_entry() { |
|
422 | 363 | Ok(Some(self.assume_entry())) |
|
423 | 364 | } else { |
|
424 | 365 | Ok(None) |
|
425 | 366 | } |
|
426 | 367 | } |
|
427 | 368 | |
|
428 | 369 | pub(super) fn children<'on_disk>( |
|
429 | 370 | &self, |
|
430 | 371 | on_disk: &'on_disk [u8], |
|
431 | 372 | ) -> Result<&'on_disk [Node], DirstateV2ParseError> { |
|
432 | 373 | read_nodes(on_disk, self.children) |
|
433 | 374 | } |
|
434 | 375 | |
|
435 | 376 | pub(super) fn to_in_memory_node<'on_disk>( |
|
436 | 377 | &self, |
|
437 | 378 | on_disk: &'on_disk [u8], |
|
438 | 379 | ) -> Result<dirstate_map::Node<'on_disk>, DirstateV2ParseError> { |
|
439 | 380 | Ok(dirstate_map::Node { |
|
440 | 381 | children: dirstate_map::ChildNodes::OnDisk( |
|
441 | 382 | self.children(on_disk)?, |
|
442 | 383 | ), |
|
443 | 384 | copy_source: self.copy_source(on_disk)?.map(Cow::Borrowed), |
|
444 | 385 | data: self.node_data()?, |
|
445 | 386 | descendants_with_entry_count: self |
|
446 | 387 | .descendants_with_entry_count |
|
447 | 388 | .get(), |
|
448 | 389 | tracked_descendants_count: self.tracked_descendants_count.get(), |
|
449 | 390 | }) |
|
450 | 391 | } |
|
451 | 392 | } |
|
452 | 393 | |
|
453 | 394 | impl Entry { |
|
454 | 395 | fn from_dirstate_entry(entry: &DirstateEntry) -> (Flags, Self) { |
|
455 | 396 | let (wdir_tracked, p1_tracked, p2_info, mode_size_opt, mtime_opt) = |
|
456 | 397 | entry.v2_data(); |
|
457 | 398 | // TODO: convert throug raw flag bits instead? |
|
458 | 399 | let mut flags = Flags::empty(); |
|
459 | 400 | flags.set(Flags::WDIR_TRACKED, wdir_tracked); |
|
460 | 401 | flags.set(Flags::P1_TRACKED, p1_tracked); |
|
461 | 402 | flags.set(Flags::P2_INFO, p2_info); |
|
462 | 403 | let (mode, size, mtime); |
|
463 | 404 | if let Some((m, s)) = mode_size_opt { |
|
464 | 405 | mode = m; |
|
465 | 406 | size = s; |
|
466 | 407 | flags.insert(Flags::HAS_MODE_AND_SIZE) |
|
467 | 408 | } else { |
|
468 | 409 | mode = 0; |
|
469 | 410 | size = 0; |
|
470 | 411 | } |
|
471 | 412 | if let Some(m) = mtime_opt { |
|
472 | 413 | mtime = m; |
|
473 | 414 | flags.insert(Flags::HAS_MTIME); |
|
474 | 415 | } else { |
|
475 | 416 | mtime = 0; |
|
476 | 417 | } |
|
477 | 418 | let raw_entry = Entry { |
|
478 | 419 | mode: mode.into(), |
|
479 | 420 | size: size.into(), |
|
480 | 421 | mtime: mtime.into(), |
|
481 | 422 | }; |
|
482 | 423 | (flags, raw_entry) |
|
483 | 424 | } |
|
484 | 425 | |
|
485 | 426 | fn from_timestamp(timestamp: Timestamp) -> Self { |
|
486 | 427 | // Safety: both types implement the `ByteCast` trait, so we could |
|
487 | 428 | // safely use `as_bytes` and `from_bytes` to do this conversion. Using |
|
488 | 429 | // `transmute` instead makes the compiler check that the two types |
|
489 | 430 | // have the same size, which eliminates the error case of |
|
490 | 431 | // `from_bytes`. |
|
491 | 432 | unsafe { std::mem::transmute::<Timestamp, Entry>(timestamp) } |
|
492 | 433 | } |
|
493 | 434 | |
|
494 | 435 | fn as_timestamp(&self) -> &Timestamp { |
|
495 | 436 | // Safety: same as above in `from_timestamp` |
|
496 | 437 | unsafe { &*(self as *const Entry as *const Timestamp) } |
|
497 | 438 | } |
|
498 | 439 | } |
|
499 | 440 | |
|
500 | 441 | impl Timestamp { |
|
501 | 442 | pub fn seconds(&self) -> i64 { |
|
502 | 443 | self.seconds.get() |
|
503 | 444 | } |
|
504 | 445 | } |
|
505 | 446 | |
|
506 | 447 | impl From<SystemTime> for Timestamp { |
|
507 | 448 | fn from(system_time: SystemTime) -> Self { |
|
508 | 449 | let (secs, nanos) = match system_time.duration_since(UNIX_EPOCH) { |
|
509 | 450 | Ok(duration) => { |
|
510 | 451 | (duration.as_secs() as i64, duration.subsec_nanos()) |
|
511 | 452 | } |
|
512 | 453 | Err(error) => { |
|
513 | 454 | let negative = error.duration(); |
|
514 | 455 | (-(negative.as_secs() as i64), negative.subsec_nanos()) |
|
515 | 456 | } |
|
516 | 457 | }; |
|
517 | 458 | Timestamp { |
|
518 | 459 | seconds: secs.into(), |
|
519 | 460 | nanoseconds: nanos.into(), |
|
520 | 461 | } |
|
521 | 462 | } |
|
522 | 463 | } |
|
523 | 464 | |
|
524 | 465 | impl From<&'_ Timestamp> for SystemTime { |
|
525 | 466 | fn from(timestamp: &'_ Timestamp) -> Self { |
|
526 | 467 | let secs = timestamp.seconds.get(); |
|
527 | 468 | let nanos = timestamp.nanoseconds.get(); |
|
528 | 469 | if secs >= 0 { |
|
529 | 470 | UNIX_EPOCH + Duration::new(secs as u64, nanos) |
|
530 | 471 | } else { |
|
531 | 472 | UNIX_EPOCH - Duration::new((-secs) as u64, nanos) |
|
532 | 473 | } |
|
533 | 474 | } |
|
534 | 475 | } |
|
535 | 476 | |
|
536 | 477 | fn read_hg_path( |
|
537 | 478 | on_disk: &[u8], |
|
538 | 479 | slice: PathSlice, |
|
539 | 480 | ) -> Result<&HgPath, DirstateV2ParseError> { |
|
540 | 481 | read_slice(on_disk, slice.start, slice.len.get()).map(HgPath::new) |
|
541 | 482 | } |
|
542 | 483 | |
|
543 | 484 | fn read_nodes( |
|
544 | 485 | on_disk: &[u8], |
|
545 | 486 | slice: ChildNodes, |
|
546 | 487 | ) -> Result<&[Node], DirstateV2ParseError> { |
|
547 | 488 | read_slice(on_disk, slice.start, slice.len.get()) |
|
548 | 489 | } |
|
549 | 490 | |
|
550 | 491 | fn read_slice<T, Len>( |
|
551 | 492 | on_disk: &[u8], |
|
552 | 493 | start: Offset, |
|
553 | 494 | len: Len, |
|
554 | 495 | ) -> Result<&[T], DirstateV2ParseError> |
|
555 | 496 | where |
|
556 | 497 | T: BytesCast, |
|
557 | 498 | Len: TryInto<usize>, |
|
558 | 499 | { |
|
559 | 500 | // Either `usize::MAX` would result in "out of bounds" error since a single |
|
560 | 501 | // `&[u8]` cannot occupy the entire addess space. |
|
561 | 502 | let start = start.get().try_into().unwrap_or(std::usize::MAX); |
|
562 | 503 | let len = len.try_into().unwrap_or(std::usize::MAX); |
|
563 | 504 | on_disk |
|
564 | 505 | .get(start..) |
|
565 | 506 | .and_then(|bytes| T::slice_from_bytes(bytes, len).ok()) |
|
566 | 507 | .map(|(slice, _rest)| slice) |
|
567 | 508 | .ok_or_else(|| DirstateV2ParseError) |
|
568 | 509 | } |
|
569 | 510 | |
|
570 | 511 | pub(crate) fn for_each_tracked_path<'on_disk>( |
|
571 | 512 | on_disk: &'on_disk [u8], |
|
572 | 513 | metadata: &[u8], |
|
573 | 514 | mut f: impl FnMut(&'on_disk HgPath), |
|
574 | 515 | ) -> Result<(), DirstateV2ParseError> { |
|
575 | 516 | let (meta, _) = TreeMetadata::from_bytes(metadata) |
|
576 | 517 | .map_err(|_| DirstateV2ParseError)?; |
|
577 | 518 | fn recur<'on_disk>( |
|
578 | 519 | on_disk: &'on_disk [u8], |
|
579 | 520 | nodes: ChildNodes, |
|
580 | 521 | f: &mut impl FnMut(&'on_disk HgPath), |
|
581 | 522 | ) -> Result<(), DirstateV2ParseError> { |
|
582 | 523 | for node in read_nodes(on_disk, nodes)? { |
|
583 | 524 | if let Some(entry) = node.entry()? { |
|
584 | 525 | if entry.state().is_tracked() { |
|
585 | 526 | f(node.full_path(on_disk)?) |
|
586 | 527 | } |
|
587 | 528 | } |
|
588 | 529 | recur(on_disk, node.children, f)? |
|
589 | 530 | } |
|
590 | 531 | Ok(()) |
|
591 | 532 | } |
|
592 | 533 | recur(on_disk, meta.root_nodes, &mut f) |
|
593 | 534 | } |
|
594 | 535 | |
|
595 | 536 | /// Returns new data and metadata, together with whether that data should be |
|
596 | 537 | /// appended to the existing data file whose content is at |
|
597 | 538 | /// `dirstate_map.on_disk` (true), instead of written to a new data file |
|
598 | 539 | /// (false). |
|
599 | 540 | pub(super) fn write( |
|
600 | 541 | dirstate_map: &mut DirstateMap, |
|
601 | 542 | can_append: bool, |
|
602 | 543 | ) -> Result<(Vec<u8>, Vec<u8>, bool), DirstateError> { |
|
603 | 544 | let append = can_append && dirstate_map.write_should_append(); |
|
604 | 545 | |
|
605 | 546 | // This ignores the space for paths, and for nodes without an entry. |
|
606 | 547 | // TODO: better estimate? Skip the `Vec` and write to a file directly? |
|
607 | 548 | let size_guess = std::mem::size_of::<Node>() |
|
608 | 549 | * dirstate_map.nodes_with_entry_count as usize; |
|
609 | 550 | |
|
610 | 551 | let mut writer = Writer { |
|
611 | 552 | dirstate_map, |
|
612 | 553 | append, |
|
613 | 554 | out: Vec::with_capacity(size_guess), |
|
614 | 555 | }; |
|
615 | 556 | |
|
616 | 557 | let root_nodes = writer.write_nodes(dirstate_map.root.as_ref())?; |
|
617 | 558 | |
|
618 | 559 | let meta = TreeMetadata { |
|
619 | 560 | root_nodes, |
|
620 | 561 | nodes_with_entry_count: dirstate_map.nodes_with_entry_count.into(), |
|
621 | 562 | nodes_with_copy_source_count: dirstate_map |
|
622 | 563 | .nodes_with_copy_source_count |
|
623 | 564 | .into(), |
|
624 | 565 | unreachable_bytes: dirstate_map.unreachable_bytes.into(), |
|
625 | 566 | unused: [0; 4], |
|
626 | 567 | ignore_patterns_hash: dirstate_map.ignore_patterns_hash, |
|
627 | 568 | }; |
|
628 | 569 | Ok((writer.out, meta.as_bytes().to_vec(), append)) |
|
629 | 570 | } |
|
630 | 571 | |
|
631 | 572 | struct Writer<'dmap, 'on_disk> { |
|
632 | 573 | dirstate_map: &'dmap DirstateMap<'on_disk>, |
|
633 | 574 | append: bool, |
|
634 | 575 | out: Vec<u8>, |
|
635 | 576 | } |
|
636 | 577 | |
|
637 | 578 | impl Writer<'_, '_> { |
|
638 | 579 | fn write_nodes( |
|
639 | 580 | &mut self, |
|
640 | 581 | nodes: dirstate_map::ChildNodesRef, |
|
641 | 582 | ) -> Result<ChildNodes, DirstateError> { |
|
642 | 583 | // Reuse already-written nodes if possible |
|
643 | 584 | if self.append { |
|
644 | 585 | if let dirstate_map::ChildNodesRef::OnDisk(nodes_slice) = nodes { |
|
645 | 586 | let start = self.on_disk_offset_of(nodes_slice).expect( |
|
646 | 587 | "dirstate-v2 OnDisk nodes not found within on_disk", |
|
647 | 588 | ); |
|
648 | 589 | let len = child_nodes_len_from_usize(nodes_slice.len()); |
|
649 | 590 | return Ok(ChildNodes { start, len }); |
|
650 | 591 | } |
|
651 | 592 | } |
|
652 | 593 | |
|
653 | 594 | // `dirstate_map::ChildNodes::InMemory` contains a `HashMap` which has |
|
654 | 595 | // undefined iteration order. Sort to enable binary search in the |
|
655 | 596 | // written file. |
|
656 | 597 | let nodes = nodes.sorted(); |
|
657 | 598 | let nodes_len = nodes.len(); |
|
658 | 599 | |
|
659 | 600 | // First accumulate serialized nodes in a `Vec` |
|
660 | 601 | let mut on_disk_nodes = Vec::with_capacity(nodes_len); |
|
661 | 602 | for node in nodes { |
|
662 | 603 | let children = |
|
663 | 604 | self.write_nodes(node.children(self.dirstate_map.on_disk)?)?; |
|
664 | 605 | let full_path = node.full_path(self.dirstate_map.on_disk)?; |
|
665 | 606 | let full_path = self.write_path(full_path.as_bytes()); |
|
666 | 607 | let copy_source = if let Some(source) = |
|
667 | 608 | node.copy_source(self.dirstate_map.on_disk)? |
|
668 | 609 | { |
|
669 | 610 | self.write_path(source.as_bytes()) |
|
670 | 611 | } else { |
|
671 | 612 | PathSlice { |
|
672 | 613 | start: 0.into(), |
|
673 | 614 | len: 0.into(), |
|
674 | 615 | } |
|
675 | 616 | }; |
|
676 | 617 | on_disk_nodes.push(match node { |
|
677 | 618 | NodeRef::InMemory(path, node) => { |
|
678 | 619 | let (flags, data) = match &node.data { |
|
679 | 620 | dirstate_map::NodeData::Entry(entry) => { |
|
680 | 621 | Entry::from_dirstate_entry(entry) |
|
681 | 622 | } |
|
682 | 623 | dirstate_map::NodeData::CachedDirectory { mtime } => { |
|
683 | 624 | (Flags::HAS_MTIME, Entry::from_timestamp(*mtime)) |
|
684 | 625 | } |
|
685 | 626 | dirstate_map::NodeData::None => ( |
|
686 | 627 | Flags::empty(), |
|
687 | 628 | Entry { |
|
688 | 629 | mode: 0.into(), |
|
689 | 630 | size: 0.into(), |
|
690 | 631 | mtime: 0.into(), |
|
691 | 632 | }, |
|
692 | 633 | ), |
|
693 | 634 | }; |
|
694 | 635 | Node { |
|
695 | 636 | children, |
|
696 | 637 | copy_source, |
|
697 | 638 | full_path, |
|
698 | 639 | base_name_start: u16::try_from(path.base_name_start()) |
|
699 | 640 | // Could only panic for paths over 64 KiB |
|
700 | 641 | .expect("dirstate-v2 path length overflow") |
|
701 | 642 | .into(), |
|
702 | 643 | descendants_with_entry_count: node |
|
703 | 644 | .descendants_with_entry_count |
|
704 | 645 | .into(), |
|
705 | 646 | tracked_descendants_count: node |
|
706 | 647 | .tracked_descendants_count |
|
707 | 648 | .into(), |
|
708 | 649 | flags, |
|
709 | 650 | data, |
|
710 | 651 | } |
|
711 | 652 | } |
|
712 | 653 | NodeRef::OnDisk(node) => Node { |
|
713 | 654 | children, |
|
714 | 655 | copy_source, |
|
715 | 656 | full_path, |
|
716 | 657 | ..*node |
|
717 | 658 | }, |
|
718 | 659 | }) |
|
719 | 660 | } |
|
720 | 661 | // β¦ so we can write them contiguously, after writing everything else |
|
721 | 662 | // they refer to. |
|
722 | 663 | let start = self.current_offset(); |
|
723 | 664 | let len = child_nodes_len_from_usize(nodes_len); |
|
724 | 665 | self.out.extend(on_disk_nodes.as_bytes()); |
|
725 | 666 | Ok(ChildNodes { start, len }) |
|
726 | 667 | } |
|
727 | 668 | |
|
728 | 669 | /// If the given slice of items is within `on_disk`, returns its offset |
|
729 | 670 | /// from the start of `on_disk`. |
|
730 | 671 | fn on_disk_offset_of<T>(&self, slice: &[T]) -> Option<Offset> |
|
731 | 672 | where |
|
732 | 673 | T: BytesCast, |
|
733 | 674 | { |
|
734 | 675 | fn address_range(slice: &[u8]) -> std::ops::RangeInclusive<usize> { |
|
735 | 676 | let start = slice.as_ptr() as usize; |
|
736 | 677 | let end = start + slice.len(); |
|
737 | 678 | start..=end |
|
738 | 679 | } |
|
739 | 680 | let slice_addresses = address_range(slice.as_bytes()); |
|
740 | 681 | let on_disk_addresses = address_range(self.dirstate_map.on_disk); |
|
741 | 682 | if on_disk_addresses.contains(slice_addresses.start()) |
|
742 | 683 | && on_disk_addresses.contains(slice_addresses.end()) |
|
743 | 684 | { |
|
744 | 685 | let offset = slice_addresses.start() - on_disk_addresses.start(); |
|
745 | 686 | Some(offset_from_usize(offset)) |
|
746 | 687 | } else { |
|
747 | 688 | None |
|
748 | 689 | } |
|
749 | 690 | } |
|
750 | 691 | |
|
751 | 692 | fn current_offset(&mut self) -> Offset { |
|
752 | 693 | let mut offset = self.out.len(); |
|
753 | 694 | if self.append { |
|
754 | 695 | offset += self.dirstate_map.on_disk.len() |
|
755 | 696 | } |
|
756 | 697 | offset_from_usize(offset) |
|
757 | 698 | } |
|
758 | 699 | |
|
759 | 700 | fn write_path(&mut self, slice: &[u8]) -> PathSlice { |
|
760 | 701 | let len = path_len_from_usize(slice.len()); |
|
761 | 702 | // Reuse an already-written path if possible |
|
762 | 703 | if self.append { |
|
763 | 704 | if let Some(start) = self.on_disk_offset_of(slice) { |
|
764 | 705 | return PathSlice { start, len }; |
|
765 | 706 | } |
|
766 | 707 | } |
|
767 | 708 | let start = self.current_offset(); |
|
768 | 709 | self.out.extend(slice.as_bytes()); |
|
769 | 710 | PathSlice { start, len } |
|
770 | 711 | } |
|
771 | 712 | } |
|
772 | 713 | |
|
773 | 714 | fn offset_from_usize(x: usize) -> Offset { |
|
774 | 715 | u32::try_from(x) |
|
775 | 716 | // Could only panic for a dirstate file larger than 4 GiB |
|
776 | 717 | .expect("dirstate-v2 offset overflow") |
|
777 | 718 | .into() |
|
778 | 719 | } |
|
779 | 720 | |
|
780 | 721 | fn child_nodes_len_from_usize(x: usize) -> Size { |
|
781 | 722 | u32::try_from(x) |
|
782 | 723 | // Could only panic with over 4 billion nodes |
|
783 | 724 | .expect("dirstate-v2 slice length overflow") |
|
784 | 725 | .into() |
|
785 | 726 | } |
|
786 | 727 | |
|
787 | 728 | fn path_len_from_usize(x: usize) -> PathSize { |
|
788 | 729 | u16::try_from(x) |
|
789 | 730 | // Could only panic for paths over 64 KiB |
|
790 | 731 | .expect("dirstate-v2 path length overflow") |
|
791 | 732 | .into() |
|
792 | 733 | } |
General Comments 0
You need to be logged in to leave comments.
Login now