Show More
@@ -1,341 +1,346 b'' | |||
|
1 | 1 | use crate::errors::HgError; |
|
2 | 2 | use crate::revlog::{Node, NodePrefix}; |
|
3 | 3 | use crate::revlog::{Revision, NULL_REVISION}; |
|
4 | 4 | use crate::revlog::{Revlog, RevlogEntry, RevlogError}; |
|
5 | 5 | use crate::utils::hg_path::HgPath; |
|
6 | 6 | use crate::vfs::Vfs; |
|
7 | 7 | use itertools::Itertools; |
|
8 | 8 | use std::ascii::escape_default; |
|
9 | 9 | use std::borrow::Cow; |
|
10 | 10 | use std::fmt::{Debug, Formatter}; |
|
11 | 11 | |
|
12 | 12 | /// A specialized `Revlog` to work with changelog data format. |
|
13 | 13 | pub struct Changelog { |
|
14 | 14 | /// The generic `revlog` format. |
|
15 | 15 | pub(crate) revlog: Revlog, |
|
16 | 16 | } |
|
17 | 17 | |
|
18 | 18 | impl Changelog { |
|
19 | 19 | /// Open the `changelog` of a repository given by its root. |
|
20 | 20 | pub fn open(store_vfs: &Vfs, use_nodemap: bool) -> Result<Self, HgError> { |
|
21 | 21 | let revlog = |
|
22 | 22 | Revlog::open(store_vfs, "00changelog.i", None, use_nodemap)?; |
|
23 | 23 | Ok(Self { revlog }) |
|
24 | 24 | } |
|
25 | 25 | |
|
26 | 26 | /// Return the `ChangelogRevisionData` for the given node ID. |
|
27 | 27 | pub fn data_for_node( |
|
28 | 28 | &self, |
|
29 | 29 | node: NodePrefix, |
|
30 | 30 | ) -> Result<ChangelogRevisionData, RevlogError> { |
|
31 | 31 | let rev = self.revlog.rev_from_node(node)?; |
|
32 | 32 | self.data_for_rev(rev) |
|
33 | 33 | } |
|
34 | 34 | |
|
35 | 35 | /// Return the [`ChangelogEntry`] for the given revision number. |
|
36 | 36 | pub fn entry_for_rev( |
|
37 | 37 | &self, |
|
38 | 38 | rev: Revision, |
|
39 | 39 | ) -> Result<ChangelogEntry, RevlogError> { |
|
40 | 40 | let revlog_entry = self.revlog.get_entry(rev)?; |
|
41 | 41 | Ok(ChangelogEntry { revlog_entry }) |
|
42 | 42 | } |
|
43 | 43 | |
|
44 | 44 | /// Return the [`ChangelogRevisionData`] for the given revision number. |
|
45 | 45 | /// |
|
46 | 46 | /// This is a useful shortcut in case the caller does not need the |
|
47 | 47 | /// generic revlog information (parents, hashes etc). Otherwise |
|
48 | 48 | /// consider taking a [`ChangelogEntry`] with |
|
49 | 49 | /// [entry_for_rev](`Self::entry_for_rev`) and doing everything from there. |
|
50 | 50 | pub fn data_for_rev( |
|
51 | 51 | &self, |
|
52 | 52 | rev: Revision, |
|
53 | 53 | ) -> Result<ChangelogRevisionData, RevlogError> { |
|
54 | 54 | if rev == NULL_REVISION { |
|
55 | 55 | return Ok(ChangelogRevisionData::null()); |
|
56 | 56 | } |
|
57 | 57 | self.entry_for_rev(rev)?.data() |
|
58 | 58 | } |
|
59 | 59 | |
|
60 | 60 | pub fn node_from_rev(&self, rev: Revision) -> Option<&Node> { |
|
61 | 61 | self.revlog.node_from_rev(rev) |
|
62 | 62 | } |
|
63 | 63 | |
|
64 | 64 | pub fn rev_from_node( |
|
65 | 65 | &self, |
|
66 | 66 | node: NodePrefix, |
|
67 | 67 | ) -> Result<Revision, RevlogError> { |
|
68 | 68 | self.revlog.rev_from_node(node) |
|
69 | 69 | } |
|
70 | 70 | } |
|
71 | 71 | |
|
72 | 72 | /// A specialized `RevlogEntry` for `changelog` data format |
|
73 | 73 | /// |
|
74 | 74 | /// This is a `RevlogEntry` with the added semantics that the associated |
|
75 | 75 | /// data should meet the requirements for `changelog`, materialized by |
|
76 | 76 | /// the fact that `data()` constructs a `ChangelogRevisionData`. |
|
77 | 77 | /// In case that promise would be broken, the `data` method returns an error. |
|
78 | 78 | #[derive(Clone)] |
|
79 | 79 | pub struct ChangelogEntry<'changelog> { |
|
80 | 80 | /// Same data, as a generic `RevlogEntry`. |
|
81 | 81 | pub(crate) revlog_entry: RevlogEntry<'changelog>, |
|
82 | 82 | } |
|
83 | 83 | |
|
84 | 84 | impl<'changelog> ChangelogEntry<'changelog> { |
|
85 | 85 | pub fn data<'a>( |
|
86 | 86 | &'a self, |
|
87 | 87 | ) -> Result<ChangelogRevisionData<'changelog>, RevlogError> { |
|
88 | 88 | let bytes = self.revlog_entry.data()?; |
|
89 | 89 | if bytes.is_empty() { |
|
90 | 90 | Ok(ChangelogRevisionData::null()) |
|
91 | 91 | } else { |
|
92 | 92 | Ok(ChangelogRevisionData::new(bytes).map_err(|err| { |
|
93 | 93 | RevlogError::Other(HgError::CorruptedRepository(format!( |
|
94 | 94 | "Invalid changelog data for revision {}: {:?}", |
|
95 | 95 | self.revlog_entry.revision(), |
|
96 | 96 | err |
|
97 | 97 | ))) |
|
98 | 98 | })?) |
|
99 | 99 | } |
|
100 | 100 | } |
|
101 | 101 | |
|
102 | 102 | /// Obtain a reference to the underlying `RevlogEntry`. |
|
103 | 103 | /// |
|
104 | 104 | /// This allows the caller to access the information that is common |
|
105 | 105 | /// to all revlog entries: revision number, node id, parent revisions etc. |
|
106 | 106 | pub fn as_revlog_entry(&self) -> &RevlogEntry { |
|
107 | 107 | &self.revlog_entry |
|
108 | 108 | } |
|
109 | 109 | |
|
110 | 110 | pub fn p1_entry(&self) -> Result<Option<ChangelogEntry>, RevlogError> { |
|
111 | 111 | Ok(self |
|
112 | 112 | .revlog_entry |
|
113 | 113 | .p1_entry()? |
|
114 | 114 | .map(|revlog_entry| Self { revlog_entry })) |
|
115 | 115 | } |
|
116 | 116 | |
|
117 | 117 | pub fn p2_entry(&self) -> Result<Option<ChangelogEntry>, RevlogError> { |
|
118 | 118 | Ok(self |
|
119 | 119 | .revlog_entry |
|
120 | 120 | .p2_entry()? |
|
121 | 121 | .map(|revlog_entry| Self { revlog_entry })) |
|
122 | 122 | } |
|
123 | 123 | } |
|
124 | 124 | |
|
125 | 125 | /// `Changelog` entry which knows how to interpret the `changelog` data bytes. |
|
126 | 126 | #[derive(PartialEq)] |
|
127 | 127 | pub struct ChangelogRevisionData<'changelog> { |
|
128 | 128 | /// The data bytes of the `changelog` entry. |
|
129 | 129 | bytes: Cow<'changelog, [u8]>, |
|
130 | 130 | /// The end offset for the hex manifest (not including the newline) |
|
131 | 131 | manifest_end: usize, |
|
132 | 132 | /// The end offset for the user+email (not including the newline) |
|
133 | 133 | user_end: usize, |
|
134 | 134 | /// The end offset for the timestamp+timezone+extras (not including the |
|
135 | 135 | /// newline) |
|
136 | 136 | timestamp_end: usize, |
|
137 | 137 | /// The end offset for the file list (not including the newline) |
|
138 | 138 | files_end: usize, |
|
139 | 139 | } |
|
140 | 140 | |
|
141 | 141 | impl<'changelog> ChangelogRevisionData<'changelog> { |
|
142 | 142 | fn new(bytes: Cow<'changelog, [u8]>) -> Result<Self, HgError> { |
|
143 | 143 | let mut line_iter = bytes.split(|b| b == &b'\n'); |
|
144 | 144 | let manifest_end = line_iter |
|
145 | 145 | .next() |
|
146 | 146 | .expect("Empty iterator from split()?") |
|
147 | 147 | .len(); |
|
148 | 148 | let user_slice = line_iter.next().ok_or_else(|| { |
|
149 | 149 | HgError::corrupted("Changeset data truncated after manifest line") |
|
150 | 150 | })?; |
|
151 | 151 | let user_end = manifest_end + 1 + user_slice.len(); |
|
152 | 152 | let timestamp_slice = line_iter.next().ok_or_else(|| { |
|
153 | 153 | HgError::corrupted("Changeset data truncated after user line") |
|
154 | 154 | })?; |
|
155 | 155 | let timestamp_end = user_end + 1 + timestamp_slice.len(); |
|
156 | 156 | let mut files_end = timestamp_end + 1; |
|
157 | 157 | loop { |
|
158 | 158 | let line = line_iter.next().ok_or_else(|| { |
|
159 | 159 | HgError::corrupted("Changeset data truncated in files list") |
|
160 | 160 | })?; |
|
161 | 161 | if line.is_empty() { |
|
162 | 162 | if files_end == bytes.len() { |
|
163 | 163 | // The list of files ended with a single newline (there |
|
164 | 164 | // should be two) |
|
165 | 165 | return Err(HgError::corrupted( |
|
166 | 166 | "Changeset data truncated after files list", |
|
167 | 167 | )); |
|
168 | 168 | } |
|
169 | 169 | files_end -= 1; |
|
170 | 170 | break; |
|
171 | 171 | } |
|
172 | 172 | files_end += line.len() + 1; |
|
173 | 173 | } |
|
174 | 174 | |
|
175 | 175 | Ok(Self { |
|
176 | 176 | bytes, |
|
177 | 177 | manifest_end, |
|
178 | 178 | user_end, |
|
179 | 179 | timestamp_end, |
|
180 | 180 | files_end, |
|
181 | 181 | }) |
|
182 | 182 | } |
|
183 | 183 | |
|
184 | 184 | fn null() -> Self { |
|
185 | 185 | Self::new(Cow::Borrowed( |
|
186 | 186 | b"0000000000000000000000000000000000000000\n\n0 0\n\n", |
|
187 | 187 | )) |
|
188 | 188 | .unwrap() |
|
189 | 189 | } |
|
190 | 190 | |
|
191 | 191 | /// Return an iterator over the lines of the entry. |
|
192 | 192 | pub fn lines(&self) -> impl Iterator<Item = &[u8]> { |
|
193 | 193 | self.bytes.split(|b| b == &b'\n') |
|
194 | 194 | } |
|
195 | 195 | |
|
196 | 196 | /// Return the node id of the `manifest` referenced by this `changelog` |
|
197 | 197 | /// entry. |
|
198 | 198 | pub fn manifest_node(&self) -> Result<Node, HgError> { |
|
199 | 199 | let manifest_node_hex = &self.bytes[..self.manifest_end]; |
|
200 | 200 | Node::from_hex_for_repo(manifest_node_hex) |
|
201 | 201 | } |
|
202 | 202 | |
|
203 | 203 | /// The full user string (usually a name followed by an email enclosed in |
|
204 | 204 | /// angle brackets) |
|
205 | 205 | pub fn user(&self) -> &[u8] { |
|
206 | 206 | &self.bytes[self.manifest_end + 1..self.user_end] |
|
207 | 207 | } |
|
208 | 208 | |
|
209 | 209 | /// The full timestamp line (timestamp in seconds, offset in seconds, and |
|
210 | 210 | /// possibly extras) |
|
211 | 211 | // TODO: We should expose this in a more useful way |
|
212 | 212 | pub fn timestamp_line(&self) -> &[u8] { |
|
213 | 213 | &self.bytes[self.user_end + 1..self.timestamp_end] |
|
214 | 214 | } |
|
215 | 215 | |
|
216 | 216 | /// The files changed in this revision. |
|
217 | 217 | pub fn files(&self) -> impl Iterator<Item = &HgPath> { |
|
218 | 218 | self.bytes[self.timestamp_end + 1..self.files_end] |
|
219 | 219 | .split(|b| b == &b'\n') |
|
220 | 220 | .map(HgPath::new) |
|
221 | 221 | } |
|
222 | 222 | |
|
223 | 223 | /// The change description. |
|
224 | 224 | pub fn description(&self) -> &[u8] { |
|
225 | 225 | &self.bytes[self.files_end + 2..] |
|
226 | 226 | } |
|
227 | 227 | } |
|
228 | 228 | |
|
229 | 229 | impl Debug for ChangelogRevisionData<'_> { |
|
230 | 230 | fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result { |
|
231 | 231 | f.debug_struct("ChangelogRevisionData") |
|
232 | 232 | .field("bytes", &debug_bytes(&self.bytes)) |
|
233 | 233 | .field("manifest", &debug_bytes(&self.bytes[..self.manifest_end])) |
|
234 | 234 | .field( |
|
235 | 235 | "user", |
|
236 | 236 | &debug_bytes( |
|
237 | 237 | &self.bytes[self.manifest_end + 1..self.user_end], |
|
238 | 238 | ), |
|
239 | 239 | ) |
|
240 | 240 | .field( |
|
241 | 241 | "timestamp", |
|
242 | 242 | &debug_bytes( |
|
243 | 243 | &self.bytes[self.user_end + 1..self.timestamp_end], |
|
244 | 244 | ), |
|
245 | 245 | ) |
|
246 | 246 | .field( |
|
247 | 247 | "files", |
|
248 | 248 | &debug_bytes( |
|
249 | 249 | &self.bytes[self.timestamp_end + 1..self.files_end], |
|
250 | 250 | ), |
|
251 | 251 | ) |
|
252 | 252 | .field( |
|
253 | 253 | "description", |
|
254 | 254 | &debug_bytes(&self.bytes[self.files_end + 2..]), |
|
255 | 255 | ) |
|
256 | 256 | .finish() |
|
257 | 257 | } |
|
258 | 258 | } |
|
259 | 259 | |
|
260 | 260 | fn debug_bytes(bytes: &[u8]) -> String { |
|
261 | 261 | String::from_utf8_lossy( |
|
262 | 262 | &bytes.iter().flat_map(|b| escape_default(*b)).collect_vec(), |
|
263 | 263 | ) |
|
264 | 264 | .to_string() |
|
265 | 265 | } |
|
266 | 266 | |
|
267 | 267 | #[cfg(test)] |
|
268 | 268 | mod tests { |
|
269 | 269 | use super::*; |
|
270 | 270 | use crate::vfs::Vfs; |
|
271 | 271 | use crate::NULL_REVISION; |
|
272 | 272 | use pretty_assertions::assert_eq; |
|
273 | 273 | |
|
274 | 274 | #[test] |
|
275 | 275 | fn test_create_changelogrevisiondata_invalid() { |
|
276 | 276 | // Completely empty |
|
277 | 277 | assert!(ChangelogRevisionData::new(Cow::Borrowed(b"abcd")).is_err()); |
|
278 | 278 | // No newline after manifest |
|
279 | 279 | assert!(ChangelogRevisionData::new(Cow::Borrowed(b"abcd")).is_err()); |
|
280 | 280 | // No newline after user |
|
281 | 281 | assert!(ChangelogRevisionData::new(Cow::Borrowed(b"abcd\n")).is_err()); |
|
282 | 282 | // No newline after timestamp |
|
283 | 283 | assert!( |
|
284 | 284 | ChangelogRevisionData::new(Cow::Borrowed(b"abcd\n\n0 0")).is_err() |
|
285 | 285 | ); |
|
286 | 286 | // Missing newline after files |
|
287 | 287 | assert!(ChangelogRevisionData::new(Cow::Borrowed( |
|
288 | 288 | b"abcd\n\n0 0\nfile1\nfile2" |
|
289 | 289 | )) |
|
290 | 290 | .is_err(),); |
|
291 | 291 | // Only one newline after files |
|
292 | 292 | assert!(ChangelogRevisionData::new(Cow::Borrowed( |
|
293 | 293 | b"abcd\n\n0 0\nfile1\nfile2\n" |
|
294 | 294 | )) |
|
295 | 295 | .is_err(),); |
|
296 | 296 | } |
|
297 | 297 | |
|
298 | 298 | #[test] |
|
299 | 299 | fn test_create_changelogrevisiondata() { |
|
300 | 300 | let data = ChangelogRevisionData::new(Cow::Borrowed( |
|
301 | 301 | b"0123456789abcdef0123456789abcdef01234567 |
|
302 | 302 | Some One <someone@example.com> |
|
303 | 303 | 0 0 |
|
304 | 304 | file1 |
|
305 | 305 | file2 |
|
306 | 306 | |
|
307 | 307 | some |
|
308 | 308 | commit |
|
309 | 309 | message", |
|
310 | 310 | )) |
|
311 | 311 | .unwrap(); |
|
312 | 312 | assert_eq!( |
|
313 | 313 | data.manifest_node().unwrap(), |
|
314 | 314 | Node::from_hex("0123456789abcdef0123456789abcdef01234567") |
|
315 | 315 | .unwrap() |
|
316 | 316 | ); |
|
317 | 317 | assert_eq!(data.user(), b"Some One <someone@example.com>"); |
|
318 | 318 | assert_eq!(data.timestamp_line(), b"0 0"); |
|
319 | 319 | assert_eq!( |
|
320 | 320 | data.files().collect_vec(), |
|
321 | 321 | vec![HgPath::new("file1"), HgPath::new("file2")] |
|
322 | 322 | ); |
|
323 | 323 | assert_eq!(data.description(), b"some\ncommit\nmessage"); |
|
324 | 324 | } |
|
325 | 325 | |
|
326 | 326 | #[test] |
|
327 | 327 | fn test_data_from_rev_null() -> Result<(), RevlogError> { |
|
328 | 328 | // an empty revlog will be enough for this case |
|
329 | 329 | let temp = tempfile::tempdir().unwrap(); |
|
330 | 330 | let vfs = Vfs { base: temp.path() }; |
|
331 | 331 | std::fs::write(temp.path().join("foo.i"), b"").unwrap(); |
|
332 | 332 | let revlog = Revlog::open(&vfs, "foo.i", None, false).unwrap(); |
|
333 | 333 | |
|
334 | 334 | let changelog = Changelog { revlog }; |
|
335 | 335 | assert_eq!( |
|
336 | 336 | changelog.data_for_rev(NULL_REVISION)?, |
|
337 | 337 | ChangelogRevisionData::null() |
|
338 | 338 | ); |
|
339 | // same with the intermediate entry object | |
|
340 | assert_eq!( | |
|
341 | changelog.entry_for_rev(NULL_REVISION)?.data()?, | |
|
342 | ChangelogRevisionData::null() | |
|
343 | ); | |
|
339 | 344 | Ok(()) |
|
340 | 345 | } |
|
341 | 346 | } |
@@ -1,823 +1,829 b'' | |||
|
1 | 1 | // Copyright 2018-2023 Georges Racinet <georges.racinet@octobus.net> |
|
2 | 2 | // and Mercurial contributors |
|
3 | 3 | // |
|
4 | 4 | // This software may be used and distributed according to the terms of the |
|
5 | 5 | // GNU General Public License version 2 or any later version. |
|
6 | 6 | //! Mercurial concepts for handling revision history |
|
7 | 7 | |
|
8 | 8 | pub mod node; |
|
9 | 9 | pub mod nodemap; |
|
10 | 10 | mod nodemap_docket; |
|
11 | 11 | pub mod path_encode; |
|
12 | 12 | pub use node::{FromHexError, Node, NodePrefix}; |
|
13 | 13 | pub mod changelog; |
|
14 | 14 | pub mod filelog; |
|
15 | 15 | pub mod index; |
|
16 | 16 | pub mod manifest; |
|
17 | 17 | pub mod patch; |
|
18 | 18 | |
|
19 | 19 | use std::borrow::Cow; |
|
20 | 20 | use std::io::Read; |
|
21 | 21 | use std::ops::Deref; |
|
22 | 22 | use std::path::Path; |
|
23 | 23 | |
|
24 | 24 | use flate2::read::ZlibDecoder; |
|
25 | 25 | use sha1::{Digest, Sha1}; |
|
26 | 26 | use std::cell::RefCell; |
|
27 | 27 | use zstd; |
|
28 | 28 | |
|
29 | 29 | use self::node::{NODE_BYTES_LENGTH, NULL_NODE}; |
|
30 | 30 | use self::nodemap_docket::NodeMapDocket; |
|
31 | 31 | use super::index::Index; |
|
32 | 32 | use super::nodemap::{NodeMap, NodeMapError}; |
|
33 | 33 | use crate::errors::HgError; |
|
34 | 34 | use crate::vfs::Vfs; |
|
35 | 35 | |
|
36 | 36 | /// Mercurial revision numbers |
|
37 | 37 | /// |
|
38 | 38 | /// As noted in revlog.c, revision numbers are actually encoded in |
|
39 | 39 | /// 4 bytes, and are liberally converted to ints, whence the i32 |
|
40 | 40 | pub type Revision = i32; |
|
41 | 41 | |
|
42 | 42 | /// Marker expressing the absence of a parent |
|
43 | 43 | /// |
|
44 | 44 | /// Independently of the actual representation, `NULL_REVISION` is guaranteed |
|
45 | 45 | /// to be smaller than all existing revisions. |
|
46 | 46 | pub const NULL_REVISION: Revision = -1; |
|
47 | 47 | |
|
48 | 48 | /// Same as `mercurial.node.wdirrev` |
|
49 | 49 | /// |
|
50 | 50 | /// This is also equal to `i32::max_value()`, but it's better to spell |
|
51 | 51 | /// it out explicitely, same as in `mercurial.node` |
|
52 | 52 | #[allow(clippy::unreadable_literal)] |
|
53 | 53 | pub const WORKING_DIRECTORY_REVISION: Revision = 0x7fffffff; |
|
54 | 54 | |
|
55 | 55 | pub const WORKING_DIRECTORY_HEX: &str = |
|
56 | 56 | "ffffffffffffffffffffffffffffffffffffffff"; |
|
57 | 57 | |
|
58 | 58 | /// The simplest expression of what we need of Mercurial DAGs. |
|
59 | 59 | pub trait Graph { |
|
60 | 60 | /// Return the two parents of the given `Revision`. |
|
61 | 61 | /// |
|
62 | 62 | /// Each of the parents can be independently `NULL_REVISION` |
|
63 | 63 | fn parents(&self, rev: Revision) -> Result<[Revision; 2], GraphError>; |
|
64 | 64 | } |
|
65 | 65 | |
|
66 | 66 | #[derive(Clone, Debug, PartialEq)] |
|
67 | 67 | pub enum GraphError { |
|
68 | 68 | ParentOutOfRange(Revision), |
|
69 | 69 | WorkingDirectoryUnsupported, |
|
70 | 70 | } |
|
71 | 71 | |
|
72 | 72 | /// The Mercurial Revlog Index |
|
73 | 73 | /// |
|
74 | 74 | /// This is currently limited to the minimal interface that is needed for |
|
75 | 75 | /// the [`nodemap`](nodemap/index.html) module |
|
76 | 76 | pub trait RevlogIndex { |
|
77 | 77 | /// Total number of Revisions referenced in this index |
|
78 | 78 | fn len(&self) -> usize; |
|
79 | 79 | |
|
80 | 80 | fn is_empty(&self) -> bool { |
|
81 | 81 | self.len() == 0 |
|
82 | 82 | } |
|
83 | 83 | |
|
84 | 84 | /// Return a reference to the Node or `None` if rev is out of bounds |
|
85 | 85 | /// |
|
86 | 86 | /// `NULL_REVISION` is not considered to be out of bounds. |
|
87 | 87 | fn node(&self, rev: Revision) -> Option<&Node>; |
|
88 | 88 | } |
|
89 | 89 | |
|
90 | 90 | const REVISION_FLAG_CENSORED: u16 = 1 << 15; |
|
91 | 91 | const REVISION_FLAG_ELLIPSIS: u16 = 1 << 14; |
|
92 | 92 | const REVISION_FLAG_EXTSTORED: u16 = 1 << 13; |
|
93 | 93 | const REVISION_FLAG_HASCOPIESINFO: u16 = 1 << 12; |
|
94 | 94 | |
|
95 | 95 | // Keep this in sync with REVIDX_KNOWN_FLAGS in |
|
96 | 96 | // mercurial/revlogutils/flagutil.py |
|
97 | 97 | const REVIDX_KNOWN_FLAGS: u16 = REVISION_FLAG_CENSORED |
|
98 | 98 | | REVISION_FLAG_ELLIPSIS |
|
99 | 99 | | REVISION_FLAG_EXTSTORED |
|
100 | 100 | | REVISION_FLAG_HASCOPIESINFO; |
|
101 | 101 | |
|
102 | 102 | const NULL_REVLOG_ENTRY_FLAGS: u16 = 0; |
|
103 | 103 | |
|
104 | 104 | #[derive(Debug, derive_more::From)] |
|
105 | 105 | pub enum RevlogError { |
|
106 | 106 | InvalidRevision, |
|
107 | 107 | /// Working directory is not supported |
|
108 | 108 | WDirUnsupported, |
|
109 | 109 | /// Found more than one entry whose ID match the requested prefix |
|
110 | 110 | AmbiguousPrefix, |
|
111 | 111 | #[from] |
|
112 | 112 | Other(HgError), |
|
113 | 113 | } |
|
114 | 114 | |
|
115 | 115 | impl From<NodeMapError> for RevlogError { |
|
116 | 116 | fn from(error: NodeMapError) -> Self { |
|
117 | 117 | match error { |
|
118 | 118 | NodeMapError::MultipleResults => RevlogError::AmbiguousPrefix, |
|
119 | 119 | NodeMapError::RevisionNotInIndex(rev) => RevlogError::corrupted( |
|
120 | 120 | format!("nodemap point to revision {} not in index", rev), |
|
121 | 121 | ), |
|
122 | 122 | } |
|
123 | 123 | } |
|
124 | 124 | } |
|
125 | 125 | |
|
126 | 126 | fn corrupted<S: AsRef<str>>(context: S) -> HgError { |
|
127 | 127 | HgError::corrupted(format!("corrupted revlog, {}", context.as_ref())) |
|
128 | 128 | } |
|
129 | 129 | |
|
130 | 130 | impl RevlogError { |
|
131 | 131 | fn corrupted<S: AsRef<str>>(context: S) -> Self { |
|
132 | 132 | RevlogError::Other(corrupted(context)) |
|
133 | 133 | } |
|
134 | 134 | } |
|
135 | 135 | |
|
136 | 136 | /// Read only implementation of revlog. |
|
137 | 137 | pub struct Revlog { |
|
138 | 138 | /// When index and data are not interleaved: bytes of the revlog index. |
|
139 | 139 | /// When index and data are interleaved: bytes of the revlog index and |
|
140 | 140 | /// data. |
|
141 | 141 | index: Index, |
|
142 | 142 | /// When index and data are not interleaved: bytes of the revlog data |
|
143 | 143 | data_bytes: Option<Box<dyn Deref<Target = [u8]> + Send>>, |
|
144 | 144 | /// When present on disk: the persistent nodemap for this revlog |
|
145 | 145 | nodemap: Option<nodemap::NodeTree>, |
|
146 | 146 | } |
|
147 | 147 | |
|
148 | 148 | impl Revlog { |
|
149 | 149 | /// Open a revlog index file. |
|
150 | 150 | /// |
|
151 | 151 | /// It will also open the associated data file if index and data are not |
|
152 | 152 | /// interleaved. |
|
153 | 153 | pub fn open( |
|
154 | 154 | store_vfs: &Vfs, |
|
155 | 155 | index_path: impl AsRef<Path>, |
|
156 | 156 | data_path: Option<&Path>, |
|
157 | 157 | use_nodemap: bool, |
|
158 | 158 | ) -> Result<Self, HgError> { |
|
159 | 159 | let index_path = index_path.as_ref(); |
|
160 | 160 | let index = { |
|
161 | 161 | match store_vfs.mmap_open_opt(&index_path)? { |
|
162 | 162 | None => Index::new(Box::new(vec![])), |
|
163 | 163 | Some(index_mmap) => { |
|
164 | 164 | let index = Index::new(Box::new(index_mmap))?; |
|
165 | 165 | Ok(index) |
|
166 | 166 | } |
|
167 | 167 | } |
|
168 | 168 | }?; |
|
169 | 169 | |
|
170 | 170 | let default_data_path = index_path.with_extension("d"); |
|
171 | 171 | |
|
172 | 172 | // type annotation required |
|
173 | 173 | // won't recognize Mmap as Deref<Target = [u8]> |
|
174 | 174 | let data_bytes: Option<Box<dyn Deref<Target = [u8]> + Send>> = |
|
175 | 175 | if index.is_inline() { |
|
176 | 176 | None |
|
177 | 177 | } else { |
|
178 | 178 | let data_path = data_path.unwrap_or(&default_data_path); |
|
179 | 179 | let data_mmap = store_vfs.mmap_open(data_path)?; |
|
180 | 180 | Some(Box::new(data_mmap)) |
|
181 | 181 | }; |
|
182 | 182 | |
|
183 | 183 | let nodemap = if index.is_inline() || !use_nodemap { |
|
184 | 184 | None |
|
185 | 185 | } else { |
|
186 | 186 | NodeMapDocket::read_from_file(store_vfs, index_path)?.map( |
|
187 | 187 | |(docket, data)| { |
|
188 | 188 | nodemap::NodeTree::load_bytes( |
|
189 | 189 | Box::new(data), |
|
190 | 190 | docket.data_length, |
|
191 | 191 | ) |
|
192 | 192 | }, |
|
193 | 193 | ) |
|
194 | 194 | }; |
|
195 | 195 | |
|
196 | 196 | Ok(Revlog { |
|
197 | 197 | index, |
|
198 | 198 | data_bytes, |
|
199 | 199 | nodemap, |
|
200 | 200 | }) |
|
201 | 201 | } |
|
202 | 202 | |
|
203 | 203 | /// Return number of entries of the `Revlog`. |
|
204 | 204 | pub fn len(&self) -> usize { |
|
205 | 205 | self.index.len() |
|
206 | 206 | } |
|
207 | 207 | |
|
208 | 208 | /// Returns `true` if the `Revlog` has zero `entries`. |
|
209 | 209 | pub fn is_empty(&self) -> bool { |
|
210 | 210 | self.index.is_empty() |
|
211 | 211 | } |
|
212 | 212 | |
|
213 | 213 | /// Returns the node ID for the given revision number, if it exists in this |
|
214 | 214 | /// revlog |
|
215 | 215 | pub fn node_from_rev(&self, rev: Revision) -> Option<&Node> { |
|
216 | 216 | if rev == NULL_REVISION { |
|
217 | 217 | return Some(&NULL_NODE); |
|
218 | 218 | } |
|
219 | 219 | Some(self.index.get_entry(rev)?.hash()) |
|
220 | 220 | } |
|
221 | 221 | |
|
222 | 222 | /// Return the revision number for the given node ID, if it exists in this |
|
223 | 223 | /// revlog |
|
224 | 224 | pub fn rev_from_node( |
|
225 | 225 | &self, |
|
226 | 226 | node: NodePrefix, |
|
227 | 227 | ) -> Result<Revision, RevlogError> { |
|
228 | 228 | let looked_up = if let Some(nodemap) = &self.nodemap { |
|
229 | 229 | nodemap |
|
230 | 230 | .find_bin(&self.index, node)? |
|
231 | 231 | .ok_or(RevlogError::InvalidRevision) |
|
232 | 232 | } else { |
|
233 | 233 | self.rev_from_node_no_persistent_nodemap(node) |
|
234 | 234 | }; |
|
235 | 235 | |
|
236 | 236 | if node.is_prefix_of(&NULL_NODE) { |
|
237 | 237 | return match looked_up { |
|
238 | 238 | Ok(_) => Err(RevlogError::AmbiguousPrefix), |
|
239 | 239 | Err(RevlogError::InvalidRevision) => Ok(NULL_REVISION), |
|
240 | 240 | res => res, |
|
241 | 241 | }; |
|
242 | 242 | }; |
|
243 | 243 | |
|
244 | 244 | looked_up |
|
245 | 245 | } |
|
246 | 246 | |
|
247 | 247 | /// Same as `rev_from_node`, without using a persistent nodemap |
|
248 | 248 | /// |
|
249 | 249 | /// This is used as fallback when a persistent nodemap is not present. |
|
250 | 250 | /// This happens when the persistent-nodemap experimental feature is not |
|
251 | 251 | /// enabled, or for small revlogs. |
|
252 | 252 | fn rev_from_node_no_persistent_nodemap( |
|
253 | 253 | &self, |
|
254 | 254 | node: NodePrefix, |
|
255 | 255 | ) -> Result<Revision, RevlogError> { |
|
256 | 256 | // Linear scan of the revlog |
|
257 | 257 | // TODO: consider building a non-persistent nodemap in memory to |
|
258 | 258 | // optimize these cases. |
|
259 | 259 | let mut found_by_prefix = None; |
|
260 | 260 | for rev in (0..self.len() as Revision).rev() { |
|
261 | 261 | let index_entry = self.index.get_entry(rev).ok_or_else(|| { |
|
262 | 262 | HgError::corrupted( |
|
263 | 263 | "revlog references a revision not in the index", |
|
264 | 264 | ) |
|
265 | 265 | })?; |
|
266 | 266 | if node == *index_entry.hash() { |
|
267 | 267 | return Ok(rev); |
|
268 | 268 | } |
|
269 | 269 | if node.is_prefix_of(index_entry.hash()) { |
|
270 | 270 | if found_by_prefix.is_some() { |
|
271 | 271 | return Err(RevlogError::AmbiguousPrefix); |
|
272 | 272 | } |
|
273 | 273 | found_by_prefix = Some(rev) |
|
274 | 274 | } |
|
275 | 275 | } |
|
276 | 276 | found_by_prefix.ok_or(RevlogError::InvalidRevision) |
|
277 | 277 | } |
|
278 | 278 | |
|
279 | 279 | /// Returns whether the given revision exists in this revlog. |
|
280 | 280 | pub fn has_rev(&self, rev: Revision) -> bool { |
|
281 | 281 | self.index.get_entry(rev).is_some() |
|
282 | 282 | } |
|
283 | 283 | |
|
284 | 284 | /// Return the full data associated to a revision. |
|
285 | 285 | /// |
|
286 | 286 | /// All entries required to build the final data out of deltas will be |
|
287 | 287 | /// retrieved as needed, and the deltas will be applied to the inital |
|
288 | 288 | /// snapshot to rebuild the final data. |
|
289 | 289 | pub fn get_rev_data( |
|
290 | 290 | &self, |
|
291 | 291 | rev: Revision, |
|
292 | 292 | ) -> Result<Cow<[u8]>, RevlogError> { |
|
293 | 293 | if rev == NULL_REVISION { |
|
294 | 294 | return Ok(Cow::Borrowed(&[])); |
|
295 | 295 | }; |
|
296 | 296 | Ok(self.get_entry(rev)?.data()?) |
|
297 | 297 | } |
|
298 | 298 | |
|
299 | 299 | /// Check the hash of some given data against the recorded hash. |
|
300 | 300 | pub fn check_hash( |
|
301 | 301 | &self, |
|
302 | 302 | p1: Revision, |
|
303 | 303 | p2: Revision, |
|
304 | 304 | expected: &[u8], |
|
305 | 305 | data: &[u8], |
|
306 | 306 | ) -> bool { |
|
307 | 307 | let e1 = self.index.get_entry(p1); |
|
308 | 308 | let h1 = match e1 { |
|
309 | 309 | Some(ref entry) => entry.hash(), |
|
310 | 310 | None => &NULL_NODE, |
|
311 | 311 | }; |
|
312 | 312 | let e2 = self.index.get_entry(p2); |
|
313 | 313 | let h2 = match e2 { |
|
314 | 314 | Some(ref entry) => entry.hash(), |
|
315 | 315 | None => &NULL_NODE, |
|
316 | 316 | }; |
|
317 | 317 | |
|
318 | 318 | hash(data, h1.as_bytes(), h2.as_bytes()) == expected |
|
319 | 319 | } |
|
320 | 320 | |
|
321 | 321 | /// Build the full data of a revision out its snapshot |
|
322 | 322 | /// and its deltas. |
|
323 | 323 | fn build_data_from_deltas( |
|
324 | 324 | snapshot: RevlogEntry, |
|
325 | 325 | deltas: &[RevlogEntry], |
|
326 | 326 | ) -> Result<Vec<u8>, HgError> { |
|
327 | 327 | let snapshot = snapshot.data_chunk()?; |
|
328 | 328 | let deltas = deltas |
|
329 | 329 | .iter() |
|
330 | 330 | .rev() |
|
331 | 331 | .map(RevlogEntry::data_chunk) |
|
332 | 332 | .collect::<Result<Vec<_>, _>>()?; |
|
333 | 333 | let patches: Vec<_> = |
|
334 | 334 | deltas.iter().map(|d| patch::PatchList::new(d)).collect(); |
|
335 | 335 | let patch = patch::fold_patch_lists(&patches); |
|
336 | 336 | Ok(patch.apply(&snapshot)) |
|
337 | 337 | } |
|
338 | 338 | |
|
339 | 339 | /// Return the revlog data. |
|
340 | 340 | fn data(&self) -> &[u8] { |
|
341 | 341 | match &self.data_bytes { |
|
342 | 342 | Some(data_bytes) => data_bytes, |
|
343 | 343 | None => panic!( |
|
344 | 344 | "forgot to load the data or trying to access inline data" |
|
345 | 345 | ), |
|
346 | 346 | } |
|
347 | 347 | } |
|
348 | 348 | |
|
349 | 349 | pub fn make_null_entry(&self) -> RevlogEntry { |
|
350 | 350 | RevlogEntry { |
|
351 | 351 | revlog: self, |
|
352 | 352 | rev: NULL_REVISION, |
|
353 | 353 | bytes: b"", |
|
354 | 354 | compressed_len: 0, |
|
355 | 355 | uncompressed_len: 0, |
|
356 | 356 | base_rev_or_base_of_delta_chain: None, |
|
357 | 357 | p1: NULL_REVISION, |
|
358 | 358 | p2: NULL_REVISION, |
|
359 | 359 | flags: NULL_REVLOG_ENTRY_FLAGS, |
|
360 | 360 | hash: NULL_NODE, |
|
361 | 361 | } |
|
362 | 362 | } |
|
363 | 363 | |
|
364 | 364 | /// Get an entry of the revlog. |
|
365 | 365 | pub fn get_entry( |
|
366 | 366 | &self, |
|
367 | 367 | rev: Revision, |
|
368 | 368 | ) -> Result<RevlogEntry, RevlogError> { |
|
369 | 369 | if rev == NULL_REVISION { |
|
370 | 370 | return Ok(self.make_null_entry()); |
|
371 | 371 | } |
|
372 | 372 | let index_entry = self |
|
373 | 373 | .index |
|
374 | 374 | .get_entry(rev) |
|
375 | 375 | .ok_or(RevlogError::InvalidRevision)?; |
|
376 | 376 | let start = index_entry.offset(); |
|
377 | 377 | let end = start + index_entry.compressed_len() as usize; |
|
378 | 378 | let data = if self.index.is_inline() { |
|
379 | 379 | self.index.data(start, end) |
|
380 | 380 | } else { |
|
381 | 381 | &self.data()[start..end] |
|
382 | 382 | }; |
|
383 | 383 | let entry = RevlogEntry { |
|
384 | 384 | revlog: self, |
|
385 | 385 | rev, |
|
386 | 386 | bytes: data, |
|
387 | 387 | compressed_len: index_entry.compressed_len(), |
|
388 | 388 | uncompressed_len: index_entry.uncompressed_len(), |
|
389 | 389 | base_rev_or_base_of_delta_chain: if index_entry |
|
390 | 390 | .base_revision_or_base_of_delta_chain() |
|
391 | 391 | == rev |
|
392 | 392 | { |
|
393 | 393 | None |
|
394 | 394 | } else { |
|
395 | 395 | Some(index_entry.base_revision_or_base_of_delta_chain()) |
|
396 | 396 | }, |
|
397 | 397 | p1: index_entry.p1(), |
|
398 | 398 | p2: index_entry.p2(), |
|
399 | 399 | flags: index_entry.flags(), |
|
400 | 400 | hash: *index_entry.hash(), |
|
401 | 401 | }; |
|
402 | 402 | Ok(entry) |
|
403 | 403 | } |
|
404 | 404 | |
|
405 | 405 | /// when resolving internal references within revlog, any errors |
|
406 | 406 | /// should be reported as corruption, instead of e.g. "invalid revision" |
|
407 | 407 | fn get_entry_internal( |
|
408 | 408 | &self, |
|
409 | 409 | rev: Revision, |
|
410 | 410 | ) -> Result<RevlogEntry, HgError> { |
|
411 | 411 | self.get_entry(rev) |
|
412 | 412 | .map_err(|_| corrupted(format!("revision {} out of range", rev))) |
|
413 | 413 | } |
|
414 | 414 | } |
|
415 | 415 | |
|
416 | 416 | /// The revlog entry's bytes and the necessary informations to extract |
|
417 | 417 | /// the entry's data. |
|
418 | 418 | #[derive(Clone)] |
|
419 | 419 | pub struct RevlogEntry<'revlog> { |
|
420 | 420 | revlog: &'revlog Revlog, |
|
421 | 421 | rev: Revision, |
|
422 | 422 | bytes: &'revlog [u8], |
|
423 | 423 | compressed_len: u32, |
|
424 | 424 | uncompressed_len: i32, |
|
425 | 425 | base_rev_or_base_of_delta_chain: Option<Revision>, |
|
426 | 426 | p1: Revision, |
|
427 | 427 | p2: Revision, |
|
428 | 428 | flags: u16, |
|
429 | 429 | hash: Node, |
|
430 | 430 | } |
|
431 | 431 | |
|
432 | 432 | thread_local! { |
|
433 | 433 | // seems fine to [unwrap] here: this can only fail due to memory allocation |
|
434 | 434 | // failing, and it's normal for that to cause panic. |
|
435 | 435 | static ZSTD_DECODER : RefCell<zstd::bulk::Decompressor<'static>> = |
|
436 | 436 | RefCell::new(zstd::bulk::Decompressor::new().ok().unwrap()); |
|
437 | 437 | } |
|
438 | 438 | |
|
439 | 439 | fn zstd_decompress_to_buffer( |
|
440 | 440 | bytes: &[u8], |
|
441 | 441 | buf: &mut Vec<u8>, |
|
442 | 442 | ) -> Result<usize, std::io::Error> { |
|
443 | 443 | ZSTD_DECODER |
|
444 | 444 | .with(|decoder| decoder.borrow_mut().decompress_to_buffer(bytes, buf)) |
|
445 | 445 | } |
|
446 | 446 | |
|
447 | 447 | impl<'revlog> RevlogEntry<'revlog> { |
|
448 | 448 | pub fn revision(&self) -> Revision { |
|
449 | 449 | self.rev |
|
450 | 450 | } |
|
451 | 451 | |
|
452 | 452 | pub fn node(&self) -> &Node { |
|
453 | 453 | &self.hash |
|
454 | 454 | } |
|
455 | 455 | |
|
456 | 456 | pub fn uncompressed_len(&self) -> Option<u32> { |
|
457 | 457 | u32::try_from(self.uncompressed_len).ok() |
|
458 | 458 | } |
|
459 | 459 | |
|
460 | 460 | pub fn has_p1(&self) -> bool { |
|
461 | 461 | self.p1 != NULL_REVISION |
|
462 | 462 | } |
|
463 | 463 | |
|
464 | 464 | pub fn p1_entry( |
|
465 | 465 | &self, |
|
466 | 466 | ) -> Result<Option<RevlogEntry<'revlog>>, RevlogError> { |
|
467 | 467 | if self.p1 == NULL_REVISION { |
|
468 | 468 | Ok(None) |
|
469 | 469 | } else { |
|
470 | 470 | Ok(Some(self.revlog.get_entry(self.p1)?)) |
|
471 | 471 | } |
|
472 | 472 | } |
|
473 | 473 | |
|
474 | 474 | pub fn p2_entry( |
|
475 | 475 | &self, |
|
476 | 476 | ) -> Result<Option<RevlogEntry<'revlog>>, RevlogError> { |
|
477 | 477 | if self.p2 == NULL_REVISION { |
|
478 | 478 | Ok(None) |
|
479 | 479 | } else { |
|
480 | 480 | Ok(Some(self.revlog.get_entry(self.p2)?)) |
|
481 | 481 | } |
|
482 | 482 | } |
|
483 | 483 | |
|
484 | 484 | pub fn p1(&self) -> Option<Revision> { |
|
485 | 485 | if self.p1 == NULL_REVISION { |
|
486 | 486 | None |
|
487 | 487 | } else { |
|
488 | 488 | Some(self.p1) |
|
489 | 489 | } |
|
490 | 490 | } |
|
491 | 491 | |
|
492 | 492 | pub fn p2(&self) -> Option<Revision> { |
|
493 | 493 | if self.p2 == NULL_REVISION { |
|
494 | 494 | None |
|
495 | 495 | } else { |
|
496 | 496 | Some(self.p2) |
|
497 | 497 | } |
|
498 | 498 | } |
|
499 | 499 | |
|
500 | 500 | pub fn is_censored(&self) -> bool { |
|
501 | 501 | (self.flags & REVISION_FLAG_CENSORED) != 0 |
|
502 | 502 | } |
|
503 | 503 | |
|
504 | 504 | pub fn has_length_affecting_flag_processor(&self) -> bool { |
|
505 | 505 | // Relevant Python code: revlog.size() |
|
506 | 506 | // note: ELLIPSIS is known to not change the content |
|
507 | 507 | (self.flags & (REVIDX_KNOWN_FLAGS ^ REVISION_FLAG_ELLIPSIS)) != 0 |
|
508 | 508 | } |
|
509 | 509 | |
|
510 | 510 | /// The data for this entry, after resolving deltas if any. |
|
511 | 511 | pub fn rawdata(&self) -> Result<Cow<'revlog, [u8]>, HgError> { |
|
512 | 512 | let mut entry = self.clone(); |
|
513 | 513 | let mut delta_chain = vec![]; |
|
514 | 514 | |
|
515 | 515 | // The meaning of `base_rev_or_base_of_delta_chain` depends on |
|
516 | 516 | // generaldelta. See the doc on `ENTRY_DELTA_BASE` in |
|
517 | 517 | // `mercurial/revlogutils/constants.py` and the code in |
|
518 | 518 | // [_chaininfo] and in [index_deltachain]. |
|
519 | 519 | let uses_generaldelta = self.revlog.index.uses_generaldelta(); |
|
520 | 520 | while let Some(base_rev) = entry.base_rev_or_base_of_delta_chain { |
|
521 | 521 | let base_rev = if uses_generaldelta { |
|
522 | 522 | base_rev |
|
523 | 523 | } else { |
|
524 | 524 | entry.rev - 1 |
|
525 | 525 | }; |
|
526 | 526 | delta_chain.push(entry); |
|
527 | 527 | entry = self.revlog.get_entry_internal(base_rev)?; |
|
528 | 528 | } |
|
529 | 529 | |
|
530 | 530 | let data = if delta_chain.is_empty() { |
|
531 | 531 | entry.data_chunk()? |
|
532 | 532 | } else { |
|
533 | 533 | Revlog::build_data_from_deltas(entry, &delta_chain)?.into() |
|
534 | 534 | }; |
|
535 | 535 | |
|
536 | 536 | Ok(data) |
|
537 | 537 | } |
|
538 | 538 | |
|
539 | 539 | fn check_data( |
|
540 | 540 | &self, |
|
541 | 541 | data: Cow<'revlog, [u8]>, |
|
542 | 542 | ) -> Result<Cow<'revlog, [u8]>, HgError> { |
|
543 | 543 | if self.revlog.check_hash( |
|
544 | 544 | self.p1, |
|
545 | 545 | self.p2, |
|
546 | 546 | self.hash.as_bytes(), |
|
547 | 547 | &data, |
|
548 | 548 | ) { |
|
549 | 549 | Ok(data) |
|
550 | 550 | } else { |
|
551 | 551 | if (self.flags & REVISION_FLAG_ELLIPSIS) != 0 { |
|
552 | 552 | return Err(HgError::unsupported( |
|
553 | 553 | "ellipsis revisions are not supported by rhg", |
|
554 | 554 | )); |
|
555 | 555 | } |
|
556 | 556 | Err(corrupted(format!( |
|
557 | 557 | "hash check failed for revision {}", |
|
558 | 558 | self.rev |
|
559 | 559 | ))) |
|
560 | 560 | } |
|
561 | 561 | } |
|
562 | 562 | |
|
563 | 563 | pub fn data(&self) -> Result<Cow<'revlog, [u8]>, HgError> { |
|
564 | 564 | let data = self.rawdata()?; |
|
565 | if self.rev == NULL_REVISION { | |
|
566 | return Ok(data); | |
|
567 | } | |
|
565 | 568 | if self.is_censored() { |
|
566 | 569 | return Err(HgError::CensoredNodeError); |
|
567 | 570 | } |
|
568 | 571 | self.check_data(data) |
|
569 | 572 | } |
|
570 | 573 | |
|
571 | 574 | /// Extract the data contained in the entry. |
|
572 | 575 | /// This may be a delta. (See `is_delta`.) |
|
573 | 576 | fn data_chunk(&self) -> Result<Cow<'revlog, [u8]>, HgError> { |
|
574 | 577 | if self.bytes.is_empty() { |
|
575 | 578 | return Ok(Cow::Borrowed(&[])); |
|
576 | 579 | } |
|
577 | 580 | match self.bytes[0] { |
|
578 | 581 | // Revision data is the entirety of the entry, including this |
|
579 | 582 | // header. |
|
580 | 583 | b'\0' => Ok(Cow::Borrowed(self.bytes)), |
|
581 | 584 | // Raw revision data follows. |
|
582 | 585 | b'u' => Ok(Cow::Borrowed(&self.bytes[1..])), |
|
583 | 586 | // zlib (RFC 1950) data. |
|
584 | 587 | b'x' => Ok(Cow::Owned(self.uncompressed_zlib_data()?)), |
|
585 | 588 | // zstd data. |
|
586 | 589 | b'\x28' => Ok(Cow::Owned(self.uncompressed_zstd_data()?)), |
|
587 | 590 | // A proper new format should have had a repo/store requirement. |
|
588 | 591 | format_type => Err(corrupted(format!( |
|
589 | 592 | "unknown compression header '{}'", |
|
590 | 593 | format_type |
|
591 | 594 | ))), |
|
592 | 595 | } |
|
593 | 596 | } |
|
594 | 597 | |
|
595 | 598 | fn uncompressed_zlib_data(&self) -> Result<Vec<u8>, HgError> { |
|
596 | 599 | let mut decoder = ZlibDecoder::new(self.bytes); |
|
597 | 600 | if self.is_delta() { |
|
598 | 601 | let mut buf = Vec::with_capacity(self.compressed_len as usize); |
|
599 | 602 | decoder |
|
600 | 603 | .read_to_end(&mut buf) |
|
601 | 604 | .map_err(|e| corrupted(e.to_string()))?; |
|
602 | 605 | Ok(buf) |
|
603 | 606 | } else { |
|
604 | 607 | let cap = self.uncompressed_len.max(0) as usize; |
|
605 | 608 | let mut buf = vec![0; cap]; |
|
606 | 609 | decoder |
|
607 | 610 | .read_exact(&mut buf) |
|
608 | 611 | .map_err(|e| corrupted(e.to_string()))?; |
|
609 | 612 | Ok(buf) |
|
610 | 613 | } |
|
611 | 614 | } |
|
612 | 615 | |
|
613 | 616 | fn uncompressed_zstd_data(&self) -> Result<Vec<u8>, HgError> { |
|
614 | 617 | let cap = self.uncompressed_len.max(0) as usize; |
|
615 | 618 | if self.is_delta() { |
|
616 | 619 | // [cap] is usually an over-estimate of the space needed because |
|
617 | 620 | // it's the length of delta-decoded data, but we're interested |
|
618 | 621 | // in the size of the delta. |
|
619 | 622 | // This means we have to [shrink_to_fit] to avoid holding on |
|
620 | 623 | // to a large chunk of memory, but it also means we must have a |
|
621 | 624 | // fallback branch, for the case when the delta is longer than |
|
622 | 625 | // the original data (surprisingly, this does happen in practice) |
|
623 | 626 | let mut buf = Vec::with_capacity(cap); |
|
624 | 627 | match zstd_decompress_to_buffer(self.bytes, &mut buf) { |
|
625 | 628 | Ok(_) => buf.shrink_to_fit(), |
|
626 | 629 | Err(_) => { |
|
627 | 630 | buf.clear(); |
|
628 | 631 | zstd::stream::copy_decode(self.bytes, &mut buf) |
|
629 | 632 | .map_err(|e| corrupted(e.to_string()))?; |
|
630 | 633 | } |
|
631 | 634 | }; |
|
632 | 635 | Ok(buf) |
|
633 | 636 | } else { |
|
634 | 637 | let mut buf = Vec::with_capacity(cap); |
|
635 | 638 | let len = zstd_decompress_to_buffer(self.bytes, &mut buf) |
|
636 | 639 | .map_err(|e| corrupted(e.to_string()))?; |
|
637 | 640 | if len != self.uncompressed_len as usize { |
|
638 | 641 | Err(corrupted("uncompressed length does not match")) |
|
639 | 642 | } else { |
|
640 | 643 | Ok(buf) |
|
641 | 644 | } |
|
642 | 645 | } |
|
643 | 646 | } |
|
644 | 647 | |
|
645 | 648 | /// Tell if the entry is a snapshot or a delta |
|
646 | 649 | /// (influences on decompression). |
|
647 | 650 | fn is_delta(&self) -> bool { |
|
648 | 651 | self.base_rev_or_base_of_delta_chain.is_some() |
|
649 | 652 | } |
|
650 | 653 | } |
|
651 | 654 | |
|
652 | 655 | /// Calculate the hash of a revision given its data and its parents. |
|
653 | 656 | fn hash( |
|
654 | 657 | data: &[u8], |
|
655 | 658 | p1_hash: &[u8], |
|
656 | 659 | p2_hash: &[u8], |
|
657 | 660 | ) -> [u8; NODE_BYTES_LENGTH] { |
|
658 | 661 | let mut hasher = Sha1::new(); |
|
659 | 662 | let (a, b) = (p1_hash, p2_hash); |
|
660 | 663 | if a > b { |
|
661 | 664 | hasher.update(b); |
|
662 | 665 | hasher.update(a); |
|
663 | 666 | } else { |
|
664 | 667 | hasher.update(a); |
|
665 | 668 | hasher.update(b); |
|
666 | 669 | } |
|
667 | 670 | hasher.update(data); |
|
668 | 671 | *hasher.finalize().as_ref() |
|
669 | 672 | } |
|
670 | 673 | |
|
671 | 674 | #[cfg(test)] |
|
672 | 675 | mod tests { |
|
673 | 676 | use super::*; |
|
674 | 677 | use crate::index::{IndexEntryBuilder, INDEX_ENTRY_SIZE}; |
|
675 | 678 | use itertools::Itertools; |
|
676 | 679 | |
|
677 | 680 | #[test] |
|
678 | 681 | fn test_empty() { |
|
679 | 682 | let temp = tempfile::tempdir().unwrap(); |
|
680 | 683 | let vfs = Vfs { base: temp.path() }; |
|
681 | 684 | std::fs::write(temp.path().join("foo.i"), b"").unwrap(); |
|
682 | 685 | let revlog = Revlog::open(&vfs, "foo.i", None, false).unwrap(); |
|
683 | 686 | assert!(revlog.is_empty()); |
|
684 | 687 | assert_eq!(revlog.len(), 0); |
|
685 | 688 | assert!(revlog.get_entry(0).is_err()); |
|
686 | 689 | assert!(!revlog.has_rev(0)); |
|
687 | 690 | assert_eq!( |
|
688 | 691 | revlog.rev_from_node(NULL_NODE.into()).unwrap(), |
|
689 | 692 | NULL_REVISION |
|
690 | 693 | ); |
|
694 | let null_entry = revlog.get_entry(NULL_REVISION).ok().unwrap(); | |
|
695 | assert_eq!(null_entry.revision(), NULL_REVISION); | |
|
696 | assert!(null_entry.data().unwrap().is_empty()); | |
|
691 | 697 | } |
|
692 | 698 | |
|
693 | 699 | #[test] |
|
694 | 700 | fn test_inline() { |
|
695 | 701 | let temp = tempfile::tempdir().unwrap(); |
|
696 | 702 | let vfs = Vfs { base: temp.path() }; |
|
697 | 703 | let node0 = Node::from_hex("2ed2a3912a0b24502043eae84ee4b279c18b90dd") |
|
698 | 704 | .unwrap(); |
|
699 | 705 | let node1 = Node::from_hex("b004912a8510032a0350a74daa2803dadfb00e12") |
|
700 | 706 | .unwrap(); |
|
701 | 707 | let node2 = Node::from_hex("dd6ad206e907be60927b5a3117b97dffb2590582") |
|
702 | 708 | .unwrap(); |
|
703 | 709 | let entry0_bytes = IndexEntryBuilder::new() |
|
704 | 710 | .is_first(true) |
|
705 | 711 | .with_version(1) |
|
706 | 712 | .with_inline(true) |
|
707 | 713 | .with_offset(INDEX_ENTRY_SIZE) |
|
708 | 714 | .with_node(node0) |
|
709 | 715 | .build(); |
|
710 | 716 | let entry1_bytes = IndexEntryBuilder::new() |
|
711 | 717 | .with_offset(INDEX_ENTRY_SIZE) |
|
712 | 718 | .with_node(node1) |
|
713 | 719 | .build(); |
|
714 | 720 | let entry2_bytes = IndexEntryBuilder::new() |
|
715 | 721 | .with_offset(INDEX_ENTRY_SIZE) |
|
716 | 722 | .with_p1(0) |
|
717 | 723 | .with_p2(1) |
|
718 | 724 | .with_node(node2) |
|
719 | 725 | .build(); |
|
720 | 726 | let contents = vec![entry0_bytes, entry1_bytes, entry2_bytes] |
|
721 | 727 | .into_iter() |
|
722 | 728 | .flatten() |
|
723 | 729 | .collect_vec(); |
|
724 | 730 | std::fs::write(temp.path().join("foo.i"), contents).unwrap(); |
|
725 | 731 | let revlog = Revlog::open(&vfs, "foo.i", None, false).unwrap(); |
|
726 | 732 | |
|
727 | 733 | let entry0 = revlog.get_entry(0).ok().unwrap(); |
|
728 | 734 | assert_eq!(entry0.revision(), 0); |
|
729 | 735 | assert_eq!(*entry0.node(), node0); |
|
730 | 736 | assert!(!entry0.has_p1()); |
|
731 | 737 | assert_eq!(entry0.p1(), None); |
|
732 | 738 | assert_eq!(entry0.p2(), None); |
|
733 | 739 | let p1_entry = entry0.p1_entry().unwrap(); |
|
734 | 740 | assert!(p1_entry.is_none()); |
|
735 | 741 | let p2_entry = entry0.p2_entry().unwrap(); |
|
736 | 742 | assert!(p2_entry.is_none()); |
|
737 | 743 | |
|
738 | 744 | let entry1 = revlog.get_entry(1).ok().unwrap(); |
|
739 | 745 | assert_eq!(entry1.revision(), 1); |
|
740 | 746 | assert_eq!(*entry1.node(), node1); |
|
741 | 747 | assert!(!entry1.has_p1()); |
|
742 | 748 | assert_eq!(entry1.p1(), None); |
|
743 | 749 | assert_eq!(entry1.p2(), None); |
|
744 | 750 | let p1_entry = entry1.p1_entry().unwrap(); |
|
745 | 751 | assert!(p1_entry.is_none()); |
|
746 | 752 | let p2_entry = entry1.p2_entry().unwrap(); |
|
747 | 753 | assert!(p2_entry.is_none()); |
|
748 | 754 | |
|
749 | 755 | let entry2 = revlog.get_entry(2).ok().unwrap(); |
|
750 | 756 | assert_eq!(entry2.revision(), 2); |
|
751 | 757 | assert_eq!(*entry2.node(), node2); |
|
752 | 758 | assert!(entry2.has_p1()); |
|
753 | 759 | assert_eq!(entry2.p1(), Some(0)); |
|
754 | 760 | assert_eq!(entry2.p2(), Some(1)); |
|
755 | 761 | let p1_entry = entry2.p1_entry().unwrap(); |
|
756 | 762 | assert!(p1_entry.is_some()); |
|
757 | 763 | assert_eq!(p1_entry.unwrap().revision(), 0); |
|
758 | 764 | let p2_entry = entry2.p2_entry().unwrap(); |
|
759 | 765 | assert!(p2_entry.is_some()); |
|
760 | 766 | assert_eq!(p2_entry.unwrap().revision(), 1); |
|
761 | 767 | } |
|
762 | 768 | |
|
763 | 769 | #[test] |
|
764 | 770 | fn test_nodemap() { |
|
765 | 771 | let temp = tempfile::tempdir().unwrap(); |
|
766 | 772 | let vfs = Vfs { base: temp.path() }; |
|
767 | 773 | |
|
768 | 774 | // building a revlog with a forced Node starting with zeros |
|
769 | 775 | // This is a corruption, but it does not preclude using the nodemap |
|
770 | 776 | // if we don't try and access the data |
|
771 | 777 | let node0 = Node::from_hex("00d2a3912a0b24502043eae84ee4b279c18b90dd") |
|
772 | 778 | .unwrap(); |
|
773 | 779 | let node1 = Node::from_hex("b004912a8510032a0350a74daa2803dadfb00e12") |
|
774 | 780 | .unwrap(); |
|
775 | 781 | let entry0_bytes = IndexEntryBuilder::new() |
|
776 | 782 | .is_first(true) |
|
777 | 783 | .with_version(1) |
|
778 | 784 | .with_inline(true) |
|
779 | 785 | .with_offset(INDEX_ENTRY_SIZE) |
|
780 | 786 | .with_node(node0) |
|
781 | 787 | .build(); |
|
782 | 788 | let entry1_bytes = IndexEntryBuilder::new() |
|
783 | 789 | .with_offset(INDEX_ENTRY_SIZE) |
|
784 | 790 | .with_node(node1) |
|
785 | 791 | .build(); |
|
786 | 792 | let contents = vec![entry0_bytes, entry1_bytes] |
|
787 | 793 | .into_iter() |
|
788 | 794 | .flatten() |
|
789 | 795 | .collect_vec(); |
|
790 | 796 | std::fs::write(temp.path().join("foo.i"), contents).unwrap(); |
|
791 | 797 | let revlog = Revlog::open(&vfs, "foo.i", None, false).unwrap(); |
|
792 | 798 | |
|
793 | 799 | // accessing the data shows the corruption |
|
794 | 800 | revlog.get_entry(0).unwrap().data().unwrap_err(); |
|
795 | 801 | |
|
796 | 802 | assert_eq!(revlog.rev_from_node(NULL_NODE.into()).unwrap(), -1); |
|
797 | 803 | assert_eq!(revlog.rev_from_node(node0.into()).unwrap(), 0); |
|
798 | 804 | assert_eq!(revlog.rev_from_node(node1.into()).unwrap(), 1); |
|
799 | 805 | assert_eq!( |
|
800 | 806 | revlog |
|
801 | 807 | .rev_from_node(NodePrefix::from_hex("000").unwrap()) |
|
802 | 808 | .unwrap(), |
|
803 | 809 | -1 |
|
804 | 810 | ); |
|
805 | 811 | assert_eq!( |
|
806 | 812 | revlog |
|
807 | 813 | .rev_from_node(NodePrefix::from_hex("b00").unwrap()) |
|
808 | 814 | .unwrap(), |
|
809 | 815 | 1 |
|
810 | 816 | ); |
|
811 | 817 | // RevlogError does not implement PartialEq |
|
812 | 818 | // (ultimately because io::Error does not) |
|
813 | 819 | match revlog |
|
814 | 820 | .rev_from_node(NodePrefix::from_hex("00").unwrap()) |
|
815 | 821 | .expect_err("Expected to give AmbiguousPrefix error") |
|
816 | 822 | { |
|
817 | 823 | RevlogError::AmbiguousPrefix => (), |
|
818 | 824 | e => { |
|
819 | 825 | panic!("Got another error than AmbiguousPrefix: {:?}", e); |
|
820 | 826 | } |
|
821 | 827 | }; |
|
822 | 828 | } |
|
823 | 829 | } |
General Comments 0
You need to be logged in to leave comments.
Login now