Show More
@@ -1,1302 +1,1303 b'' | |||
|
1 | 1 | # __init__.py - remotefilelog extension |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2013 Facebook, Inc. |
|
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 | """remotefilelog causes Mercurial to lazilly fetch file contents (EXPERIMENTAL) |
|
8 | 8 | |
|
9 | 9 | This extension is HIGHLY EXPERIMENTAL. There are NO BACKWARDS COMPATIBILITY |
|
10 | 10 | GUARANTEES. This means that repositories created with this extension may |
|
11 | 11 | only be usable with the exact version of this extension/Mercurial that was |
|
12 | 12 | used. The extension attempts to enforce this in order to prevent repository |
|
13 | 13 | corruption. |
|
14 | 14 | |
|
15 | 15 | remotefilelog works by fetching file contents lazily and storing them |
|
16 | 16 | in a cache on the client rather than in revlogs. This allows enormous |
|
17 | 17 | histories to be transferred only partially, making them easier to |
|
18 | 18 | operate on. |
|
19 | 19 | |
|
20 | 20 | Configs: |
|
21 | 21 | |
|
22 | 22 | ``packs.maxchainlen`` specifies the maximum delta chain length in pack files |
|
23 | 23 | |
|
24 | 24 | ``packs.maxpacksize`` specifies the maximum pack file size |
|
25 | 25 | |
|
26 | 26 | ``packs.maxpackfilecount`` specifies the maximum number of packs in the |
|
27 | 27 | shared cache (trees only for now) |
|
28 | 28 | |
|
29 | 29 | ``remotefilelog.backgroundprefetch`` runs prefetch in background when True |
|
30 | 30 | |
|
31 | 31 | ``remotefilelog.bgprefetchrevs`` specifies revisions to fetch on commit and |
|
32 | 32 | update, and on other commands that use them. Different from pullprefetch. |
|
33 | 33 | |
|
34 | 34 | ``remotefilelog.gcrepack`` does garbage collection during repack when True |
|
35 | 35 | |
|
36 | 36 | ``remotefilelog.nodettl`` specifies maximum TTL of a node in seconds before |
|
37 | 37 | it is garbage collected |
|
38 | 38 | |
|
39 | 39 | ``remotefilelog.repackonhggc`` runs repack on hg gc when True |
|
40 | 40 | |
|
41 | 41 | ``remotefilelog.prefetchdays`` specifies the maximum age of a commit in |
|
42 | 42 | days after which it is no longer prefetched. |
|
43 | 43 | |
|
44 | 44 | ``remotefilelog.prefetchdelay`` specifies delay between background |
|
45 | 45 | prefetches in seconds after operations that change the working copy parent |
|
46 | 46 | |
|
47 | 47 | ``remotefilelog.data.gencountlimit`` constraints the minimum number of data |
|
48 | 48 | pack files required to be considered part of a generation. In particular, |
|
49 | 49 | minimum number of packs files > gencountlimit. |
|
50 | 50 | |
|
51 | 51 | ``remotefilelog.data.generations`` list for specifying the lower bound of |
|
52 | 52 | each generation of the data pack files. For example, list ['100MB','1MB'] |
|
53 | 53 | or ['1MB', '100MB'] will lead to three generations: [0, 1MB), [ |
|
54 | 54 | 1MB, 100MB) and [100MB, infinity). |
|
55 | 55 | |
|
56 | 56 | ``remotefilelog.data.maxrepackpacks`` the maximum number of pack files to |
|
57 | 57 | include in an incremental data repack. |
|
58 | 58 | |
|
59 | 59 | ``remotefilelog.data.repackmaxpacksize`` the maximum size of a pack file for |
|
60 | 60 | it to be considered for an incremental data repack. |
|
61 | 61 | |
|
62 | 62 | ``remotefilelog.data.repacksizelimit`` the maximum total size of pack files |
|
63 | 63 | to include in an incremental data repack. |
|
64 | 64 | |
|
65 | 65 | ``remotefilelog.history.gencountlimit`` constraints the minimum number of |
|
66 | 66 | history pack files required to be considered part of a generation. In |
|
67 | 67 | particular, minimum number of packs files > gencountlimit. |
|
68 | 68 | |
|
69 | 69 | ``remotefilelog.history.generations`` list for specifying the lower bound of |
|
70 | 70 | each generation of the history pack files. For example, list [ |
|
71 | 71 | '100MB', '1MB'] or ['1MB', '100MB'] will lead to three generations: [ |
|
72 | 72 | 0, 1MB), [1MB, 100MB) and [100MB, infinity). |
|
73 | 73 | |
|
74 | 74 | ``remotefilelog.history.maxrepackpacks`` the maximum number of pack files to |
|
75 | 75 | include in an incremental history repack. |
|
76 | 76 | |
|
77 | 77 | ``remotefilelog.history.repackmaxpacksize`` the maximum size of a pack file |
|
78 | 78 | for it to be considered for an incremental history repack. |
|
79 | 79 | |
|
80 | 80 | ``remotefilelog.history.repacksizelimit`` the maximum total size of pack |
|
81 | 81 | files to include in an incremental history repack. |
|
82 | 82 | |
|
83 | 83 | ``remotefilelog.backgroundrepack`` automatically consolidate packs in the |
|
84 | 84 | background |
|
85 | 85 | |
|
86 | 86 | ``remotefilelog.cachepath`` path to cache |
|
87 | 87 | |
|
88 | 88 | ``remotefilelog.cachegroup`` if set, make cache directory sgid to this |
|
89 | 89 | group |
|
90 | 90 | |
|
91 | 91 | ``remotefilelog.cacheprocess`` binary to invoke for fetching file data |
|
92 | 92 | |
|
93 | 93 | ``remotefilelog.debug`` turn on remotefilelog-specific debug output |
|
94 | 94 | |
|
95 | 95 | ``remotefilelog.excludepattern`` pattern of files to exclude from pulls |
|
96 | 96 | |
|
97 | 97 | ``remotefilelog.includepattern`` pattern of files to include in pulls |
|
98 | 98 | |
|
99 | 99 | ``remotefilelog.fetchwarning``: message to print when too many |
|
100 | 100 | single-file fetches occur |
|
101 | 101 | |
|
102 | 102 | ``remotefilelog.getfilesstep`` number of files to request in a single RPC |
|
103 | 103 | |
|
104 | 104 | ``remotefilelog.getfilestype`` if set to 'threaded' use threads to fetch |
|
105 | 105 | files, otherwise use optimistic fetching |
|
106 | 106 | |
|
107 | 107 | ``remotefilelog.pullprefetch`` revset for selecting files that should be |
|
108 | 108 | eagerly downloaded rather than lazily |
|
109 | 109 | |
|
110 | 110 | ``remotefilelog.reponame`` name of the repo. If set, used to partition |
|
111 | 111 | data from other repos in a shared store. |
|
112 | 112 | |
|
113 | 113 | ``remotefilelog.server`` if true, enable server-side functionality |
|
114 | 114 | |
|
115 | 115 | ``remotefilelog.servercachepath`` path for caching blobs on the server |
|
116 | 116 | |
|
117 | 117 | ``remotefilelog.serverexpiration`` number of days to keep cached server |
|
118 | 118 | blobs |
|
119 | 119 | |
|
120 | 120 | ``remotefilelog.validatecache`` if set, check cache entries for corruption |
|
121 | 121 | before returning blobs |
|
122 | 122 | |
|
123 | 123 | ``remotefilelog.validatecachelog`` if set, check cache entries for |
|
124 | 124 | corruption before returning metadata |
|
125 | 125 | |
|
126 | 126 | """ |
|
127 | 127 | from __future__ import absolute_import |
|
128 | 128 | |
|
129 | 129 | import os |
|
130 | 130 | import time |
|
131 | 131 | import traceback |
|
132 | 132 | |
|
133 | 133 | from mercurial.node import hex |
|
134 | 134 | from mercurial.i18n import _ |
|
135 | 135 | from mercurial.pycompat import open |
|
136 | 136 | from mercurial import ( |
|
137 | 137 | changegroup, |
|
138 | 138 | changelog, |
|
139 | 139 | cmdutil, |
|
140 | 140 | commands, |
|
141 | 141 | configitems, |
|
142 | 142 | context, |
|
143 | 143 | copies, |
|
144 | 144 | debugcommands as hgdebugcommands, |
|
145 | 145 | dispatch, |
|
146 | 146 | error, |
|
147 | 147 | exchange, |
|
148 | 148 | extensions, |
|
149 | 149 | hg, |
|
150 | 150 | localrepo, |
|
151 | 151 | match, |
|
152 | 152 | merge, |
|
153 | 153 | node as nodemod, |
|
154 | 154 | patch, |
|
155 | 155 | pycompat, |
|
156 | 156 | registrar, |
|
157 | 157 | repair, |
|
158 | 158 | repoview, |
|
159 | 159 | revset, |
|
160 | 160 | scmutil, |
|
161 | 161 | smartset, |
|
162 | 162 | streamclone, |
|
163 | 163 | util, |
|
164 | 164 | ) |
|
165 | 165 | from . import ( |
|
166 | 166 | constants, |
|
167 | 167 | debugcommands, |
|
168 | 168 | fileserverclient, |
|
169 | 169 | remotefilectx, |
|
170 | 170 | remotefilelog, |
|
171 | 171 | remotefilelogserver, |
|
172 | 172 | repack as repackmod, |
|
173 | 173 | shallowbundle, |
|
174 | 174 | shallowrepo, |
|
175 | 175 | shallowstore, |
|
176 | 176 | shallowutil, |
|
177 | 177 | shallowverifier, |
|
178 | 178 | ) |
|
179 | 179 | |
|
180 | 180 | # ensures debug commands are registered |
|
181 | 181 | hgdebugcommands.command |
|
182 | 182 | |
|
183 | 183 | cmdtable = {} |
|
184 | 184 | command = registrar.command(cmdtable) |
|
185 | 185 | |
|
186 | 186 | configtable = {} |
|
187 | 187 | configitem = registrar.configitem(configtable) |
|
188 | 188 | |
|
189 | 189 | configitem(b'remotefilelog', b'debug', default=False) |
|
190 | 190 | |
|
191 | 191 | configitem(b'remotefilelog', b'reponame', default=b'') |
|
192 | 192 | configitem(b'remotefilelog', b'cachepath', default=None) |
|
193 | 193 | configitem(b'remotefilelog', b'cachegroup', default=None) |
|
194 | 194 | configitem(b'remotefilelog', b'cacheprocess', default=None) |
|
195 | 195 | configitem(b'remotefilelog', b'cacheprocess.includepath', default=None) |
|
196 | 196 | configitem(b"remotefilelog", b"cachelimit", default=b"1000 GB") |
|
197 | 197 | |
|
198 | 198 | configitem( |
|
199 | 199 | b'remotefilelog', |
|
200 | 200 | b'fallbackpath', |
|
201 | 201 | default=configitems.dynamicdefault, |
|
202 | 202 | alias=[(b'remotefilelog', b'fallbackrepo')], |
|
203 | 203 | ) |
|
204 | 204 | |
|
205 | 205 | configitem(b'remotefilelog', b'validatecachelog', default=None) |
|
206 | 206 | configitem(b'remotefilelog', b'validatecache', default=b'on') |
|
207 | 207 | configitem(b'remotefilelog', b'server', default=None) |
|
208 | 208 | configitem(b'remotefilelog', b'servercachepath', default=None) |
|
209 | 209 | configitem(b"remotefilelog", b"serverexpiration", default=30) |
|
210 | 210 | configitem(b'remotefilelog', b'backgroundrepack', default=False) |
|
211 | 211 | configitem(b'remotefilelog', b'bgprefetchrevs', default=None) |
|
212 | 212 | configitem(b'remotefilelog', b'pullprefetch', default=None) |
|
213 | 213 | configitem(b'remotefilelog', b'backgroundprefetch', default=False) |
|
214 | 214 | configitem(b'remotefilelog', b'prefetchdelay', default=120) |
|
215 | 215 | configitem(b'remotefilelog', b'prefetchdays', default=14) |
|
216 | 216 | |
|
217 | 217 | configitem(b'remotefilelog', b'getfilesstep', default=10000) |
|
218 | 218 | configitem(b'remotefilelog', b'getfilestype', default=b'optimistic') |
|
219 | 219 | configitem(b'remotefilelog', b'batchsize', configitems.dynamicdefault) |
|
220 | 220 | configitem(b'remotefilelog', b'fetchwarning', default=b'') |
|
221 | 221 | |
|
222 | 222 | configitem(b'remotefilelog', b'includepattern', default=None) |
|
223 | 223 | configitem(b'remotefilelog', b'excludepattern', default=None) |
|
224 | 224 | |
|
225 | 225 | configitem(b'remotefilelog', b'gcrepack', default=False) |
|
226 | 226 | configitem(b'remotefilelog', b'repackonhggc', default=False) |
|
227 | 227 | configitem(b'repack', b'chainorphansbysize', default=True, experimental=True) |
|
228 | 228 | |
|
229 | 229 | configitem(b'packs', b'maxpacksize', default=0) |
|
230 | 230 | configitem(b'packs', b'maxchainlen', default=1000) |
|
231 | 231 | |
|
232 | 232 | configitem(b'devel', b'remotefilelog.ensurestart', default=False) |
|
233 | configitem(b'devel', b'remotefilelog.bg-wait', default=False) | |
|
233 | 234 | |
|
234 | 235 | # default TTL limit is 30 days |
|
235 | 236 | _defaultlimit = 60 * 60 * 24 * 30 |
|
236 | 237 | configitem(b'remotefilelog', b'nodettl', default=_defaultlimit) |
|
237 | 238 | |
|
238 | 239 | configitem(b'remotefilelog', b'data.gencountlimit', default=2), |
|
239 | 240 | configitem( |
|
240 | 241 | b'remotefilelog', b'data.generations', default=[b'1GB', b'100MB', b'1MB'] |
|
241 | 242 | ) |
|
242 | 243 | configitem(b'remotefilelog', b'data.maxrepackpacks', default=50) |
|
243 | 244 | configitem(b'remotefilelog', b'data.repackmaxpacksize', default=b'4GB') |
|
244 | 245 | configitem(b'remotefilelog', b'data.repacksizelimit', default=b'100MB') |
|
245 | 246 | |
|
246 | 247 | configitem(b'remotefilelog', b'history.gencountlimit', default=2), |
|
247 | 248 | configitem(b'remotefilelog', b'history.generations', default=[b'100MB']) |
|
248 | 249 | configitem(b'remotefilelog', b'history.maxrepackpacks', default=50) |
|
249 | 250 | configitem(b'remotefilelog', b'history.repackmaxpacksize', default=b'400MB') |
|
250 | 251 | configitem(b'remotefilelog', b'history.repacksizelimit', default=b'100MB') |
|
251 | 252 | |
|
252 | 253 | # Note for extension authors: ONLY specify testedwith = 'ships-with-hg-core' for |
|
253 | 254 | # extensions which SHIP WITH MERCURIAL. Non-mainline extensions should |
|
254 | 255 | # be specifying the version(s) of Mercurial they are tested with, or |
|
255 | 256 | # leave the attribute unspecified. |
|
256 | 257 | testedwith = b'ships-with-hg-core' |
|
257 | 258 | |
|
258 | 259 | repoclass = localrepo.localrepository |
|
259 | 260 | repoclass._basesupported.add(constants.SHALLOWREPO_REQUIREMENT) |
|
260 | 261 | |
|
261 | 262 | isenabled = shallowutil.isenabled |
|
262 | 263 | |
|
263 | 264 | |
|
264 | 265 | def uisetup(ui): |
|
265 | 266 | """Wraps user facing Mercurial commands to swap them out with shallow |
|
266 | 267 | versions. |
|
267 | 268 | """ |
|
268 | 269 | hg.wirepeersetupfuncs.append(fileserverclient.peersetup) |
|
269 | 270 | |
|
270 | 271 | entry = extensions.wrapcommand(commands.table, b'clone', cloneshallow) |
|
271 | 272 | entry[1].append( |
|
272 | 273 | ( |
|
273 | 274 | b'', |
|
274 | 275 | b'shallow', |
|
275 | 276 | None, |
|
276 | 277 | _(b"create a shallow clone which uses remote file history"), |
|
277 | 278 | ) |
|
278 | 279 | ) |
|
279 | 280 | |
|
280 | 281 | extensions.wrapcommand( |
|
281 | 282 | commands.table, b'debugindex', debugcommands.debugindex |
|
282 | 283 | ) |
|
283 | 284 | extensions.wrapcommand( |
|
284 | 285 | commands.table, b'debugindexdot', debugcommands.debugindexdot |
|
285 | 286 | ) |
|
286 | 287 | extensions.wrapcommand(commands.table, b'log', log) |
|
287 | 288 | extensions.wrapcommand(commands.table, b'pull', pull) |
|
288 | 289 | |
|
289 | 290 | # Prevent 'hg manifest --all' |
|
290 | 291 | def _manifest(orig, ui, repo, *args, **opts): |
|
291 | 292 | if isenabled(repo) and opts.get(r'all'): |
|
292 | 293 | raise error.Abort(_(b"--all is not supported in a shallow repo")) |
|
293 | 294 | |
|
294 | 295 | return orig(ui, repo, *args, **opts) |
|
295 | 296 | |
|
296 | 297 | extensions.wrapcommand(commands.table, b"manifest", _manifest) |
|
297 | 298 | |
|
298 | 299 | # Wrap remotefilelog with lfs code |
|
299 | 300 | def _lfsloaded(loaded=False): |
|
300 | 301 | lfsmod = None |
|
301 | 302 | try: |
|
302 | 303 | lfsmod = extensions.find(b'lfs') |
|
303 | 304 | except KeyError: |
|
304 | 305 | pass |
|
305 | 306 | if lfsmod: |
|
306 | 307 | lfsmod.wrapfilelog(remotefilelog.remotefilelog) |
|
307 | 308 | fileserverclient._lfsmod = lfsmod |
|
308 | 309 | |
|
309 | 310 | extensions.afterloaded(b'lfs', _lfsloaded) |
|
310 | 311 | |
|
311 | 312 | # debugdata needs remotefilelog.len to work |
|
312 | 313 | extensions.wrapcommand(commands.table, b'debugdata', debugdatashallow) |
|
313 | 314 | |
|
314 | 315 | changegroup.cgpacker = shallowbundle.shallowcg1packer |
|
315 | 316 | |
|
316 | 317 | extensions.wrapfunction( |
|
317 | 318 | changegroup, b'_addchangegroupfiles', shallowbundle.addchangegroupfiles |
|
318 | 319 | ) |
|
319 | 320 | extensions.wrapfunction( |
|
320 | 321 | changegroup, b'makechangegroup', shallowbundle.makechangegroup |
|
321 | 322 | ) |
|
322 | 323 | extensions.wrapfunction(localrepo, b'makestore', storewrapper) |
|
323 | 324 | extensions.wrapfunction(exchange, b'pull', exchangepull) |
|
324 | 325 | extensions.wrapfunction(merge, b'applyupdates', applyupdates) |
|
325 | 326 | extensions.wrapfunction(merge, b'_checkunknownfiles', checkunknownfiles) |
|
326 | 327 | extensions.wrapfunction(context.workingctx, b'_checklookup', checklookup) |
|
327 | 328 | extensions.wrapfunction(scmutil, b'_findrenames', findrenames) |
|
328 | 329 | extensions.wrapfunction( |
|
329 | 330 | copies, b'_computeforwardmissing', computeforwardmissing |
|
330 | 331 | ) |
|
331 | 332 | extensions.wrapfunction(dispatch, b'runcommand', runcommand) |
|
332 | 333 | extensions.wrapfunction(repair, b'_collectbrokencsets', _collectbrokencsets) |
|
333 | 334 | extensions.wrapfunction(context.changectx, b'filectx', filectx) |
|
334 | 335 | extensions.wrapfunction(context.workingctx, b'filectx', workingfilectx) |
|
335 | 336 | extensions.wrapfunction(patch, b'trydiff', trydiff) |
|
336 | 337 | extensions.wrapfunction(hg, b'verify', _verify) |
|
337 | 338 | scmutil.fileprefetchhooks.add(b'remotefilelog', _fileprefetchhook) |
|
338 | 339 | |
|
339 | 340 | # disappointing hacks below |
|
340 | 341 | extensions.wrapfunction(scmutil, b'getrenamedfn', getrenamedfn) |
|
341 | 342 | extensions.wrapfunction(revset, b'filelog', filelogrevset) |
|
342 | 343 | revset.symbols[b'filelog'] = revset.filelog |
|
343 | 344 | extensions.wrapfunction(cmdutil, b'walkfilerevs', walkfilerevs) |
|
344 | 345 | |
|
345 | 346 | |
|
346 | 347 | def cloneshallow(orig, ui, repo, *args, **opts): |
|
347 | 348 | if opts.get(r'shallow'): |
|
348 | 349 | repos = [] |
|
349 | 350 | |
|
350 | 351 | def pull_shallow(orig, self, *args, **kwargs): |
|
351 | 352 | if not isenabled(self): |
|
352 | 353 | repos.append(self.unfiltered()) |
|
353 | 354 | # set up the client hooks so the post-clone update works |
|
354 | 355 | setupclient(self.ui, self.unfiltered()) |
|
355 | 356 | |
|
356 | 357 | # setupclient fixed the class on the repo itself |
|
357 | 358 | # but we also need to fix it on the repoview |
|
358 | 359 | if isinstance(self, repoview.repoview): |
|
359 | 360 | self.__class__.__bases__ = ( |
|
360 | 361 | self.__class__.__bases__[0], |
|
361 | 362 | self.unfiltered().__class__, |
|
362 | 363 | ) |
|
363 | 364 | self.requirements.add(constants.SHALLOWREPO_REQUIREMENT) |
|
364 | 365 | self._writerequirements() |
|
365 | 366 | |
|
366 | 367 | # Since setupclient hadn't been called, exchange.pull was not |
|
367 | 368 | # wrapped. So we need to manually invoke our version of it. |
|
368 | 369 | return exchangepull(orig, self, *args, **kwargs) |
|
369 | 370 | else: |
|
370 | 371 | return orig(self, *args, **kwargs) |
|
371 | 372 | |
|
372 | 373 | extensions.wrapfunction(exchange, b'pull', pull_shallow) |
|
373 | 374 | |
|
374 | 375 | # Wrap the stream logic to add requirements and to pass include/exclude |
|
375 | 376 | # patterns around. |
|
376 | 377 | def setup_streamout(repo, remote): |
|
377 | 378 | # Replace remote.stream_out with a version that sends file |
|
378 | 379 | # patterns. |
|
379 | 380 | def stream_out_shallow(orig): |
|
380 | 381 | caps = remote.capabilities() |
|
381 | 382 | if constants.NETWORK_CAP_LEGACY_SSH_GETFILES in caps: |
|
382 | 383 | opts = {} |
|
383 | 384 | if repo.includepattern: |
|
384 | 385 | opts[r'includepattern'] = b'\0'.join( |
|
385 | 386 | repo.includepattern |
|
386 | 387 | ) |
|
387 | 388 | if repo.excludepattern: |
|
388 | 389 | opts[r'excludepattern'] = b'\0'.join( |
|
389 | 390 | repo.excludepattern |
|
390 | 391 | ) |
|
391 | 392 | return remote._callstream(b'stream_out_shallow', **opts) |
|
392 | 393 | else: |
|
393 | 394 | return orig() |
|
394 | 395 | |
|
395 | 396 | extensions.wrapfunction(remote, b'stream_out', stream_out_shallow) |
|
396 | 397 | |
|
397 | 398 | def stream_wrap(orig, op): |
|
398 | 399 | setup_streamout(op.repo, op.remote) |
|
399 | 400 | return orig(op) |
|
400 | 401 | |
|
401 | 402 | extensions.wrapfunction( |
|
402 | 403 | streamclone, b'maybeperformlegacystreamclone', stream_wrap |
|
403 | 404 | ) |
|
404 | 405 | |
|
405 | 406 | def canperformstreamclone(orig, pullop, bundle2=False): |
|
406 | 407 | # remotefilelog is currently incompatible with the |
|
407 | 408 | # bundle2 flavor of streamclones, so force us to use |
|
408 | 409 | # v1 instead. |
|
409 | 410 | if b'v2' in pullop.remotebundle2caps.get(b'stream', []): |
|
410 | 411 | pullop.remotebundle2caps[b'stream'] = [ |
|
411 | 412 | c for c in pullop.remotebundle2caps[b'stream'] if c != b'v2' |
|
412 | 413 | ] |
|
413 | 414 | if bundle2: |
|
414 | 415 | return False, None |
|
415 | 416 | supported, requirements = orig(pullop, bundle2=bundle2) |
|
416 | 417 | if requirements is not None: |
|
417 | 418 | requirements.add(constants.SHALLOWREPO_REQUIREMENT) |
|
418 | 419 | return supported, requirements |
|
419 | 420 | |
|
420 | 421 | extensions.wrapfunction( |
|
421 | 422 | streamclone, b'canperformstreamclone', canperformstreamclone |
|
422 | 423 | ) |
|
423 | 424 | |
|
424 | 425 | try: |
|
425 | 426 | orig(ui, repo, *args, **opts) |
|
426 | 427 | finally: |
|
427 | 428 | if opts.get(r'shallow'): |
|
428 | 429 | for r in repos: |
|
429 | 430 | if util.safehasattr(r, b'fileservice'): |
|
430 | 431 | r.fileservice.close() |
|
431 | 432 | |
|
432 | 433 | |
|
433 | 434 | def debugdatashallow(orig, *args, **kwds): |
|
434 | 435 | oldlen = remotefilelog.remotefilelog.__len__ |
|
435 | 436 | try: |
|
436 | 437 | remotefilelog.remotefilelog.__len__ = lambda x: 1 |
|
437 | 438 | return orig(*args, **kwds) |
|
438 | 439 | finally: |
|
439 | 440 | remotefilelog.remotefilelog.__len__ = oldlen |
|
440 | 441 | |
|
441 | 442 | |
|
442 | 443 | def reposetup(ui, repo): |
|
443 | 444 | if not repo.local(): |
|
444 | 445 | return |
|
445 | 446 | |
|
446 | 447 | # put here intentionally bc doesnt work in uisetup |
|
447 | 448 | ui.setconfig(b'hooks', b'update.prefetch', wcpprefetch) |
|
448 | 449 | ui.setconfig(b'hooks', b'commit.prefetch', wcpprefetch) |
|
449 | 450 | |
|
450 | 451 | isserverenabled = ui.configbool(b'remotefilelog', b'server') |
|
451 | 452 | isshallowclient = isenabled(repo) |
|
452 | 453 | |
|
453 | 454 | if isserverenabled and isshallowclient: |
|
454 | 455 | raise RuntimeError(b"Cannot be both a server and shallow client.") |
|
455 | 456 | |
|
456 | 457 | if isshallowclient: |
|
457 | 458 | setupclient(ui, repo) |
|
458 | 459 | |
|
459 | 460 | if isserverenabled: |
|
460 | 461 | remotefilelogserver.setupserver(ui, repo) |
|
461 | 462 | |
|
462 | 463 | |
|
463 | 464 | def setupclient(ui, repo): |
|
464 | 465 | if not isinstance(repo, localrepo.localrepository): |
|
465 | 466 | return |
|
466 | 467 | |
|
467 | 468 | # Even clients get the server setup since they need to have the |
|
468 | 469 | # wireprotocol endpoints registered. |
|
469 | 470 | remotefilelogserver.onetimesetup(ui) |
|
470 | 471 | onetimeclientsetup(ui) |
|
471 | 472 | |
|
472 | 473 | shallowrepo.wraprepo(repo) |
|
473 | 474 | repo.store = shallowstore.wrapstore(repo.store) |
|
474 | 475 | |
|
475 | 476 | |
|
476 | 477 | def storewrapper(orig, requirements, path, vfstype): |
|
477 | 478 | s = orig(requirements, path, vfstype) |
|
478 | 479 | if constants.SHALLOWREPO_REQUIREMENT in requirements: |
|
479 | 480 | s = shallowstore.wrapstore(s) |
|
480 | 481 | |
|
481 | 482 | return s |
|
482 | 483 | |
|
483 | 484 | |
|
484 | 485 | # prefetch files before update |
|
485 | 486 | def applyupdates( |
|
486 | 487 | orig, repo, actions, wctx, mctx, overwrite, wantfiledata, labels=None |
|
487 | 488 | ): |
|
488 | 489 | if isenabled(repo): |
|
489 | 490 | manifest = mctx.manifest() |
|
490 | 491 | files = [] |
|
491 | 492 | for f, args, msg in actions[b'g']: |
|
492 | 493 | files.append((f, hex(manifest[f]))) |
|
493 | 494 | # batch fetch the needed files from the server |
|
494 | 495 | repo.fileservice.prefetch(files) |
|
495 | 496 | return orig( |
|
496 | 497 | repo, actions, wctx, mctx, overwrite, wantfiledata, labels=labels |
|
497 | 498 | ) |
|
498 | 499 | |
|
499 | 500 | |
|
500 | 501 | # Prefetch merge checkunknownfiles |
|
501 | 502 | def checkunknownfiles(orig, repo, wctx, mctx, force, actions, *args, **kwargs): |
|
502 | 503 | if isenabled(repo): |
|
503 | 504 | files = [] |
|
504 | 505 | sparsematch = repo.maybesparsematch(mctx.rev()) |
|
505 | 506 | for f, (m, actionargs, msg) in pycompat.iteritems(actions): |
|
506 | 507 | if sparsematch and not sparsematch(f): |
|
507 | 508 | continue |
|
508 | 509 | if m in (b'c', b'dc', b'cm'): |
|
509 | 510 | files.append((f, hex(mctx.filenode(f)))) |
|
510 | 511 | elif m == b'dg': |
|
511 | 512 | f2 = actionargs[0] |
|
512 | 513 | files.append((f2, hex(mctx.filenode(f2)))) |
|
513 | 514 | # batch fetch the needed files from the server |
|
514 | 515 | repo.fileservice.prefetch(files) |
|
515 | 516 | return orig(repo, wctx, mctx, force, actions, *args, **kwargs) |
|
516 | 517 | |
|
517 | 518 | |
|
518 | 519 | # Prefetch files before status attempts to look at their size and contents |
|
519 | 520 | def checklookup(orig, self, files): |
|
520 | 521 | repo = self._repo |
|
521 | 522 | if isenabled(repo): |
|
522 | 523 | prefetchfiles = [] |
|
523 | 524 | for parent in self._parents: |
|
524 | 525 | for f in files: |
|
525 | 526 | if f in parent: |
|
526 | 527 | prefetchfiles.append((f, hex(parent.filenode(f)))) |
|
527 | 528 | # batch fetch the needed files from the server |
|
528 | 529 | repo.fileservice.prefetch(prefetchfiles) |
|
529 | 530 | return orig(self, files) |
|
530 | 531 | |
|
531 | 532 | |
|
532 | 533 | # Prefetch the logic that compares added and removed files for renames |
|
533 | 534 | def findrenames(orig, repo, matcher, added, removed, *args, **kwargs): |
|
534 | 535 | if isenabled(repo): |
|
535 | 536 | files = [] |
|
536 | 537 | pmf = repo[b'.'].manifest() |
|
537 | 538 | for f in removed: |
|
538 | 539 | if f in pmf: |
|
539 | 540 | files.append((f, hex(pmf[f]))) |
|
540 | 541 | # batch fetch the needed files from the server |
|
541 | 542 | repo.fileservice.prefetch(files) |
|
542 | 543 | return orig(repo, matcher, added, removed, *args, **kwargs) |
|
543 | 544 | |
|
544 | 545 | |
|
545 | 546 | # prefetch files before pathcopies check |
|
546 | 547 | def computeforwardmissing(orig, a, b, match=None): |
|
547 | 548 | missing = orig(a, b, match=match) |
|
548 | 549 | repo = a._repo |
|
549 | 550 | if isenabled(repo): |
|
550 | 551 | mb = b.manifest() |
|
551 | 552 | |
|
552 | 553 | files = [] |
|
553 | 554 | sparsematch = repo.maybesparsematch(b.rev()) |
|
554 | 555 | if sparsematch: |
|
555 | 556 | sparsemissing = set() |
|
556 | 557 | for f in missing: |
|
557 | 558 | if sparsematch(f): |
|
558 | 559 | files.append((f, hex(mb[f]))) |
|
559 | 560 | sparsemissing.add(f) |
|
560 | 561 | missing = sparsemissing |
|
561 | 562 | |
|
562 | 563 | # batch fetch the needed files from the server |
|
563 | 564 | repo.fileservice.prefetch(files) |
|
564 | 565 | return missing |
|
565 | 566 | |
|
566 | 567 | |
|
567 | 568 | # close cache miss server connection after the command has finished |
|
568 | 569 | def runcommand(orig, lui, repo, *args, **kwargs): |
|
569 | 570 | fileservice = None |
|
570 | 571 | # repo can be None when running in chg: |
|
571 | 572 | # - at startup, reposetup was called because serve is not norepo |
|
572 | 573 | # - a norepo command like "help" is called |
|
573 | 574 | if repo and isenabled(repo): |
|
574 | 575 | fileservice = repo.fileservice |
|
575 | 576 | try: |
|
576 | 577 | return orig(lui, repo, *args, **kwargs) |
|
577 | 578 | finally: |
|
578 | 579 | if fileservice: |
|
579 | 580 | fileservice.close() |
|
580 | 581 | |
|
581 | 582 | |
|
582 | 583 | # prevent strip from stripping remotefilelogs |
|
583 | 584 | def _collectbrokencsets(orig, repo, files, striprev): |
|
584 | 585 | if isenabled(repo): |
|
585 | 586 | files = list([f for f in files if not repo.shallowmatch(f)]) |
|
586 | 587 | return orig(repo, files, striprev) |
|
587 | 588 | |
|
588 | 589 | |
|
589 | 590 | # changectx wrappers |
|
590 | 591 | def filectx(orig, self, path, fileid=None, filelog=None): |
|
591 | 592 | if fileid is None: |
|
592 | 593 | fileid = self.filenode(path) |
|
593 | 594 | if isenabled(self._repo) and self._repo.shallowmatch(path): |
|
594 | 595 | return remotefilectx.remotefilectx( |
|
595 | 596 | self._repo, path, fileid=fileid, changectx=self, filelog=filelog |
|
596 | 597 | ) |
|
597 | 598 | return orig(self, path, fileid=fileid, filelog=filelog) |
|
598 | 599 | |
|
599 | 600 | |
|
600 | 601 | def workingfilectx(orig, self, path, filelog=None): |
|
601 | 602 | if isenabled(self._repo) and self._repo.shallowmatch(path): |
|
602 | 603 | return remotefilectx.remoteworkingfilectx( |
|
603 | 604 | self._repo, path, workingctx=self, filelog=filelog |
|
604 | 605 | ) |
|
605 | 606 | return orig(self, path, filelog=filelog) |
|
606 | 607 | |
|
607 | 608 | |
|
608 | 609 | # prefetch required revisions before a diff |
|
609 | 610 | def trydiff( |
|
610 | 611 | orig, |
|
611 | 612 | repo, |
|
612 | 613 | revs, |
|
613 | 614 | ctx1, |
|
614 | 615 | ctx2, |
|
615 | 616 | modified, |
|
616 | 617 | added, |
|
617 | 618 | removed, |
|
618 | 619 | copy, |
|
619 | 620 | getfilectx, |
|
620 | 621 | *args, |
|
621 | 622 | **kwargs |
|
622 | 623 | ): |
|
623 | 624 | if isenabled(repo): |
|
624 | 625 | prefetch = [] |
|
625 | 626 | mf1 = ctx1.manifest() |
|
626 | 627 | for fname in modified + added + removed: |
|
627 | 628 | if fname in mf1: |
|
628 | 629 | fnode = getfilectx(fname, ctx1).filenode() |
|
629 | 630 | # fnode can be None if it's a edited working ctx file |
|
630 | 631 | if fnode: |
|
631 | 632 | prefetch.append((fname, hex(fnode))) |
|
632 | 633 | if fname not in removed: |
|
633 | 634 | fnode = getfilectx(fname, ctx2).filenode() |
|
634 | 635 | if fnode: |
|
635 | 636 | prefetch.append((fname, hex(fnode))) |
|
636 | 637 | |
|
637 | 638 | repo.fileservice.prefetch(prefetch) |
|
638 | 639 | |
|
639 | 640 | return orig( |
|
640 | 641 | repo, |
|
641 | 642 | revs, |
|
642 | 643 | ctx1, |
|
643 | 644 | ctx2, |
|
644 | 645 | modified, |
|
645 | 646 | added, |
|
646 | 647 | removed, |
|
647 | 648 | copy, |
|
648 | 649 | getfilectx, |
|
649 | 650 | *args, |
|
650 | 651 | **kwargs |
|
651 | 652 | ) |
|
652 | 653 | |
|
653 | 654 | |
|
654 | 655 | # Prevent verify from processing files |
|
655 | 656 | # a stub for mercurial.hg.verify() |
|
656 | 657 | def _verify(orig, repo, level=None): |
|
657 | 658 | lock = repo.lock() |
|
658 | 659 | try: |
|
659 | 660 | return shallowverifier.shallowverifier(repo).verify() |
|
660 | 661 | finally: |
|
661 | 662 | lock.release() |
|
662 | 663 | |
|
663 | 664 | |
|
664 | 665 | clientonetime = False |
|
665 | 666 | |
|
666 | 667 | |
|
667 | 668 | def onetimeclientsetup(ui): |
|
668 | 669 | global clientonetime |
|
669 | 670 | if clientonetime: |
|
670 | 671 | return |
|
671 | 672 | clientonetime = True |
|
672 | 673 | |
|
673 | 674 | # Don't commit filelogs until we know the commit hash, since the hash |
|
674 | 675 | # is present in the filelog blob. |
|
675 | 676 | # This violates Mercurial's filelog->manifest->changelog write order, |
|
676 | 677 | # but is generally fine for client repos. |
|
677 | 678 | pendingfilecommits = [] |
|
678 | 679 | |
|
679 | 680 | def addrawrevision( |
|
680 | 681 | orig, |
|
681 | 682 | self, |
|
682 | 683 | rawtext, |
|
683 | 684 | transaction, |
|
684 | 685 | link, |
|
685 | 686 | p1, |
|
686 | 687 | p2, |
|
687 | 688 | node, |
|
688 | 689 | flags, |
|
689 | 690 | cachedelta=None, |
|
690 | 691 | _metatuple=None, |
|
691 | 692 | ): |
|
692 | 693 | if isinstance(link, int): |
|
693 | 694 | pendingfilecommits.append( |
|
694 | 695 | ( |
|
695 | 696 | self, |
|
696 | 697 | rawtext, |
|
697 | 698 | transaction, |
|
698 | 699 | link, |
|
699 | 700 | p1, |
|
700 | 701 | p2, |
|
701 | 702 | node, |
|
702 | 703 | flags, |
|
703 | 704 | cachedelta, |
|
704 | 705 | _metatuple, |
|
705 | 706 | ) |
|
706 | 707 | ) |
|
707 | 708 | return node |
|
708 | 709 | else: |
|
709 | 710 | return orig( |
|
710 | 711 | self, |
|
711 | 712 | rawtext, |
|
712 | 713 | transaction, |
|
713 | 714 | link, |
|
714 | 715 | p1, |
|
715 | 716 | p2, |
|
716 | 717 | node, |
|
717 | 718 | flags, |
|
718 | 719 | cachedelta, |
|
719 | 720 | _metatuple=_metatuple, |
|
720 | 721 | ) |
|
721 | 722 | |
|
722 | 723 | extensions.wrapfunction( |
|
723 | 724 | remotefilelog.remotefilelog, b'addrawrevision', addrawrevision |
|
724 | 725 | ) |
|
725 | 726 | |
|
726 | 727 | def changelogadd(orig, self, *args): |
|
727 | 728 | oldlen = len(self) |
|
728 | 729 | node = orig(self, *args) |
|
729 | 730 | newlen = len(self) |
|
730 | 731 | if oldlen != newlen: |
|
731 | 732 | for oldargs in pendingfilecommits: |
|
732 | 733 | log, rt, tr, link, p1, p2, n, fl, c, m = oldargs |
|
733 | 734 | linknode = self.node(link) |
|
734 | 735 | if linknode == node: |
|
735 | 736 | log.addrawrevision(rt, tr, linknode, p1, p2, n, fl, c, m) |
|
736 | 737 | else: |
|
737 | 738 | raise error.ProgrammingError( |
|
738 | 739 | b'pending multiple integer revisions are not supported' |
|
739 | 740 | ) |
|
740 | 741 | else: |
|
741 | 742 | # "link" is actually wrong here (it is set to len(changelog)) |
|
742 | 743 | # if changelog remains unchanged, skip writing file revisions |
|
743 | 744 | # but still do a sanity check about pending multiple revisions |
|
744 | 745 | if len(set(x[3] for x in pendingfilecommits)) > 1: |
|
745 | 746 | raise error.ProgrammingError( |
|
746 | 747 | b'pending multiple integer revisions are not supported' |
|
747 | 748 | ) |
|
748 | 749 | del pendingfilecommits[:] |
|
749 | 750 | return node |
|
750 | 751 | |
|
751 | 752 | extensions.wrapfunction(changelog.changelog, b'add', changelogadd) |
|
752 | 753 | |
|
753 | 754 | |
|
754 | 755 | def getrenamedfn(orig, repo, endrev=None): |
|
755 | 756 | if not isenabled(repo) or copies.usechangesetcentricalgo(repo): |
|
756 | 757 | return orig(repo, endrev) |
|
757 | 758 | |
|
758 | 759 | rcache = {} |
|
759 | 760 | |
|
760 | 761 | def getrenamed(fn, rev): |
|
761 | 762 | '''looks up all renames for a file (up to endrev) the first |
|
762 | 763 | time the file is given. It indexes on the changerev and only |
|
763 | 764 | parses the manifest if linkrev != changerev. |
|
764 | 765 | Returns rename info for fn at changerev rev.''' |
|
765 | 766 | if rev in rcache.setdefault(fn, {}): |
|
766 | 767 | return rcache[fn][rev] |
|
767 | 768 | |
|
768 | 769 | try: |
|
769 | 770 | fctx = repo[rev].filectx(fn) |
|
770 | 771 | for ancestor in fctx.ancestors(): |
|
771 | 772 | if ancestor.path() == fn: |
|
772 | 773 | renamed = ancestor.renamed() |
|
773 | 774 | rcache[fn][ancestor.rev()] = renamed and renamed[0] |
|
774 | 775 | |
|
775 | 776 | renamed = fctx.renamed() |
|
776 | 777 | return renamed and renamed[0] |
|
777 | 778 | except error.LookupError: |
|
778 | 779 | return None |
|
779 | 780 | |
|
780 | 781 | return getrenamed |
|
781 | 782 | |
|
782 | 783 | |
|
783 | 784 | def walkfilerevs(orig, repo, match, follow, revs, fncache): |
|
784 | 785 | if not isenabled(repo): |
|
785 | 786 | return orig(repo, match, follow, revs, fncache) |
|
786 | 787 | |
|
787 | 788 | # remotefilelog's can't be walked in rev order, so throw. |
|
788 | 789 | # The caller will see the exception and walk the commit tree instead. |
|
789 | 790 | if not follow: |
|
790 | 791 | raise cmdutil.FileWalkError(b"Cannot walk via filelog") |
|
791 | 792 | |
|
792 | 793 | wanted = set() |
|
793 | 794 | minrev, maxrev = min(revs), max(revs) |
|
794 | 795 | |
|
795 | 796 | pctx = repo[b'.'] |
|
796 | 797 | for filename in match.files(): |
|
797 | 798 | if filename not in pctx: |
|
798 | 799 | raise error.Abort( |
|
799 | 800 | _(b'cannot follow file not in parent revision: "%s"') % filename |
|
800 | 801 | ) |
|
801 | 802 | fctx = pctx[filename] |
|
802 | 803 | |
|
803 | 804 | linkrev = fctx.linkrev() |
|
804 | 805 | if linkrev >= minrev and linkrev <= maxrev: |
|
805 | 806 | fncache.setdefault(linkrev, []).append(filename) |
|
806 | 807 | wanted.add(linkrev) |
|
807 | 808 | |
|
808 | 809 | for ancestor in fctx.ancestors(): |
|
809 | 810 | linkrev = ancestor.linkrev() |
|
810 | 811 | if linkrev >= minrev and linkrev <= maxrev: |
|
811 | 812 | fncache.setdefault(linkrev, []).append(ancestor.path()) |
|
812 | 813 | wanted.add(linkrev) |
|
813 | 814 | |
|
814 | 815 | return wanted |
|
815 | 816 | |
|
816 | 817 | |
|
817 | 818 | def filelogrevset(orig, repo, subset, x): |
|
818 | 819 | """``filelog(pattern)`` |
|
819 | 820 | Changesets connected to the specified filelog. |
|
820 | 821 | |
|
821 | 822 | For performance reasons, ``filelog()`` does not show every changeset |
|
822 | 823 | that affects the requested file(s). See :hg:`help log` for details. For |
|
823 | 824 | a slower, more accurate result, use ``file()``. |
|
824 | 825 | """ |
|
825 | 826 | |
|
826 | 827 | if not isenabled(repo): |
|
827 | 828 | return orig(repo, subset, x) |
|
828 | 829 | |
|
829 | 830 | # i18n: "filelog" is a keyword |
|
830 | 831 | pat = revset.getstring(x, _(b"filelog requires a pattern")) |
|
831 | 832 | m = match.match( |
|
832 | 833 | repo.root, repo.getcwd(), [pat], default=b'relpath', ctx=repo[None] |
|
833 | 834 | ) |
|
834 | 835 | s = set() |
|
835 | 836 | |
|
836 | 837 | if not match.patkind(pat): |
|
837 | 838 | # slow |
|
838 | 839 | for r in subset: |
|
839 | 840 | ctx = repo[r] |
|
840 | 841 | cfiles = ctx.files() |
|
841 | 842 | for f in m.files(): |
|
842 | 843 | if f in cfiles: |
|
843 | 844 | s.add(ctx.rev()) |
|
844 | 845 | break |
|
845 | 846 | else: |
|
846 | 847 | # partial |
|
847 | 848 | files = (f for f in repo[None] if m(f)) |
|
848 | 849 | for f in files: |
|
849 | 850 | fctx = repo[None].filectx(f) |
|
850 | 851 | s.add(fctx.linkrev()) |
|
851 | 852 | for actx in fctx.ancestors(): |
|
852 | 853 | s.add(actx.linkrev()) |
|
853 | 854 | |
|
854 | 855 | return smartset.baseset([r for r in subset if r in s]) |
|
855 | 856 | |
|
856 | 857 | |
|
857 | 858 | @command(b'gc', [], _(b'hg gc [REPO...]'), norepo=True) |
|
858 | 859 | def gc(ui, *args, **opts): |
|
859 | 860 | '''garbage collect the client and server filelog caches |
|
860 | 861 | ''' |
|
861 | 862 | cachepaths = set() |
|
862 | 863 | |
|
863 | 864 | # get the system client cache |
|
864 | 865 | systemcache = shallowutil.getcachepath(ui, allowempty=True) |
|
865 | 866 | if systemcache: |
|
866 | 867 | cachepaths.add(systemcache) |
|
867 | 868 | |
|
868 | 869 | # get repo client and server cache |
|
869 | 870 | repopaths = [] |
|
870 | 871 | pwd = ui.environ.get(b'PWD') |
|
871 | 872 | if pwd: |
|
872 | 873 | repopaths.append(pwd) |
|
873 | 874 | |
|
874 | 875 | repopaths.extend(args) |
|
875 | 876 | repos = [] |
|
876 | 877 | for repopath in repopaths: |
|
877 | 878 | try: |
|
878 | 879 | repo = hg.peer(ui, {}, repopath) |
|
879 | 880 | repos.append(repo) |
|
880 | 881 | |
|
881 | 882 | repocache = shallowutil.getcachepath(repo.ui, allowempty=True) |
|
882 | 883 | if repocache: |
|
883 | 884 | cachepaths.add(repocache) |
|
884 | 885 | except error.RepoError: |
|
885 | 886 | pass |
|
886 | 887 | |
|
887 | 888 | # gc client cache |
|
888 | 889 | for cachepath in cachepaths: |
|
889 | 890 | gcclient(ui, cachepath) |
|
890 | 891 | |
|
891 | 892 | # gc server cache |
|
892 | 893 | for repo in repos: |
|
893 | 894 | remotefilelogserver.gcserver(ui, repo._repo) |
|
894 | 895 | |
|
895 | 896 | |
|
896 | 897 | def gcclient(ui, cachepath): |
|
897 | 898 | # get list of repos that use this cache |
|
898 | 899 | repospath = os.path.join(cachepath, b'repos') |
|
899 | 900 | if not os.path.exists(repospath): |
|
900 | 901 | ui.warn(_(b"no known cache at %s\n") % cachepath) |
|
901 | 902 | return |
|
902 | 903 | |
|
903 | 904 | reposfile = open(repospath, b'rb') |
|
904 | 905 | repos = {r[:-1] for r in reposfile.readlines()} |
|
905 | 906 | reposfile.close() |
|
906 | 907 | |
|
907 | 908 | # build list of useful files |
|
908 | 909 | validrepos = [] |
|
909 | 910 | keepkeys = set() |
|
910 | 911 | |
|
911 | 912 | sharedcache = None |
|
912 | 913 | filesrepacked = False |
|
913 | 914 | |
|
914 | 915 | count = 0 |
|
915 | 916 | progress = ui.makeprogress( |
|
916 | 917 | _(b"analyzing repositories"), unit=b"repos", total=len(repos) |
|
917 | 918 | ) |
|
918 | 919 | for path in repos: |
|
919 | 920 | progress.update(count) |
|
920 | 921 | count += 1 |
|
921 | 922 | try: |
|
922 | 923 | path = ui.expandpath(os.path.normpath(path)) |
|
923 | 924 | except TypeError as e: |
|
924 | 925 | ui.warn(_(b"warning: malformed path: %r:%s\n") % (path, e)) |
|
925 | 926 | traceback.print_exc() |
|
926 | 927 | continue |
|
927 | 928 | try: |
|
928 | 929 | peer = hg.peer(ui, {}, path) |
|
929 | 930 | repo = peer._repo |
|
930 | 931 | except error.RepoError: |
|
931 | 932 | continue |
|
932 | 933 | |
|
933 | 934 | validrepos.append(path) |
|
934 | 935 | |
|
935 | 936 | # Protect against any repo or config changes that have happened since |
|
936 | 937 | # this repo was added to the repos file. We'd rather this loop succeed |
|
937 | 938 | # and too much be deleted, than the loop fail and nothing gets deleted. |
|
938 | 939 | if not isenabled(repo): |
|
939 | 940 | continue |
|
940 | 941 | |
|
941 | 942 | if not util.safehasattr(repo, b'name'): |
|
942 | 943 | ui.warn( |
|
943 | 944 | _(b"repo %s is a misconfigured remotefilelog repo\n") % path |
|
944 | 945 | ) |
|
945 | 946 | continue |
|
946 | 947 | |
|
947 | 948 | # If garbage collection on repack and repack on hg gc are enabled |
|
948 | 949 | # then loose files are repacked and garbage collected. |
|
949 | 950 | # Otherwise regular garbage collection is performed. |
|
950 | 951 | repackonhggc = repo.ui.configbool(b'remotefilelog', b'repackonhggc') |
|
951 | 952 | gcrepack = repo.ui.configbool(b'remotefilelog', b'gcrepack') |
|
952 | 953 | if repackonhggc and gcrepack: |
|
953 | 954 | try: |
|
954 | 955 | repackmod.incrementalrepack(repo) |
|
955 | 956 | filesrepacked = True |
|
956 | 957 | continue |
|
957 | 958 | except (IOError, repackmod.RepackAlreadyRunning): |
|
958 | 959 | # If repack cannot be performed due to not enough disk space |
|
959 | 960 | # continue doing garbage collection of loose files w/o repack |
|
960 | 961 | pass |
|
961 | 962 | |
|
962 | 963 | reponame = repo.name |
|
963 | 964 | if not sharedcache: |
|
964 | 965 | sharedcache = repo.sharedstore |
|
965 | 966 | |
|
966 | 967 | # Compute a keepset which is not garbage collected |
|
967 | 968 | def keyfn(fname, fnode): |
|
968 | 969 | return fileserverclient.getcachekey(reponame, fname, hex(fnode)) |
|
969 | 970 | |
|
970 | 971 | keepkeys = repackmod.keepset(repo, keyfn=keyfn, lastkeepkeys=keepkeys) |
|
971 | 972 | |
|
972 | 973 | progress.complete() |
|
973 | 974 | |
|
974 | 975 | # write list of valid repos back |
|
975 | 976 | oldumask = os.umask(0o002) |
|
976 | 977 | try: |
|
977 | 978 | reposfile = open(repospath, b'wb') |
|
978 | 979 | reposfile.writelines([(b"%s\n" % r) for r in validrepos]) |
|
979 | 980 | reposfile.close() |
|
980 | 981 | finally: |
|
981 | 982 | os.umask(oldumask) |
|
982 | 983 | |
|
983 | 984 | # prune cache |
|
984 | 985 | if sharedcache is not None: |
|
985 | 986 | sharedcache.gc(keepkeys) |
|
986 | 987 | elif not filesrepacked: |
|
987 | 988 | ui.warn(_(b"warning: no valid repos in repofile\n")) |
|
988 | 989 | |
|
989 | 990 | |
|
990 | 991 | def log(orig, ui, repo, *pats, **opts): |
|
991 | 992 | if not isenabled(repo): |
|
992 | 993 | return orig(ui, repo, *pats, **opts) |
|
993 | 994 | |
|
994 | 995 | follow = opts.get(r'follow') |
|
995 | 996 | revs = opts.get(r'rev') |
|
996 | 997 | if pats: |
|
997 | 998 | # Force slowpath for non-follow patterns and follows that start from |
|
998 | 999 | # non-working-copy-parent revs. |
|
999 | 1000 | if not follow or revs: |
|
1000 | 1001 | # This forces the slowpath |
|
1001 | 1002 | opts[r'removed'] = True |
|
1002 | 1003 | |
|
1003 | 1004 | # If this is a non-follow log without any revs specified, recommend that |
|
1004 | 1005 | # the user add -f to speed it up. |
|
1005 | 1006 | if not follow and not revs: |
|
1006 | 1007 | match = scmutil.match(repo[b'.'], pats, pycompat.byteskwargs(opts)) |
|
1007 | 1008 | isfile = not match.anypats() |
|
1008 | 1009 | if isfile: |
|
1009 | 1010 | for file in match.files(): |
|
1010 | 1011 | if not os.path.isfile(repo.wjoin(file)): |
|
1011 | 1012 | isfile = False |
|
1012 | 1013 | break |
|
1013 | 1014 | |
|
1014 | 1015 | if isfile: |
|
1015 | 1016 | ui.warn( |
|
1016 | 1017 | _( |
|
1017 | 1018 | b"warning: file log can be slow on large repos - " |
|
1018 | 1019 | + b"use -f to speed it up\n" |
|
1019 | 1020 | ) |
|
1020 | 1021 | ) |
|
1021 | 1022 | |
|
1022 | 1023 | return orig(ui, repo, *pats, **opts) |
|
1023 | 1024 | |
|
1024 | 1025 | |
|
1025 | 1026 | def revdatelimit(ui, revset): |
|
1026 | 1027 | """Update revset so that only changesets no older than 'prefetchdays' days |
|
1027 | 1028 | are included. The default value is set to 14 days. If 'prefetchdays' is set |
|
1028 | 1029 | to zero or negative value then date restriction is not applied. |
|
1029 | 1030 | """ |
|
1030 | 1031 | days = ui.configint(b'remotefilelog', b'prefetchdays') |
|
1031 | 1032 | if days > 0: |
|
1032 | 1033 | revset = b'(%s) & date(-%s)' % (revset, days) |
|
1033 | 1034 | return revset |
|
1034 | 1035 | |
|
1035 | 1036 | |
|
1036 | 1037 | def readytofetch(repo): |
|
1037 | 1038 | """Check that enough time has passed since the last background prefetch. |
|
1038 | 1039 | This only relates to prefetches after operations that change the working |
|
1039 | 1040 | copy parent. Default delay between background prefetches is 2 minutes. |
|
1040 | 1041 | """ |
|
1041 | 1042 | timeout = repo.ui.configint(b'remotefilelog', b'prefetchdelay') |
|
1042 | 1043 | fname = repo.vfs.join(b'lastprefetch') |
|
1043 | 1044 | |
|
1044 | 1045 | ready = False |
|
1045 | 1046 | with open(fname, b'a'): |
|
1046 | 1047 | # the with construct above is used to avoid race conditions |
|
1047 | 1048 | modtime = os.path.getmtime(fname) |
|
1048 | 1049 | if (time.time() - modtime) > timeout: |
|
1049 | 1050 | os.utime(fname, None) |
|
1050 | 1051 | ready = True |
|
1051 | 1052 | |
|
1052 | 1053 | return ready |
|
1053 | 1054 | |
|
1054 | 1055 | |
|
1055 | 1056 | def wcpprefetch(ui, repo, **kwargs): |
|
1056 | 1057 | """Prefetches in background revisions specified by bgprefetchrevs revset. |
|
1057 | 1058 | Does background repack if backgroundrepack flag is set in config. |
|
1058 | 1059 | """ |
|
1059 | 1060 | shallow = isenabled(repo) |
|
1060 | 1061 | bgprefetchrevs = ui.config(b'remotefilelog', b'bgprefetchrevs') |
|
1061 | 1062 | isready = readytofetch(repo) |
|
1062 | 1063 | |
|
1063 | 1064 | if not (shallow and bgprefetchrevs and isready): |
|
1064 | 1065 | return |
|
1065 | 1066 | |
|
1066 | 1067 | bgrepack = repo.ui.configbool(b'remotefilelog', b'backgroundrepack') |
|
1067 | 1068 | # update a revset with a date limit |
|
1068 | 1069 | bgprefetchrevs = revdatelimit(ui, bgprefetchrevs) |
|
1069 | 1070 | |
|
1070 | 1071 | def anon(): |
|
1071 | 1072 | if util.safehasattr(repo, b'ranprefetch') and repo.ranprefetch: |
|
1072 | 1073 | return |
|
1073 | 1074 | repo.ranprefetch = True |
|
1074 | 1075 | repo.backgroundprefetch(bgprefetchrevs, repack=bgrepack) |
|
1075 | 1076 | |
|
1076 | 1077 | repo._afterlock(anon) |
|
1077 | 1078 | |
|
1078 | 1079 | |
|
1079 | 1080 | def pull(orig, ui, repo, *pats, **opts): |
|
1080 | 1081 | result = orig(ui, repo, *pats, **opts) |
|
1081 | 1082 | |
|
1082 | 1083 | if isenabled(repo): |
|
1083 | 1084 | # prefetch if it's configured |
|
1084 | 1085 | prefetchrevset = ui.config(b'remotefilelog', b'pullprefetch') |
|
1085 | 1086 | bgrepack = repo.ui.configbool(b'remotefilelog', b'backgroundrepack') |
|
1086 | 1087 | bgprefetch = repo.ui.configbool(b'remotefilelog', b'backgroundprefetch') |
|
1087 | 1088 | ensurestart = repo.ui.configbool(b'devel', b'remotefilelog.ensurestart') |
|
1088 | 1089 | |
|
1089 | 1090 | if prefetchrevset: |
|
1090 | 1091 | ui.status(_(b"prefetching file contents\n")) |
|
1091 | 1092 | revs = scmutil.revrange(repo, [prefetchrevset]) |
|
1092 | 1093 | base = repo[b'.'].rev() |
|
1093 | 1094 | if bgprefetch: |
|
1094 | 1095 | repo.backgroundprefetch( |
|
1095 | 1096 | prefetchrevset, repack=bgrepack, ensurestart=ensurestart |
|
1096 | 1097 | ) |
|
1097 | 1098 | else: |
|
1098 | 1099 | repo.prefetch(revs, base=base) |
|
1099 | 1100 | if bgrepack: |
|
1100 | 1101 | repackmod.backgroundrepack( |
|
1101 | 1102 | repo, incremental=True, ensurestart=ensurestart |
|
1102 | 1103 | ) |
|
1103 | 1104 | elif bgrepack: |
|
1104 | 1105 | repackmod.backgroundrepack( |
|
1105 | 1106 | repo, incremental=True, ensurestart=ensurestart |
|
1106 | 1107 | ) |
|
1107 | 1108 | |
|
1108 | 1109 | return result |
|
1109 | 1110 | |
|
1110 | 1111 | |
|
1111 | 1112 | def exchangepull(orig, repo, remote, *args, **kwargs): |
|
1112 | 1113 | # Hook into the callstream/getbundle to insert bundle capabilities |
|
1113 | 1114 | # during a pull. |
|
1114 | 1115 | def localgetbundle( |
|
1115 | 1116 | orig, source, heads=None, common=None, bundlecaps=None, **kwargs |
|
1116 | 1117 | ): |
|
1117 | 1118 | if not bundlecaps: |
|
1118 | 1119 | bundlecaps = set() |
|
1119 | 1120 | bundlecaps.add(constants.BUNDLE2_CAPABLITY) |
|
1120 | 1121 | return orig( |
|
1121 | 1122 | source, heads=heads, common=common, bundlecaps=bundlecaps, **kwargs |
|
1122 | 1123 | ) |
|
1123 | 1124 | |
|
1124 | 1125 | if util.safehasattr(remote, b'_callstream'): |
|
1125 | 1126 | remote._localrepo = repo |
|
1126 | 1127 | elif util.safehasattr(remote, b'getbundle'): |
|
1127 | 1128 | extensions.wrapfunction(remote, b'getbundle', localgetbundle) |
|
1128 | 1129 | |
|
1129 | 1130 | return orig(repo, remote, *args, **kwargs) |
|
1130 | 1131 | |
|
1131 | 1132 | |
|
1132 | 1133 | def _fileprefetchhook(repo, revs, match): |
|
1133 | 1134 | if isenabled(repo): |
|
1134 | 1135 | allfiles = [] |
|
1135 | 1136 | for rev in revs: |
|
1136 | 1137 | if rev == nodemod.wdirrev or rev is None: |
|
1137 | 1138 | continue |
|
1138 | 1139 | ctx = repo[rev] |
|
1139 | 1140 | mf = ctx.manifest() |
|
1140 | 1141 | sparsematch = repo.maybesparsematch(ctx.rev()) |
|
1141 | 1142 | for path in ctx.walk(match): |
|
1142 | 1143 | if (not sparsematch or sparsematch(path)) and path in mf: |
|
1143 | 1144 | allfiles.append((path, hex(mf[path]))) |
|
1144 | 1145 | repo.fileservice.prefetch(allfiles) |
|
1145 | 1146 | |
|
1146 | 1147 | |
|
1147 | 1148 | @command( |
|
1148 | 1149 | b'debugremotefilelog', |
|
1149 | 1150 | [(b'd', b'decompress', None, _(b'decompress the filelog first')),], |
|
1150 | 1151 | _(b'hg debugremotefilelog <path>'), |
|
1151 | 1152 | norepo=True, |
|
1152 | 1153 | ) |
|
1153 | 1154 | def debugremotefilelog(ui, path, **opts): |
|
1154 | 1155 | return debugcommands.debugremotefilelog(ui, path, **opts) |
|
1155 | 1156 | |
|
1156 | 1157 | |
|
1157 | 1158 | @command( |
|
1158 | 1159 | b'verifyremotefilelog', |
|
1159 | 1160 | [(b'd', b'decompress', None, _(b'decompress the filelogs first')),], |
|
1160 | 1161 | _(b'hg verifyremotefilelogs <directory>'), |
|
1161 | 1162 | norepo=True, |
|
1162 | 1163 | ) |
|
1163 | 1164 | def verifyremotefilelog(ui, path, **opts): |
|
1164 | 1165 | return debugcommands.verifyremotefilelog(ui, path, **opts) |
|
1165 | 1166 | |
|
1166 | 1167 | |
|
1167 | 1168 | @command( |
|
1168 | 1169 | b'debugdatapack', |
|
1169 | 1170 | [ |
|
1170 | 1171 | (b'', b'long', None, _(b'print the long hashes')), |
|
1171 | 1172 | (b'', b'node', b'', _(b'dump the contents of node'), b'NODE'), |
|
1172 | 1173 | ], |
|
1173 | 1174 | _(b'hg debugdatapack <paths>'), |
|
1174 | 1175 | norepo=True, |
|
1175 | 1176 | ) |
|
1176 | 1177 | def debugdatapack(ui, *paths, **opts): |
|
1177 | 1178 | return debugcommands.debugdatapack(ui, *paths, **opts) |
|
1178 | 1179 | |
|
1179 | 1180 | |
|
1180 | 1181 | @command(b'debughistorypack', [], _(b'hg debughistorypack <path>'), norepo=True) |
|
1181 | 1182 | def debughistorypack(ui, path, **opts): |
|
1182 | 1183 | return debugcommands.debughistorypack(ui, path) |
|
1183 | 1184 | |
|
1184 | 1185 | |
|
1185 | 1186 | @command(b'debugkeepset', [], _(b'hg debugkeepset')) |
|
1186 | 1187 | def debugkeepset(ui, repo, **opts): |
|
1187 | 1188 | # The command is used to measure keepset computation time |
|
1188 | 1189 | def keyfn(fname, fnode): |
|
1189 | 1190 | return fileserverclient.getcachekey(repo.name, fname, hex(fnode)) |
|
1190 | 1191 | |
|
1191 | 1192 | repackmod.keepset(repo, keyfn) |
|
1192 | 1193 | return |
|
1193 | 1194 | |
|
1194 | 1195 | |
|
1195 | 1196 | @command(b'debugwaitonrepack', [], _(b'hg debugwaitonrepack')) |
|
1196 | 1197 | def debugwaitonrepack(ui, repo, **opts): |
|
1197 | 1198 | return debugcommands.debugwaitonrepack(repo) |
|
1198 | 1199 | |
|
1199 | 1200 | |
|
1200 | 1201 | @command(b'debugwaitonprefetch', [], _(b'hg debugwaitonprefetch')) |
|
1201 | 1202 | def debugwaitonprefetch(ui, repo, **opts): |
|
1202 | 1203 | return debugcommands.debugwaitonprefetch(repo) |
|
1203 | 1204 | |
|
1204 | 1205 | |
|
1205 | 1206 | def resolveprefetchopts(ui, opts): |
|
1206 | 1207 | if not opts.get(b'rev'): |
|
1207 | 1208 | revset = [b'.', b'draft()'] |
|
1208 | 1209 | |
|
1209 | 1210 | prefetchrevset = ui.config(b'remotefilelog', b'pullprefetch', None) |
|
1210 | 1211 | if prefetchrevset: |
|
1211 | 1212 | revset.append(b'(%s)' % prefetchrevset) |
|
1212 | 1213 | bgprefetchrevs = ui.config(b'remotefilelog', b'bgprefetchrevs', None) |
|
1213 | 1214 | if bgprefetchrevs: |
|
1214 | 1215 | revset.append(b'(%s)' % bgprefetchrevs) |
|
1215 | 1216 | revset = b'+'.join(revset) |
|
1216 | 1217 | |
|
1217 | 1218 | # update a revset with a date limit |
|
1218 | 1219 | revset = revdatelimit(ui, revset) |
|
1219 | 1220 | |
|
1220 | 1221 | opts[b'rev'] = [revset] |
|
1221 | 1222 | |
|
1222 | 1223 | if not opts.get(b'base'): |
|
1223 | 1224 | opts[b'base'] = None |
|
1224 | 1225 | |
|
1225 | 1226 | return opts |
|
1226 | 1227 | |
|
1227 | 1228 | |
|
1228 | 1229 | @command( |
|
1229 | 1230 | b'prefetch', |
|
1230 | 1231 | [ |
|
1231 | 1232 | (b'r', b'rev', [], _(b'prefetch the specified revisions'), _(b'REV')), |
|
1232 | 1233 | (b'', b'repack', False, _(b'run repack after prefetch')), |
|
1233 | 1234 | (b'b', b'base', b'', _(b"rev that is assumed to already be local")), |
|
1234 | 1235 | ] |
|
1235 | 1236 | + commands.walkopts, |
|
1236 | 1237 | _(b'hg prefetch [OPTIONS] [FILE...]'), |
|
1237 | 1238 | helpcategory=command.CATEGORY_MAINTENANCE, |
|
1238 | 1239 | ) |
|
1239 | 1240 | def prefetch(ui, repo, *pats, **opts): |
|
1240 | 1241 | """prefetch file revisions from the server |
|
1241 | 1242 | |
|
1242 | 1243 | Prefetchs file revisions for the specified revs and stores them in the |
|
1243 | 1244 | local remotefilelog cache. If no rev is specified, the default rev is |
|
1244 | 1245 | used which is the union of dot, draft, pullprefetch and bgprefetchrev. |
|
1245 | 1246 | File names or patterns can be used to limit which files are downloaded. |
|
1246 | 1247 | |
|
1247 | 1248 | Return 0 on success. |
|
1248 | 1249 | """ |
|
1249 | 1250 | opts = pycompat.byteskwargs(opts) |
|
1250 | 1251 | if not isenabled(repo): |
|
1251 | 1252 | raise error.Abort(_(b"repo is not shallow")) |
|
1252 | 1253 | |
|
1253 | 1254 | opts = resolveprefetchopts(ui, opts) |
|
1254 | 1255 | revs = scmutil.revrange(repo, opts.get(b'rev')) |
|
1255 | 1256 | repo.prefetch(revs, opts.get(b'base'), pats, opts) |
|
1256 | 1257 | |
|
1257 | 1258 | ensurestart = repo.ui.configbool(b'devel', b'remotefilelog.ensurestart') |
|
1258 | 1259 | |
|
1259 | 1260 | # Run repack in background |
|
1260 | 1261 | if opts.get(b'repack'): |
|
1261 | 1262 | repackmod.backgroundrepack( |
|
1262 | 1263 | repo, incremental=True, ensurestart=ensurestart |
|
1263 | 1264 | ) |
|
1264 | 1265 | |
|
1265 | 1266 | |
|
1266 | 1267 | @command( |
|
1267 | 1268 | b'repack', |
|
1268 | 1269 | [ |
|
1269 | 1270 | (b'', b'background', None, _(b'run in a background process'), None), |
|
1270 | 1271 | (b'', b'incremental', None, _(b'do an incremental repack'), None), |
|
1271 | 1272 | ( |
|
1272 | 1273 | b'', |
|
1273 | 1274 | b'packsonly', |
|
1274 | 1275 | None, |
|
1275 | 1276 | _(b'only repack packs (skip loose objects)'), |
|
1276 | 1277 | None, |
|
1277 | 1278 | ), |
|
1278 | 1279 | ], |
|
1279 | 1280 | _(b'hg repack [OPTIONS]'), |
|
1280 | 1281 | ) |
|
1281 | 1282 | def repack_(ui, repo, *pats, **opts): |
|
1282 | 1283 | if opts.get(r'background'): |
|
1283 | 1284 | ensurestart = repo.ui.configbool(b'devel', b'remotefilelog.ensurestart') |
|
1284 | 1285 | repackmod.backgroundrepack( |
|
1285 | 1286 | repo, |
|
1286 | 1287 | incremental=opts.get(r'incremental'), |
|
1287 | 1288 | packsonly=opts.get(r'packsonly', False), |
|
1288 | 1289 | ensurestart=ensurestart, |
|
1289 | 1290 | ) |
|
1290 | 1291 | return |
|
1291 | 1292 | |
|
1292 | 1293 | options = {b'packsonly': opts.get(r'packsonly')} |
|
1293 | 1294 | |
|
1294 | 1295 | try: |
|
1295 | 1296 | if opts.get(r'incremental'): |
|
1296 | 1297 | repackmod.incrementalrepack(repo, options=options) |
|
1297 | 1298 | else: |
|
1298 | 1299 | repackmod.fullrepack(repo, options=options) |
|
1299 | 1300 | except repackmod.RepackAlreadyRunning as ex: |
|
1300 | 1301 | # Don't propogate the exception if the repack is already in |
|
1301 | 1302 | # progress, since we want the command to exit 0. |
|
1302 | 1303 | repo.ui.warn(b'%s\n' % ex) |
@@ -1,912 +1,918 b'' | |||
|
1 | 1 | from __future__ import absolute_import |
|
2 | 2 | |
|
3 | 3 | import os |
|
4 | 4 | import time |
|
5 | 5 | |
|
6 | 6 | from mercurial.i18n import _ |
|
7 | 7 | from mercurial.node import ( |
|
8 | 8 | nullid, |
|
9 | 9 | short, |
|
10 | 10 | ) |
|
11 | 11 | from mercurial import ( |
|
12 | 12 | encoding, |
|
13 | 13 | error, |
|
14 | 14 | lock as lockmod, |
|
15 | 15 | mdiff, |
|
16 | 16 | policy, |
|
17 | 17 | pycompat, |
|
18 | 18 | scmutil, |
|
19 | 19 | util, |
|
20 | 20 | vfs, |
|
21 | 21 | ) |
|
22 | 22 | from mercurial.utils import procutil |
|
23 | 23 | from . import ( |
|
24 | 24 | constants, |
|
25 | 25 | contentstore, |
|
26 | 26 | datapack, |
|
27 | 27 | historypack, |
|
28 | 28 | metadatastore, |
|
29 | 29 | shallowutil, |
|
30 | 30 | ) |
|
31 | 31 | |
|
32 | 32 | osutil = policy.importmod(r'osutil') |
|
33 | 33 | |
|
34 | 34 | |
|
35 | 35 | class RepackAlreadyRunning(error.Abort): |
|
36 | 36 | pass |
|
37 | 37 | |
|
38 | 38 | |
|
39 | 39 | def backgroundrepack( |
|
40 | 40 | repo, incremental=True, packsonly=False, ensurestart=False |
|
41 | 41 | ): |
|
42 | 42 | cmd = [procutil.hgexecutable(), b'-R', repo.origroot, b'repack'] |
|
43 | 43 | msg = _(b"(running background repack)\n") |
|
44 | 44 | if incremental: |
|
45 | 45 | cmd.append(b'--incremental') |
|
46 | 46 | msg = _(b"(running background incremental repack)\n") |
|
47 | 47 | if packsonly: |
|
48 | 48 | cmd.append(b'--packsonly') |
|
49 | 49 | repo.ui.warn(msg) |
|
50 | 50 | # We know this command will find a binary, so don't block on it starting. |
|
51 | procutil.runbgcommand(cmd, encoding.environ, ensurestart=ensurestart) | |
|
51 | kwargs = {} | |
|
52 | if repo.ui.configbool(b'devel', b'remotefilelog.bg-wait'): | |
|
53 | kwargs['record_wait'] = repo.ui.atexit | |
|
54 | ||
|
55 | procutil.runbgcommand( | |
|
56 | cmd, encoding.environ, ensurestart=ensurestart, **kwargs | |
|
57 | ) | |
|
52 | 58 | |
|
53 | 59 | |
|
54 | 60 | def fullrepack(repo, options=None): |
|
55 | 61 | """If ``packsonly`` is True, stores creating only loose objects are skipped. |
|
56 | 62 | """ |
|
57 | 63 | if util.safehasattr(repo, 'shareddatastores'): |
|
58 | 64 | datasource = contentstore.unioncontentstore(*repo.shareddatastores) |
|
59 | 65 | historysource = metadatastore.unionmetadatastore( |
|
60 | 66 | *repo.sharedhistorystores, allowincomplete=True |
|
61 | 67 | ) |
|
62 | 68 | |
|
63 | 69 | packpath = shallowutil.getcachepackpath( |
|
64 | 70 | repo, constants.FILEPACK_CATEGORY |
|
65 | 71 | ) |
|
66 | 72 | _runrepack( |
|
67 | 73 | repo, |
|
68 | 74 | datasource, |
|
69 | 75 | historysource, |
|
70 | 76 | packpath, |
|
71 | 77 | constants.FILEPACK_CATEGORY, |
|
72 | 78 | options=options, |
|
73 | 79 | ) |
|
74 | 80 | |
|
75 | 81 | if util.safehasattr(repo.manifestlog, 'datastore'): |
|
76 | 82 | localdata, shareddata = _getmanifeststores(repo) |
|
77 | 83 | lpackpath, ldstores, lhstores = localdata |
|
78 | 84 | spackpath, sdstores, shstores = shareddata |
|
79 | 85 | |
|
80 | 86 | # Repack the shared manifest store |
|
81 | 87 | datasource = contentstore.unioncontentstore(*sdstores) |
|
82 | 88 | historysource = metadatastore.unionmetadatastore( |
|
83 | 89 | *shstores, allowincomplete=True |
|
84 | 90 | ) |
|
85 | 91 | _runrepack( |
|
86 | 92 | repo, |
|
87 | 93 | datasource, |
|
88 | 94 | historysource, |
|
89 | 95 | spackpath, |
|
90 | 96 | constants.TREEPACK_CATEGORY, |
|
91 | 97 | options=options, |
|
92 | 98 | ) |
|
93 | 99 | |
|
94 | 100 | # Repack the local manifest store |
|
95 | 101 | datasource = contentstore.unioncontentstore( |
|
96 | 102 | *ldstores, allowincomplete=True |
|
97 | 103 | ) |
|
98 | 104 | historysource = metadatastore.unionmetadatastore( |
|
99 | 105 | *lhstores, allowincomplete=True |
|
100 | 106 | ) |
|
101 | 107 | _runrepack( |
|
102 | 108 | repo, |
|
103 | 109 | datasource, |
|
104 | 110 | historysource, |
|
105 | 111 | lpackpath, |
|
106 | 112 | constants.TREEPACK_CATEGORY, |
|
107 | 113 | options=options, |
|
108 | 114 | ) |
|
109 | 115 | |
|
110 | 116 | |
|
111 | 117 | def incrementalrepack(repo, options=None): |
|
112 | 118 | """This repacks the repo by looking at the distribution of pack files in the |
|
113 | 119 | repo and performing the most minimal repack to keep the repo in good shape. |
|
114 | 120 | """ |
|
115 | 121 | if util.safehasattr(repo, 'shareddatastores'): |
|
116 | 122 | packpath = shallowutil.getcachepackpath( |
|
117 | 123 | repo, constants.FILEPACK_CATEGORY |
|
118 | 124 | ) |
|
119 | 125 | _incrementalrepack( |
|
120 | 126 | repo, |
|
121 | 127 | repo.shareddatastores, |
|
122 | 128 | repo.sharedhistorystores, |
|
123 | 129 | packpath, |
|
124 | 130 | constants.FILEPACK_CATEGORY, |
|
125 | 131 | options=options, |
|
126 | 132 | ) |
|
127 | 133 | |
|
128 | 134 | if util.safehasattr(repo.manifestlog, 'datastore'): |
|
129 | 135 | localdata, shareddata = _getmanifeststores(repo) |
|
130 | 136 | lpackpath, ldstores, lhstores = localdata |
|
131 | 137 | spackpath, sdstores, shstores = shareddata |
|
132 | 138 | |
|
133 | 139 | # Repack the shared manifest store |
|
134 | 140 | _incrementalrepack( |
|
135 | 141 | repo, |
|
136 | 142 | sdstores, |
|
137 | 143 | shstores, |
|
138 | 144 | spackpath, |
|
139 | 145 | constants.TREEPACK_CATEGORY, |
|
140 | 146 | options=options, |
|
141 | 147 | ) |
|
142 | 148 | |
|
143 | 149 | # Repack the local manifest store |
|
144 | 150 | _incrementalrepack( |
|
145 | 151 | repo, |
|
146 | 152 | ldstores, |
|
147 | 153 | lhstores, |
|
148 | 154 | lpackpath, |
|
149 | 155 | constants.TREEPACK_CATEGORY, |
|
150 | 156 | allowincompletedata=True, |
|
151 | 157 | options=options, |
|
152 | 158 | ) |
|
153 | 159 | |
|
154 | 160 | |
|
155 | 161 | def _getmanifeststores(repo): |
|
156 | 162 | shareddatastores = repo.manifestlog.shareddatastores |
|
157 | 163 | localdatastores = repo.manifestlog.localdatastores |
|
158 | 164 | sharedhistorystores = repo.manifestlog.sharedhistorystores |
|
159 | 165 | localhistorystores = repo.manifestlog.localhistorystores |
|
160 | 166 | |
|
161 | 167 | sharedpackpath = shallowutil.getcachepackpath( |
|
162 | 168 | repo, constants.TREEPACK_CATEGORY |
|
163 | 169 | ) |
|
164 | 170 | localpackpath = shallowutil.getlocalpackpath( |
|
165 | 171 | repo.svfs.vfs.base, constants.TREEPACK_CATEGORY |
|
166 | 172 | ) |
|
167 | 173 | |
|
168 | 174 | return ( |
|
169 | 175 | (localpackpath, localdatastores, localhistorystores), |
|
170 | 176 | (sharedpackpath, shareddatastores, sharedhistorystores), |
|
171 | 177 | ) |
|
172 | 178 | |
|
173 | 179 | |
|
174 | 180 | def _topacks(packpath, files, constructor): |
|
175 | 181 | paths = list(os.path.join(packpath, p) for p in files) |
|
176 | 182 | packs = list(constructor(p) for p in paths) |
|
177 | 183 | return packs |
|
178 | 184 | |
|
179 | 185 | |
|
180 | 186 | def _deletebigpacks(repo, folder, files): |
|
181 | 187 | """Deletes packfiles that are bigger than ``packs.maxpacksize``. |
|
182 | 188 | |
|
183 | 189 | Returns ``files` with the removed files omitted.""" |
|
184 | 190 | maxsize = repo.ui.configbytes(b"packs", b"maxpacksize") |
|
185 | 191 | if maxsize <= 0: |
|
186 | 192 | return files |
|
187 | 193 | |
|
188 | 194 | # This only considers datapacks today, but we could broaden it to include |
|
189 | 195 | # historypacks. |
|
190 | 196 | VALIDEXTS = [b".datapack", b".dataidx"] |
|
191 | 197 | |
|
192 | 198 | # Either an oversize index or datapack will trigger cleanup of the whole |
|
193 | 199 | # pack: |
|
194 | 200 | oversized = { |
|
195 | 201 | os.path.splitext(path)[0] |
|
196 | 202 | for path, ftype, stat in files |
|
197 | 203 | if (stat.st_size > maxsize and (os.path.splitext(path)[1] in VALIDEXTS)) |
|
198 | 204 | } |
|
199 | 205 | |
|
200 | 206 | for rootfname in oversized: |
|
201 | 207 | rootpath = os.path.join(folder, rootfname) |
|
202 | 208 | for ext in VALIDEXTS: |
|
203 | 209 | path = rootpath + ext |
|
204 | 210 | repo.ui.debug( |
|
205 | 211 | b'removing oversize packfile %s (%s)\n' |
|
206 | 212 | % (path, util.bytecount(os.stat(path).st_size)) |
|
207 | 213 | ) |
|
208 | 214 | os.unlink(path) |
|
209 | 215 | return [row for row in files if os.path.basename(row[0]) not in oversized] |
|
210 | 216 | |
|
211 | 217 | |
|
212 | 218 | def _incrementalrepack( |
|
213 | 219 | repo, |
|
214 | 220 | datastore, |
|
215 | 221 | historystore, |
|
216 | 222 | packpath, |
|
217 | 223 | category, |
|
218 | 224 | allowincompletedata=False, |
|
219 | 225 | options=None, |
|
220 | 226 | ): |
|
221 | 227 | shallowutil.mkstickygroupdir(repo.ui, packpath) |
|
222 | 228 | |
|
223 | 229 | files = osutil.listdir(packpath, stat=True) |
|
224 | 230 | files = _deletebigpacks(repo, packpath, files) |
|
225 | 231 | datapacks = _topacks( |
|
226 | 232 | packpath, _computeincrementaldatapack(repo.ui, files), datapack.datapack |
|
227 | 233 | ) |
|
228 | 234 | datapacks.extend( |
|
229 | 235 | s for s in datastore if not isinstance(s, datapack.datapackstore) |
|
230 | 236 | ) |
|
231 | 237 | |
|
232 | 238 | historypacks = _topacks( |
|
233 | 239 | packpath, |
|
234 | 240 | _computeincrementalhistorypack(repo.ui, files), |
|
235 | 241 | historypack.historypack, |
|
236 | 242 | ) |
|
237 | 243 | historypacks.extend( |
|
238 | 244 | s |
|
239 | 245 | for s in historystore |
|
240 | 246 | if not isinstance(s, historypack.historypackstore) |
|
241 | 247 | ) |
|
242 | 248 | |
|
243 | 249 | # ``allhistory{files,packs}`` contains all known history packs, even ones we |
|
244 | 250 | # don't plan to repack. They are used during the datapack repack to ensure |
|
245 | 251 | # good ordering of nodes. |
|
246 | 252 | allhistoryfiles = _allpackfileswithsuffix( |
|
247 | 253 | files, historypack.PACKSUFFIX, historypack.INDEXSUFFIX |
|
248 | 254 | ) |
|
249 | 255 | allhistorypacks = _topacks( |
|
250 | 256 | packpath, |
|
251 | 257 | (f for f, mode, stat in allhistoryfiles), |
|
252 | 258 | historypack.historypack, |
|
253 | 259 | ) |
|
254 | 260 | allhistorypacks.extend( |
|
255 | 261 | s |
|
256 | 262 | for s in historystore |
|
257 | 263 | if not isinstance(s, historypack.historypackstore) |
|
258 | 264 | ) |
|
259 | 265 | _runrepack( |
|
260 | 266 | repo, |
|
261 | 267 | contentstore.unioncontentstore( |
|
262 | 268 | *datapacks, allowincomplete=allowincompletedata |
|
263 | 269 | ), |
|
264 | 270 | metadatastore.unionmetadatastore(*historypacks, allowincomplete=True), |
|
265 | 271 | packpath, |
|
266 | 272 | category, |
|
267 | 273 | fullhistory=metadatastore.unionmetadatastore( |
|
268 | 274 | *allhistorypacks, allowincomplete=True |
|
269 | 275 | ), |
|
270 | 276 | options=options, |
|
271 | 277 | ) |
|
272 | 278 | |
|
273 | 279 | |
|
274 | 280 | def _computeincrementaldatapack(ui, files): |
|
275 | 281 | opts = { |
|
276 | 282 | b'gencountlimit': ui.configint(b'remotefilelog', b'data.gencountlimit'), |
|
277 | 283 | b'generations': ui.configlist(b'remotefilelog', b'data.generations'), |
|
278 | 284 | b'maxrepackpacks': ui.configint( |
|
279 | 285 | b'remotefilelog', b'data.maxrepackpacks' |
|
280 | 286 | ), |
|
281 | 287 | b'repackmaxpacksize': ui.configbytes( |
|
282 | 288 | b'remotefilelog', b'data.repackmaxpacksize' |
|
283 | 289 | ), |
|
284 | 290 | b'repacksizelimit': ui.configbytes( |
|
285 | 291 | b'remotefilelog', b'data.repacksizelimit' |
|
286 | 292 | ), |
|
287 | 293 | } |
|
288 | 294 | |
|
289 | 295 | packfiles = _allpackfileswithsuffix( |
|
290 | 296 | files, datapack.PACKSUFFIX, datapack.INDEXSUFFIX |
|
291 | 297 | ) |
|
292 | 298 | return _computeincrementalpack(packfiles, opts) |
|
293 | 299 | |
|
294 | 300 | |
|
295 | 301 | def _computeincrementalhistorypack(ui, files): |
|
296 | 302 | opts = { |
|
297 | 303 | b'gencountlimit': ui.configint( |
|
298 | 304 | b'remotefilelog', b'history.gencountlimit' |
|
299 | 305 | ), |
|
300 | 306 | b'generations': ui.configlist( |
|
301 | 307 | b'remotefilelog', b'history.generations', [b'100MB'] |
|
302 | 308 | ), |
|
303 | 309 | b'maxrepackpacks': ui.configint( |
|
304 | 310 | b'remotefilelog', b'history.maxrepackpacks' |
|
305 | 311 | ), |
|
306 | 312 | b'repackmaxpacksize': ui.configbytes( |
|
307 | 313 | b'remotefilelog', b'history.repackmaxpacksize', b'400MB' |
|
308 | 314 | ), |
|
309 | 315 | b'repacksizelimit': ui.configbytes( |
|
310 | 316 | b'remotefilelog', b'history.repacksizelimit' |
|
311 | 317 | ), |
|
312 | 318 | } |
|
313 | 319 | |
|
314 | 320 | packfiles = _allpackfileswithsuffix( |
|
315 | 321 | files, historypack.PACKSUFFIX, historypack.INDEXSUFFIX |
|
316 | 322 | ) |
|
317 | 323 | return _computeincrementalpack(packfiles, opts) |
|
318 | 324 | |
|
319 | 325 | |
|
320 | 326 | def _allpackfileswithsuffix(files, packsuffix, indexsuffix): |
|
321 | 327 | result = [] |
|
322 | 328 | fileset = set(fn for fn, mode, stat in files) |
|
323 | 329 | for filename, mode, stat in files: |
|
324 | 330 | if not filename.endswith(packsuffix): |
|
325 | 331 | continue |
|
326 | 332 | |
|
327 | 333 | prefix = filename[: -len(packsuffix)] |
|
328 | 334 | |
|
329 | 335 | # Don't process a pack if it doesn't have an index. |
|
330 | 336 | if (prefix + indexsuffix) not in fileset: |
|
331 | 337 | continue |
|
332 | 338 | result.append((prefix, mode, stat)) |
|
333 | 339 | |
|
334 | 340 | return result |
|
335 | 341 | |
|
336 | 342 | |
|
337 | 343 | def _computeincrementalpack(files, opts): |
|
338 | 344 | """Given a set of pack files along with the configuration options, this |
|
339 | 345 | function computes the list of files that should be packed as part of an |
|
340 | 346 | incremental repack. |
|
341 | 347 | |
|
342 | 348 | It tries to strike a balance between keeping incremental repacks cheap (i.e. |
|
343 | 349 | packing small things when possible, and rolling the packs up to the big ones |
|
344 | 350 | over time). |
|
345 | 351 | """ |
|
346 | 352 | |
|
347 | 353 | limits = list( |
|
348 | 354 | sorted((util.sizetoint(s) for s in opts[b'generations']), reverse=True) |
|
349 | 355 | ) |
|
350 | 356 | limits.append(0) |
|
351 | 357 | |
|
352 | 358 | # Group the packs by generation (i.e. by size) |
|
353 | 359 | generations = [] |
|
354 | 360 | for i in pycompat.xrange(len(limits)): |
|
355 | 361 | generations.append([]) |
|
356 | 362 | |
|
357 | 363 | sizes = {} |
|
358 | 364 | for prefix, mode, stat in files: |
|
359 | 365 | size = stat.st_size |
|
360 | 366 | if size > opts[b'repackmaxpacksize']: |
|
361 | 367 | continue |
|
362 | 368 | |
|
363 | 369 | sizes[prefix] = size |
|
364 | 370 | for i, limit in enumerate(limits): |
|
365 | 371 | if size > limit: |
|
366 | 372 | generations[i].append(prefix) |
|
367 | 373 | break |
|
368 | 374 | |
|
369 | 375 | # Steps for picking what packs to repack: |
|
370 | 376 | # 1. Pick the largest generation with > gencountlimit pack files. |
|
371 | 377 | # 2. Take the smallest three packs. |
|
372 | 378 | # 3. While total-size-of-packs < repacksizelimit: add another pack |
|
373 | 379 | |
|
374 | 380 | # Find the largest generation with more than gencountlimit packs |
|
375 | 381 | genpacks = [] |
|
376 | 382 | for i, limit in enumerate(limits): |
|
377 | 383 | if len(generations[i]) > opts[b'gencountlimit']: |
|
378 | 384 | # Sort to be smallest last, for easy popping later |
|
379 | 385 | genpacks.extend( |
|
380 | 386 | sorted(generations[i], reverse=True, key=lambda x: sizes[x]) |
|
381 | 387 | ) |
|
382 | 388 | break |
|
383 | 389 | |
|
384 | 390 | # Take as many packs from the generation as we can |
|
385 | 391 | chosenpacks = genpacks[-3:] |
|
386 | 392 | genpacks = genpacks[:-3] |
|
387 | 393 | repacksize = sum(sizes[n] for n in chosenpacks) |
|
388 | 394 | while ( |
|
389 | 395 | repacksize < opts[b'repacksizelimit'] |
|
390 | 396 | and genpacks |
|
391 | 397 | and len(chosenpacks) < opts[b'maxrepackpacks'] |
|
392 | 398 | ): |
|
393 | 399 | chosenpacks.append(genpacks.pop()) |
|
394 | 400 | repacksize += sizes[chosenpacks[-1]] |
|
395 | 401 | |
|
396 | 402 | return chosenpacks |
|
397 | 403 | |
|
398 | 404 | |
|
399 | 405 | def _runrepack( |
|
400 | 406 | repo, data, history, packpath, category, fullhistory=None, options=None |
|
401 | 407 | ): |
|
402 | 408 | shallowutil.mkstickygroupdir(repo.ui, packpath) |
|
403 | 409 | |
|
404 | 410 | def isold(repo, filename, node): |
|
405 | 411 | """Check if the file node is older than a limit. |
|
406 | 412 | Unless a limit is specified in the config the default limit is taken. |
|
407 | 413 | """ |
|
408 | 414 | filectx = repo.filectx(filename, fileid=node) |
|
409 | 415 | filetime = repo[filectx.linkrev()].date() |
|
410 | 416 | |
|
411 | 417 | ttl = repo.ui.configint(b'remotefilelog', b'nodettl') |
|
412 | 418 | |
|
413 | 419 | limit = time.time() - ttl |
|
414 | 420 | return filetime[0] < limit |
|
415 | 421 | |
|
416 | 422 | garbagecollect = repo.ui.configbool(b'remotefilelog', b'gcrepack') |
|
417 | 423 | if not fullhistory: |
|
418 | 424 | fullhistory = history |
|
419 | 425 | packer = repacker( |
|
420 | 426 | repo, |
|
421 | 427 | data, |
|
422 | 428 | history, |
|
423 | 429 | fullhistory, |
|
424 | 430 | category, |
|
425 | 431 | gc=garbagecollect, |
|
426 | 432 | isold=isold, |
|
427 | 433 | options=options, |
|
428 | 434 | ) |
|
429 | 435 | |
|
430 | 436 | with datapack.mutabledatapack(repo.ui, packpath) as dpack: |
|
431 | 437 | with historypack.mutablehistorypack(repo.ui, packpath) as hpack: |
|
432 | 438 | try: |
|
433 | 439 | packer.run(dpack, hpack) |
|
434 | 440 | except error.LockHeld: |
|
435 | 441 | raise RepackAlreadyRunning( |
|
436 | 442 | _( |
|
437 | 443 | b"skipping repack - another repack " |
|
438 | 444 | b"is already running" |
|
439 | 445 | ) |
|
440 | 446 | ) |
|
441 | 447 | |
|
442 | 448 | |
|
443 | 449 | def keepset(repo, keyfn, lastkeepkeys=None): |
|
444 | 450 | """Computes a keepset which is not garbage collected. |
|
445 | 451 | 'keyfn' is a function that maps filename, node to a unique key. |
|
446 | 452 | 'lastkeepkeys' is an optional argument and if provided the keepset |
|
447 | 453 | function updates lastkeepkeys with more keys and returns the result. |
|
448 | 454 | """ |
|
449 | 455 | if not lastkeepkeys: |
|
450 | 456 | keepkeys = set() |
|
451 | 457 | else: |
|
452 | 458 | keepkeys = lastkeepkeys |
|
453 | 459 | |
|
454 | 460 | # We want to keep: |
|
455 | 461 | # 1. Working copy parent |
|
456 | 462 | # 2. Draft commits |
|
457 | 463 | # 3. Parents of draft commits |
|
458 | 464 | # 4. Pullprefetch and bgprefetchrevs revsets if specified |
|
459 | 465 | revs = [b'.', b'draft()', b'parents(draft())'] |
|
460 | 466 | prefetchrevs = repo.ui.config(b'remotefilelog', b'pullprefetch', None) |
|
461 | 467 | if prefetchrevs: |
|
462 | 468 | revs.append(b'(%s)' % prefetchrevs) |
|
463 | 469 | prefetchrevs = repo.ui.config(b'remotefilelog', b'bgprefetchrevs', None) |
|
464 | 470 | if prefetchrevs: |
|
465 | 471 | revs.append(b'(%s)' % prefetchrevs) |
|
466 | 472 | revs = b'+'.join(revs) |
|
467 | 473 | |
|
468 | 474 | revs = [b'sort((%s), "topo")' % revs] |
|
469 | 475 | keep = scmutil.revrange(repo, revs) |
|
470 | 476 | |
|
471 | 477 | processed = set() |
|
472 | 478 | lastmanifest = None |
|
473 | 479 | |
|
474 | 480 | # process the commits in toposorted order starting from the oldest |
|
475 | 481 | for r in reversed(keep._list): |
|
476 | 482 | if repo[r].p1().rev() in processed: |
|
477 | 483 | # if the direct parent has already been processed |
|
478 | 484 | # then we only need to process the delta |
|
479 | 485 | m = repo[r].manifestctx().readdelta() |
|
480 | 486 | else: |
|
481 | 487 | # otherwise take the manifest and diff it |
|
482 | 488 | # with the previous manifest if one exists |
|
483 | 489 | if lastmanifest: |
|
484 | 490 | m = repo[r].manifest().diff(lastmanifest) |
|
485 | 491 | else: |
|
486 | 492 | m = repo[r].manifest() |
|
487 | 493 | lastmanifest = repo[r].manifest() |
|
488 | 494 | processed.add(r) |
|
489 | 495 | |
|
490 | 496 | # populate keepkeys with keys from the current manifest |
|
491 | 497 | if type(m) is dict: |
|
492 | 498 | # m is a result of diff of two manifests and is a dictionary that |
|
493 | 499 | # maps filename to ((newnode, newflag), (oldnode, oldflag)) tuple |
|
494 | 500 | for filename, diff in pycompat.iteritems(m): |
|
495 | 501 | if diff[0][0] is not None: |
|
496 | 502 | keepkeys.add(keyfn(filename, diff[0][0])) |
|
497 | 503 | else: |
|
498 | 504 | # m is a manifest object |
|
499 | 505 | for filename, filenode in pycompat.iteritems(m): |
|
500 | 506 | keepkeys.add(keyfn(filename, filenode)) |
|
501 | 507 | |
|
502 | 508 | return keepkeys |
|
503 | 509 | |
|
504 | 510 | |
|
505 | 511 | class repacker(object): |
|
506 | 512 | """Class for orchestrating the repack of data and history information into a |
|
507 | 513 | new format. |
|
508 | 514 | """ |
|
509 | 515 | |
|
510 | 516 | def __init__( |
|
511 | 517 | self, |
|
512 | 518 | repo, |
|
513 | 519 | data, |
|
514 | 520 | history, |
|
515 | 521 | fullhistory, |
|
516 | 522 | category, |
|
517 | 523 | gc=False, |
|
518 | 524 | isold=None, |
|
519 | 525 | options=None, |
|
520 | 526 | ): |
|
521 | 527 | self.repo = repo |
|
522 | 528 | self.data = data |
|
523 | 529 | self.history = history |
|
524 | 530 | self.fullhistory = fullhistory |
|
525 | 531 | self.unit = constants.getunits(category) |
|
526 | 532 | self.garbagecollect = gc |
|
527 | 533 | self.options = options |
|
528 | 534 | if self.garbagecollect: |
|
529 | 535 | if not isold: |
|
530 | 536 | raise ValueError(b"Function 'isold' is not properly specified") |
|
531 | 537 | # use (filename, node) tuple as a keepset key |
|
532 | 538 | self.keepkeys = keepset(repo, lambda f, n: (f, n)) |
|
533 | 539 | self.isold = isold |
|
534 | 540 | |
|
535 | 541 | def run(self, targetdata, targethistory): |
|
536 | 542 | ledger = repackledger() |
|
537 | 543 | |
|
538 | 544 | with lockmod.lock( |
|
539 | 545 | repacklockvfs(self.repo), b"repacklock", desc=None, timeout=0 |
|
540 | 546 | ): |
|
541 | 547 | self.repo.hook(b'prerepack') |
|
542 | 548 | |
|
543 | 549 | # Populate ledger from source |
|
544 | 550 | self.data.markledger(ledger, options=self.options) |
|
545 | 551 | self.history.markledger(ledger, options=self.options) |
|
546 | 552 | |
|
547 | 553 | # Run repack |
|
548 | 554 | self.repackdata(ledger, targetdata) |
|
549 | 555 | self.repackhistory(ledger, targethistory) |
|
550 | 556 | |
|
551 | 557 | # Call cleanup on each source |
|
552 | 558 | for source in ledger.sources: |
|
553 | 559 | source.cleanup(ledger) |
|
554 | 560 | |
|
555 | 561 | def _chainorphans(self, ui, filename, nodes, orphans, deltabases): |
|
556 | 562 | """Reorderes ``orphans`` into a single chain inside ``nodes`` and |
|
557 | 563 | ``deltabases``. |
|
558 | 564 | |
|
559 | 565 | We often have orphan entries (nodes without a base that aren't |
|
560 | 566 | referenced by other nodes -- i.e., part of a chain) due to gaps in |
|
561 | 567 | history. Rather than store them as individual fulltexts, we prefer to |
|
562 | 568 | insert them as one chain sorted by size. |
|
563 | 569 | """ |
|
564 | 570 | if not orphans: |
|
565 | 571 | return nodes |
|
566 | 572 | |
|
567 | 573 | def getsize(node, default=0): |
|
568 | 574 | meta = self.data.getmeta(filename, node) |
|
569 | 575 | if constants.METAKEYSIZE in meta: |
|
570 | 576 | return meta[constants.METAKEYSIZE] |
|
571 | 577 | else: |
|
572 | 578 | return default |
|
573 | 579 | |
|
574 | 580 | # Sort orphans by size; biggest first is preferred, since it's more |
|
575 | 581 | # likely to be the newest version assuming files grow over time. |
|
576 | 582 | # (Sort by node first to ensure the sort is stable.) |
|
577 | 583 | orphans = sorted(orphans) |
|
578 | 584 | orphans = list(sorted(orphans, key=getsize, reverse=True)) |
|
579 | 585 | if ui.debugflag: |
|
580 | 586 | ui.debug( |
|
581 | 587 | b"%s: orphan chain: %s\n" |
|
582 | 588 | % (filename, b", ".join([short(s) for s in orphans])) |
|
583 | 589 | ) |
|
584 | 590 | |
|
585 | 591 | # Create one contiguous chain and reassign deltabases. |
|
586 | 592 | for i, node in enumerate(orphans): |
|
587 | 593 | if i == 0: |
|
588 | 594 | deltabases[node] = (nullid, 0) |
|
589 | 595 | else: |
|
590 | 596 | parent = orphans[i - 1] |
|
591 | 597 | deltabases[node] = (parent, deltabases[parent][1] + 1) |
|
592 | 598 | nodes = [n for n in nodes if n not in orphans] |
|
593 | 599 | nodes += orphans |
|
594 | 600 | return nodes |
|
595 | 601 | |
|
596 | 602 | def repackdata(self, ledger, target): |
|
597 | 603 | ui = self.repo.ui |
|
598 | 604 | maxchainlen = ui.configint(b'packs', b'maxchainlen', 1000) |
|
599 | 605 | |
|
600 | 606 | byfile = {} |
|
601 | 607 | for entry in pycompat.itervalues(ledger.entries): |
|
602 | 608 | if entry.datasource: |
|
603 | 609 | byfile.setdefault(entry.filename, {})[entry.node] = entry |
|
604 | 610 | |
|
605 | 611 | count = 0 |
|
606 | 612 | repackprogress = ui.makeprogress( |
|
607 | 613 | _(b"repacking data"), unit=self.unit, total=len(byfile) |
|
608 | 614 | ) |
|
609 | 615 | for filename, entries in sorted(pycompat.iteritems(byfile)): |
|
610 | 616 | repackprogress.update(count) |
|
611 | 617 | |
|
612 | 618 | ancestors = {} |
|
613 | 619 | nodes = list(node for node in entries) |
|
614 | 620 | nohistory = [] |
|
615 | 621 | buildprogress = ui.makeprogress( |
|
616 | 622 | _(b"building history"), unit=b'nodes', total=len(nodes) |
|
617 | 623 | ) |
|
618 | 624 | for i, node in enumerate(nodes): |
|
619 | 625 | if node in ancestors: |
|
620 | 626 | continue |
|
621 | 627 | buildprogress.update(i) |
|
622 | 628 | try: |
|
623 | 629 | ancestors.update( |
|
624 | 630 | self.fullhistory.getancestors( |
|
625 | 631 | filename, node, known=ancestors |
|
626 | 632 | ) |
|
627 | 633 | ) |
|
628 | 634 | except KeyError: |
|
629 | 635 | # Since we're packing data entries, we may not have the |
|
630 | 636 | # corresponding history entries for them. It's not a big |
|
631 | 637 | # deal, but the entries won't be delta'd perfectly. |
|
632 | 638 | nohistory.append(node) |
|
633 | 639 | buildprogress.complete() |
|
634 | 640 | |
|
635 | 641 | # Order the nodes children first, so we can produce reverse deltas |
|
636 | 642 | orderednodes = list(reversed(self._toposort(ancestors))) |
|
637 | 643 | if len(nohistory) > 0: |
|
638 | 644 | ui.debug( |
|
639 | 645 | b'repackdata: %d nodes without history\n' % len(nohistory) |
|
640 | 646 | ) |
|
641 | 647 | orderednodes.extend(sorted(nohistory)) |
|
642 | 648 | |
|
643 | 649 | # Filter orderednodes to just the nodes we want to serialize (it |
|
644 | 650 | # currently also has the edge nodes' ancestors). |
|
645 | 651 | orderednodes = list( |
|
646 | 652 | filter(lambda node: node in nodes, orderednodes) |
|
647 | 653 | ) |
|
648 | 654 | |
|
649 | 655 | # Garbage collect old nodes: |
|
650 | 656 | if self.garbagecollect: |
|
651 | 657 | neworderednodes = [] |
|
652 | 658 | for node in orderednodes: |
|
653 | 659 | # If the node is old and is not in the keepset, we skip it, |
|
654 | 660 | # and mark as garbage collected |
|
655 | 661 | if (filename, node) not in self.keepkeys and self.isold( |
|
656 | 662 | self.repo, filename, node |
|
657 | 663 | ): |
|
658 | 664 | entries[node].gced = True |
|
659 | 665 | continue |
|
660 | 666 | neworderednodes.append(node) |
|
661 | 667 | orderednodes = neworderednodes |
|
662 | 668 | |
|
663 | 669 | # Compute delta bases for nodes: |
|
664 | 670 | deltabases = {} |
|
665 | 671 | nobase = set() |
|
666 | 672 | referenced = set() |
|
667 | 673 | nodes = set(nodes) |
|
668 | 674 | processprogress = ui.makeprogress( |
|
669 | 675 | _(b"processing nodes"), unit=b'nodes', total=len(orderednodes) |
|
670 | 676 | ) |
|
671 | 677 | for i, node in enumerate(orderednodes): |
|
672 | 678 | processprogress.update(i) |
|
673 | 679 | # Find delta base |
|
674 | 680 | # TODO: allow delta'ing against most recent descendant instead |
|
675 | 681 | # of immediate child |
|
676 | 682 | deltatuple = deltabases.get(node, None) |
|
677 | 683 | if deltatuple is None: |
|
678 | 684 | deltabase, chainlen = nullid, 0 |
|
679 | 685 | deltabases[node] = (nullid, 0) |
|
680 | 686 | nobase.add(node) |
|
681 | 687 | else: |
|
682 | 688 | deltabase, chainlen = deltatuple |
|
683 | 689 | referenced.add(deltabase) |
|
684 | 690 | |
|
685 | 691 | # Use available ancestor information to inform our delta choices |
|
686 | 692 | ancestorinfo = ancestors.get(node) |
|
687 | 693 | if ancestorinfo: |
|
688 | 694 | p1, p2, linknode, copyfrom = ancestorinfo |
|
689 | 695 | |
|
690 | 696 | # The presence of copyfrom means we're at a point where the |
|
691 | 697 | # file was copied from elsewhere. So don't attempt to do any |
|
692 | 698 | # deltas with the other file. |
|
693 | 699 | if copyfrom: |
|
694 | 700 | p1 = nullid |
|
695 | 701 | |
|
696 | 702 | if chainlen < maxchainlen: |
|
697 | 703 | # Record this child as the delta base for its parents. |
|
698 | 704 | # This may be non optimal, since the parents may have |
|
699 | 705 | # many children, and this will only choose the last one. |
|
700 | 706 | # TODO: record all children and try all deltas to find |
|
701 | 707 | # best |
|
702 | 708 | if p1 != nullid: |
|
703 | 709 | deltabases[p1] = (node, chainlen + 1) |
|
704 | 710 | if p2 != nullid: |
|
705 | 711 | deltabases[p2] = (node, chainlen + 1) |
|
706 | 712 | |
|
707 | 713 | # experimental config: repack.chainorphansbysize |
|
708 | 714 | if ui.configbool(b'repack', b'chainorphansbysize'): |
|
709 | 715 | orphans = nobase - referenced |
|
710 | 716 | orderednodes = self._chainorphans( |
|
711 | 717 | ui, filename, orderednodes, orphans, deltabases |
|
712 | 718 | ) |
|
713 | 719 | |
|
714 | 720 | # Compute deltas and write to the pack |
|
715 | 721 | for i, node in enumerate(orderednodes): |
|
716 | 722 | deltabase, chainlen = deltabases[node] |
|
717 | 723 | # Compute delta |
|
718 | 724 | # TODO: Optimize the deltachain fetching. Since we're |
|
719 | 725 | # iterating over the different version of the file, we may |
|
720 | 726 | # be fetching the same deltachain over and over again. |
|
721 | 727 | if deltabase != nullid: |
|
722 | 728 | deltaentry = self.data.getdelta(filename, node) |
|
723 | 729 | delta, deltabasename, origdeltabase, meta = deltaentry |
|
724 | 730 | size = meta.get(constants.METAKEYSIZE) |
|
725 | 731 | if ( |
|
726 | 732 | deltabasename != filename |
|
727 | 733 | or origdeltabase != deltabase |
|
728 | 734 | or size is None |
|
729 | 735 | ): |
|
730 | 736 | deltabasetext = self.data.get(filename, deltabase) |
|
731 | 737 | original = self.data.get(filename, node) |
|
732 | 738 | size = len(original) |
|
733 | 739 | delta = mdiff.textdiff(deltabasetext, original) |
|
734 | 740 | else: |
|
735 | 741 | delta = self.data.get(filename, node) |
|
736 | 742 | size = len(delta) |
|
737 | 743 | meta = self.data.getmeta(filename, node) |
|
738 | 744 | |
|
739 | 745 | # TODO: don't use the delta if it's larger than the fulltext |
|
740 | 746 | if constants.METAKEYSIZE not in meta: |
|
741 | 747 | meta[constants.METAKEYSIZE] = size |
|
742 | 748 | target.add(filename, node, deltabase, delta, meta) |
|
743 | 749 | |
|
744 | 750 | entries[node].datarepacked = True |
|
745 | 751 | |
|
746 | 752 | processprogress.complete() |
|
747 | 753 | count += 1 |
|
748 | 754 | |
|
749 | 755 | repackprogress.complete() |
|
750 | 756 | target.close(ledger=ledger) |
|
751 | 757 | |
|
752 | 758 | def repackhistory(self, ledger, target): |
|
753 | 759 | ui = self.repo.ui |
|
754 | 760 | |
|
755 | 761 | byfile = {} |
|
756 | 762 | for entry in pycompat.itervalues(ledger.entries): |
|
757 | 763 | if entry.historysource: |
|
758 | 764 | byfile.setdefault(entry.filename, {})[entry.node] = entry |
|
759 | 765 | |
|
760 | 766 | progress = ui.makeprogress( |
|
761 | 767 | _(b"repacking history"), unit=self.unit, total=len(byfile) |
|
762 | 768 | ) |
|
763 | 769 | for filename, entries in sorted(pycompat.iteritems(byfile)): |
|
764 | 770 | ancestors = {} |
|
765 | 771 | nodes = list(node for node in entries) |
|
766 | 772 | |
|
767 | 773 | for node in nodes: |
|
768 | 774 | if node in ancestors: |
|
769 | 775 | continue |
|
770 | 776 | ancestors.update( |
|
771 | 777 | self.history.getancestors(filename, node, known=ancestors) |
|
772 | 778 | ) |
|
773 | 779 | |
|
774 | 780 | # Order the nodes children first |
|
775 | 781 | orderednodes = reversed(self._toposort(ancestors)) |
|
776 | 782 | |
|
777 | 783 | # Write to the pack |
|
778 | 784 | dontprocess = set() |
|
779 | 785 | for node in orderednodes: |
|
780 | 786 | p1, p2, linknode, copyfrom = ancestors[node] |
|
781 | 787 | |
|
782 | 788 | # If the node is marked dontprocess, but it's also in the |
|
783 | 789 | # explicit entries set, that means the node exists both in this |
|
784 | 790 | # file and in another file that was copied to this file. |
|
785 | 791 | # Usually this happens if the file was copied to another file, |
|
786 | 792 | # then the copy was deleted, then reintroduced without copy |
|
787 | 793 | # metadata. The original add and the new add have the same hash |
|
788 | 794 | # since the content is identical and the parents are null. |
|
789 | 795 | if node in dontprocess and node not in entries: |
|
790 | 796 | # If copyfrom == filename, it means the copy history |
|
791 | 797 | # went to come other file, then came back to this one, so we |
|
792 | 798 | # should continue processing it. |
|
793 | 799 | if p1 != nullid and copyfrom != filename: |
|
794 | 800 | dontprocess.add(p1) |
|
795 | 801 | if p2 != nullid: |
|
796 | 802 | dontprocess.add(p2) |
|
797 | 803 | continue |
|
798 | 804 | |
|
799 | 805 | if copyfrom: |
|
800 | 806 | dontprocess.add(p1) |
|
801 | 807 | |
|
802 | 808 | target.add(filename, node, p1, p2, linknode, copyfrom) |
|
803 | 809 | |
|
804 | 810 | if node in entries: |
|
805 | 811 | entries[node].historyrepacked = True |
|
806 | 812 | |
|
807 | 813 | progress.increment() |
|
808 | 814 | |
|
809 | 815 | progress.complete() |
|
810 | 816 | target.close(ledger=ledger) |
|
811 | 817 | |
|
812 | 818 | def _toposort(self, ancestors): |
|
813 | 819 | def parentfunc(node): |
|
814 | 820 | p1, p2, linknode, copyfrom = ancestors[node] |
|
815 | 821 | parents = [] |
|
816 | 822 | if p1 != nullid: |
|
817 | 823 | parents.append(p1) |
|
818 | 824 | if p2 != nullid: |
|
819 | 825 | parents.append(p2) |
|
820 | 826 | return parents |
|
821 | 827 | |
|
822 | 828 | sortednodes = shallowutil.sortnodes(ancestors.keys(), parentfunc) |
|
823 | 829 | return sortednodes |
|
824 | 830 | |
|
825 | 831 | |
|
826 | 832 | class repackledger(object): |
|
827 | 833 | """Storage for all the bookkeeping that happens during a repack. It contains |
|
828 | 834 | the list of revisions being repacked, what happened to each revision, and |
|
829 | 835 | which source store contained which revision originally (for later cleanup). |
|
830 | 836 | """ |
|
831 | 837 | |
|
832 | 838 | def __init__(self): |
|
833 | 839 | self.entries = {} |
|
834 | 840 | self.sources = {} |
|
835 | 841 | self.created = set() |
|
836 | 842 | |
|
837 | 843 | def markdataentry(self, source, filename, node): |
|
838 | 844 | """Mark the given filename+node revision as having a data rev in the |
|
839 | 845 | given source. |
|
840 | 846 | """ |
|
841 | 847 | entry = self._getorcreateentry(filename, node) |
|
842 | 848 | entry.datasource = True |
|
843 | 849 | entries = self.sources.get(source) |
|
844 | 850 | if not entries: |
|
845 | 851 | entries = set() |
|
846 | 852 | self.sources[source] = entries |
|
847 | 853 | entries.add(entry) |
|
848 | 854 | |
|
849 | 855 | def markhistoryentry(self, source, filename, node): |
|
850 | 856 | """Mark the given filename+node revision as having a history rev in the |
|
851 | 857 | given source. |
|
852 | 858 | """ |
|
853 | 859 | entry = self._getorcreateentry(filename, node) |
|
854 | 860 | entry.historysource = True |
|
855 | 861 | entries = self.sources.get(source) |
|
856 | 862 | if not entries: |
|
857 | 863 | entries = set() |
|
858 | 864 | self.sources[source] = entries |
|
859 | 865 | entries.add(entry) |
|
860 | 866 | |
|
861 | 867 | def _getorcreateentry(self, filename, node): |
|
862 | 868 | key = (filename, node) |
|
863 | 869 | value = self.entries.get(key) |
|
864 | 870 | if not value: |
|
865 | 871 | value = repackentry(filename, node) |
|
866 | 872 | self.entries[key] = value |
|
867 | 873 | |
|
868 | 874 | return value |
|
869 | 875 | |
|
870 | 876 | def addcreated(self, value): |
|
871 | 877 | self.created.add(value) |
|
872 | 878 | |
|
873 | 879 | |
|
874 | 880 | class repackentry(object): |
|
875 | 881 | """Simple class representing a single revision entry in the repackledger. |
|
876 | 882 | """ |
|
877 | 883 | |
|
878 | 884 | __slots__ = ( |
|
879 | 885 | r'filename', |
|
880 | 886 | r'node', |
|
881 | 887 | r'datasource', |
|
882 | 888 | r'historysource', |
|
883 | 889 | r'datarepacked', |
|
884 | 890 | r'historyrepacked', |
|
885 | 891 | r'gced', |
|
886 | 892 | ) |
|
887 | 893 | |
|
888 | 894 | def __init__(self, filename, node): |
|
889 | 895 | self.filename = filename |
|
890 | 896 | self.node = node |
|
891 | 897 | # If the revision has a data entry in the source |
|
892 | 898 | self.datasource = False |
|
893 | 899 | # If the revision has a history entry in the source |
|
894 | 900 | self.historysource = False |
|
895 | 901 | # If the revision's data entry was repacked into the repack target |
|
896 | 902 | self.datarepacked = False |
|
897 | 903 | # If the revision's history entry was repacked into the repack target |
|
898 | 904 | self.historyrepacked = False |
|
899 | 905 | # If garbage collected |
|
900 | 906 | self.gced = False |
|
901 | 907 | |
|
902 | 908 | |
|
903 | 909 | def repacklockvfs(repo): |
|
904 | 910 | if util.safehasattr(repo, 'name'): |
|
905 | 911 | # Lock in the shared cache so repacks across multiple copies of the same |
|
906 | 912 | # repo are coordinated. |
|
907 | 913 | sharedcachepath = shallowutil.getcachepackpath( |
|
908 | 914 | repo, constants.FILEPACK_CATEGORY |
|
909 | 915 | ) |
|
910 | 916 | return vfs.vfs(sharedcachepath) |
|
911 | 917 | else: |
|
912 | 918 | return repo.svfs |
@@ -1,354 +1,358 b'' | |||
|
1 | 1 | # shallowrepo.py - shallow repository that uses remote filelogs |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2013 Facebook, Inc. |
|
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 | from __future__ import absolute_import |
|
8 | 8 | |
|
9 | 9 | import os |
|
10 | 10 | |
|
11 | 11 | from mercurial.i18n import _ |
|
12 | 12 | from mercurial.node import hex, nullid, nullrev |
|
13 | 13 | from mercurial import ( |
|
14 | 14 | encoding, |
|
15 | 15 | error, |
|
16 | 16 | localrepo, |
|
17 | 17 | match, |
|
18 | 18 | pycompat, |
|
19 | 19 | scmutil, |
|
20 | 20 | sparse, |
|
21 | 21 | util, |
|
22 | 22 | ) |
|
23 | 23 | from mercurial.utils import procutil |
|
24 | 24 | from . import ( |
|
25 | 25 | connectionpool, |
|
26 | 26 | constants, |
|
27 | 27 | contentstore, |
|
28 | 28 | datapack, |
|
29 | 29 | fileserverclient, |
|
30 | 30 | historypack, |
|
31 | 31 | metadatastore, |
|
32 | 32 | remotefilectx, |
|
33 | 33 | remotefilelog, |
|
34 | 34 | shallowutil, |
|
35 | 35 | ) |
|
36 | 36 | |
|
37 | 37 | # These make*stores functions are global so that other extensions can replace |
|
38 | 38 | # them. |
|
39 | 39 | def makelocalstores(repo): |
|
40 | 40 | """In-repo stores, like .hg/store/data; can not be discarded.""" |
|
41 | 41 | localpath = os.path.join(repo.svfs.vfs.base, b'data') |
|
42 | 42 | if not os.path.exists(localpath): |
|
43 | 43 | os.makedirs(localpath) |
|
44 | 44 | |
|
45 | 45 | # Instantiate local data stores |
|
46 | 46 | localcontent = contentstore.remotefilelogcontentstore( |
|
47 | 47 | repo, localpath, repo.name, shared=False |
|
48 | 48 | ) |
|
49 | 49 | localmetadata = metadatastore.remotefilelogmetadatastore( |
|
50 | 50 | repo, localpath, repo.name, shared=False |
|
51 | 51 | ) |
|
52 | 52 | return localcontent, localmetadata |
|
53 | 53 | |
|
54 | 54 | |
|
55 | 55 | def makecachestores(repo): |
|
56 | 56 | """Typically machine-wide, cache of remote data; can be discarded.""" |
|
57 | 57 | # Instantiate shared cache stores |
|
58 | 58 | cachepath = shallowutil.getcachepath(repo.ui) |
|
59 | 59 | cachecontent = contentstore.remotefilelogcontentstore( |
|
60 | 60 | repo, cachepath, repo.name, shared=True |
|
61 | 61 | ) |
|
62 | 62 | cachemetadata = metadatastore.remotefilelogmetadatastore( |
|
63 | 63 | repo, cachepath, repo.name, shared=True |
|
64 | 64 | ) |
|
65 | 65 | |
|
66 | 66 | repo.sharedstore = cachecontent |
|
67 | 67 | repo.shareddatastores.append(cachecontent) |
|
68 | 68 | repo.sharedhistorystores.append(cachemetadata) |
|
69 | 69 | |
|
70 | 70 | return cachecontent, cachemetadata |
|
71 | 71 | |
|
72 | 72 | |
|
73 | 73 | def makeremotestores(repo, cachecontent, cachemetadata): |
|
74 | 74 | """These stores fetch data from a remote server.""" |
|
75 | 75 | # Instantiate remote stores |
|
76 | 76 | repo.fileservice = fileserverclient.fileserverclient(repo) |
|
77 | 77 | remotecontent = contentstore.remotecontentstore( |
|
78 | 78 | repo.ui, repo.fileservice, cachecontent |
|
79 | 79 | ) |
|
80 | 80 | remotemetadata = metadatastore.remotemetadatastore( |
|
81 | 81 | repo.ui, repo.fileservice, cachemetadata |
|
82 | 82 | ) |
|
83 | 83 | return remotecontent, remotemetadata |
|
84 | 84 | |
|
85 | 85 | |
|
86 | 86 | def makepackstores(repo): |
|
87 | 87 | """Packs are more efficient (to read from) cache stores.""" |
|
88 | 88 | # Instantiate pack stores |
|
89 | 89 | packpath = shallowutil.getcachepackpath(repo, constants.FILEPACK_CATEGORY) |
|
90 | 90 | packcontentstore = datapack.datapackstore(repo.ui, packpath) |
|
91 | 91 | packmetadatastore = historypack.historypackstore(repo.ui, packpath) |
|
92 | 92 | |
|
93 | 93 | repo.shareddatastores.append(packcontentstore) |
|
94 | 94 | repo.sharedhistorystores.append(packmetadatastore) |
|
95 | 95 | shallowutil.reportpackmetrics( |
|
96 | 96 | repo.ui, b'filestore', packcontentstore, packmetadatastore |
|
97 | 97 | ) |
|
98 | 98 | return packcontentstore, packmetadatastore |
|
99 | 99 | |
|
100 | 100 | |
|
101 | 101 | def makeunionstores(repo): |
|
102 | 102 | """Union stores iterate the other stores and return the first result.""" |
|
103 | 103 | repo.shareddatastores = [] |
|
104 | 104 | repo.sharedhistorystores = [] |
|
105 | 105 | |
|
106 | 106 | packcontentstore, packmetadatastore = makepackstores(repo) |
|
107 | 107 | cachecontent, cachemetadata = makecachestores(repo) |
|
108 | 108 | localcontent, localmetadata = makelocalstores(repo) |
|
109 | 109 | remotecontent, remotemetadata = makeremotestores( |
|
110 | 110 | repo, cachecontent, cachemetadata |
|
111 | 111 | ) |
|
112 | 112 | |
|
113 | 113 | # Instantiate union stores |
|
114 | 114 | repo.contentstore = contentstore.unioncontentstore( |
|
115 | 115 | packcontentstore, |
|
116 | 116 | cachecontent, |
|
117 | 117 | localcontent, |
|
118 | 118 | remotecontent, |
|
119 | 119 | writestore=localcontent, |
|
120 | 120 | ) |
|
121 | 121 | repo.metadatastore = metadatastore.unionmetadatastore( |
|
122 | 122 | packmetadatastore, |
|
123 | 123 | cachemetadata, |
|
124 | 124 | localmetadata, |
|
125 | 125 | remotemetadata, |
|
126 | 126 | writestore=localmetadata, |
|
127 | 127 | ) |
|
128 | 128 | |
|
129 | 129 | fileservicedatawrite = cachecontent |
|
130 | 130 | fileservicehistorywrite = cachemetadata |
|
131 | 131 | repo.fileservice.setstore( |
|
132 | 132 | repo.contentstore, |
|
133 | 133 | repo.metadatastore, |
|
134 | 134 | fileservicedatawrite, |
|
135 | 135 | fileservicehistorywrite, |
|
136 | 136 | ) |
|
137 | 137 | shallowutil.reportpackmetrics( |
|
138 | 138 | repo.ui, b'filestore', packcontentstore, packmetadatastore |
|
139 | 139 | ) |
|
140 | 140 | |
|
141 | 141 | |
|
142 | 142 | def wraprepo(repo): |
|
143 | 143 | class shallowrepository(repo.__class__): |
|
144 | 144 | @util.propertycache |
|
145 | 145 | def name(self): |
|
146 | 146 | return self.ui.config(b'remotefilelog', b'reponame') |
|
147 | 147 | |
|
148 | 148 | @util.propertycache |
|
149 | 149 | def fallbackpath(self): |
|
150 | 150 | path = repo.ui.config( |
|
151 | 151 | b"remotefilelog", |
|
152 | 152 | b"fallbackpath", |
|
153 | 153 | repo.ui.config(b'paths', b'default'), |
|
154 | 154 | ) |
|
155 | 155 | if not path: |
|
156 | 156 | raise error.Abort( |
|
157 | 157 | b"no remotefilelog server " |
|
158 | 158 | b"configured - is your .hg/hgrc trusted?" |
|
159 | 159 | ) |
|
160 | 160 | |
|
161 | 161 | return path |
|
162 | 162 | |
|
163 | 163 | def maybesparsematch(self, *revs, **kwargs): |
|
164 | 164 | ''' |
|
165 | 165 | A wrapper that allows the remotefilelog to invoke sparsematch() if |
|
166 | 166 | this is a sparse repository, or returns None if this is not a |
|
167 | 167 | sparse repository. |
|
168 | 168 | ''' |
|
169 | 169 | if revs: |
|
170 | 170 | ret = sparse.matcher(repo, revs=revs) |
|
171 | 171 | else: |
|
172 | 172 | ret = sparse.matcher(repo) |
|
173 | 173 | |
|
174 | 174 | if ret.always(): |
|
175 | 175 | return None |
|
176 | 176 | return ret |
|
177 | 177 | |
|
178 | 178 | def file(self, f): |
|
179 | 179 | if f[0] == b'/': |
|
180 | 180 | f = f[1:] |
|
181 | 181 | |
|
182 | 182 | if self.shallowmatch(f): |
|
183 | 183 | return remotefilelog.remotefilelog(self.svfs, f, self) |
|
184 | 184 | else: |
|
185 | 185 | return super(shallowrepository, self).file(f) |
|
186 | 186 | |
|
187 | 187 | def filectx(self, path, *args, **kwargs): |
|
188 | 188 | if self.shallowmatch(path): |
|
189 | 189 | return remotefilectx.remotefilectx(self, path, *args, **kwargs) |
|
190 | 190 | else: |
|
191 | 191 | return super(shallowrepository, self).filectx( |
|
192 | 192 | path, *args, **kwargs |
|
193 | 193 | ) |
|
194 | 194 | |
|
195 | 195 | @localrepo.unfilteredmethod |
|
196 | 196 | def commitctx(self, ctx, error=False, origctx=None): |
|
197 | 197 | """Add a new revision to current repository. |
|
198 | 198 | Revision information is passed via the context argument. |
|
199 | 199 | """ |
|
200 | 200 | |
|
201 | 201 | # some contexts already have manifest nodes, they don't need any |
|
202 | 202 | # prefetching (for example if we're just editing a commit message |
|
203 | 203 | # we can reuse manifest |
|
204 | 204 | if not ctx.manifestnode(): |
|
205 | 205 | # prefetch files that will likely be compared |
|
206 | 206 | m1 = ctx.p1().manifest() |
|
207 | 207 | files = [] |
|
208 | 208 | for f in ctx.modified() + ctx.added(): |
|
209 | 209 | fparent1 = m1.get(f, nullid) |
|
210 | 210 | if fparent1 != nullid: |
|
211 | 211 | files.append((f, hex(fparent1))) |
|
212 | 212 | self.fileservice.prefetch(files) |
|
213 | 213 | return super(shallowrepository, self).commitctx( |
|
214 | 214 | ctx, error=error, origctx=origctx |
|
215 | 215 | ) |
|
216 | 216 | |
|
217 | 217 | def backgroundprefetch( |
|
218 | 218 | self, |
|
219 | 219 | revs, |
|
220 | 220 | base=None, |
|
221 | 221 | repack=False, |
|
222 | 222 | pats=None, |
|
223 | 223 | opts=None, |
|
224 | 224 | ensurestart=False, |
|
225 | 225 | ): |
|
226 | 226 | """Runs prefetch in background with optional repack |
|
227 | 227 | """ |
|
228 | 228 | cmd = [procutil.hgexecutable(), b'-R', repo.origroot, b'prefetch'] |
|
229 | 229 | if repack: |
|
230 | 230 | cmd.append(b'--repack') |
|
231 | 231 | if revs: |
|
232 | 232 | cmd += [b'-r', revs] |
|
233 | 233 | # We know this command will find a binary, so don't block |
|
234 | 234 | # on it starting. |
|
235 | kwargs = {} | |
|
236 | if repo.ui.configbool(b'devel', b'remotefilelog.bg-wait'): | |
|
237 | kwargs['record_wait'] = repo.ui.atexit | |
|
238 | ||
|
235 | 239 | procutil.runbgcommand( |
|
236 | cmd, encoding.environ, ensurestart=ensurestart | |
|
240 | cmd, encoding.environ, ensurestart=ensurestart, **kwargs | |
|
237 | 241 | ) |
|
238 | 242 | |
|
239 | 243 | def prefetch(self, revs, base=None, pats=None, opts=None): |
|
240 | 244 | """Prefetches all the necessary file revisions for the given revs |
|
241 | 245 | Optionally runs repack in background |
|
242 | 246 | """ |
|
243 | 247 | with repo._lock( |
|
244 | 248 | repo.svfs, |
|
245 | 249 | b'prefetchlock', |
|
246 | 250 | True, |
|
247 | 251 | None, |
|
248 | 252 | None, |
|
249 | 253 | _(b'prefetching in %s') % repo.origroot, |
|
250 | 254 | ): |
|
251 | 255 | self._prefetch(revs, base, pats, opts) |
|
252 | 256 | |
|
253 | 257 | def _prefetch(self, revs, base=None, pats=None, opts=None): |
|
254 | 258 | fallbackpath = self.fallbackpath |
|
255 | 259 | if fallbackpath: |
|
256 | 260 | # If we know a rev is on the server, we should fetch the server |
|
257 | 261 | # version of those files, since our local file versions might |
|
258 | 262 | # become obsolete if the local commits are stripped. |
|
259 | 263 | localrevs = repo.revs(b'outgoing(%s)', fallbackpath) |
|
260 | 264 | if base is not None and base != nullrev: |
|
261 | 265 | serverbase = list( |
|
262 | 266 | repo.revs( |
|
263 | 267 | b'first(reverse(::%s) - %ld)', base, localrevs |
|
264 | 268 | ) |
|
265 | 269 | ) |
|
266 | 270 | if serverbase: |
|
267 | 271 | base = serverbase[0] |
|
268 | 272 | else: |
|
269 | 273 | localrevs = repo |
|
270 | 274 | |
|
271 | 275 | mfl = repo.manifestlog |
|
272 | 276 | mfrevlog = mfl.getstorage(b'') |
|
273 | 277 | if base is not None: |
|
274 | 278 | mfdict = mfl[repo[base].manifestnode()].read() |
|
275 | 279 | skip = set(pycompat.iteritems(mfdict)) |
|
276 | 280 | else: |
|
277 | 281 | skip = set() |
|
278 | 282 | |
|
279 | 283 | # Copy the skip set to start large and avoid constant resizing, |
|
280 | 284 | # and since it's likely to be very similar to the prefetch set. |
|
281 | 285 | files = skip.copy() |
|
282 | 286 | serverfiles = skip.copy() |
|
283 | 287 | visited = set() |
|
284 | 288 | visited.add(nullrev) |
|
285 | 289 | revcount = len(revs) |
|
286 | 290 | progress = self.ui.makeprogress(_(b'prefetching'), total=revcount) |
|
287 | 291 | progress.update(0) |
|
288 | 292 | for rev in sorted(revs): |
|
289 | 293 | ctx = repo[rev] |
|
290 | 294 | if pats: |
|
291 | 295 | m = scmutil.match(ctx, pats, opts) |
|
292 | 296 | sparsematch = repo.maybesparsematch(rev) |
|
293 | 297 | |
|
294 | 298 | mfnode = ctx.manifestnode() |
|
295 | 299 | mfrev = mfrevlog.rev(mfnode) |
|
296 | 300 | |
|
297 | 301 | # Decompressing manifests is expensive. |
|
298 | 302 | # When possible, only read the deltas. |
|
299 | 303 | p1, p2 = mfrevlog.parentrevs(mfrev) |
|
300 | 304 | if p1 in visited and p2 in visited: |
|
301 | 305 | mfdict = mfl[mfnode].readfast() |
|
302 | 306 | else: |
|
303 | 307 | mfdict = mfl[mfnode].read() |
|
304 | 308 | |
|
305 | 309 | diff = pycompat.iteritems(mfdict) |
|
306 | 310 | if pats: |
|
307 | 311 | diff = (pf for pf in diff if m(pf[0])) |
|
308 | 312 | if sparsematch: |
|
309 | 313 | diff = (pf for pf in diff if sparsematch(pf[0])) |
|
310 | 314 | if rev not in localrevs: |
|
311 | 315 | serverfiles.update(diff) |
|
312 | 316 | else: |
|
313 | 317 | files.update(diff) |
|
314 | 318 | |
|
315 | 319 | visited.add(mfrev) |
|
316 | 320 | progress.increment() |
|
317 | 321 | |
|
318 | 322 | files.difference_update(skip) |
|
319 | 323 | serverfiles.difference_update(skip) |
|
320 | 324 | progress.complete() |
|
321 | 325 | |
|
322 | 326 | # Fetch files known to be on the server |
|
323 | 327 | if serverfiles: |
|
324 | 328 | results = [(path, hex(fnode)) for (path, fnode) in serverfiles] |
|
325 | 329 | repo.fileservice.prefetch(results, force=True) |
|
326 | 330 | |
|
327 | 331 | # Fetch files that may or may not be on the server |
|
328 | 332 | if files: |
|
329 | 333 | results = [(path, hex(fnode)) for (path, fnode) in files] |
|
330 | 334 | repo.fileservice.prefetch(results) |
|
331 | 335 | |
|
332 | 336 | def close(self): |
|
333 | 337 | super(shallowrepository, self).close() |
|
334 | 338 | self.connectionpool.close() |
|
335 | 339 | |
|
336 | 340 | repo.__class__ = shallowrepository |
|
337 | 341 | |
|
338 | 342 | repo.shallowmatch = match.always() |
|
339 | 343 | |
|
340 | 344 | makeunionstores(repo) |
|
341 | 345 | |
|
342 | 346 | repo.includepattern = repo.ui.configlist( |
|
343 | 347 | b"remotefilelog", b"includepattern", None |
|
344 | 348 | ) |
|
345 | 349 | repo.excludepattern = repo.ui.configlist( |
|
346 | 350 | b"remotefilelog", b"excludepattern", None |
|
347 | 351 | ) |
|
348 | 352 | if not util.safehasattr(repo, 'connectionpool'): |
|
349 | 353 | repo.connectionpool = connectionpool.connectionpool(repo) |
|
350 | 354 | |
|
351 | 355 | if repo.includepattern or repo.excludepattern: |
|
352 | 356 | repo.shallowmatch = match.match( |
|
353 | 357 | repo.root, b'', None, repo.includepattern, repo.excludepattern |
|
354 | 358 | ) |
General Comments 0
You need to be logged in to leave comments.
Login now