Show More
@@ -1,765 +1,765 | |||
|
1 | 1 | # blobstore.py - local and remote (speaking Git-LFS protocol) blob storages |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2017 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 | |
|
8 | 8 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import contextlib |
|
11 | 11 | import errno |
|
12 | 12 | import hashlib |
|
13 | 13 | import json |
|
14 | 14 | import os |
|
15 | 15 | import re |
|
16 | 16 | import socket |
|
17 | 17 | |
|
18 | 18 | from mercurial.i18n import _ |
|
19 | 19 | from mercurial.pycompat import getattr |
|
20 | 20 | |
|
21 | 21 | from mercurial import ( |
|
22 | 22 | encoding, |
|
23 | 23 | error, |
|
24 | 24 | httpconnection as httpconnectionmod, |
|
25 | 25 | node, |
|
26 | 26 | pathutil, |
|
27 | 27 | pycompat, |
|
28 | 28 | url as urlmod, |
|
29 | 29 | util, |
|
30 | 30 | vfs as vfsmod, |
|
31 | 31 | worker, |
|
32 | 32 | ) |
|
33 | 33 | |
|
34 | 34 | from mercurial.utils import stringutil |
|
35 | 35 | |
|
36 | 36 | from ..largefiles import lfutil |
|
37 | 37 | |
|
38 | 38 | # 64 bytes for SHA256 |
|
39 | 39 | _lfsre = re.compile(br'\A[a-f0-9]{64}\Z') |
|
40 | 40 | |
|
41 | 41 | |
|
42 | 42 | class lfsvfs(vfsmod.vfs): |
|
43 | 43 | def join(self, path): |
|
44 | 44 | """split the path at first two characters, like: XX/XXXXX...""" |
|
45 | 45 | if not _lfsre.match(path): |
|
46 | 46 | raise error.ProgrammingError(b'unexpected lfs path: %s' % path) |
|
47 | 47 | return super(lfsvfs, self).join(path[0:2], path[2:]) |
|
48 | 48 | |
|
49 | 49 | def walk(self, path=None, onerror=None): |
|
50 | 50 | """Yield (dirpath, [], oids) tuple for blobs under path |
|
51 | 51 | |
|
52 | 52 | Oids only exist in the root of this vfs, so dirpath is always ''. |
|
53 | 53 | """ |
|
54 | 54 | root = os.path.normpath(self.base) |
|
55 | 55 | # when dirpath == root, dirpath[prefixlen:] becomes empty |
|
56 | 56 | # because len(dirpath) < prefixlen. |
|
57 | 57 | prefixlen = len(pathutil.normasprefix(root)) |
|
58 | 58 | oids = [] |
|
59 | 59 | |
|
60 | 60 | for dirpath, dirs, files in os.walk( |
|
61 | 61 | self.reljoin(self.base, path or b''), onerror=onerror |
|
62 | 62 | ): |
|
63 | 63 | dirpath = dirpath[prefixlen:] |
|
64 | 64 | |
|
65 | 65 | # Silently skip unexpected files and directories |
|
66 | 66 | if len(dirpath) == 2: |
|
67 | 67 | oids.extend( |
|
68 | 68 | [dirpath + f for f in files if _lfsre.match(dirpath + f)] |
|
69 | 69 | ) |
|
70 | 70 | |
|
71 | 71 | yield (b'', [], oids) |
|
72 | 72 | |
|
73 | 73 | |
|
74 | 74 | class nullvfs(lfsvfs): |
|
75 | 75 | def __init__(self): |
|
76 | 76 | pass |
|
77 | 77 | |
|
78 | 78 | def exists(self, oid): |
|
79 | 79 | return False |
|
80 | 80 | |
|
81 | 81 | def read(self, oid): |
|
82 | 82 | # store.read() calls into here if the blob doesn't exist in its |
|
83 | 83 | # self.vfs. Raise the same error as a normal vfs when asked to read a |
|
84 | 84 | # file that doesn't exist. The only difference is the full file path |
|
85 | 85 | # isn't available in the error. |
|
86 | 86 | raise IOError( |
|
87 | 87 | errno.ENOENT, |
|
88 | 88 | pycompat.sysstr(b'%s: No such file or directory' % oid), |
|
89 | 89 | ) |
|
90 | 90 | |
|
91 | 91 | def walk(self, path=None, onerror=None): |
|
92 | 92 | return (b'', [], []) |
|
93 | 93 | |
|
94 | 94 | def write(self, oid, data): |
|
95 | 95 | pass |
|
96 | 96 | |
|
97 | 97 | |
|
98 | 98 | class lfsuploadfile(httpconnectionmod.httpsendfile): |
|
99 | 99 | """a file-like object that supports keepalive. |
|
100 | 100 | """ |
|
101 | 101 | |
|
102 | 102 | def __init__(self, ui, filename): |
|
103 | 103 | super(lfsuploadfile, self).__init__(ui, filename, b'rb') |
|
104 | 104 | self.read = self._data.read |
|
105 | 105 | |
|
106 | 106 | def _makeprogress(self): |
|
107 | 107 | return None # progress is handled by the worker client |
|
108 | 108 | |
|
109 | 109 | |
|
110 | 110 | class local(object): |
|
111 | 111 | """Local blobstore for large file contents. |
|
112 | 112 | |
|
113 | 113 | This blobstore is used both as a cache and as a staging area for large blobs |
|
114 | 114 | to be uploaded to the remote blobstore. |
|
115 | 115 | """ |
|
116 | 116 | |
|
117 | 117 | def __init__(self, repo): |
|
118 | 118 | fullpath = repo.svfs.join(b'lfs/objects') |
|
119 | 119 | self.vfs = lfsvfs(fullpath) |
|
120 | 120 | |
|
121 | 121 | if repo.ui.configbool(b'experimental', b'lfs.disableusercache'): |
|
122 | 122 | self.cachevfs = nullvfs() |
|
123 | 123 | else: |
|
124 | 124 | usercache = lfutil._usercachedir(repo.ui, b'lfs') |
|
125 | 125 | self.cachevfs = lfsvfs(usercache) |
|
126 | 126 | self.ui = repo.ui |
|
127 | 127 | |
|
128 | 128 | def open(self, oid): |
|
129 | 129 | """Open a read-only file descriptor to the named blob, in either the |
|
130 | 130 | usercache or the local store.""" |
|
131 |
return open(self.path(oid), |
|
|
131 | return open(self.path(oid), 'rb') | |
|
132 | 132 | |
|
133 | 133 | def path(self, oid): |
|
134 | 134 | """Build the path for the given blob ``oid``. |
|
135 | 135 | |
|
136 | 136 | If the blob exists locally, the path may point to either the usercache |
|
137 | 137 | or the local store. If it doesn't, it will point to the local store. |
|
138 | 138 | This is meant for situations where existing code that isn't LFS aware |
|
139 | 139 | needs to open a blob. Generally, prefer the ``open`` method on this |
|
140 | 140 | class. |
|
141 | 141 | """ |
|
142 | 142 | # The usercache is the most likely place to hold the file. Commit will |
|
143 | 143 | # write to both it and the local store, as will anything that downloads |
|
144 | 144 | # the blobs. However, things like clone without an update won't |
|
145 | 145 | # populate the local store. For an init + push of a local clone, |
|
146 | 146 | # the usercache is the only place it _could_ be. If not present, the |
|
147 | 147 | # missing file msg here will indicate the local repo, not the usercache. |
|
148 | 148 | if self.cachevfs.exists(oid): |
|
149 | 149 | return self.cachevfs.join(oid) |
|
150 | 150 | |
|
151 | 151 | return self.vfs.join(oid) |
|
152 | 152 | |
|
153 | 153 | def download(self, oid, src, content_length): |
|
154 | 154 | """Read the blob from the remote source in chunks, verify the content, |
|
155 | 155 | and write to this local blobstore.""" |
|
156 | 156 | sha256 = hashlib.sha256() |
|
157 | 157 | size = 0 |
|
158 | 158 | |
|
159 | 159 | with self.vfs(oid, b'wb', atomictemp=True) as fp: |
|
160 | 160 | for chunk in util.filechunkiter(src, size=1048576): |
|
161 | 161 | fp.write(chunk) |
|
162 | 162 | sha256.update(chunk) |
|
163 | 163 | size += len(chunk) |
|
164 | 164 | |
|
165 | 165 | # If the server advertised a length longer than what we actually |
|
166 | 166 | # received, then we should expect that the server crashed while |
|
167 | 167 | # producing the response (but the server has no way of telling us |
|
168 | 168 | # that), and we really don't need to try to write the response to |
|
169 | 169 | # the localstore, because it's not going to match the expected. |
|
170 | 170 | if content_length is not None and int(content_length) != size: |
|
171 | 171 | msg = ( |
|
172 | 172 | b"Response length (%s) does not match Content-Length " |
|
173 | 173 | b"header (%d): likely server-side crash" |
|
174 | 174 | ) |
|
175 | 175 | raise LfsRemoteError(_(msg) % (size, int(content_length))) |
|
176 | 176 | |
|
177 | 177 | realoid = node.hex(sha256.digest()) |
|
178 | 178 | if realoid != oid: |
|
179 | 179 | raise LfsCorruptionError( |
|
180 | 180 | _(b'corrupt remote lfs object: %s') % oid |
|
181 | 181 | ) |
|
182 | 182 | |
|
183 | 183 | self._linktousercache(oid) |
|
184 | 184 | |
|
185 | 185 | def write(self, oid, data): |
|
186 | 186 | """Write blob to local blobstore. |
|
187 | 187 | |
|
188 | 188 | This should only be called from the filelog during a commit or similar. |
|
189 | 189 | As such, there is no need to verify the data. Imports from a remote |
|
190 | 190 | store must use ``download()`` instead.""" |
|
191 | 191 | with self.vfs(oid, b'wb', atomictemp=True) as fp: |
|
192 | 192 | fp.write(data) |
|
193 | 193 | |
|
194 | 194 | self._linktousercache(oid) |
|
195 | 195 | |
|
196 | 196 | def linkfromusercache(self, oid): |
|
197 | 197 | """Link blobs found in the user cache into this store. |
|
198 | 198 | |
|
199 | 199 | The server module needs to do this when it lets the client know not to |
|
200 | 200 | upload the blob, to ensure it is always available in this store. |
|
201 | 201 | Normally this is done implicitly when the client reads or writes the |
|
202 | 202 | blob, but that doesn't happen when the server tells the client that it |
|
203 | 203 | already has the blob. |
|
204 | 204 | """ |
|
205 | 205 | if not isinstance(self.cachevfs, nullvfs) and not self.vfs.exists(oid): |
|
206 | 206 | self.ui.note(_(b'lfs: found %s in the usercache\n') % oid) |
|
207 | 207 | lfutil.link(self.cachevfs.join(oid), self.vfs.join(oid)) |
|
208 | 208 | |
|
209 | 209 | def _linktousercache(self, oid): |
|
210 | 210 | # XXX: should we verify the content of the cache, and hardlink back to |
|
211 | 211 | # the local store on success, but truncate, write and link on failure? |
|
212 | 212 | if not self.cachevfs.exists(oid) and not isinstance( |
|
213 | 213 | self.cachevfs, nullvfs |
|
214 | 214 | ): |
|
215 | 215 | self.ui.note(_(b'lfs: adding %s to the usercache\n') % oid) |
|
216 | 216 | lfutil.link(self.vfs.join(oid), self.cachevfs.join(oid)) |
|
217 | 217 | |
|
218 | 218 | def read(self, oid, verify=True): |
|
219 | 219 | """Read blob from local blobstore.""" |
|
220 | 220 | if not self.vfs.exists(oid): |
|
221 | 221 | blob = self._read(self.cachevfs, oid, verify) |
|
222 | 222 | |
|
223 | 223 | # Even if revlog will verify the content, it needs to be verified |
|
224 | 224 | # now before making the hardlink to avoid propagating corrupt blobs. |
|
225 | 225 | # Don't abort if corruption is detected, because `hg verify` will |
|
226 | 226 | # give more useful info about the corruption- simply don't add the |
|
227 | 227 | # hardlink. |
|
228 | 228 | if verify or node.hex(hashlib.sha256(blob).digest()) == oid: |
|
229 | 229 | self.ui.note(_(b'lfs: found %s in the usercache\n') % oid) |
|
230 | 230 | lfutil.link(self.cachevfs.join(oid), self.vfs.join(oid)) |
|
231 | 231 | else: |
|
232 | 232 | self.ui.note(_(b'lfs: found %s in the local lfs store\n') % oid) |
|
233 | 233 | blob = self._read(self.vfs, oid, verify) |
|
234 | 234 | return blob |
|
235 | 235 | |
|
236 | 236 | def _read(self, vfs, oid, verify): |
|
237 | 237 | """Read blob (after verifying) from the given store""" |
|
238 | 238 | blob = vfs.read(oid) |
|
239 | 239 | if verify: |
|
240 | 240 | _verify(oid, blob) |
|
241 | 241 | return blob |
|
242 | 242 | |
|
243 | 243 | def verify(self, oid): |
|
244 | 244 | """Indicate whether or not the hash of the underlying file matches its |
|
245 | 245 | name.""" |
|
246 | 246 | sha256 = hashlib.sha256() |
|
247 | 247 | |
|
248 | 248 | with self.open(oid) as fp: |
|
249 | 249 | for chunk in util.filechunkiter(fp, size=1048576): |
|
250 | 250 | sha256.update(chunk) |
|
251 | 251 | |
|
252 | 252 | return oid == node.hex(sha256.digest()) |
|
253 | 253 | |
|
254 | 254 | def has(self, oid): |
|
255 | 255 | """Returns True if the local blobstore contains the requested blob, |
|
256 | 256 | False otherwise.""" |
|
257 | 257 | return self.cachevfs.exists(oid) or self.vfs.exists(oid) |
|
258 | 258 | |
|
259 | 259 | |
|
260 | 260 | def _urlerrorreason(urlerror): |
|
261 | 261 | '''Create a friendly message for the given URLError to be used in an |
|
262 | 262 | LfsRemoteError message. |
|
263 | 263 | ''' |
|
264 | 264 | inst = urlerror |
|
265 | 265 | |
|
266 | 266 | if isinstance(urlerror.reason, Exception): |
|
267 | 267 | inst = urlerror.reason |
|
268 | 268 | |
|
269 | 269 | if util.safehasattr(inst, b'reason'): |
|
270 | 270 | try: # usually it is in the form (errno, strerror) |
|
271 | 271 | reason = inst.reason.args[1] |
|
272 | 272 | except (AttributeError, IndexError): |
|
273 | 273 | # it might be anything, for example a string |
|
274 | 274 | reason = inst.reason |
|
275 | 275 | if isinstance(reason, pycompat.unicode): |
|
276 | 276 | # SSLError of Python 2.7.9 contains a unicode |
|
277 | 277 | reason = encoding.unitolocal(reason) |
|
278 | 278 | return reason |
|
279 | 279 | elif getattr(inst, "strerror", None): |
|
280 | 280 | return encoding.strtolocal(inst.strerror) |
|
281 | 281 | else: |
|
282 | 282 | return stringutil.forcebytestr(urlerror) |
|
283 | 283 | |
|
284 | 284 | |
|
285 | 285 | class lfsauthhandler(util.urlreq.basehandler): |
|
286 | 286 | handler_order = 480 # Before HTTPDigestAuthHandler (== 490) |
|
287 | 287 | |
|
288 | 288 | def http_error_401(self, req, fp, code, msg, headers): |
|
289 | 289 | """Enforces that any authentication performed is HTTP Basic |
|
290 | 290 | Authentication. No authentication is also acceptable. |
|
291 | 291 | """ |
|
292 | 292 | authreq = headers.get('www-authenticate', None) |
|
293 | 293 | if authreq: |
|
294 | 294 | scheme = authreq.split()[0] |
|
295 | 295 | |
|
296 | 296 | if scheme.lower() != 'basic': |
|
297 | 297 | msg = _(b'the server must support Basic Authentication') |
|
298 | 298 | raise util.urlerr.httperror( |
|
299 | 299 | req.get_full_url(), |
|
300 | 300 | code, |
|
301 | 301 | encoding.strfromlocal(msg), |
|
302 | 302 | headers, |
|
303 | 303 | fp, |
|
304 | 304 | ) |
|
305 | 305 | return None |
|
306 | 306 | |
|
307 | 307 | |
|
308 | 308 | class _gitlfsremote(object): |
|
309 | 309 | def __init__(self, repo, url): |
|
310 | 310 | ui = repo.ui |
|
311 | 311 | self.ui = ui |
|
312 | 312 | baseurl, authinfo = url.authinfo() |
|
313 | 313 | self.baseurl = baseurl.rstrip(b'/') |
|
314 | 314 | useragent = repo.ui.config(b'experimental', b'lfs.user-agent') |
|
315 | 315 | if not useragent: |
|
316 | 316 | useragent = b'git-lfs/2.3.4 (Mercurial %s)' % util.version() |
|
317 | 317 | self.urlopener = urlmod.opener(ui, authinfo, useragent) |
|
318 | 318 | self.urlopener.add_handler(lfsauthhandler()) |
|
319 | 319 | self.retry = ui.configint(b'lfs', b'retry') |
|
320 | 320 | |
|
321 | 321 | def writebatch(self, pointers, fromstore): |
|
322 | 322 | """Batch upload from local to remote blobstore.""" |
|
323 | 323 | self._batch(_deduplicate(pointers), fromstore, b'upload') |
|
324 | 324 | |
|
325 | 325 | def readbatch(self, pointers, tostore): |
|
326 | 326 | """Batch download from remote to local blostore.""" |
|
327 | 327 | self._batch(_deduplicate(pointers), tostore, b'download') |
|
328 | 328 | |
|
329 | 329 | def _batchrequest(self, pointers, action): |
|
330 | 330 | """Get metadata about objects pointed by pointers for given action |
|
331 | 331 | |
|
332 | 332 | Return decoded JSON object like {'objects': [{'oid': '', 'size': 1}]} |
|
333 | 333 | See https://github.com/git-lfs/git-lfs/blob/master/docs/api/batch.md |
|
334 | 334 | """ |
|
335 | 335 | objects = [ |
|
336 | 336 | {'oid': pycompat.strurl(p.oid()), 'size': p.size()} |
|
337 | 337 | for p in pointers |
|
338 | 338 | ] |
|
339 | 339 | requestdata = pycompat.bytesurl( |
|
340 | 340 | json.dumps( |
|
341 | 341 | {'objects': objects, 'operation': pycompat.strurl(action),} |
|
342 | 342 | ) |
|
343 | 343 | ) |
|
344 | 344 | url = b'%s/objects/batch' % self.baseurl |
|
345 | 345 | batchreq = util.urlreq.request(pycompat.strurl(url), data=requestdata) |
|
346 | 346 | batchreq.add_header('Accept', 'application/vnd.git-lfs+json') |
|
347 | 347 | batchreq.add_header('Content-Type', 'application/vnd.git-lfs+json') |
|
348 | 348 | try: |
|
349 | 349 | with contextlib.closing(self.urlopener.open(batchreq)) as rsp: |
|
350 | 350 | rawjson = rsp.read() |
|
351 | 351 | except util.urlerr.httperror as ex: |
|
352 | 352 | hints = { |
|
353 | 353 | 400: _( |
|
354 | 354 | b'check that lfs serving is enabled on %s and "%s" is ' |
|
355 | 355 | b'supported' |
|
356 | 356 | ) |
|
357 | 357 | % (self.baseurl, action), |
|
358 | 358 | 404: _(b'the "lfs.url" config may be used to override %s') |
|
359 | 359 | % self.baseurl, |
|
360 | 360 | } |
|
361 | 361 | hint = hints.get(ex.code, _(b'api=%s, action=%s') % (url, action)) |
|
362 | 362 | raise LfsRemoteError( |
|
363 | 363 | _(b'LFS HTTP error: %s') % stringutil.forcebytestr(ex), |
|
364 | 364 | hint=hint, |
|
365 | 365 | ) |
|
366 | 366 | except util.urlerr.urlerror as ex: |
|
367 | 367 | hint = ( |
|
368 | 368 | _(b'the "lfs.url" config may be used to override %s') |
|
369 | 369 | % self.baseurl |
|
370 | 370 | ) |
|
371 | 371 | raise LfsRemoteError( |
|
372 | 372 | _(b'LFS error: %s') % _urlerrorreason(ex), hint=hint |
|
373 | 373 | ) |
|
374 | 374 | try: |
|
375 | 375 | response = pycompat.json_loads(rawjson) |
|
376 | 376 | except ValueError: |
|
377 | 377 | raise LfsRemoteError( |
|
378 | 378 | _(b'LFS server returns invalid JSON: %s') |
|
379 | 379 | % rawjson.encode("utf-8") |
|
380 | 380 | ) |
|
381 | 381 | |
|
382 | 382 | if self.ui.debugflag: |
|
383 | 383 | self.ui.debug(b'Status: %d\n' % rsp.status) |
|
384 | 384 | # lfs-test-server and hg serve return headers in different order |
|
385 | 385 | headers = pycompat.bytestr(rsp.info()).strip() |
|
386 | 386 | self.ui.debug(b'%s\n' % b'\n'.join(sorted(headers.splitlines()))) |
|
387 | 387 | |
|
388 | 388 | if 'objects' in response: |
|
389 | 389 | response['objects'] = sorted( |
|
390 | 390 | response['objects'], key=lambda p: p['oid'] |
|
391 | 391 | ) |
|
392 | 392 | self.ui.debug( |
|
393 | 393 | b'%s\n' |
|
394 | 394 | % pycompat.bytesurl( |
|
395 | 395 | json.dumps( |
|
396 | 396 | response, |
|
397 | 397 | indent=2, |
|
398 | 398 | separators=('', ': '), |
|
399 | 399 | sort_keys=True, |
|
400 | 400 | ) |
|
401 | 401 | ) |
|
402 | 402 | ) |
|
403 | 403 | |
|
404 | 404 | def encodestr(x): |
|
405 | 405 | if isinstance(x, pycompat.unicode): |
|
406 | 406 | return x.encode('utf-8') |
|
407 | 407 | return x |
|
408 | 408 | |
|
409 | 409 | return pycompat.rapply(encodestr, response) |
|
410 | 410 | |
|
411 | 411 | def _checkforservererror(self, pointers, responses, action): |
|
412 | 412 | """Scans errors from objects |
|
413 | 413 | |
|
414 | 414 | Raises LfsRemoteError if any objects have an error""" |
|
415 | 415 | for response in responses: |
|
416 | 416 | # The server should return 404 when objects cannot be found. Some |
|
417 | 417 | # server implementation (ex. lfs-test-server) does not set "error" |
|
418 | 418 | # but just removes "download" from "actions". Treat that case |
|
419 | 419 | # as the same as 404 error. |
|
420 | 420 | if b'error' not in response: |
|
421 | 421 | if action == b'download' and action not in response.get( |
|
422 | 422 | b'actions', [] |
|
423 | 423 | ): |
|
424 | 424 | code = 404 |
|
425 | 425 | else: |
|
426 | 426 | continue |
|
427 | 427 | else: |
|
428 | 428 | # An error dict without a code doesn't make much sense, so |
|
429 | 429 | # treat as a server error. |
|
430 | 430 | code = response.get(b'error').get(b'code', 500) |
|
431 | 431 | |
|
432 | 432 | ptrmap = {p.oid(): p for p in pointers} |
|
433 | 433 | p = ptrmap.get(response[b'oid'], None) |
|
434 | 434 | if p: |
|
435 | 435 | filename = getattr(p, 'filename', b'unknown') |
|
436 | 436 | errors = { |
|
437 | 437 | 404: b'The object does not exist', |
|
438 | 438 | 410: b'The object was removed by the owner', |
|
439 | 439 | 422: b'Validation error', |
|
440 | 440 | 500: b'Internal server error', |
|
441 | 441 | } |
|
442 | 442 | msg = errors.get(code, b'status code %d' % code) |
|
443 | 443 | raise LfsRemoteError( |
|
444 | 444 | _(b'LFS server error for "%s": %s') % (filename, msg) |
|
445 | 445 | ) |
|
446 | 446 | else: |
|
447 | 447 | raise LfsRemoteError( |
|
448 | 448 | _(b'LFS server error. Unsolicited response for oid %s') |
|
449 | 449 | % response[b'oid'] |
|
450 | 450 | ) |
|
451 | 451 | |
|
452 | 452 | def _extractobjects(self, response, pointers, action): |
|
453 | 453 | """extract objects from response of the batch API |
|
454 | 454 | |
|
455 | 455 | response: parsed JSON object returned by batch API |
|
456 | 456 | return response['objects'] filtered by action |
|
457 | 457 | raise if any object has an error |
|
458 | 458 | """ |
|
459 | 459 | # Scan errors from objects - fail early |
|
460 | 460 | objects = response.get(b'objects', []) |
|
461 | 461 | self._checkforservererror(pointers, objects, action) |
|
462 | 462 | |
|
463 | 463 | # Filter objects with given action. Practically, this skips uploading |
|
464 | 464 | # objects which exist in the server. |
|
465 | 465 | filteredobjects = [ |
|
466 | 466 | o for o in objects if action in o.get(b'actions', []) |
|
467 | 467 | ] |
|
468 | 468 | |
|
469 | 469 | return filteredobjects |
|
470 | 470 | |
|
471 | 471 | def _basictransfer(self, obj, action, localstore): |
|
472 | 472 | """Download or upload a single object using basic transfer protocol |
|
473 | 473 | |
|
474 | 474 | obj: dict, an object description returned by batch API |
|
475 | 475 | action: string, one of ['upload', 'download'] |
|
476 | 476 | localstore: blobstore.local |
|
477 | 477 | |
|
478 | 478 | See https://github.com/git-lfs/git-lfs/blob/master/docs/api/\ |
|
479 | 479 | basic-transfers.md |
|
480 | 480 | """ |
|
481 | 481 | oid = obj[b'oid'] |
|
482 | 482 | href = obj[b'actions'][action].get(b'href') |
|
483 | 483 | headers = obj[b'actions'][action].get(b'header', {}).items() |
|
484 | 484 | |
|
485 | 485 | request = util.urlreq.request(pycompat.strurl(href)) |
|
486 | 486 | if action == b'upload': |
|
487 | 487 | # If uploading blobs, read data from local blobstore. |
|
488 | 488 | if not localstore.verify(oid): |
|
489 | 489 | raise error.Abort( |
|
490 | 490 | _(b'detected corrupt lfs object: %s') % oid, |
|
491 | 491 | hint=_(b'run hg verify'), |
|
492 | 492 | ) |
|
493 | 493 | |
|
494 | 494 | for k, v in headers: |
|
495 | 495 | request.add_header(pycompat.strurl(k), pycompat.strurl(v)) |
|
496 | 496 | |
|
497 | 497 | try: |
|
498 | 498 | if action == b'upload': |
|
499 | 499 | request.data = lfsuploadfile(self.ui, localstore.path(oid)) |
|
500 | 500 | request.get_method = lambda: 'PUT' |
|
501 | 501 | request.add_header('Content-Type', 'application/octet-stream') |
|
502 | 502 | request.add_header('Content-Length', request.data.length) |
|
503 | 503 | |
|
504 | 504 | with contextlib.closing(self.urlopener.open(request)) as res: |
|
505 | 505 | contentlength = res.info().get(b"content-length") |
|
506 | 506 | ui = self.ui # Shorten debug lines |
|
507 | 507 | if self.ui.debugflag: |
|
508 | 508 | ui.debug(b'Status: %d\n' % res.status) |
|
509 | 509 | # lfs-test-server and hg serve return headers in different |
|
510 | 510 | # order |
|
511 | 511 | headers = pycompat.bytestr(res.info()).strip() |
|
512 | 512 | ui.debug(b'%s\n' % b'\n'.join(sorted(headers.splitlines()))) |
|
513 | 513 | |
|
514 | 514 | if action == b'download': |
|
515 | 515 | # If downloading blobs, store downloaded data to local |
|
516 | 516 | # blobstore |
|
517 | 517 | localstore.download(oid, res, contentlength) |
|
518 | 518 | else: |
|
519 | 519 | blocks = [] |
|
520 | 520 | while True: |
|
521 | 521 | data = res.read(1048576) |
|
522 | 522 | if not data: |
|
523 | 523 | break |
|
524 | 524 | blocks.append(data) |
|
525 | 525 | |
|
526 | 526 | response = b"".join(blocks) |
|
527 | 527 | if response: |
|
528 | 528 | ui.debug(b'lfs %s response: %s' % (action, response)) |
|
529 | 529 | except util.urlerr.httperror as ex: |
|
530 | 530 | if self.ui.debugflag: |
|
531 | 531 | self.ui.debug( |
|
532 | 532 | b'%s: %s\n' % (oid, ex.read()) |
|
533 | 533 | ) # XXX: also bytes? |
|
534 | 534 | raise LfsRemoteError( |
|
535 | 535 | _(b'LFS HTTP error: %s (oid=%s, action=%s)') |
|
536 | 536 | % (stringutil.forcebytestr(ex), oid, action) |
|
537 | 537 | ) |
|
538 | 538 | except util.urlerr.urlerror as ex: |
|
539 | 539 | hint = _(b'attempted connection to %s') % pycompat.bytesurl( |
|
540 | 540 | util.urllibcompat.getfullurl(request) |
|
541 | 541 | ) |
|
542 | 542 | raise LfsRemoteError( |
|
543 | 543 | _(b'LFS error: %s') % _urlerrorreason(ex), hint=hint |
|
544 | 544 | ) |
|
545 | 545 | finally: |
|
546 | 546 | if request.data: |
|
547 | 547 | request.data.close() |
|
548 | 548 | |
|
549 | 549 | def _batch(self, pointers, localstore, action): |
|
550 | 550 | if action not in [b'upload', b'download']: |
|
551 | 551 | raise error.ProgrammingError(b'invalid Git-LFS action: %s' % action) |
|
552 | 552 | |
|
553 | 553 | response = self._batchrequest(pointers, action) |
|
554 | 554 | objects = self._extractobjects(response, pointers, action) |
|
555 | 555 | total = sum(x.get(b'size', 0) for x in objects) |
|
556 | 556 | sizes = {} |
|
557 | 557 | for obj in objects: |
|
558 | 558 | sizes[obj.get(b'oid')] = obj.get(b'size', 0) |
|
559 | 559 | topic = { |
|
560 | 560 | b'upload': _(b'lfs uploading'), |
|
561 | 561 | b'download': _(b'lfs downloading'), |
|
562 | 562 | }[action] |
|
563 | 563 | if len(objects) > 1: |
|
564 | 564 | self.ui.note( |
|
565 | 565 | _(b'lfs: need to transfer %d objects (%s)\n') |
|
566 | 566 | % (len(objects), util.bytecount(total)) |
|
567 | 567 | ) |
|
568 | 568 | |
|
569 | 569 | def transfer(chunk): |
|
570 | 570 | for obj in chunk: |
|
571 | 571 | objsize = obj.get(b'size', 0) |
|
572 | 572 | if self.ui.verbose: |
|
573 | 573 | if action == b'download': |
|
574 | 574 | msg = _(b'lfs: downloading %s (%s)\n') |
|
575 | 575 | elif action == b'upload': |
|
576 | 576 | msg = _(b'lfs: uploading %s (%s)\n') |
|
577 | 577 | self.ui.note( |
|
578 | 578 | msg % (obj.get(b'oid'), util.bytecount(objsize)) |
|
579 | 579 | ) |
|
580 | 580 | retry = self.retry |
|
581 | 581 | while True: |
|
582 | 582 | try: |
|
583 | 583 | self._basictransfer(obj, action, localstore) |
|
584 | 584 | yield 1, obj.get(b'oid') |
|
585 | 585 | break |
|
586 | 586 | except socket.error as ex: |
|
587 | 587 | if retry > 0: |
|
588 | 588 | self.ui.note( |
|
589 | 589 | _(b'lfs: failed: %r (remaining retry %d)\n') |
|
590 | 590 | % (stringutil.forcebytestr(ex), retry) |
|
591 | 591 | ) |
|
592 | 592 | retry -= 1 |
|
593 | 593 | continue |
|
594 | 594 | raise |
|
595 | 595 | |
|
596 | 596 | # Until https multiplexing gets sorted out |
|
597 | 597 | if self.ui.configbool(b'experimental', b'lfs.worker-enable'): |
|
598 | 598 | oids = worker.worker( |
|
599 | 599 | self.ui, |
|
600 | 600 | 0.1, |
|
601 | 601 | transfer, |
|
602 | 602 | (), |
|
603 | 603 | sorted(objects, key=lambda o: o.get(b'oid')), |
|
604 | 604 | ) |
|
605 | 605 | else: |
|
606 | 606 | oids = transfer(sorted(objects, key=lambda o: o.get(b'oid'))) |
|
607 | 607 | |
|
608 | 608 | with self.ui.makeprogress( |
|
609 | 609 | topic, unit=_(b"bytes"), total=total |
|
610 | 610 | ) as progress: |
|
611 | 611 | progress.update(0) |
|
612 | 612 | processed = 0 |
|
613 | 613 | blobs = 0 |
|
614 | 614 | for _one, oid in oids: |
|
615 | 615 | processed += sizes[oid] |
|
616 | 616 | blobs += 1 |
|
617 | 617 | progress.update(processed) |
|
618 | 618 | self.ui.note(_(b'lfs: processed: %s\n') % oid) |
|
619 | 619 | |
|
620 | 620 | if blobs > 0: |
|
621 | 621 | if action == b'upload': |
|
622 | 622 | self.ui.status( |
|
623 | 623 | _(b'lfs: uploaded %d files (%s)\n') |
|
624 | 624 | % (blobs, util.bytecount(processed)) |
|
625 | 625 | ) |
|
626 | 626 | elif action == b'download': |
|
627 | 627 | self.ui.status( |
|
628 | 628 | _(b'lfs: downloaded %d files (%s)\n') |
|
629 | 629 | % (blobs, util.bytecount(processed)) |
|
630 | 630 | ) |
|
631 | 631 | |
|
632 | 632 | def __del__(self): |
|
633 | 633 | # copied from mercurial/httppeer.py |
|
634 | 634 | urlopener = getattr(self, 'urlopener', None) |
|
635 | 635 | if urlopener: |
|
636 | 636 | for h in urlopener.handlers: |
|
637 | 637 | h.close() |
|
638 | 638 | getattr(h, "close_all", lambda: None)() |
|
639 | 639 | |
|
640 | 640 | |
|
641 | 641 | class _dummyremote(object): |
|
642 | 642 | """Dummy store storing blobs to temp directory.""" |
|
643 | 643 | |
|
644 | 644 | def __init__(self, repo, url): |
|
645 | 645 | fullpath = repo.vfs.join(b'lfs', url.path) |
|
646 | 646 | self.vfs = lfsvfs(fullpath) |
|
647 | 647 | |
|
648 | 648 | def writebatch(self, pointers, fromstore): |
|
649 | 649 | for p in _deduplicate(pointers): |
|
650 | 650 | content = fromstore.read(p.oid(), verify=True) |
|
651 | 651 | with self.vfs(p.oid(), b'wb', atomictemp=True) as fp: |
|
652 | 652 | fp.write(content) |
|
653 | 653 | |
|
654 | 654 | def readbatch(self, pointers, tostore): |
|
655 | 655 | for p in _deduplicate(pointers): |
|
656 | 656 | with self.vfs(p.oid(), b'rb') as fp: |
|
657 | 657 | tostore.download(p.oid(), fp, None) |
|
658 | 658 | |
|
659 | 659 | |
|
660 | 660 | class _nullremote(object): |
|
661 | 661 | """Null store storing blobs to /dev/null.""" |
|
662 | 662 | |
|
663 | 663 | def __init__(self, repo, url): |
|
664 | 664 | pass |
|
665 | 665 | |
|
666 | 666 | def writebatch(self, pointers, fromstore): |
|
667 | 667 | pass |
|
668 | 668 | |
|
669 | 669 | def readbatch(self, pointers, tostore): |
|
670 | 670 | pass |
|
671 | 671 | |
|
672 | 672 | |
|
673 | 673 | class _promptremote(object): |
|
674 | 674 | """Prompt user to set lfs.url when accessed.""" |
|
675 | 675 | |
|
676 | 676 | def __init__(self, repo, url): |
|
677 | 677 | pass |
|
678 | 678 | |
|
679 | 679 | def writebatch(self, pointers, fromstore, ui=None): |
|
680 | 680 | self._prompt() |
|
681 | 681 | |
|
682 | 682 | def readbatch(self, pointers, tostore, ui=None): |
|
683 | 683 | self._prompt() |
|
684 | 684 | |
|
685 | 685 | def _prompt(self): |
|
686 | 686 | raise error.Abort(_(b'lfs.url needs to be configured')) |
|
687 | 687 | |
|
688 | 688 | |
|
689 | 689 | _storemap = { |
|
690 | 690 | b'https': _gitlfsremote, |
|
691 | 691 | b'http': _gitlfsremote, |
|
692 | 692 | b'file': _dummyremote, |
|
693 | 693 | b'null': _nullremote, |
|
694 | 694 | None: _promptremote, |
|
695 | 695 | } |
|
696 | 696 | |
|
697 | 697 | |
|
698 | 698 | def _deduplicate(pointers): |
|
699 | 699 | """Remove any duplicate oids that exist in the list""" |
|
700 | 700 | reduced = util.sortdict() |
|
701 | 701 | for p in pointers: |
|
702 | 702 | reduced[p.oid()] = p |
|
703 | 703 | return reduced.values() |
|
704 | 704 | |
|
705 | 705 | |
|
706 | 706 | def _verify(oid, content): |
|
707 | 707 | realoid = node.hex(hashlib.sha256(content).digest()) |
|
708 | 708 | if realoid != oid: |
|
709 | 709 | raise LfsCorruptionError( |
|
710 | 710 | _(b'detected corrupt lfs object: %s') % oid, |
|
711 | 711 | hint=_(b'run hg verify'), |
|
712 | 712 | ) |
|
713 | 713 | |
|
714 | 714 | |
|
715 | 715 | def remote(repo, remote=None): |
|
716 | 716 | """remotestore factory. return a store in _storemap depending on config |
|
717 | 717 | |
|
718 | 718 | If ``lfs.url`` is specified, use that remote endpoint. Otherwise, try to |
|
719 | 719 | infer the endpoint, based on the remote repository using the same path |
|
720 | 720 | adjustments as git. As an extension, 'http' is supported as well so that |
|
721 | 721 | ``hg serve`` works out of the box. |
|
722 | 722 | |
|
723 | 723 | https://github.com/git-lfs/git-lfs/blob/master/docs/api/server-discovery.md |
|
724 | 724 | """ |
|
725 | 725 | lfsurl = repo.ui.config(b'lfs', b'url') |
|
726 | 726 | url = util.url(lfsurl or b'') |
|
727 | 727 | if lfsurl is None: |
|
728 | 728 | if remote: |
|
729 | 729 | path = remote |
|
730 | 730 | elif util.safehasattr(repo, b'_subtoppath'): |
|
731 | 731 | # The pull command sets this during the optional update phase, which |
|
732 | 732 | # tells exactly where the pull originated, whether 'paths.default' |
|
733 | 733 | # or explicit. |
|
734 | 734 | path = repo._subtoppath |
|
735 | 735 | else: |
|
736 | 736 | # TODO: investigate 'paths.remote:lfsurl' style path customization, |
|
737 | 737 | # and fall back to inferring from 'paths.remote' if unspecified. |
|
738 | 738 | path = repo.ui.config(b'paths', b'default') or b'' |
|
739 | 739 | |
|
740 | 740 | defaulturl = util.url(path) |
|
741 | 741 | |
|
742 | 742 | # TODO: support local paths as well. |
|
743 | 743 | # TODO: consider the ssh -> https transformation that git applies |
|
744 | 744 | if defaulturl.scheme in (b'http', b'https'): |
|
745 | 745 | if defaulturl.path and defaulturl.path[:-1] != b'/': |
|
746 | 746 | defaulturl.path += b'/' |
|
747 | 747 | defaulturl.path = (defaulturl.path or b'') + b'.git/info/lfs' |
|
748 | 748 | |
|
749 | 749 | url = util.url(bytes(defaulturl)) |
|
750 | 750 | repo.ui.note(_(b'lfs: assuming remote store: %s\n') % url) |
|
751 | 751 | |
|
752 | 752 | scheme = url.scheme |
|
753 | 753 | if scheme not in _storemap: |
|
754 | 754 | raise error.Abort(_(b'lfs: unknown url scheme: %s') % scheme) |
|
755 | 755 | return _storemap[scheme](repo, url) |
|
756 | 756 | |
|
757 | 757 | |
|
758 | 758 | class LfsRemoteError(error.StorageError): |
|
759 | 759 | pass |
|
760 | 760 | |
|
761 | 761 | |
|
762 | 762 | class LfsCorruptionError(error.Abort): |
|
763 | 763 | """Raised when a corrupt blob is detected, aborting an operation |
|
764 | 764 | |
|
765 | 765 | It exists to allow specialized handling on the server side.""" |
General Comments 0
You need to be logged in to leave comments.
Login now