Show More
@@ -1,1532 +1,1534 | |||||
1 | # bundle2.py - generic container format to transmit arbitrary data. |
|
1 | # bundle2.py - generic container format to transmit arbitrary data. | |
2 | # |
|
2 | # | |
3 | # Copyright 2013 Facebook, Inc. |
|
3 | # Copyright 2013 Facebook, Inc. | |
4 | # |
|
4 | # | |
5 | # This software may be used and distributed according to the terms of the |
|
5 | # This software may be used and distributed according to the terms of the | |
6 | # GNU General Public License version 2 or any later version. |
|
6 | # GNU General Public License version 2 or any later version. | |
7 | """Handling of the new bundle2 format |
|
7 | """Handling of the new bundle2 format | |
8 |
|
8 | |||
9 | The goal of bundle2 is to act as an atomically packet to transmit a set of |
|
9 | The goal of bundle2 is to act as an atomically packet to transmit a set of | |
10 | payloads in an application agnostic way. It consist in a sequence of "parts" |
|
10 | payloads in an application agnostic way. It consist in a sequence of "parts" | |
11 | that will be handed to and processed by the application layer. |
|
11 | that will be handed to and processed by the application layer. | |
12 |
|
12 | |||
13 |
|
13 | |||
14 | General format architecture |
|
14 | General format architecture | |
15 | =========================== |
|
15 | =========================== | |
16 |
|
16 | |||
17 | The format is architectured as follow |
|
17 | The format is architectured as follow | |
18 |
|
18 | |||
19 | - magic string |
|
19 | - magic string | |
20 | - stream level parameters |
|
20 | - stream level parameters | |
21 | - payload parts (any number) |
|
21 | - payload parts (any number) | |
22 | - end of stream marker. |
|
22 | - end of stream marker. | |
23 |
|
23 | |||
24 | the Binary format |
|
24 | the Binary format | |
25 | ============================ |
|
25 | ============================ | |
26 |
|
26 | |||
27 | All numbers are unsigned and big-endian. |
|
27 | All numbers are unsigned and big-endian. | |
28 |
|
28 | |||
29 | stream level parameters |
|
29 | stream level parameters | |
30 | ------------------------ |
|
30 | ------------------------ | |
31 |
|
31 | |||
32 | Binary format is as follow |
|
32 | Binary format is as follow | |
33 |
|
33 | |||
34 | :params size: int32 |
|
34 | :params size: int32 | |
35 |
|
35 | |||
36 | The total number of Bytes used by the parameters |
|
36 | The total number of Bytes used by the parameters | |
37 |
|
37 | |||
38 | :params value: arbitrary number of Bytes |
|
38 | :params value: arbitrary number of Bytes | |
39 |
|
39 | |||
40 | A blob of `params size` containing the serialized version of all stream level |
|
40 | A blob of `params size` containing the serialized version of all stream level | |
41 | parameters. |
|
41 | parameters. | |
42 |
|
42 | |||
43 | The blob contains a space separated list of parameters. Parameters with value |
|
43 | The blob contains a space separated list of parameters. Parameters with value | |
44 | are stored in the form `<name>=<value>`. Both name and value are urlquoted. |
|
44 | are stored in the form `<name>=<value>`. Both name and value are urlquoted. | |
45 |
|
45 | |||
46 | Empty name are obviously forbidden. |
|
46 | Empty name are obviously forbidden. | |
47 |
|
47 | |||
48 | Name MUST start with a letter. If this first letter is lower case, the |
|
48 | Name MUST start with a letter. If this first letter is lower case, the | |
49 | parameter is advisory and can be safely ignored. However when the first |
|
49 | parameter is advisory and can be safely ignored. However when the first | |
50 | letter is capital, the parameter is mandatory and the bundling process MUST |
|
50 | letter is capital, the parameter is mandatory and the bundling process MUST | |
51 | stop if he is not able to proceed it. |
|
51 | stop if he is not able to proceed it. | |
52 |
|
52 | |||
53 | Stream parameters use a simple textual format for two main reasons: |
|
53 | Stream parameters use a simple textual format for two main reasons: | |
54 |
|
54 | |||
55 | - Stream level parameters should remain simple and we want to discourage any |
|
55 | - Stream level parameters should remain simple and we want to discourage any | |
56 | crazy usage. |
|
56 | crazy usage. | |
57 | - Textual data allow easy human inspection of a bundle2 header in case of |
|
57 | - Textual data allow easy human inspection of a bundle2 header in case of | |
58 | troubles. |
|
58 | troubles. | |
59 |
|
59 | |||
60 | Any Applicative level options MUST go into a bundle2 part instead. |
|
60 | Any Applicative level options MUST go into a bundle2 part instead. | |
61 |
|
61 | |||
62 | Payload part |
|
62 | Payload part | |
63 | ------------------------ |
|
63 | ------------------------ | |
64 |
|
64 | |||
65 | Binary format is as follow |
|
65 | Binary format is as follow | |
66 |
|
66 | |||
67 | :header size: int32 |
|
67 | :header size: int32 | |
68 |
|
68 | |||
69 | The total number of Bytes used by the part header. When the header is empty |
|
69 | The total number of Bytes used by the part header. When the header is empty | |
70 | (size = 0) this is interpreted as the end of stream marker. |
|
70 | (size = 0) this is interpreted as the end of stream marker. | |
71 |
|
71 | |||
72 | :header: |
|
72 | :header: | |
73 |
|
73 | |||
74 | The header defines how to interpret the part. It contains two piece of |
|
74 | The header defines how to interpret the part. It contains two piece of | |
75 | data: the part type, and the part parameters. |
|
75 | data: the part type, and the part parameters. | |
76 |
|
76 | |||
77 | The part type is used to route an application level handler, that can |
|
77 | The part type is used to route an application level handler, that can | |
78 | interpret payload. |
|
78 | interpret payload. | |
79 |
|
79 | |||
80 | Part parameters are passed to the application level handler. They are |
|
80 | Part parameters are passed to the application level handler. They are | |
81 | meant to convey information that will help the application level object to |
|
81 | meant to convey information that will help the application level object to | |
82 | interpret the part payload. |
|
82 | interpret the part payload. | |
83 |
|
83 | |||
84 | The binary format of the header is has follow |
|
84 | The binary format of the header is has follow | |
85 |
|
85 | |||
86 | :typesize: (one byte) |
|
86 | :typesize: (one byte) | |
87 |
|
87 | |||
88 | :parttype: alphanumerical part name (restricted to [a-zA-Z0-9_:-]*) |
|
88 | :parttype: alphanumerical part name (restricted to [a-zA-Z0-9_:-]*) | |
89 |
|
89 | |||
90 | :partid: A 32bits integer (unique in the bundle) that can be used to refer |
|
90 | :partid: A 32bits integer (unique in the bundle) that can be used to refer | |
91 | to this part. |
|
91 | to this part. | |
92 |
|
92 | |||
93 | :parameters: |
|
93 | :parameters: | |
94 |
|
94 | |||
95 | Part's parameter may have arbitrary content, the binary structure is:: |
|
95 | Part's parameter may have arbitrary content, the binary structure is:: | |
96 |
|
96 | |||
97 | <mandatory-count><advisory-count><param-sizes><param-data> |
|
97 | <mandatory-count><advisory-count><param-sizes><param-data> | |
98 |
|
98 | |||
99 | :mandatory-count: 1 byte, number of mandatory parameters |
|
99 | :mandatory-count: 1 byte, number of mandatory parameters | |
100 |
|
100 | |||
101 | :advisory-count: 1 byte, number of advisory parameters |
|
101 | :advisory-count: 1 byte, number of advisory parameters | |
102 |
|
102 | |||
103 | :param-sizes: |
|
103 | :param-sizes: | |
104 |
|
104 | |||
105 | N couple of bytes, where N is the total number of parameters. Each |
|
105 | N couple of bytes, where N is the total number of parameters. Each | |
106 | couple contains (<size-of-key>, <size-of-value) for one parameter. |
|
106 | couple contains (<size-of-key>, <size-of-value) for one parameter. | |
107 |
|
107 | |||
108 | :param-data: |
|
108 | :param-data: | |
109 |
|
109 | |||
110 | A blob of bytes from which each parameter key and value can be |
|
110 | A blob of bytes from which each parameter key and value can be | |
111 | retrieved using the list of size couples stored in the previous |
|
111 | retrieved using the list of size couples stored in the previous | |
112 | field. |
|
112 | field. | |
113 |
|
113 | |||
114 | Mandatory parameters comes first, then the advisory ones. |
|
114 | Mandatory parameters comes first, then the advisory ones. | |
115 |
|
115 | |||
116 | Each parameter's key MUST be unique within the part. |
|
116 | Each parameter's key MUST be unique within the part. | |
117 |
|
117 | |||
118 | :payload: |
|
118 | :payload: | |
119 |
|
119 | |||
120 | payload is a series of `<chunksize><chunkdata>`. |
|
120 | payload is a series of `<chunksize><chunkdata>`. | |
121 |
|
121 | |||
122 | `chunksize` is an int32, `chunkdata` are plain bytes (as much as |
|
122 | `chunksize` is an int32, `chunkdata` are plain bytes (as much as | |
123 | `chunksize` says)` The payload part is concluded by a zero size chunk. |
|
123 | `chunksize` says)` The payload part is concluded by a zero size chunk. | |
124 |
|
124 | |||
125 | The current implementation always produces either zero or one chunk. |
|
125 | The current implementation always produces either zero or one chunk. | |
126 | This is an implementation limitation that will ultimately be lifted. |
|
126 | This is an implementation limitation that will ultimately be lifted. | |
127 |
|
127 | |||
128 | `chunksize` can be negative to trigger special case processing. No such |
|
128 | `chunksize` can be negative to trigger special case processing. No such | |
129 | processing is in place yet. |
|
129 | processing is in place yet. | |
130 |
|
130 | |||
131 | Bundle processing |
|
131 | Bundle processing | |
132 | ============================ |
|
132 | ============================ | |
133 |
|
133 | |||
134 | Each part is processed in order using a "part handler". Handler are registered |
|
134 | Each part is processed in order using a "part handler". Handler are registered | |
135 | for a certain part type. |
|
135 | for a certain part type. | |
136 |
|
136 | |||
137 | The matching of a part to its handler is case insensitive. The case of the |
|
137 | The matching of a part to its handler is case insensitive. The case of the | |
138 | part type is used to know if a part is mandatory or advisory. If the Part type |
|
138 | part type is used to know if a part is mandatory or advisory. If the Part type | |
139 | contains any uppercase char it is considered mandatory. When no handler is |
|
139 | contains any uppercase char it is considered mandatory. When no handler is | |
140 | known for a Mandatory part, the process is aborted and an exception is raised. |
|
140 | known for a Mandatory part, the process is aborted and an exception is raised. | |
141 | If the part is advisory and no handler is known, the part is ignored. When the |
|
141 | If the part is advisory and no handler is known, the part is ignored. When the | |
142 | process is aborted, the full bundle is still read from the stream to keep the |
|
142 | process is aborted, the full bundle is still read from the stream to keep the | |
143 | channel usable. But none of the part read from an abort are processed. In the |
|
143 | channel usable. But none of the part read from an abort are processed. In the | |
144 | future, dropping the stream may become an option for channel we do not care to |
|
144 | future, dropping the stream may become an option for channel we do not care to | |
145 | preserve. |
|
145 | preserve. | |
146 | """ |
|
146 | """ | |
147 |
|
147 | |||
148 | from __future__ import absolute_import |
|
148 | from __future__ import absolute_import | |
149 |
|
149 | |||
150 | import errno |
|
150 | import errno | |
151 | import re |
|
151 | import re | |
152 | import string |
|
152 | import string | |
153 | import struct |
|
153 | import struct | |
154 | import sys |
|
154 | import sys | |
155 | import urllib |
|
155 | import urllib | |
156 |
|
156 | |||
157 | from .i18n import _ |
|
157 | from .i18n import _ | |
158 | from . import ( |
|
158 | from . import ( | |
159 | changegroup, |
|
159 | changegroup, | |
160 | error, |
|
160 | error, | |
161 | obsolete, |
|
161 | obsolete, | |
162 | pushkey, |
|
162 | pushkey, | |
163 | tags, |
|
163 | tags, | |
164 | url, |
|
164 | url, | |
165 | util, |
|
165 | util, | |
166 | ) |
|
166 | ) | |
167 |
|
167 | |||
168 | _pack = struct.pack |
|
168 | _pack = struct.pack | |
169 | _unpack = struct.unpack |
|
169 | _unpack = struct.unpack | |
170 |
|
170 | |||
171 | _fstreamparamsize = '>i' |
|
171 | _fstreamparamsize = '>i' | |
172 | _fpartheadersize = '>i' |
|
172 | _fpartheadersize = '>i' | |
173 | _fparttypesize = '>B' |
|
173 | _fparttypesize = '>B' | |
174 | _fpartid = '>I' |
|
174 | _fpartid = '>I' | |
175 | _fpayloadsize = '>i' |
|
175 | _fpayloadsize = '>i' | |
176 | _fpartparamcount = '>BB' |
|
176 | _fpartparamcount = '>BB' | |
177 |
|
177 | |||
178 | preferedchunksize = 4096 |
|
178 | preferedchunksize = 4096 | |
179 |
|
179 | |||
180 | _parttypeforbidden = re.compile('[^a-zA-Z0-9_:-]') |
|
180 | _parttypeforbidden = re.compile('[^a-zA-Z0-9_:-]') | |
181 |
|
181 | |||
182 | def outdebug(ui, message): |
|
182 | def outdebug(ui, message): | |
183 | """debug regarding output stream (bundling)""" |
|
183 | """debug regarding output stream (bundling)""" | |
184 | if ui.configbool('devel', 'bundle2.debug', False): |
|
184 | if ui.configbool('devel', 'bundle2.debug', False): | |
185 | ui.debug('bundle2-output: %s\n' % message) |
|
185 | ui.debug('bundle2-output: %s\n' % message) | |
186 |
|
186 | |||
187 | def indebug(ui, message): |
|
187 | def indebug(ui, message): | |
188 | """debug on input stream (unbundling)""" |
|
188 | """debug on input stream (unbundling)""" | |
189 | if ui.configbool('devel', 'bundle2.debug', False): |
|
189 | if ui.configbool('devel', 'bundle2.debug', False): | |
190 | ui.debug('bundle2-input: %s\n' % message) |
|
190 | ui.debug('bundle2-input: %s\n' % message) | |
191 |
|
191 | |||
192 | def validateparttype(parttype): |
|
192 | def validateparttype(parttype): | |
193 | """raise ValueError if a parttype contains invalid character""" |
|
193 | """raise ValueError if a parttype contains invalid character""" | |
194 | if _parttypeforbidden.search(parttype): |
|
194 | if _parttypeforbidden.search(parttype): | |
195 | raise ValueError(parttype) |
|
195 | raise ValueError(parttype) | |
196 |
|
196 | |||
197 | def _makefpartparamsizes(nbparams): |
|
197 | def _makefpartparamsizes(nbparams): | |
198 | """return a struct format to read part parameter sizes |
|
198 | """return a struct format to read part parameter sizes | |
199 |
|
199 | |||
200 | The number parameters is variable so we need to build that format |
|
200 | The number parameters is variable so we need to build that format | |
201 | dynamically. |
|
201 | dynamically. | |
202 | """ |
|
202 | """ | |
203 | return '>'+('BB'*nbparams) |
|
203 | return '>'+('BB'*nbparams) | |
204 |
|
204 | |||
205 | parthandlermapping = {} |
|
205 | parthandlermapping = {} | |
206 |
|
206 | |||
207 | def parthandler(parttype, params=()): |
|
207 | def parthandler(parttype, params=()): | |
208 | """decorator that register a function as a bundle2 part handler |
|
208 | """decorator that register a function as a bundle2 part handler | |
209 |
|
209 | |||
210 | eg:: |
|
210 | eg:: | |
211 |
|
211 | |||
212 | @parthandler('myparttype', ('mandatory', 'param', 'handled')) |
|
212 | @parthandler('myparttype', ('mandatory', 'param', 'handled')) | |
213 | def myparttypehandler(...): |
|
213 | def myparttypehandler(...): | |
214 | '''process a part of type "my part".''' |
|
214 | '''process a part of type "my part".''' | |
215 | ... |
|
215 | ... | |
216 | """ |
|
216 | """ | |
217 | validateparttype(parttype) |
|
217 | validateparttype(parttype) | |
218 | def _decorator(func): |
|
218 | def _decorator(func): | |
219 | lparttype = parttype.lower() # enforce lower case matching. |
|
219 | lparttype = parttype.lower() # enforce lower case matching. | |
220 | assert lparttype not in parthandlermapping |
|
220 | assert lparttype not in parthandlermapping | |
221 | parthandlermapping[lparttype] = func |
|
221 | parthandlermapping[lparttype] = func | |
222 | func.params = frozenset(params) |
|
222 | func.params = frozenset(params) | |
223 | return func |
|
223 | return func | |
224 | return _decorator |
|
224 | return _decorator | |
225 |
|
225 | |||
226 | class unbundlerecords(object): |
|
226 | class unbundlerecords(object): | |
227 | """keep record of what happens during and unbundle |
|
227 | """keep record of what happens during and unbundle | |
228 |
|
228 | |||
229 | New records are added using `records.add('cat', obj)`. Where 'cat' is a |
|
229 | New records are added using `records.add('cat', obj)`. Where 'cat' is a | |
230 | category of record and obj is an arbitrary object. |
|
230 | category of record and obj is an arbitrary object. | |
231 |
|
231 | |||
232 | `records['cat']` will return all entries of this category 'cat'. |
|
232 | `records['cat']` will return all entries of this category 'cat'. | |
233 |
|
233 | |||
234 | Iterating on the object itself will yield `('category', obj)` tuples |
|
234 | Iterating on the object itself will yield `('category', obj)` tuples | |
235 | for all entries. |
|
235 | for all entries. | |
236 |
|
236 | |||
237 | All iterations happens in chronological order. |
|
237 | All iterations happens in chronological order. | |
238 | """ |
|
238 | """ | |
239 |
|
239 | |||
240 | def __init__(self): |
|
240 | def __init__(self): | |
241 | self._categories = {} |
|
241 | self._categories = {} | |
242 | self._sequences = [] |
|
242 | self._sequences = [] | |
243 | self._replies = {} |
|
243 | self._replies = {} | |
244 |
|
244 | |||
245 | def add(self, category, entry, inreplyto=None): |
|
245 | def add(self, category, entry, inreplyto=None): | |
246 | """add a new record of a given category. |
|
246 | """add a new record of a given category. | |
247 |
|
247 | |||
248 | The entry can then be retrieved in the list returned by |
|
248 | The entry can then be retrieved in the list returned by | |
249 | self['category'].""" |
|
249 | self['category'].""" | |
250 | self._categories.setdefault(category, []).append(entry) |
|
250 | self._categories.setdefault(category, []).append(entry) | |
251 | self._sequences.append((category, entry)) |
|
251 | self._sequences.append((category, entry)) | |
252 | if inreplyto is not None: |
|
252 | if inreplyto is not None: | |
253 | self.getreplies(inreplyto).add(category, entry) |
|
253 | self.getreplies(inreplyto).add(category, entry) | |
254 |
|
254 | |||
255 | def getreplies(self, partid): |
|
255 | def getreplies(self, partid): | |
256 | """get the records that are replies to a specific part""" |
|
256 | """get the records that are replies to a specific part""" | |
257 | return self._replies.setdefault(partid, unbundlerecords()) |
|
257 | return self._replies.setdefault(partid, unbundlerecords()) | |
258 |
|
258 | |||
259 | def __getitem__(self, cat): |
|
259 | def __getitem__(self, cat): | |
260 | return tuple(self._categories.get(cat, ())) |
|
260 | return tuple(self._categories.get(cat, ())) | |
261 |
|
261 | |||
262 | def __iter__(self): |
|
262 | def __iter__(self): | |
263 | return iter(self._sequences) |
|
263 | return iter(self._sequences) | |
264 |
|
264 | |||
265 | def __len__(self): |
|
265 | def __len__(self): | |
266 | return len(self._sequences) |
|
266 | return len(self._sequences) | |
267 |
|
267 | |||
268 | def __nonzero__(self): |
|
268 | def __nonzero__(self): | |
269 | return bool(self._sequences) |
|
269 | return bool(self._sequences) | |
270 |
|
270 | |||
271 | class bundleoperation(object): |
|
271 | class bundleoperation(object): | |
272 | """an object that represents a single bundling process |
|
272 | """an object that represents a single bundling process | |
273 |
|
273 | |||
274 | Its purpose is to carry unbundle-related objects and states. |
|
274 | Its purpose is to carry unbundle-related objects and states. | |
275 |
|
275 | |||
276 | A new object should be created at the beginning of each bundle processing. |
|
276 | A new object should be created at the beginning of each bundle processing. | |
277 | The object is to be returned by the processing function. |
|
277 | The object is to be returned by the processing function. | |
278 |
|
278 | |||
279 | The object has very little content now it will ultimately contain: |
|
279 | The object has very little content now it will ultimately contain: | |
280 | * an access to the repo the bundle is applied to, |
|
280 | * an access to the repo the bundle is applied to, | |
281 | * a ui object, |
|
281 | * a ui object, | |
282 | * a way to retrieve a transaction to add changes to the repo, |
|
282 | * a way to retrieve a transaction to add changes to the repo, | |
283 | * a way to record the result of processing each part, |
|
283 | * a way to record the result of processing each part, | |
284 | * a way to construct a bundle response when applicable. |
|
284 | * a way to construct a bundle response when applicable. | |
285 | """ |
|
285 | """ | |
286 |
|
286 | |||
287 | def __init__(self, repo, transactiongetter, captureoutput=True): |
|
287 | def __init__(self, repo, transactiongetter, captureoutput=True): | |
288 | self.repo = repo |
|
288 | self.repo = repo | |
289 | self.ui = repo.ui |
|
289 | self.ui = repo.ui | |
290 | self.records = unbundlerecords() |
|
290 | self.records = unbundlerecords() | |
291 | self.gettransaction = transactiongetter |
|
291 | self.gettransaction = transactiongetter | |
292 | self.reply = None |
|
292 | self.reply = None | |
293 | self.captureoutput = captureoutput |
|
293 | self.captureoutput = captureoutput | |
294 |
|
294 | |||
295 | class TransactionUnavailable(RuntimeError): |
|
295 | class TransactionUnavailable(RuntimeError): | |
296 | pass |
|
296 | pass | |
297 |
|
297 | |||
298 | def _notransaction(): |
|
298 | def _notransaction(): | |
299 | """default method to get a transaction while processing a bundle |
|
299 | """default method to get a transaction while processing a bundle | |
300 |
|
300 | |||
301 | Raise an exception to highlight the fact that no transaction was expected |
|
301 | Raise an exception to highlight the fact that no transaction was expected | |
302 | to be created""" |
|
302 | to be created""" | |
303 | raise TransactionUnavailable() |
|
303 | raise TransactionUnavailable() | |
304 |
|
304 | |||
305 | def applybundle(repo, unbundler, tr, op=None): |
|
305 | def applybundle(repo, unbundler, tr, source=None, op=None): | |
306 | # transform me into unbundler.apply() as soon as the freeze is lifted |
|
306 | # transform me into unbundler.apply() as soon as the freeze is lifted | |
307 | tr.hookargs['bundle2'] = '1' |
|
307 | tr.hookargs['bundle2'] = '1' | |
|
308 | if source is not None and 'source' not in tr.hookargs: | |||
|
309 | tr.hookargs['source'] = source | |||
308 | return processbundle(repo, unbundler, lambda: tr, op=op) |
|
310 | return processbundle(repo, unbundler, lambda: tr, op=op) | |
309 |
|
311 | |||
310 | def processbundle(repo, unbundler, transactiongetter=None, op=None): |
|
312 | def processbundle(repo, unbundler, transactiongetter=None, op=None): | |
311 | """This function process a bundle, apply effect to/from a repo |
|
313 | """This function process a bundle, apply effect to/from a repo | |
312 |
|
314 | |||
313 | It iterates over each part then searches for and uses the proper handling |
|
315 | It iterates over each part then searches for and uses the proper handling | |
314 | code to process the part. Parts are processed in order. |
|
316 | code to process the part. Parts are processed in order. | |
315 |
|
317 | |||
316 | This is very early version of this function that will be strongly reworked |
|
318 | This is very early version of this function that will be strongly reworked | |
317 | before final usage. |
|
319 | before final usage. | |
318 |
|
320 | |||
319 | Unknown Mandatory part will abort the process. |
|
321 | Unknown Mandatory part will abort the process. | |
320 |
|
322 | |||
321 | It is temporarily possible to provide a prebuilt bundleoperation to the |
|
323 | It is temporarily possible to provide a prebuilt bundleoperation to the | |
322 | function. This is used to ensure output is properly propagated in case of |
|
324 | function. This is used to ensure output is properly propagated in case of | |
323 | an error during the unbundling. This output capturing part will likely be |
|
325 | an error during the unbundling. This output capturing part will likely be | |
324 | reworked and this ability will probably go away in the process. |
|
326 | reworked and this ability will probably go away in the process. | |
325 | """ |
|
327 | """ | |
326 | if op is None: |
|
328 | if op is None: | |
327 | if transactiongetter is None: |
|
329 | if transactiongetter is None: | |
328 | transactiongetter = _notransaction |
|
330 | transactiongetter = _notransaction | |
329 | op = bundleoperation(repo, transactiongetter) |
|
331 | op = bundleoperation(repo, transactiongetter) | |
330 | # todo: |
|
332 | # todo: | |
331 | # - replace this is a init function soon. |
|
333 | # - replace this is a init function soon. | |
332 | # - exception catching |
|
334 | # - exception catching | |
333 | unbundler.params |
|
335 | unbundler.params | |
334 | if repo.ui.debugflag: |
|
336 | if repo.ui.debugflag: | |
335 | msg = ['bundle2-input-bundle:'] |
|
337 | msg = ['bundle2-input-bundle:'] | |
336 | if unbundler.params: |
|
338 | if unbundler.params: | |
337 | msg.append(' %i params') |
|
339 | msg.append(' %i params') | |
338 | if op.gettransaction is None: |
|
340 | if op.gettransaction is None: | |
339 | msg.append(' no-transaction') |
|
341 | msg.append(' no-transaction') | |
340 | else: |
|
342 | else: | |
341 | msg.append(' with-transaction') |
|
343 | msg.append(' with-transaction') | |
342 | msg.append('\n') |
|
344 | msg.append('\n') | |
343 | repo.ui.debug(''.join(msg)) |
|
345 | repo.ui.debug(''.join(msg)) | |
344 | iterparts = enumerate(unbundler.iterparts()) |
|
346 | iterparts = enumerate(unbundler.iterparts()) | |
345 | part = None |
|
347 | part = None | |
346 | nbpart = 0 |
|
348 | nbpart = 0 | |
347 | try: |
|
349 | try: | |
348 | for nbpart, part in iterparts: |
|
350 | for nbpart, part in iterparts: | |
349 | _processpart(op, part) |
|
351 | _processpart(op, part) | |
350 | except BaseException as exc: |
|
352 | except BaseException as exc: | |
351 | for nbpart, part in iterparts: |
|
353 | for nbpart, part in iterparts: | |
352 | # consume the bundle content |
|
354 | # consume the bundle content | |
353 | part.seek(0, 2) |
|
355 | part.seek(0, 2) | |
354 | # Small hack to let caller code distinguish exceptions from bundle2 |
|
356 | # Small hack to let caller code distinguish exceptions from bundle2 | |
355 | # processing from processing the old format. This is mostly |
|
357 | # processing from processing the old format. This is mostly | |
356 | # needed to handle different return codes to unbundle according to the |
|
358 | # needed to handle different return codes to unbundle according to the | |
357 | # type of bundle. We should probably clean up or drop this return code |
|
359 | # type of bundle. We should probably clean up or drop this return code | |
358 | # craziness in a future version. |
|
360 | # craziness in a future version. | |
359 | exc.duringunbundle2 = True |
|
361 | exc.duringunbundle2 = True | |
360 | salvaged = [] |
|
362 | salvaged = [] | |
361 | replycaps = None |
|
363 | replycaps = None | |
362 | if op.reply is not None: |
|
364 | if op.reply is not None: | |
363 | salvaged = op.reply.salvageoutput() |
|
365 | salvaged = op.reply.salvageoutput() | |
364 | replycaps = op.reply.capabilities |
|
366 | replycaps = op.reply.capabilities | |
365 | exc._replycaps = replycaps |
|
367 | exc._replycaps = replycaps | |
366 | exc._bundle2salvagedoutput = salvaged |
|
368 | exc._bundle2salvagedoutput = salvaged | |
367 | raise |
|
369 | raise | |
368 | finally: |
|
370 | finally: | |
369 | repo.ui.debug('bundle2-input-bundle: %i parts total\n' % nbpart) |
|
371 | repo.ui.debug('bundle2-input-bundle: %i parts total\n' % nbpart) | |
370 |
|
372 | |||
371 | return op |
|
373 | return op | |
372 |
|
374 | |||
373 | def _processpart(op, part): |
|
375 | def _processpart(op, part): | |
374 | """process a single part from a bundle |
|
376 | """process a single part from a bundle | |
375 |
|
377 | |||
376 | The part is guaranteed to have been fully consumed when the function exits |
|
378 | The part is guaranteed to have been fully consumed when the function exits | |
377 | (even if an exception is raised).""" |
|
379 | (even if an exception is raised).""" | |
378 | status = 'unknown' # used by debug output |
|
380 | status = 'unknown' # used by debug output | |
379 | try: |
|
381 | try: | |
380 | try: |
|
382 | try: | |
381 | handler = parthandlermapping.get(part.type) |
|
383 | handler = parthandlermapping.get(part.type) | |
382 | if handler is None: |
|
384 | if handler is None: | |
383 | status = 'unsupported-type' |
|
385 | status = 'unsupported-type' | |
384 | raise error.BundleUnknownFeatureError(parttype=part.type) |
|
386 | raise error.BundleUnknownFeatureError(parttype=part.type) | |
385 | indebug(op.ui, 'found a handler for part %r' % part.type) |
|
387 | indebug(op.ui, 'found a handler for part %r' % part.type) | |
386 | unknownparams = part.mandatorykeys - handler.params |
|
388 | unknownparams = part.mandatorykeys - handler.params | |
387 | if unknownparams: |
|
389 | if unknownparams: | |
388 | unknownparams = list(unknownparams) |
|
390 | unknownparams = list(unknownparams) | |
389 | unknownparams.sort() |
|
391 | unknownparams.sort() | |
390 | status = 'unsupported-params (%s)' % unknownparams |
|
392 | status = 'unsupported-params (%s)' % unknownparams | |
391 | raise error.BundleUnknownFeatureError(parttype=part.type, |
|
393 | raise error.BundleUnknownFeatureError(parttype=part.type, | |
392 | params=unknownparams) |
|
394 | params=unknownparams) | |
393 | status = 'supported' |
|
395 | status = 'supported' | |
394 | except error.BundleUnknownFeatureError as exc: |
|
396 | except error.BundleUnknownFeatureError as exc: | |
395 | if part.mandatory: # mandatory parts |
|
397 | if part.mandatory: # mandatory parts | |
396 | raise |
|
398 | raise | |
397 | indebug(op.ui, 'ignoring unsupported advisory part %s' % exc) |
|
399 | indebug(op.ui, 'ignoring unsupported advisory part %s' % exc) | |
398 | return # skip to part processing |
|
400 | return # skip to part processing | |
399 | finally: |
|
401 | finally: | |
400 | if op.ui.debugflag: |
|
402 | if op.ui.debugflag: | |
401 | msg = ['bundle2-input-part: "%s"' % part.type] |
|
403 | msg = ['bundle2-input-part: "%s"' % part.type] | |
402 | if not part.mandatory: |
|
404 | if not part.mandatory: | |
403 | msg.append(' (advisory)') |
|
405 | msg.append(' (advisory)') | |
404 | nbmp = len(part.mandatorykeys) |
|
406 | nbmp = len(part.mandatorykeys) | |
405 | nbap = len(part.params) - nbmp |
|
407 | nbap = len(part.params) - nbmp | |
406 | if nbmp or nbap: |
|
408 | if nbmp or nbap: | |
407 | msg.append(' (params:') |
|
409 | msg.append(' (params:') | |
408 | if nbmp: |
|
410 | if nbmp: | |
409 | msg.append(' %i mandatory' % nbmp) |
|
411 | msg.append(' %i mandatory' % nbmp) | |
410 | if nbap: |
|
412 | if nbap: | |
411 | msg.append(' %i advisory' % nbmp) |
|
413 | msg.append(' %i advisory' % nbmp) | |
412 | msg.append(')') |
|
414 | msg.append(')') | |
413 | msg.append(' %s\n' % status) |
|
415 | msg.append(' %s\n' % status) | |
414 | op.ui.debug(''.join(msg)) |
|
416 | op.ui.debug(''.join(msg)) | |
415 |
|
417 | |||
416 | # handler is called outside the above try block so that we don't |
|
418 | # handler is called outside the above try block so that we don't | |
417 | # risk catching KeyErrors from anything other than the |
|
419 | # risk catching KeyErrors from anything other than the | |
418 | # parthandlermapping lookup (any KeyError raised by handler() |
|
420 | # parthandlermapping lookup (any KeyError raised by handler() | |
419 | # itself represents a defect of a different variety). |
|
421 | # itself represents a defect of a different variety). | |
420 | output = None |
|
422 | output = None | |
421 | if op.captureoutput and op.reply is not None: |
|
423 | if op.captureoutput and op.reply is not None: | |
422 | op.ui.pushbuffer(error=True, subproc=True) |
|
424 | op.ui.pushbuffer(error=True, subproc=True) | |
423 | output = '' |
|
425 | output = '' | |
424 | try: |
|
426 | try: | |
425 | handler(op, part) |
|
427 | handler(op, part) | |
426 | finally: |
|
428 | finally: | |
427 | if output is not None: |
|
429 | if output is not None: | |
428 | output = op.ui.popbuffer() |
|
430 | output = op.ui.popbuffer() | |
429 | if output: |
|
431 | if output: | |
430 | outpart = op.reply.newpart('output', data=output, |
|
432 | outpart = op.reply.newpart('output', data=output, | |
431 | mandatory=False) |
|
433 | mandatory=False) | |
432 | outpart.addparam('in-reply-to', str(part.id), mandatory=False) |
|
434 | outpart.addparam('in-reply-to', str(part.id), mandatory=False) | |
433 | finally: |
|
435 | finally: | |
434 | # consume the part content to not corrupt the stream. |
|
436 | # consume the part content to not corrupt the stream. | |
435 | part.seek(0, 2) |
|
437 | part.seek(0, 2) | |
436 |
|
438 | |||
437 |
|
439 | |||
438 | def decodecaps(blob): |
|
440 | def decodecaps(blob): | |
439 | """decode a bundle2 caps bytes blob into a dictionary |
|
441 | """decode a bundle2 caps bytes blob into a dictionary | |
440 |
|
442 | |||
441 | The blob is a list of capabilities (one per line) |
|
443 | The blob is a list of capabilities (one per line) | |
442 | Capabilities may have values using a line of the form:: |
|
444 | Capabilities may have values using a line of the form:: | |
443 |
|
445 | |||
444 | capability=value1,value2,value3 |
|
446 | capability=value1,value2,value3 | |
445 |
|
447 | |||
446 | The values are always a list.""" |
|
448 | The values are always a list.""" | |
447 | caps = {} |
|
449 | caps = {} | |
448 | for line in blob.splitlines(): |
|
450 | for line in blob.splitlines(): | |
449 | if not line: |
|
451 | if not line: | |
450 | continue |
|
452 | continue | |
451 | if '=' not in line: |
|
453 | if '=' not in line: | |
452 | key, vals = line, () |
|
454 | key, vals = line, () | |
453 | else: |
|
455 | else: | |
454 | key, vals = line.split('=', 1) |
|
456 | key, vals = line.split('=', 1) | |
455 | vals = vals.split(',') |
|
457 | vals = vals.split(',') | |
456 | key = urllib.unquote(key) |
|
458 | key = urllib.unquote(key) | |
457 | vals = [urllib.unquote(v) for v in vals] |
|
459 | vals = [urllib.unquote(v) for v in vals] | |
458 | caps[key] = vals |
|
460 | caps[key] = vals | |
459 | return caps |
|
461 | return caps | |
460 |
|
462 | |||
461 | def encodecaps(caps): |
|
463 | def encodecaps(caps): | |
462 | """encode a bundle2 caps dictionary into a bytes blob""" |
|
464 | """encode a bundle2 caps dictionary into a bytes blob""" | |
463 | chunks = [] |
|
465 | chunks = [] | |
464 | for ca in sorted(caps): |
|
466 | for ca in sorted(caps): | |
465 | vals = caps[ca] |
|
467 | vals = caps[ca] | |
466 | ca = urllib.quote(ca) |
|
468 | ca = urllib.quote(ca) | |
467 | vals = [urllib.quote(v) for v in vals] |
|
469 | vals = [urllib.quote(v) for v in vals] | |
468 | if vals: |
|
470 | if vals: | |
469 | ca = "%s=%s" % (ca, ','.join(vals)) |
|
471 | ca = "%s=%s" % (ca, ','.join(vals)) | |
470 | chunks.append(ca) |
|
472 | chunks.append(ca) | |
471 | return '\n'.join(chunks) |
|
473 | return '\n'.join(chunks) | |
472 |
|
474 | |||
473 | class bundle20(object): |
|
475 | class bundle20(object): | |
474 | """represent an outgoing bundle2 container |
|
476 | """represent an outgoing bundle2 container | |
475 |
|
477 | |||
476 | Use the `addparam` method to add stream level parameter. and `newpart` to |
|
478 | Use the `addparam` method to add stream level parameter. and `newpart` to | |
477 | populate it. Then call `getchunks` to retrieve all the binary chunks of |
|
479 | populate it. Then call `getchunks` to retrieve all the binary chunks of | |
478 | data that compose the bundle2 container.""" |
|
480 | data that compose the bundle2 container.""" | |
479 |
|
481 | |||
480 | _magicstring = 'HG20' |
|
482 | _magicstring = 'HG20' | |
481 |
|
483 | |||
482 | def __init__(self, ui, capabilities=()): |
|
484 | def __init__(self, ui, capabilities=()): | |
483 | self.ui = ui |
|
485 | self.ui = ui | |
484 | self._params = [] |
|
486 | self._params = [] | |
485 | self._parts = [] |
|
487 | self._parts = [] | |
486 | self.capabilities = dict(capabilities) |
|
488 | self.capabilities = dict(capabilities) | |
487 | self._compressor = util.compressors[None]() |
|
489 | self._compressor = util.compressors[None]() | |
488 |
|
490 | |||
489 | def setcompression(self, alg): |
|
491 | def setcompression(self, alg): | |
490 | """setup core part compression to <alg>""" |
|
492 | """setup core part compression to <alg>""" | |
491 | if alg is None: |
|
493 | if alg is None: | |
492 | return |
|
494 | return | |
493 | assert not any(n.lower() == 'Compression' for n, v in self._params) |
|
495 | assert not any(n.lower() == 'Compression' for n, v in self._params) | |
494 | self.addparam('Compression', alg) |
|
496 | self.addparam('Compression', alg) | |
495 | self._compressor = util.compressors[alg]() |
|
497 | self._compressor = util.compressors[alg]() | |
496 |
|
498 | |||
497 | @property |
|
499 | @property | |
498 | def nbparts(self): |
|
500 | def nbparts(self): | |
499 | """total number of parts added to the bundler""" |
|
501 | """total number of parts added to the bundler""" | |
500 | return len(self._parts) |
|
502 | return len(self._parts) | |
501 |
|
503 | |||
502 | # methods used to defines the bundle2 content |
|
504 | # methods used to defines the bundle2 content | |
503 | def addparam(self, name, value=None): |
|
505 | def addparam(self, name, value=None): | |
504 | """add a stream level parameter""" |
|
506 | """add a stream level parameter""" | |
505 | if not name: |
|
507 | if not name: | |
506 | raise ValueError('empty parameter name') |
|
508 | raise ValueError('empty parameter name') | |
507 | if name[0] not in string.letters: |
|
509 | if name[0] not in string.letters: | |
508 | raise ValueError('non letter first character: %r' % name) |
|
510 | raise ValueError('non letter first character: %r' % name) | |
509 | self._params.append((name, value)) |
|
511 | self._params.append((name, value)) | |
510 |
|
512 | |||
511 | def addpart(self, part): |
|
513 | def addpart(self, part): | |
512 | """add a new part to the bundle2 container |
|
514 | """add a new part to the bundle2 container | |
513 |
|
515 | |||
514 | Parts contains the actual applicative payload.""" |
|
516 | Parts contains the actual applicative payload.""" | |
515 | assert part.id is None |
|
517 | assert part.id is None | |
516 | part.id = len(self._parts) # very cheap counter |
|
518 | part.id = len(self._parts) # very cheap counter | |
517 | self._parts.append(part) |
|
519 | self._parts.append(part) | |
518 |
|
520 | |||
519 | def newpart(self, typeid, *args, **kwargs): |
|
521 | def newpart(self, typeid, *args, **kwargs): | |
520 | """create a new part and add it to the containers |
|
522 | """create a new part and add it to the containers | |
521 |
|
523 | |||
522 | As the part is directly added to the containers. For now, this means |
|
524 | As the part is directly added to the containers. For now, this means | |
523 | that any failure to properly initialize the part after calling |
|
525 | that any failure to properly initialize the part after calling | |
524 | ``newpart`` should result in a failure of the whole bundling process. |
|
526 | ``newpart`` should result in a failure of the whole bundling process. | |
525 |
|
527 | |||
526 | You can still fall back to manually create and add if you need better |
|
528 | You can still fall back to manually create and add if you need better | |
527 | control.""" |
|
529 | control.""" | |
528 | part = bundlepart(typeid, *args, **kwargs) |
|
530 | part = bundlepart(typeid, *args, **kwargs) | |
529 | self.addpart(part) |
|
531 | self.addpart(part) | |
530 | return part |
|
532 | return part | |
531 |
|
533 | |||
532 | # methods used to generate the bundle2 stream |
|
534 | # methods used to generate the bundle2 stream | |
533 | def getchunks(self): |
|
535 | def getchunks(self): | |
534 | if self.ui.debugflag: |
|
536 | if self.ui.debugflag: | |
535 | msg = ['bundle2-output-bundle: "%s",' % self._magicstring] |
|
537 | msg = ['bundle2-output-bundle: "%s",' % self._magicstring] | |
536 | if self._params: |
|
538 | if self._params: | |
537 | msg.append(' (%i params)' % len(self._params)) |
|
539 | msg.append(' (%i params)' % len(self._params)) | |
538 | msg.append(' %i parts total\n' % len(self._parts)) |
|
540 | msg.append(' %i parts total\n' % len(self._parts)) | |
539 | self.ui.debug(''.join(msg)) |
|
541 | self.ui.debug(''.join(msg)) | |
540 | outdebug(self.ui, 'start emission of %s stream' % self._magicstring) |
|
542 | outdebug(self.ui, 'start emission of %s stream' % self._magicstring) | |
541 | yield self._magicstring |
|
543 | yield self._magicstring | |
542 | param = self._paramchunk() |
|
544 | param = self._paramchunk() | |
543 | outdebug(self.ui, 'bundle parameter: %s' % param) |
|
545 | outdebug(self.ui, 'bundle parameter: %s' % param) | |
544 | yield _pack(_fstreamparamsize, len(param)) |
|
546 | yield _pack(_fstreamparamsize, len(param)) | |
545 | if param: |
|
547 | if param: | |
546 | yield param |
|
548 | yield param | |
547 | # starting compression |
|
549 | # starting compression | |
548 | for chunk in self._getcorechunk(): |
|
550 | for chunk in self._getcorechunk(): | |
549 | yield self._compressor.compress(chunk) |
|
551 | yield self._compressor.compress(chunk) | |
550 | yield self._compressor.flush() |
|
552 | yield self._compressor.flush() | |
551 |
|
553 | |||
552 | def _paramchunk(self): |
|
554 | def _paramchunk(self): | |
553 | """return a encoded version of all stream parameters""" |
|
555 | """return a encoded version of all stream parameters""" | |
554 | blocks = [] |
|
556 | blocks = [] | |
555 | for par, value in self._params: |
|
557 | for par, value in self._params: | |
556 | par = urllib.quote(par) |
|
558 | par = urllib.quote(par) | |
557 | if value is not None: |
|
559 | if value is not None: | |
558 | value = urllib.quote(value) |
|
560 | value = urllib.quote(value) | |
559 | par = '%s=%s' % (par, value) |
|
561 | par = '%s=%s' % (par, value) | |
560 | blocks.append(par) |
|
562 | blocks.append(par) | |
561 | return ' '.join(blocks) |
|
563 | return ' '.join(blocks) | |
562 |
|
564 | |||
563 | def _getcorechunk(self): |
|
565 | def _getcorechunk(self): | |
564 | """yield chunk for the core part of the bundle |
|
566 | """yield chunk for the core part of the bundle | |
565 |
|
567 | |||
566 | (all but headers and parameters)""" |
|
568 | (all but headers and parameters)""" | |
567 | outdebug(self.ui, 'start of parts') |
|
569 | outdebug(self.ui, 'start of parts') | |
568 | for part in self._parts: |
|
570 | for part in self._parts: | |
569 | outdebug(self.ui, 'bundle part: "%s"' % part.type) |
|
571 | outdebug(self.ui, 'bundle part: "%s"' % part.type) | |
570 | for chunk in part.getchunks(ui=self.ui): |
|
572 | for chunk in part.getchunks(ui=self.ui): | |
571 | yield chunk |
|
573 | yield chunk | |
572 | outdebug(self.ui, 'end of bundle') |
|
574 | outdebug(self.ui, 'end of bundle') | |
573 | yield _pack(_fpartheadersize, 0) |
|
575 | yield _pack(_fpartheadersize, 0) | |
574 |
|
576 | |||
575 |
|
577 | |||
576 | def salvageoutput(self): |
|
578 | def salvageoutput(self): | |
577 | """return a list with a copy of all output parts in the bundle |
|
579 | """return a list with a copy of all output parts in the bundle | |
578 |
|
580 | |||
579 | This is meant to be used during error handling to make sure we preserve |
|
581 | This is meant to be used during error handling to make sure we preserve | |
580 | server output""" |
|
582 | server output""" | |
581 | salvaged = [] |
|
583 | salvaged = [] | |
582 | for part in self._parts: |
|
584 | for part in self._parts: | |
583 | if part.type.startswith('output'): |
|
585 | if part.type.startswith('output'): | |
584 | salvaged.append(part.copy()) |
|
586 | salvaged.append(part.copy()) | |
585 | return salvaged |
|
587 | return salvaged | |
586 |
|
588 | |||
587 |
|
589 | |||
588 | class unpackermixin(object): |
|
590 | class unpackermixin(object): | |
589 | """A mixin to extract bytes and struct data from a stream""" |
|
591 | """A mixin to extract bytes and struct data from a stream""" | |
590 |
|
592 | |||
591 | def __init__(self, fp): |
|
593 | def __init__(self, fp): | |
592 | self._fp = fp |
|
594 | self._fp = fp | |
593 | self._seekable = (util.safehasattr(fp, 'seek') and |
|
595 | self._seekable = (util.safehasattr(fp, 'seek') and | |
594 | util.safehasattr(fp, 'tell')) |
|
596 | util.safehasattr(fp, 'tell')) | |
595 |
|
597 | |||
596 | def _unpack(self, format): |
|
598 | def _unpack(self, format): | |
597 | """unpack this struct format from the stream""" |
|
599 | """unpack this struct format from the stream""" | |
598 | data = self._readexact(struct.calcsize(format)) |
|
600 | data = self._readexact(struct.calcsize(format)) | |
599 | return _unpack(format, data) |
|
601 | return _unpack(format, data) | |
600 |
|
602 | |||
601 | def _readexact(self, size): |
|
603 | def _readexact(self, size): | |
602 | """read exactly <size> bytes from the stream""" |
|
604 | """read exactly <size> bytes from the stream""" | |
603 | return changegroup.readexactly(self._fp, size) |
|
605 | return changegroup.readexactly(self._fp, size) | |
604 |
|
606 | |||
605 | def seek(self, offset, whence=0): |
|
607 | def seek(self, offset, whence=0): | |
606 | """move the underlying file pointer""" |
|
608 | """move the underlying file pointer""" | |
607 | if self._seekable: |
|
609 | if self._seekable: | |
608 | return self._fp.seek(offset, whence) |
|
610 | return self._fp.seek(offset, whence) | |
609 | else: |
|
611 | else: | |
610 | raise NotImplementedError(_('File pointer is not seekable')) |
|
612 | raise NotImplementedError(_('File pointer is not seekable')) | |
611 |
|
613 | |||
612 | def tell(self): |
|
614 | def tell(self): | |
613 | """return the file offset, or None if file is not seekable""" |
|
615 | """return the file offset, or None if file is not seekable""" | |
614 | if self._seekable: |
|
616 | if self._seekable: | |
615 | try: |
|
617 | try: | |
616 | return self._fp.tell() |
|
618 | return self._fp.tell() | |
617 | except IOError as e: |
|
619 | except IOError as e: | |
618 | if e.errno == errno.ESPIPE: |
|
620 | if e.errno == errno.ESPIPE: | |
619 | self._seekable = False |
|
621 | self._seekable = False | |
620 | else: |
|
622 | else: | |
621 | raise |
|
623 | raise | |
622 | return None |
|
624 | return None | |
623 |
|
625 | |||
624 | def close(self): |
|
626 | def close(self): | |
625 | """close underlying file""" |
|
627 | """close underlying file""" | |
626 | if util.safehasattr(self._fp, 'close'): |
|
628 | if util.safehasattr(self._fp, 'close'): | |
627 | return self._fp.close() |
|
629 | return self._fp.close() | |
628 |
|
630 | |||
629 | def getunbundler(ui, fp, magicstring=None): |
|
631 | def getunbundler(ui, fp, magicstring=None): | |
630 | """return a valid unbundler object for a given magicstring""" |
|
632 | """return a valid unbundler object for a given magicstring""" | |
631 | if magicstring is None: |
|
633 | if magicstring is None: | |
632 | magicstring = changegroup.readexactly(fp, 4) |
|
634 | magicstring = changegroup.readexactly(fp, 4) | |
633 | magic, version = magicstring[0:2], magicstring[2:4] |
|
635 | magic, version = magicstring[0:2], magicstring[2:4] | |
634 | if magic != 'HG': |
|
636 | if magic != 'HG': | |
635 | raise error.Abort(_('not a Mercurial bundle')) |
|
637 | raise error.Abort(_('not a Mercurial bundle')) | |
636 | unbundlerclass = formatmap.get(version) |
|
638 | unbundlerclass = formatmap.get(version) | |
637 | if unbundlerclass is None: |
|
639 | if unbundlerclass is None: | |
638 | raise error.Abort(_('unknown bundle version %s') % version) |
|
640 | raise error.Abort(_('unknown bundle version %s') % version) | |
639 | unbundler = unbundlerclass(ui, fp) |
|
641 | unbundler = unbundlerclass(ui, fp) | |
640 | indebug(ui, 'start processing of %s stream' % magicstring) |
|
642 | indebug(ui, 'start processing of %s stream' % magicstring) | |
641 | return unbundler |
|
643 | return unbundler | |
642 |
|
644 | |||
643 | class unbundle20(unpackermixin): |
|
645 | class unbundle20(unpackermixin): | |
644 | """interpret a bundle2 stream |
|
646 | """interpret a bundle2 stream | |
645 |
|
647 | |||
646 | This class is fed with a binary stream and yields parts through its |
|
648 | This class is fed with a binary stream and yields parts through its | |
647 | `iterparts` methods.""" |
|
649 | `iterparts` methods.""" | |
648 |
|
650 | |||
649 | _magicstring = 'HG20' |
|
651 | _magicstring = 'HG20' | |
650 |
|
652 | |||
651 | def __init__(self, ui, fp): |
|
653 | def __init__(self, ui, fp): | |
652 | """If header is specified, we do not read it out of the stream.""" |
|
654 | """If header is specified, we do not read it out of the stream.""" | |
653 | self.ui = ui |
|
655 | self.ui = ui | |
654 | self._decompressor = util.decompressors[None] |
|
656 | self._decompressor = util.decompressors[None] | |
655 | super(unbundle20, self).__init__(fp) |
|
657 | super(unbundle20, self).__init__(fp) | |
656 |
|
658 | |||
657 | @util.propertycache |
|
659 | @util.propertycache | |
658 | def params(self): |
|
660 | def params(self): | |
659 | """dictionary of stream level parameters""" |
|
661 | """dictionary of stream level parameters""" | |
660 | indebug(self.ui, 'reading bundle2 stream parameters') |
|
662 | indebug(self.ui, 'reading bundle2 stream parameters') | |
661 | params = {} |
|
663 | params = {} | |
662 | paramssize = self._unpack(_fstreamparamsize)[0] |
|
664 | paramssize = self._unpack(_fstreamparamsize)[0] | |
663 | if paramssize < 0: |
|
665 | if paramssize < 0: | |
664 | raise error.BundleValueError('negative bundle param size: %i' |
|
666 | raise error.BundleValueError('negative bundle param size: %i' | |
665 | % paramssize) |
|
667 | % paramssize) | |
666 | if paramssize: |
|
668 | if paramssize: | |
667 | params = self._readexact(paramssize) |
|
669 | params = self._readexact(paramssize) | |
668 | params = self._processallparams(params) |
|
670 | params = self._processallparams(params) | |
669 | return params |
|
671 | return params | |
670 |
|
672 | |||
671 | def _processallparams(self, paramsblock): |
|
673 | def _processallparams(self, paramsblock): | |
672 | """""" |
|
674 | """""" | |
673 | params = {} |
|
675 | params = {} | |
674 | for p in paramsblock.split(' '): |
|
676 | for p in paramsblock.split(' '): | |
675 | p = p.split('=', 1) |
|
677 | p = p.split('=', 1) | |
676 | p = [urllib.unquote(i) for i in p] |
|
678 | p = [urllib.unquote(i) for i in p] | |
677 | if len(p) < 2: |
|
679 | if len(p) < 2: | |
678 | p.append(None) |
|
680 | p.append(None) | |
679 | self._processparam(*p) |
|
681 | self._processparam(*p) | |
680 | params[p[0]] = p[1] |
|
682 | params[p[0]] = p[1] | |
681 | return params |
|
683 | return params | |
682 |
|
684 | |||
683 |
|
685 | |||
684 | def _processparam(self, name, value): |
|
686 | def _processparam(self, name, value): | |
685 | """process a parameter, applying its effect if needed |
|
687 | """process a parameter, applying its effect if needed | |
686 |
|
688 | |||
687 | Parameter starting with a lower case letter are advisory and will be |
|
689 | Parameter starting with a lower case letter are advisory and will be | |
688 | ignored when unknown. Those starting with an upper case letter are |
|
690 | ignored when unknown. Those starting with an upper case letter are | |
689 | mandatory and will this function will raise a KeyError when unknown. |
|
691 | mandatory and will this function will raise a KeyError when unknown. | |
690 |
|
692 | |||
691 | Note: no option are currently supported. Any input will be either |
|
693 | Note: no option are currently supported. Any input will be either | |
692 | ignored or failing. |
|
694 | ignored or failing. | |
693 | """ |
|
695 | """ | |
694 | if not name: |
|
696 | if not name: | |
695 | raise ValueError('empty parameter name') |
|
697 | raise ValueError('empty parameter name') | |
696 | if name[0] not in string.letters: |
|
698 | if name[0] not in string.letters: | |
697 | raise ValueError('non letter first character: %r' % name) |
|
699 | raise ValueError('non letter first character: %r' % name) | |
698 | try: |
|
700 | try: | |
699 | handler = b2streamparamsmap[name.lower()] |
|
701 | handler = b2streamparamsmap[name.lower()] | |
700 | except KeyError: |
|
702 | except KeyError: | |
701 | if name[0].islower(): |
|
703 | if name[0].islower(): | |
702 | indebug(self.ui, "ignoring unknown parameter %r" % name) |
|
704 | indebug(self.ui, "ignoring unknown parameter %r" % name) | |
703 | else: |
|
705 | else: | |
704 | raise error.BundleUnknownFeatureError(params=(name,)) |
|
706 | raise error.BundleUnknownFeatureError(params=(name,)) | |
705 | else: |
|
707 | else: | |
706 | handler(self, name, value) |
|
708 | handler(self, name, value) | |
707 |
|
709 | |||
708 | def _forwardchunks(self): |
|
710 | def _forwardchunks(self): | |
709 | """utility to transfer a bundle2 as binary |
|
711 | """utility to transfer a bundle2 as binary | |
710 |
|
712 | |||
711 | This is made necessary by the fact the 'getbundle' command over 'ssh' |
|
713 | This is made necessary by the fact the 'getbundle' command over 'ssh' | |
712 | have no way to know then the reply end, relying on the bundle to be |
|
714 | have no way to know then the reply end, relying on the bundle to be | |
713 | interpreted to know its end. This is terrible and we are sorry, but we |
|
715 | interpreted to know its end. This is terrible and we are sorry, but we | |
714 | needed to move forward to get general delta enabled. |
|
716 | needed to move forward to get general delta enabled. | |
715 | """ |
|
717 | """ | |
716 | yield self._magicstring |
|
718 | yield self._magicstring | |
717 | assert 'params' not in vars(self) |
|
719 | assert 'params' not in vars(self) | |
718 | paramssize = self._unpack(_fstreamparamsize)[0] |
|
720 | paramssize = self._unpack(_fstreamparamsize)[0] | |
719 | if paramssize < 0: |
|
721 | if paramssize < 0: | |
720 | raise error.BundleValueError('negative bundle param size: %i' |
|
722 | raise error.BundleValueError('negative bundle param size: %i' | |
721 | % paramssize) |
|
723 | % paramssize) | |
722 | yield _pack(_fstreamparamsize, paramssize) |
|
724 | yield _pack(_fstreamparamsize, paramssize) | |
723 | if paramssize: |
|
725 | if paramssize: | |
724 | params = self._readexact(paramssize) |
|
726 | params = self._readexact(paramssize) | |
725 | self._processallparams(params) |
|
727 | self._processallparams(params) | |
726 | yield params |
|
728 | yield params | |
727 | assert self._decompressor is util.decompressors[None] |
|
729 | assert self._decompressor is util.decompressors[None] | |
728 | # From there, payload might need to be decompressed |
|
730 | # From there, payload might need to be decompressed | |
729 | self._fp = self._decompressor(self._fp) |
|
731 | self._fp = self._decompressor(self._fp) | |
730 | emptycount = 0 |
|
732 | emptycount = 0 | |
731 | while emptycount < 2: |
|
733 | while emptycount < 2: | |
732 | # so we can brainlessly loop |
|
734 | # so we can brainlessly loop | |
733 | assert _fpartheadersize == _fpayloadsize |
|
735 | assert _fpartheadersize == _fpayloadsize | |
734 | size = self._unpack(_fpartheadersize)[0] |
|
736 | size = self._unpack(_fpartheadersize)[0] | |
735 | yield _pack(_fpartheadersize, size) |
|
737 | yield _pack(_fpartheadersize, size) | |
736 | if size: |
|
738 | if size: | |
737 | emptycount = 0 |
|
739 | emptycount = 0 | |
738 | else: |
|
740 | else: | |
739 | emptycount += 1 |
|
741 | emptycount += 1 | |
740 | continue |
|
742 | continue | |
741 | if size == flaginterrupt: |
|
743 | if size == flaginterrupt: | |
742 | continue |
|
744 | continue | |
743 | elif size < 0: |
|
745 | elif size < 0: | |
744 | raise error.BundleValueError('negative chunk size: %i') |
|
746 | raise error.BundleValueError('negative chunk size: %i') | |
745 | yield self._readexact(size) |
|
747 | yield self._readexact(size) | |
746 |
|
748 | |||
747 |
|
749 | |||
748 | def iterparts(self): |
|
750 | def iterparts(self): | |
749 | """yield all parts contained in the stream""" |
|
751 | """yield all parts contained in the stream""" | |
750 | # make sure param have been loaded |
|
752 | # make sure param have been loaded | |
751 | self.params |
|
753 | self.params | |
752 | # From there, payload need to be decompressed |
|
754 | # From there, payload need to be decompressed | |
753 | self._fp = self._decompressor(self._fp) |
|
755 | self._fp = self._decompressor(self._fp) | |
754 | indebug(self.ui, 'start extraction of bundle2 parts') |
|
756 | indebug(self.ui, 'start extraction of bundle2 parts') | |
755 | headerblock = self._readpartheader() |
|
757 | headerblock = self._readpartheader() | |
756 | while headerblock is not None: |
|
758 | while headerblock is not None: | |
757 | part = unbundlepart(self.ui, headerblock, self._fp) |
|
759 | part = unbundlepart(self.ui, headerblock, self._fp) | |
758 | yield part |
|
760 | yield part | |
759 | part.seek(0, 2) |
|
761 | part.seek(0, 2) | |
760 | headerblock = self._readpartheader() |
|
762 | headerblock = self._readpartheader() | |
761 | indebug(self.ui, 'end of bundle2 stream') |
|
763 | indebug(self.ui, 'end of bundle2 stream') | |
762 |
|
764 | |||
763 | def _readpartheader(self): |
|
765 | def _readpartheader(self): | |
764 | """reads a part header size and return the bytes blob |
|
766 | """reads a part header size and return the bytes blob | |
765 |
|
767 | |||
766 | returns None if empty""" |
|
768 | returns None if empty""" | |
767 | headersize = self._unpack(_fpartheadersize)[0] |
|
769 | headersize = self._unpack(_fpartheadersize)[0] | |
768 | if headersize < 0: |
|
770 | if headersize < 0: | |
769 | raise error.BundleValueError('negative part header size: %i' |
|
771 | raise error.BundleValueError('negative part header size: %i' | |
770 | % headersize) |
|
772 | % headersize) | |
771 | indebug(self.ui, 'part header size: %i' % headersize) |
|
773 | indebug(self.ui, 'part header size: %i' % headersize) | |
772 | if headersize: |
|
774 | if headersize: | |
773 | return self._readexact(headersize) |
|
775 | return self._readexact(headersize) | |
774 | return None |
|
776 | return None | |
775 |
|
777 | |||
776 | def compressed(self): |
|
778 | def compressed(self): | |
777 | return False |
|
779 | return False | |
778 |
|
780 | |||
779 | formatmap = {'20': unbundle20} |
|
781 | formatmap = {'20': unbundle20} | |
780 |
|
782 | |||
781 | b2streamparamsmap = {} |
|
783 | b2streamparamsmap = {} | |
782 |
|
784 | |||
783 | def b2streamparamhandler(name): |
|
785 | def b2streamparamhandler(name): | |
784 | """register a handler for a stream level parameter""" |
|
786 | """register a handler for a stream level parameter""" | |
785 | def decorator(func): |
|
787 | def decorator(func): | |
786 | assert name not in formatmap |
|
788 | assert name not in formatmap | |
787 | b2streamparamsmap[name] = func |
|
789 | b2streamparamsmap[name] = func | |
788 | return func |
|
790 | return func | |
789 | return decorator |
|
791 | return decorator | |
790 |
|
792 | |||
791 | @b2streamparamhandler('compression') |
|
793 | @b2streamparamhandler('compression') | |
792 | def processcompression(unbundler, param, value): |
|
794 | def processcompression(unbundler, param, value): | |
793 | """read compression parameter and install payload decompression""" |
|
795 | """read compression parameter and install payload decompression""" | |
794 | if value not in util.decompressors: |
|
796 | if value not in util.decompressors: | |
795 | raise error.BundleUnknownFeatureError(params=(param,), |
|
797 | raise error.BundleUnknownFeatureError(params=(param,), | |
796 | values=(value,)) |
|
798 | values=(value,)) | |
797 | unbundler._decompressor = util.decompressors[value] |
|
799 | unbundler._decompressor = util.decompressors[value] | |
798 |
|
800 | |||
799 | class bundlepart(object): |
|
801 | class bundlepart(object): | |
800 | """A bundle2 part contains application level payload |
|
802 | """A bundle2 part contains application level payload | |
801 |
|
803 | |||
802 | The part `type` is used to route the part to the application level |
|
804 | The part `type` is used to route the part to the application level | |
803 | handler. |
|
805 | handler. | |
804 |
|
806 | |||
805 | The part payload is contained in ``part.data``. It could be raw bytes or a |
|
807 | The part payload is contained in ``part.data``. It could be raw bytes or a | |
806 | generator of byte chunks. |
|
808 | generator of byte chunks. | |
807 |
|
809 | |||
808 | You can add parameters to the part using the ``addparam`` method. |
|
810 | You can add parameters to the part using the ``addparam`` method. | |
809 | Parameters can be either mandatory (default) or advisory. Remote side |
|
811 | Parameters can be either mandatory (default) or advisory. Remote side | |
810 | should be able to safely ignore the advisory ones. |
|
812 | should be able to safely ignore the advisory ones. | |
811 |
|
813 | |||
812 | Both data and parameters cannot be modified after the generation has begun. |
|
814 | Both data and parameters cannot be modified after the generation has begun. | |
813 | """ |
|
815 | """ | |
814 |
|
816 | |||
815 | def __init__(self, parttype, mandatoryparams=(), advisoryparams=(), |
|
817 | def __init__(self, parttype, mandatoryparams=(), advisoryparams=(), | |
816 | data='', mandatory=True): |
|
818 | data='', mandatory=True): | |
817 | validateparttype(parttype) |
|
819 | validateparttype(parttype) | |
818 | self.id = None |
|
820 | self.id = None | |
819 | self.type = parttype |
|
821 | self.type = parttype | |
820 | self._data = data |
|
822 | self._data = data | |
821 | self._mandatoryparams = list(mandatoryparams) |
|
823 | self._mandatoryparams = list(mandatoryparams) | |
822 | self._advisoryparams = list(advisoryparams) |
|
824 | self._advisoryparams = list(advisoryparams) | |
823 | # checking for duplicated entries |
|
825 | # checking for duplicated entries | |
824 | self._seenparams = set() |
|
826 | self._seenparams = set() | |
825 | for pname, __ in self._mandatoryparams + self._advisoryparams: |
|
827 | for pname, __ in self._mandatoryparams + self._advisoryparams: | |
826 | if pname in self._seenparams: |
|
828 | if pname in self._seenparams: | |
827 | raise RuntimeError('duplicated params: %s' % pname) |
|
829 | raise RuntimeError('duplicated params: %s' % pname) | |
828 | self._seenparams.add(pname) |
|
830 | self._seenparams.add(pname) | |
829 | # status of the part's generation: |
|
831 | # status of the part's generation: | |
830 | # - None: not started, |
|
832 | # - None: not started, | |
831 | # - False: currently generated, |
|
833 | # - False: currently generated, | |
832 | # - True: generation done. |
|
834 | # - True: generation done. | |
833 | self._generated = None |
|
835 | self._generated = None | |
834 | self.mandatory = mandatory |
|
836 | self.mandatory = mandatory | |
835 |
|
837 | |||
836 | def copy(self): |
|
838 | def copy(self): | |
837 | """return a copy of the part |
|
839 | """return a copy of the part | |
838 |
|
840 | |||
839 | The new part have the very same content but no partid assigned yet. |
|
841 | The new part have the very same content but no partid assigned yet. | |
840 | Parts with generated data cannot be copied.""" |
|
842 | Parts with generated data cannot be copied.""" | |
841 | assert not util.safehasattr(self.data, 'next') |
|
843 | assert not util.safehasattr(self.data, 'next') | |
842 | return self.__class__(self.type, self._mandatoryparams, |
|
844 | return self.__class__(self.type, self._mandatoryparams, | |
843 | self._advisoryparams, self._data, self.mandatory) |
|
845 | self._advisoryparams, self._data, self.mandatory) | |
844 |
|
846 | |||
845 | # methods used to defines the part content |
|
847 | # methods used to defines the part content | |
846 | def __setdata(self, data): |
|
848 | def __setdata(self, data): | |
847 | if self._generated is not None: |
|
849 | if self._generated is not None: | |
848 | raise error.ReadOnlyPartError('part is being generated') |
|
850 | raise error.ReadOnlyPartError('part is being generated') | |
849 | self._data = data |
|
851 | self._data = data | |
850 | def __getdata(self): |
|
852 | def __getdata(self): | |
851 | return self._data |
|
853 | return self._data | |
852 | data = property(__getdata, __setdata) |
|
854 | data = property(__getdata, __setdata) | |
853 |
|
855 | |||
854 | @property |
|
856 | @property | |
855 | def mandatoryparams(self): |
|
857 | def mandatoryparams(self): | |
856 | # make it an immutable tuple to force people through ``addparam`` |
|
858 | # make it an immutable tuple to force people through ``addparam`` | |
857 | return tuple(self._mandatoryparams) |
|
859 | return tuple(self._mandatoryparams) | |
858 |
|
860 | |||
859 | @property |
|
861 | @property | |
860 | def advisoryparams(self): |
|
862 | def advisoryparams(self): | |
861 | # make it an immutable tuple to force people through ``addparam`` |
|
863 | # make it an immutable tuple to force people through ``addparam`` | |
862 | return tuple(self._advisoryparams) |
|
864 | return tuple(self._advisoryparams) | |
863 |
|
865 | |||
864 | def addparam(self, name, value='', mandatory=True): |
|
866 | def addparam(self, name, value='', mandatory=True): | |
865 | if self._generated is not None: |
|
867 | if self._generated is not None: | |
866 | raise error.ReadOnlyPartError('part is being generated') |
|
868 | raise error.ReadOnlyPartError('part is being generated') | |
867 | if name in self._seenparams: |
|
869 | if name in self._seenparams: | |
868 | raise ValueError('duplicated params: %s' % name) |
|
870 | raise ValueError('duplicated params: %s' % name) | |
869 | self._seenparams.add(name) |
|
871 | self._seenparams.add(name) | |
870 | params = self._advisoryparams |
|
872 | params = self._advisoryparams | |
871 | if mandatory: |
|
873 | if mandatory: | |
872 | params = self._mandatoryparams |
|
874 | params = self._mandatoryparams | |
873 | params.append((name, value)) |
|
875 | params.append((name, value)) | |
874 |
|
876 | |||
875 | # methods used to generates the bundle2 stream |
|
877 | # methods used to generates the bundle2 stream | |
876 | def getchunks(self, ui): |
|
878 | def getchunks(self, ui): | |
877 | if self._generated is not None: |
|
879 | if self._generated is not None: | |
878 | raise RuntimeError('part can only be consumed once') |
|
880 | raise RuntimeError('part can only be consumed once') | |
879 | self._generated = False |
|
881 | self._generated = False | |
880 |
|
882 | |||
881 | if ui.debugflag: |
|
883 | if ui.debugflag: | |
882 | msg = ['bundle2-output-part: "%s"' % self.type] |
|
884 | msg = ['bundle2-output-part: "%s"' % self.type] | |
883 | if not self.mandatory: |
|
885 | if not self.mandatory: | |
884 | msg.append(' (advisory)') |
|
886 | msg.append(' (advisory)') | |
885 | nbmp = len(self.mandatoryparams) |
|
887 | nbmp = len(self.mandatoryparams) | |
886 | nbap = len(self.advisoryparams) |
|
888 | nbap = len(self.advisoryparams) | |
887 | if nbmp or nbap: |
|
889 | if nbmp or nbap: | |
888 | msg.append(' (params:') |
|
890 | msg.append(' (params:') | |
889 | if nbmp: |
|
891 | if nbmp: | |
890 | msg.append(' %i mandatory' % nbmp) |
|
892 | msg.append(' %i mandatory' % nbmp) | |
891 | if nbap: |
|
893 | if nbap: | |
892 | msg.append(' %i advisory' % nbmp) |
|
894 | msg.append(' %i advisory' % nbmp) | |
893 | msg.append(')') |
|
895 | msg.append(')') | |
894 | if not self.data: |
|
896 | if not self.data: | |
895 | msg.append(' empty payload') |
|
897 | msg.append(' empty payload') | |
896 | elif util.safehasattr(self.data, 'next'): |
|
898 | elif util.safehasattr(self.data, 'next'): | |
897 | msg.append(' streamed payload') |
|
899 | msg.append(' streamed payload') | |
898 | else: |
|
900 | else: | |
899 | msg.append(' %i bytes payload' % len(self.data)) |
|
901 | msg.append(' %i bytes payload' % len(self.data)) | |
900 | msg.append('\n') |
|
902 | msg.append('\n') | |
901 | ui.debug(''.join(msg)) |
|
903 | ui.debug(''.join(msg)) | |
902 |
|
904 | |||
903 | #### header |
|
905 | #### header | |
904 | if self.mandatory: |
|
906 | if self.mandatory: | |
905 | parttype = self.type.upper() |
|
907 | parttype = self.type.upper() | |
906 | else: |
|
908 | else: | |
907 | parttype = self.type.lower() |
|
909 | parttype = self.type.lower() | |
908 | outdebug(ui, 'part %s: "%s"' % (self.id, parttype)) |
|
910 | outdebug(ui, 'part %s: "%s"' % (self.id, parttype)) | |
909 | ## parttype |
|
911 | ## parttype | |
910 | header = [_pack(_fparttypesize, len(parttype)), |
|
912 | header = [_pack(_fparttypesize, len(parttype)), | |
911 | parttype, _pack(_fpartid, self.id), |
|
913 | parttype, _pack(_fpartid, self.id), | |
912 | ] |
|
914 | ] | |
913 | ## parameters |
|
915 | ## parameters | |
914 | # count |
|
916 | # count | |
915 | manpar = self.mandatoryparams |
|
917 | manpar = self.mandatoryparams | |
916 | advpar = self.advisoryparams |
|
918 | advpar = self.advisoryparams | |
917 | header.append(_pack(_fpartparamcount, len(manpar), len(advpar))) |
|
919 | header.append(_pack(_fpartparamcount, len(manpar), len(advpar))) | |
918 | # size |
|
920 | # size | |
919 | parsizes = [] |
|
921 | parsizes = [] | |
920 | for key, value in manpar: |
|
922 | for key, value in manpar: | |
921 | parsizes.append(len(key)) |
|
923 | parsizes.append(len(key)) | |
922 | parsizes.append(len(value)) |
|
924 | parsizes.append(len(value)) | |
923 | for key, value in advpar: |
|
925 | for key, value in advpar: | |
924 | parsizes.append(len(key)) |
|
926 | parsizes.append(len(key)) | |
925 | parsizes.append(len(value)) |
|
927 | parsizes.append(len(value)) | |
926 | paramsizes = _pack(_makefpartparamsizes(len(parsizes) / 2), *parsizes) |
|
928 | paramsizes = _pack(_makefpartparamsizes(len(parsizes) / 2), *parsizes) | |
927 | header.append(paramsizes) |
|
929 | header.append(paramsizes) | |
928 | # key, value |
|
930 | # key, value | |
929 | for key, value in manpar: |
|
931 | for key, value in manpar: | |
930 | header.append(key) |
|
932 | header.append(key) | |
931 | header.append(value) |
|
933 | header.append(value) | |
932 | for key, value in advpar: |
|
934 | for key, value in advpar: | |
933 | header.append(key) |
|
935 | header.append(key) | |
934 | header.append(value) |
|
936 | header.append(value) | |
935 | ## finalize header |
|
937 | ## finalize header | |
936 | headerchunk = ''.join(header) |
|
938 | headerchunk = ''.join(header) | |
937 | outdebug(ui, 'header chunk size: %i' % len(headerchunk)) |
|
939 | outdebug(ui, 'header chunk size: %i' % len(headerchunk)) | |
938 | yield _pack(_fpartheadersize, len(headerchunk)) |
|
940 | yield _pack(_fpartheadersize, len(headerchunk)) | |
939 | yield headerchunk |
|
941 | yield headerchunk | |
940 | ## payload |
|
942 | ## payload | |
941 | try: |
|
943 | try: | |
942 | for chunk in self._payloadchunks(): |
|
944 | for chunk in self._payloadchunks(): | |
943 | outdebug(ui, 'payload chunk size: %i' % len(chunk)) |
|
945 | outdebug(ui, 'payload chunk size: %i' % len(chunk)) | |
944 | yield _pack(_fpayloadsize, len(chunk)) |
|
946 | yield _pack(_fpayloadsize, len(chunk)) | |
945 | yield chunk |
|
947 | yield chunk | |
946 | except GeneratorExit: |
|
948 | except GeneratorExit: | |
947 | # GeneratorExit means that nobody is listening for our |
|
949 | # GeneratorExit means that nobody is listening for our | |
948 | # results anyway, so just bail quickly rather than trying |
|
950 | # results anyway, so just bail quickly rather than trying | |
949 | # to produce an error part. |
|
951 | # to produce an error part. | |
950 | ui.debug('bundle2-generatorexit\n') |
|
952 | ui.debug('bundle2-generatorexit\n') | |
951 | raise |
|
953 | raise | |
952 | except BaseException as exc: |
|
954 | except BaseException as exc: | |
953 | # backup exception data for later |
|
955 | # backup exception data for later | |
954 | ui.debug('bundle2-input-stream-interrupt: encoding exception %s' |
|
956 | ui.debug('bundle2-input-stream-interrupt: encoding exception %s' | |
955 | % exc) |
|
957 | % exc) | |
956 | exc_info = sys.exc_info() |
|
958 | exc_info = sys.exc_info() | |
957 | msg = 'unexpected error: %s' % exc |
|
959 | msg = 'unexpected error: %s' % exc | |
958 | interpart = bundlepart('error:abort', [('message', msg)], |
|
960 | interpart = bundlepart('error:abort', [('message', msg)], | |
959 | mandatory=False) |
|
961 | mandatory=False) | |
960 | interpart.id = 0 |
|
962 | interpart.id = 0 | |
961 | yield _pack(_fpayloadsize, -1) |
|
963 | yield _pack(_fpayloadsize, -1) | |
962 | for chunk in interpart.getchunks(ui=ui): |
|
964 | for chunk in interpart.getchunks(ui=ui): | |
963 | yield chunk |
|
965 | yield chunk | |
964 | outdebug(ui, 'closing payload chunk') |
|
966 | outdebug(ui, 'closing payload chunk') | |
965 | # abort current part payload |
|
967 | # abort current part payload | |
966 | yield _pack(_fpayloadsize, 0) |
|
968 | yield _pack(_fpayloadsize, 0) | |
967 | raise exc_info[0], exc_info[1], exc_info[2] |
|
969 | raise exc_info[0], exc_info[1], exc_info[2] | |
968 | # end of payload |
|
970 | # end of payload | |
969 | outdebug(ui, 'closing payload chunk') |
|
971 | outdebug(ui, 'closing payload chunk') | |
970 | yield _pack(_fpayloadsize, 0) |
|
972 | yield _pack(_fpayloadsize, 0) | |
971 | self._generated = True |
|
973 | self._generated = True | |
972 |
|
974 | |||
973 | def _payloadchunks(self): |
|
975 | def _payloadchunks(self): | |
974 | """yield chunks of a the part payload |
|
976 | """yield chunks of a the part payload | |
975 |
|
977 | |||
976 | Exists to handle the different methods to provide data to a part.""" |
|
978 | Exists to handle the different methods to provide data to a part.""" | |
977 | # we only support fixed size data now. |
|
979 | # we only support fixed size data now. | |
978 | # This will be improved in the future. |
|
980 | # This will be improved in the future. | |
979 | if util.safehasattr(self.data, 'next'): |
|
981 | if util.safehasattr(self.data, 'next'): | |
980 | buff = util.chunkbuffer(self.data) |
|
982 | buff = util.chunkbuffer(self.data) | |
981 | chunk = buff.read(preferedchunksize) |
|
983 | chunk = buff.read(preferedchunksize) | |
982 | while chunk: |
|
984 | while chunk: | |
983 | yield chunk |
|
985 | yield chunk | |
984 | chunk = buff.read(preferedchunksize) |
|
986 | chunk = buff.read(preferedchunksize) | |
985 | elif len(self.data): |
|
987 | elif len(self.data): | |
986 | yield self.data |
|
988 | yield self.data | |
987 |
|
989 | |||
988 |
|
990 | |||
989 | flaginterrupt = -1 |
|
991 | flaginterrupt = -1 | |
990 |
|
992 | |||
991 | class interrupthandler(unpackermixin): |
|
993 | class interrupthandler(unpackermixin): | |
992 | """read one part and process it with restricted capability |
|
994 | """read one part and process it with restricted capability | |
993 |
|
995 | |||
994 | This allows to transmit exception raised on the producer size during part |
|
996 | This allows to transmit exception raised on the producer size during part | |
995 | iteration while the consumer is reading a part. |
|
997 | iteration while the consumer is reading a part. | |
996 |
|
998 | |||
997 | Part processed in this manner only have access to a ui object,""" |
|
999 | Part processed in this manner only have access to a ui object,""" | |
998 |
|
1000 | |||
999 | def __init__(self, ui, fp): |
|
1001 | def __init__(self, ui, fp): | |
1000 | super(interrupthandler, self).__init__(fp) |
|
1002 | super(interrupthandler, self).__init__(fp) | |
1001 | self.ui = ui |
|
1003 | self.ui = ui | |
1002 |
|
1004 | |||
1003 | def _readpartheader(self): |
|
1005 | def _readpartheader(self): | |
1004 | """reads a part header size and return the bytes blob |
|
1006 | """reads a part header size and return the bytes blob | |
1005 |
|
1007 | |||
1006 | returns None if empty""" |
|
1008 | returns None if empty""" | |
1007 | headersize = self._unpack(_fpartheadersize)[0] |
|
1009 | headersize = self._unpack(_fpartheadersize)[0] | |
1008 | if headersize < 0: |
|
1010 | if headersize < 0: | |
1009 | raise error.BundleValueError('negative part header size: %i' |
|
1011 | raise error.BundleValueError('negative part header size: %i' | |
1010 | % headersize) |
|
1012 | % headersize) | |
1011 | indebug(self.ui, 'part header size: %i\n' % headersize) |
|
1013 | indebug(self.ui, 'part header size: %i\n' % headersize) | |
1012 | if headersize: |
|
1014 | if headersize: | |
1013 | return self._readexact(headersize) |
|
1015 | return self._readexact(headersize) | |
1014 | return None |
|
1016 | return None | |
1015 |
|
1017 | |||
1016 | def __call__(self): |
|
1018 | def __call__(self): | |
1017 |
|
1019 | |||
1018 | self.ui.debug('bundle2-input-stream-interrupt:' |
|
1020 | self.ui.debug('bundle2-input-stream-interrupt:' | |
1019 | ' opening out of band context\n') |
|
1021 | ' opening out of band context\n') | |
1020 | indebug(self.ui, 'bundle2 stream interruption, looking for a part.') |
|
1022 | indebug(self.ui, 'bundle2 stream interruption, looking for a part.') | |
1021 | headerblock = self._readpartheader() |
|
1023 | headerblock = self._readpartheader() | |
1022 | if headerblock is None: |
|
1024 | if headerblock is None: | |
1023 | indebug(self.ui, 'no part found during interruption.') |
|
1025 | indebug(self.ui, 'no part found during interruption.') | |
1024 | return |
|
1026 | return | |
1025 | part = unbundlepart(self.ui, headerblock, self._fp) |
|
1027 | part = unbundlepart(self.ui, headerblock, self._fp) | |
1026 | op = interruptoperation(self.ui) |
|
1028 | op = interruptoperation(self.ui) | |
1027 | _processpart(op, part) |
|
1029 | _processpart(op, part) | |
1028 | self.ui.debug('bundle2-input-stream-interrupt:' |
|
1030 | self.ui.debug('bundle2-input-stream-interrupt:' | |
1029 | ' closing out of band context\n') |
|
1031 | ' closing out of band context\n') | |
1030 |
|
1032 | |||
1031 | class interruptoperation(object): |
|
1033 | class interruptoperation(object): | |
1032 | """A limited operation to be use by part handler during interruption |
|
1034 | """A limited operation to be use by part handler during interruption | |
1033 |
|
1035 | |||
1034 | It only have access to an ui object. |
|
1036 | It only have access to an ui object. | |
1035 | """ |
|
1037 | """ | |
1036 |
|
1038 | |||
1037 | def __init__(self, ui): |
|
1039 | def __init__(self, ui): | |
1038 | self.ui = ui |
|
1040 | self.ui = ui | |
1039 | self.reply = None |
|
1041 | self.reply = None | |
1040 | self.captureoutput = False |
|
1042 | self.captureoutput = False | |
1041 |
|
1043 | |||
1042 | @property |
|
1044 | @property | |
1043 | def repo(self): |
|
1045 | def repo(self): | |
1044 | raise RuntimeError('no repo access from stream interruption') |
|
1046 | raise RuntimeError('no repo access from stream interruption') | |
1045 |
|
1047 | |||
1046 | def gettransaction(self): |
|
1048 | def gettransaction(self): | |
1047 | raise TransactionUnavailable('no repo access from stream interruption') |
|
1049 | raise TransactionUnavailable('no repo access from stream interruption') | |
1048 |
|
1050 | |||
1049 | class unbundlepart(unpackermixin): |
|
1051 | class unbundlepart(unpackermixin): | |
1050 | """a bundle part read from a bundle""" |
|
1052 | """a bundle part read from a bundle""" | |
1051 |
|
1053 | |||
1052 | def __init__(self, ui, header, fp): |
|
1054 | def __init__(self, ui, header, fp): | |
1053 | super(unbundlepart, self).__init__(fp) |
|
1055 | super(unbundlepart, self).__init__(fp) | |
1054 | self.ui = ui |
|
1056 | self.ui = ui | |
1055 | # unbundle state attr |
|
1057 | # unbundle state attr | |
1056 | self._headerdata = header |
|
1058 | self._headerdata = header | |
1057 | self._headeroffset = 0 |
|
1059 | self._headeroffset = 0 | |
1058 | self._initialized = False |
|
1060 | self._initialized = False | |
1059 | self.consumed = False |
|
1061 | self.consumed = False | |
1060 | # part data |
|
1062 | # part data | |
1061 | self.id = None |
|
1063 | self.id = None | |
1062 | self.type = None |
|
1064 | self.type = None | |
1063 | self.mandatoryparams = None |
|
1065 | self.mandatoryparams = None | |
1064 | self.advisoryparams = None |
|
1066 | self.advisoryparams = None | |
1065 | self.params = None |
|
1067 | self.params = None | |
1066 | self.mandatorykeys = () |
|
1068 | self.mandatorykeys = () | |
1067 | self._payloadstream = None |
|
1069 | self._payloadstream = None | |
1068 | self._readheader() |
|
1070 | self._readheader() | |
1069 | self._mandatory = None |
|
1071 | self._mandatory = None | |
1070 | self._chunkindex = [] #(payload, file) position tuples for chunk starts |
|
1072 | self._chunkindex = [] #(payload, file) position tuples for chunk starts | |
1071 | self._pos = 0 |
|
1073 | self._pos = 0 | |
1072 |
|
1074 | |||
1073 | def _fromheader(self, size): |
|
1075 | def _fromheader(self, size): | |
1074 | """return the next <size> byte from the header""" |
|
1076 | """return the next <size> byte from the header""" | |
1075 | offset = self._headeroffset |
|
1077 | offset = self._headeroffset | |
1076 | data = self._headerdata[offset:(offset + size)] |
|
1078 | data = self._headerdata[offset:(offset + size)] | |
1077 | self._headeroffset = offset + size |
|
1079 | self._headeroffset = offset + size | |
1078 | return data |
|
1080 | return data | |
1079 |
|
1081 | |||
1080 | def _unpackheader(self, format): |
|
1082 | def _unpackheader(self, format): | |
1081 | """read given format from header |
|
1083 | """read given format from header | |
1082 |
|
1084 | |||
1083 | This automatically compute the size of the format to read.""" |
|
1085 | This automatically compute the size of the format to read.""" | |
1084 | data = self._fromheader(struct.calcsize(format)) |
|
1086 | data = self._fromheader(struct.calcsize(format)) | |
1085 | return _unpack(format, data) |
|
1087 | return _unpack(format, data) | |
1086 |
|
1088 | |||
1087 | def _initparams(self, mandatoryparams, advisoryparams): |
|
1089 | def _initparams(self, mandatoryparams, advisoryparams): | |
1088 | """internal function to setup all logic related parameters""" |
|
1090 | """internal function to setup all logic related parameters""" | |
1089 | # make it read only to prevent people touching it by mistake. |
|
1091 | # make it read only to prevent people touching it by mistake. | |
1090 | self.mandatoryparams = tuple(mandatoryparams) |
|
1092 | self.mandatoryparams = tuple(mandatoryparams) | |
1091 | self.advisoryparams = tuple(advisoryparams) |
|
1093 | self.advisoryparams = tuple(advisoryparams) | |
1092 | # user friendly UI |
|
1094 | # user friendly UI | |
1093 | self.params = dict(self.mandatoryparams) |
|
1095 | self.params = dict(self.mandatoryparams) | |
1094 | self.params.update(dict(self.advisoryparams)) |
|
1096 | self.params.update(dict(self.advisoryparams)) | |
1095 | self.mandatorykeys = frozenset(p[0] for p in mandatoryparams) |
|
1097 | self.mandatorykeys = frozenset(p[0] for p in mandatoryparams) | |
1096 |
|
1098 | |||
1097 | def _payloadchunks(self, chunknum=0): |
|
1099 | def _payloadchunks(self, chunknum=0): | |
1098 | '''seek to specified chunk and start yielding data''' |
|
1100 | '''seek to specified chunk and start yielding data''' | |
1099 | if len(self._chunkindex) == 0: |
|
1101 | if len(self._chunkindex) == 0: | |
1100 | assert chunknum == 0, 'Must start with chunk 0' |
|
1102 | assert chunknum == 0, 'Must start with chunk 0' | |
1101 | self._chunkindex.append((0, super(unbundlepart, self).tell())) |
|
1103 | self._chunkindex.append((0, super(unbundlepart, self).tell())) | |
1102 | else: |
|
1104 | else: | |
1103 | assert chunknum < len(self._chunkindex), \ |
|
1105 | assert chunknum < len(self._chunkindex), \ | |
1104 | 'Unknown chunk %d' % chunknum |
|
1106 | 'Unknown chunk %d' % chunknum | |
1105 | super(unbundlepart, self).seek(self._chunkindex[chunknum][1]) |
|
1107 | super(unbundlepart, self).seek(self._chunkindex[chunknum][1]) | |
1106 |
|
1108 | |||
1107 | pos = self._chunkindex[chunknum][0] |
|
1109 | pos = self._chunkindex[chunknum][0] | |
1108 | payloadsize = self._unpack(_fpayloadsize)[0] |
|
1110 | payloadsize = self._unpack(_fpayloadsize)[0] | |
1109 | indebug(self.ui, 'payload chunk size: %i' % payloadsize) |
|
1111 | indebug(self.ui, 'payload chunk size: %i' % payloadsize) | |
1110 | while payloadsize: |
|
1112 | while payloadsize: | |
1111 | if payloadsize == flaginterrupt: |
|
1113 | if payloadsize == flaginterrupt: | |
1112 | # interruption detection, the handler will now read a |
|
1114 | # interruption detection, the handler will now read a | |
1113 | # single part and process it. |
|
1115 | # single part and process it. | |
1114 | interrupthandler(self.ui, self._fp)() |
|
1116 | interrupthandler(self.ui, self._fp)() | |
1115 | elif payloadsize < 0: |
|
1117 | elif payloadsize < 0: | |
1116 | msg = 'negative payload chunk size: %i' % payloadsize |
|
1118 | msg = 'negative payload chunk size: %i' % payloadsize | |
1117 | raise error.BundleValueError(msg) |
|
1119 | raise error.BundleValueError(msg) | |
1118 | else: |
|
1120 | else: | |
1119 | result = self._readexact(payloadsize) |
|
1121 | result = self._readexact(payloadsize) | |
1120 | chunknum += 1 |
|
1122 | chunknum += 1 | |
1121 | pos += payloadsize |
|
1123 | pos += payloadsize | |
1122 | if chunknum == len(self._chunkindex): |
|
1124 | if chunknum == len(self._chunkindex): | |
1123 | self._chunkindex.append((pos, |
|
1125 | self._chunkindex.append((pos, | |
1124 | super(unbundlepart, self).tell())) |
|
1126 | super(unbundlepart, self).tell())) | |
1125 | yield result |
|
1127 | yield result | |
1126 | payloadsize = self._unpack(_fpayloadsize)[0] |
|
1128 | payloadsize = self._unpack(_fpayloadsize)[0] | |
1127 | indebug(self.ui, 'payload chunk size: %i' % payloadsize) |
|
1129 | indebug(self.ui, 'payload chunk size: %i' % payloadsize) | |
1128 |
|
1130 | |||
1129 | def _findchunk(self, pos): |
|
1131 | def _findchunk(self, pos): | |
1130 | '''for a given payload position, return a chunk number and offset''' |
|
1132 | '''for a given payload position, return a chunk number and offset''' | |
1131 | for chunk, (ppos, fpos) in enumerate(self._chunkindex): |
|
1133 | for chunk, (ppos, fpos) in enumerate(self._chunkindex): | |
1132 | if ppos == pos: |
|
1134 | if ppos == pos: | |
1133 | return chunk, 0 |
|
1135 | return chunk, 0 | |
1134 | elif ppos > pos: |
|
1136 | elif ppos > pos: | |
1135 | return chunk - 1, pos - self._chunkindex[chunk - 1][0] |
|
1137 | return chunk - 1, pos - self._chunkindex[chunk - 1][0] | |
1136 | raise ValueError('Unknown chunk') |
|
1138 | raise ValueError('Unknown chunk') | |
1137 |
|
1139 | |||
1138 | def _readheader(self): |
|
1140 | def _readheader(self): | |
1139 | """read the header and setup the object""" |
|
1141 | """read the header and setup the object""" | |
1140 | typesize = self._unpackheader(_fparttypesize)[0] |
|
1142 | typesize = self._unpackheader(_fparttypesize)[0] | |
1141 | self.type = self._fromheader(typesize) |
|
1143 | self.type = self._fromheader(typesize) | |
1142 | indebug(self.ui, 'part type: "%s"' % self.type) |
|
1144 | indebug(self.ui, 'part type: "%s"' % self.type) | |
1143 | self.id = self._unpackheader(_fpartid)[0] |
|
1145 | self.id = self._unpackheader(_fpartid)[0] | |
1144 | indebug(self.ui, 'part id: "%s"' % self.id) |
|
1146 | indebug(self.ui, 'part id: "%s"' % self.id) | |
1145 | # extract mandatory bit from type |
|
1147 | # extract mandatory bit from type | |
1146 | self.mandatory = (self.type != self.type.lower()) |
|
1148 | self.mandatory = (self.type != self.type.lower()) | |
1147 | self.type = self.type.lower() |
|
1149 | self.type = self.type.lower() | |
1148 | ## reading parameters |
|
1150 | ## reading parameters | |
1149 | # param count |
|
1151 | # param count | |
1150 | mancount, advcount = self._unpackheader(_fpartparamcount) |
|
1152 | mancount, advcount = self._unpackheader(_fpartparamcount) | |
1151 | indebug(self.ui, 'part parameters: %i' % (mancount + advcount)) |
|
1153 | indebug(self.ui, 'part parameters: %i' % (mancount + advcount)) | |
1152 | # param size |
|
1154 | # param size | |
1153 | fparamsizes = _makefpartparamsizes(mancount + advcount) |
|
1155 | fparamsizes = _makefpartparamsizes(mancount + advcount) | |
1154 | paramsizes = self._unpackheader(fparamsizes) |
|
1156 | paramsizes = self._unpackheader(fparamsizes) | |
1155 | # make it a list of couple again |
|
1157 | # make it a list of couple again | |
1156 | paramsizes = zip(paramsizes[::2], paramsizes[1::2]) |
|
1158 | paramsizes = zip(paramsizes[::2], paramsizes[1::2]) | |
1157 | # split mandatory from advisory |
|
1159 | # split mandatory from advisory | |
1158 | mansizes = paramsizes[:mancount] |
|
1160 | mansizes = paramsizes[:mancount] | |
1159 | advsizes = paramsizes[mancount:] |
|
1161 | advsizes = paramsizes[mancount:] | |
1160 | # retrieve param value |
|
1162 | # retrieve param value | |
1161 | manparams = [] |
|
1163 | manparams = [] | |
1162 | for key, value in mansizes: |
|
1164 | for key, value in mansizes: | |
1163 | manparams.append((self._fromheader(key), self._fromheader(value))) |
|
1165 | manparams.append((self._fromheader(key), self._fromheader(value))) | |
1164 | advparams = [] |
|
1166 | advparams = [] | |
1165 | for key, value in advsizes: |
|
1167 | for key, value in advsizes: | |
1166 | advparams.append((self._fromheader(key), self._fromheader(value))) |
|
1168 | advparams.append((self._fromheader(key), self._fromheader(value))) | |
1167 | self._initparams(manparams, advparams) |
|
1169 | self._initparams(manparams, advparams) | |
1168 | ## part payload |
|
1170 | ## part payload | |
1169 | self._payloadstream = util.chunkbuffer(self._payloadchunks()) |
|
1171 | self._payloadstream = util.chunkbuffer(self._payloadchunks()) | |
1170 | # we read the data, tell it |
|
1172 | # we read the data, tell it | |
1171 | self._initialized = True |
|
1173 | self._initialized = True | |
1172 |
|
1174 | |||
1173 | def read(self, size=None): |
|
1175 | def read(self, size=None): | |
1174 | """read payload data""" |
|
1176 | """read payload data""" | |
1175 | if not self._initialized: |
|
1177 | if not self._initialized: | |
1176 | self._readheader() |
|
1178 | self._readheader() | |
1177 | if size is None: |
|
1179 | if size is None: | |
1178 | data = self._payloadstream.read() |
|
1180 | data = self._payloadstream.read() | |
1179 | else: |
|
1181 | else: | |
1180 | data = self._payloadstream.read(size) |
|
1182 | data = self._payloadstream.read(size) | |
1181 | self._pos += len(data) |
|
1183 | self._pos += len(data) | |
1182 | if size is None or len(data) < size: |
|
1184 | if size is None or len(data) < size: | |
1183 | if not self.consumed and self._pos: |
|
1185 | if not self.consumed and self._pos: | |
1184 | self.ui.debug('bundle2-input-part: total payload size %i\n' |
|
1186 | self.ui.debug('bundle2-input-part: total payload size %i\n' | |
1185 | % self._pos) |
|
1187 | % self._pos) | |
1186 | self.consumed = True |
|
1188 | self.consumed = True | |
1187 | return data |
|
1189 | return data | |
1188 |
|
1190 | |||
1189 | def tell(self): |
|
1191 | def tell(self): | |
1190 | return self._pos |
|
1192 | return self._pos | |
1191 |
|
1193 | |||
1192 | def seek(self, offset, whence=0): |
|
1194 | def seek(self, offset, whence=0): | |
1193 | if whence == 0: |
|
1195 | if whence == 0: | |
1194 | newpos = offset |
|
1196 | newpos = offset | |
1195 | elif whence == 1: |
|
1197 | elif whence == 1: | |
1196 | newpos = self._pos + offset |
|
1198 | newpos = self._pos + offset | |
1197 | elif whence == 2: |
|
1199 | elif whence == 2: | |
1198 | if not self.consumed: |
|
1200 | if not self.consumed: | |
1199 | self.read() |
|
1201 | self.read() | |
1200 | newpos = self._chunkindex[-1][0] - offset |
|
1202 | newpos = self._chunkindex[-1][0] - offset | |
1201 | else: |
|
1203 | else: | |
1202 | raise ValueError('Unknown whence value: %r' % (whence,)) |
|
1204 | raise ValueError('Unknown whence value: %r' % (whence,)) | |
1203 |
|
1205 | |||
1204 | if newpos > self._chunkindex[-1][0] and not self.consumed: |
|
1206 | if newpos > self._chunkindex[-1][0] and not self.consumed: | |
1205 | self.read() |
|
1207 | self.read() | |
1206 | if not 0 <= newpos <= self._chunkindex[-1][0]: |
|
1208 | if not 0 <= newpos <= self._chunkindex[-1][0]: | |
1207 | raise ValueError('Offset out of range') |
|
1209 | raise ValueError('Offset out of range') | |
1208 |
|
1210 | |||
1209 | if self._pos != newpos: |
|
1211 | if self._pos != newpos: | |
1210 | chunk, internaloffset = self._findchunk(newpos) |
|
1212 | chunk, internaloffset = self._findchunk(newpos) | |
1211 | self._payloadstream = util.chunkbuffer(self._payloadchunks(chunk)) |
|
1213 | self._payloadstream = util.chunkbuffer(self._payloadchunks(chunk)) | |
1212 | adjust = self.read(internaloffset) |
|
1214 | adjust = self.read(internaloffset) | |
1213 | if len(adjust) != internaloffset: |
|
1215 | if len(adjust) != internaloffset: | |
1214 | raise error.Abort(_('Seek failed\n')) |
|
1216 | raise error.Abort(_('Seek failed\n')) | |
1215 | self._pos = newpos |
|
1217 | self._pos = newpos | |
1216 |
|
1218 | |||
1217 | # These are only the static capabilities. |
|
1219 | # These are only the static capabilities. | |
1218 | # Check the 'getrepocaps' function for the rest. |
|
1220 | # Check the 'getrepocaps' function for the rest. | |
1219 | capabilities = {'HG20': (), |
|
1221 | capabilities = {'HG20': (), | |
1220 | 'error': ('abort', 'unsupportedcontent', 'pushraced', |
|
1222 | 'error': ('abort', 'unsupportedcontent', 'pushraced', | |
1221 | 'pushkey'), |
|
1223 | 'pushkey'), | |
1222 | 'listkeys': (), |
|
1224 | 'listkeys': (), | |
1223 | 'pushkey': (), |
|
1225 | 'pushkey': (), | |
1224 | 'digests': tuple(sorted(util.DIGESTS.keys())), |
|
1226 | 'digests': tuple(sorted(util.DIGESTS.keys())), | |
1225 | 'remote-changegroup': ('http', 'https'), |
|
1227 | 'remote-changegroup': ('http', 'https'), | |
1226 | 'hgtagsfnodes': (), |
|
1228 | 'hgtagsfnodes': (), | |
1227 | } |
|
1229 | } | |
1228 |
|
1230 | |||
1229 | def getrepocaps(repo, allowpushback=False): |
|
1231 | def getrepocaps(repo, allowpushback=False): | |
1230 | """return the bundle2 capabilities for a given repo |
|
1232 | """return the bundle2 capabilities for a given repo | |
1231 |
|
1233 | |||
1232 | Exists to allow extensions (like evolution) to mutate the capabilities. |
|
1234 | Exists to allow extensions (like evolution) to mutate the capabilities. | |
1233 | """ |
|
1235 | """ | |
1234 | caps = capabilities.copy() |
|
1236 | caps = capabilities.copy() | |
1235 | caps['changegroup'] = tuple(sorted(changegroup.packermap.keys())) |
|
1237 | caps['changegroup'] = tuple(sorted(changegroup.packermap.keys())) | |
1236 | if obsolete.isenabled(repo, obsolete.exchangeopt): |
|
1238 | if obsolete.isenabled(repo, obsolete.exchangeopt): | |
1237 | supportedformat = tuple('V%i' % v for v in obsolete.formats) |
|
1239 | supportedformat = tuple('V%i' % v for v in obsolete.formats) | |
1238 | caps['obsmarkers'] = supportedformat |
|
1240 | caps['obsmarkers'] = supportedformat | |
1239 | if allowpushback: |
|
1241 | if allowpushback: | |
1240 | caps['pushback'] = () |
|
1242 | caps['pushback'] = () | |
1241 | return caps |
|
1243 | return caps | |
1242 |
|
1244 | |||
1243 | def bundle2caps(remote): |
|
1245 | def bundle2caps(remote): | |
1244 | """return the bundle capabilities of a peer as dict""" |
|
1246 | """return the bundle capabilities of a peer as dict""" | |
1245 | raw = remote.capable('bundle2') |
|
1247 | raw = remote.capable('bundle2') | |
1246 | if not raw and raw != '': |
|
1248 | if not raw and raw != '': | |
1247 | return {} |
|
1249 | return {} | |
1248 | capsblob = urllib.unquote(remote.capable('bundle2')) |
|
1250 | capsblob = urllib.unquote(remote.capable('bundle2')) | |
1249 | return decodecaps(capsblob) |
|
1251 | return decodecaps(capsblob) | |
1250 |
|
1252 | |||
1251 | def obsmarkersversion(caps): |
|
1253 | def obsmarkersversion(caps): | |
1252 | """extract the list of supported obsmarkers versions from a bundle2caps dict |
|
1254 | """extract the list of supported obsmarkers versions from a bundle2caps dict | |
1253 | """ |
|
1255 | """ | |
1254 | obscaps = caps.get('obsmarkers', ()) |
|
1256 | obscaps = caps.get('obsmarkers', ()) | |
1255 | return [int(c[1:]) for c in obscaps if c.startswith('V')] |
|
1257 | return [int(c[1:]) for c in obscaps if c.startswith('V')] | |
1256 |
|
1258 | |||
1257 | @parthandler('changegroup', ('version', 'nbchanges')) |
|
1259 | @parthandler('changegroup', ('version', 'nbchanges')) | |
1258 | def handlechangegroup(op, inpart): |
|
1260 | def handlechangegroup(op, inpart): | |
1259 | """apply a changegroup part on the repo |
|
1261 | """apply a changegroup part on the repo | |
1260 |
|
1262 | |||
1261 | This is a very early implementation that will massive rework before being |
|
1263 | This is a very early implementation that will massive rework before being | |
1262 | inflicted to any end-user. |
|
1264 | inflicted to any end-user. | |
1263 | """ |
|
1265 | """ | |
1264 | # Make sure we trigger a transaction creation |
|
1266 | # Make sure we trigger a transaction creation | |
1265 | # |
|
1267 | # | |
1266 | # The addchangegroup function will get a transaction object by itself, but |
|
1268 | # The addchangegroup function will get a transaction object by itself, but | |
1267 | # we need to make sure we trigger the creation of a transaction object used |
|
1269 | # we need to make sure we trigger the creation of a transaction object used | |
1268 | # for the whole processing scope. |
|
1270 | # for the whole processing scope. | |
1269 | op.gettransaction() |
|
1271 | op.gettransaction() | |
1270 | unpackerversion = inpart.params.get('version', '01') |
|
1272 | unpackerversion = inpart.params.get('version', '01') | |
1271 | # We should raise an appropriate exception here |
|
1273 | # We should raise an appropriate exception here | |
1272 | unpacker = changegroup.packermap[unpackerversion][1] |
|
1274 | unpacker = changegroup.packermap[unpackerversion][1] | |
1273 | cg = unpacker(inpart, None) |
|
1275 | cg = unpacker(inpart, None) | |
1274 | # the source and url passed here are overwritten by the one contained in |
|
1276 | # the source and url passed here are overwritten by the one contained in | |
1275 | # the transaction.hookargs argument. So 'bundle2' is a placeholder |
|
1277 | # the transaction.hookargs argument. So 'bundle2' is a placeholder | |
1276 | nbchangesets = None |
|
1278 | nbchangesets = None | |
1277 | if 'nbchanges' in inpart.params: |
|
1279 | if 'nbchanges' in inpart.params: | |
1278 | nbchangesets = int(inpart.params.get('nbchanges')) |
|
1280 | nbchangesets = int(inpart.params.get('nbchanges')) | |
1279 | ret = cg.apply(op.repo, 'bundle2', 'bundle2', expectedtotal=nbchangesets) |
|
1281 | ret = cg.apply(op.repo, 'bundle2', 'bundle2', expectedtotal=nbchangesets) | |
1280 | op.records.add('changegroup', {'return': ret}) |
|
1282 | op.records.add('changegroup', {'return': ret}) | |
1281 | if op.reply is not None: |
|
1283 | if op.reply is not None: | |
1282 | # This is definitely not the final form of this |
|
1284 | # This is definitely not the final form of this | |
1283 | # return. But one need to start somewhere. |
|
1285 | # return. But one need to start somewhere. | |
1284 | part = op.reply.newpart('reply:changegroup', mandatory=False) |
|
1286 | part = op.reply.newpart('reply:changegroup', mandatory=False) | |
1285 | part.addparam('in-reply-to', str(inpart.id), mandatory=False) |
|
1287 | part.addparam('in-reply-to', str(inpart.id), mandatory=False) | |
1286 | part.addparam('return', '%i' % ret, mandatory=False) |
|
1288 | part.addparam('return', '%i' % ret, mandatory=False) | |
1287 | assert not inpart.read() |
|
1289 | assert not inpart.read() | |
1288 |
|
1290 | |||
1289 | _remotechangegroupparams = tuple(['url', 'size', 'digests'] + |
|
1291 | _remotechangegroupparams = tuple(['url', 'size', 'digests'] + | |
1290 | ['digest:%s' % k for k in util.DIGESTS.keys()]) |
|
1292 | ['digest:%s' % k for k in util.DIGESTS.keys()]) | |
1291 | @parthandler('remote-changegroup', _remotechangegroupparams) |
|
1293 | @parthandler('remote-changegroup', _remotechangegroupparams) | |
1292 | def handleremotechangegroup(op, inpart): |
|
1294 | def handleremotechangegroup(op, inpart): | |
1293 | """apply a bundle10 on the repo, given an url and validation information |
|
1295 | """apply a bundle10 on the repo, given an url and validation information | |
1294 |
|
1296 | |||
1295 | All the information about the remote bundle to import are given as |
|
1297 | All the information about the remote bundle to import are given as | |
1296 | parameters. The parameters include: |
|
1298 | parameters. The parameters include: | |
1297 | - url: the url to the bundle10. |
|
1299 | - url: the url to the bundle10. | |
1298 | - size: the bundle10 file size. It is used to validate what was |
|
1300 | - size: the bundle10 file size. It is used to validate what was | |
1299 | retrieved by the client matches the server knowledge about the bundle. |
|
1301 | retrieved by the client matches the server knowledge about the bundle. | |
1300 | - digests: a space separated list of the digest types provided as |
|
1302 | - digests: a space separated list of the digest types provided as | |
1301 | parameters. |
|
1303 | parameters. | |
1302 | - digest:<digest-type>: the hexadecimal representation of the digest with |
|
1304 | - digest:<digest-type>: the hexadecimal representation of the digest with | |
1303 | that name. Like the size, it is used to validate what was retrieved by |
|
1305 | that name. Like the size, it is used to validate what was retrieved by | |
1304 | the client matches what the server knows about the bundle. |
|
1306 | the client matches what the server knows about the bundle. | |
1305 |
|
1307 | |||
1306 | When multiple digest types are given, all of them are checked. |
|
1308 | When multiple digest types are given, all of them are checked. | |
1307 | """ |
|
1309 | """ | |
1308 | try: |
|
1310 | try: | |
1309 | raw_url = inpart.params['url'] |
|
1311 | raw_url = inpart.params['url'] | |
1310 | except KeyError: |
|
1312 | except KeyError: | |
1311 | raise error.Abort(_('remote-changegroup: missing "%s" param') % 'url') |
|
1313 | raise error.Abort(_('remote-changegroup: missing "%s" param') % 'url') | |
1312 | parsed_url = util.url(raw_url) |
|
1314 | parsed_url = util.url(raw_url) | |
1313 | if parsed_url.scheme not in capabilities['remote-changegroup']: |
|
1315 | if parsed_url.scheme not in capabilities['remote-changegroup']: | |
1314 | raise error.Abort(_('remote-changegroup does not support %s urls') % |
|
1316 | raise error.Abort(_('remote-changegroup does not support %s urls') % | |
1315 | parsed_url.scheme) |
|
1317 | parsed_url.scheme) | |
1316 |
|
1318 | |||
1317 | try: |
|
1319 | try: | |
1318 | size = int(inpart.params['size']) |
|
1320 | size = int(inpart.params['size']) | |
1319 | except ValueError: |
|
1321 | except ValueError: | |
1320 | raise error.Abort(_('remote-changegroup: invalid value for param "%s"') |
|
1322 | raise error.Abort(_('remote-changegroup: invalid value for param "%s"') | |
1321 | % 'size') |
|
1323 | % 'size') | |
1322 | except KeyError: |
|
1324 | except KeyError: | |
1323 | raise error.Abort(_('remote-changegroup: missing "%s" param') % 'size') |
|
1325 | raise error.Abort(_('remote-changegroup: missing "%s" param') % 'size') | |
1324 |
|
1326 | |||
1325 | digests = {} |
|
1327 | digests = {} | |
1326 | for typ in inpart.params.get('digests', '').split(): |
|
1328 | for typ in inpart.params.get('digests', '').split(): | |
1327 | param = 'digest:%s' % typ |
|
1329 | param = 'digest:%s' % typ | |
1328 | try: |
|
1330 | try: | |
1329 | value = inpart.params[param] |
|
1331 | value = inpart.params[param] | |
1330 | except KeyError: |
|
1332 | except KeyError: | |
1331 | raise error.Abort(_('remote-changegroup: missing "%s" param') % |
|
1333 | raise error.Abort(_('remote-changegroup: missing "%s" param') % | |
1332 | param) |
|
1334 | param) | |
1333 | digests[typ] = value |
|
1335 | digests[typ] = value | |
1334 |
|
1336 | |||
1335 | real_part = util.digestchecker(url.open(op.ui, raw_url), size, digests) |
|
1337 | real_part = util.digestchecker(url.open(op.ui, raw_url), size, digests) | |
1336 |
|
1338 | |||
1337 | # Make sure we trigger a transaction creation |
|
1339 | # Make sure we trigger a transaction creation | |
1338 | # |
|
1340 | # | |
1339 | # The addchangegroup function will get a transaction object by itself, but |
|
1341 | # The addchangegroup function will get a transaction object by itself, but | |
1340 | # we need to make sure we trigger the creation of a transaction object used |
|
1342 | # we need to make sure we trigger the creation of a transaction object used | |
1341 | # for the whole processing scope. |
|
1343 | # for the whole processing scope. | |
1342 | op.gettransaction() |
|
1344 | op.gettransaction() | |
1343 | from . import exchange |
|
1345 | from . import exchange | |
1344 | cg = exchange.readbundle(op.repo.ui, real_part, raw_url) |
|
1346 | cg = exchange.readbundle(op.repo.ui, real_part, raw_url) | |
1345 | if not isinstance(cg, changegroup.cg1unpacker): |
|
1347 | if not isinstance(cg, changegroup.cg1unpacker): | |
1346 | raise error.Abort(_('%s: not a bundle version 1.0') % |
|
1348 | raise error.Abort(_('%s: not a bundle version 1.0') % | |
1347 | util.hidepassword(raw_url)) |
|
1349 | util.hidepassword(raw_url)) | |
1348 | ret = cg.apply(op.repo, 'bundle2', 'bundle2') |
|
1350 | ret = cg.apply(op.repo, 'bundle2', 'bundle2') | |
1349 | op.records.add('changegroup', {'return': ret}) |
|
1351 | op.records.add('changegroup', {'return': ret}) | |
1350 | if op.reply is not None: |
|
1352 | if op.reply is not None: | |
1351 | # This is definitely not the final form of this |
|
1353 | # This is definitely not the final form of this | |
1352 | # return. But one need to start somewhere. |
|
1354 | # return. But one need to start somewhere. | |
1353 | part = op.reply.newpart('reply:changegroup') |
|
1355 | part = op.reply.newpart('reply:changegroup') | |
1354 | part.addparam('in-reply-to', str(inpart.id), mandatory=False) |
|
1356 | part.addparam('in-reply-to', str(inpart.id), mandatory=False) | |
1355 | part.addparam('return', '%i' % ret, mandatory=False) |
|
1357 | part.addparam('return', '%i' % ret, mandatory=False) | |
1356 | try: |
|
1358 | try: | |
1357 | real_part.validate() |
|
1359 | real_part.validate() | |
1358 | except error.Abort as e: |
|
1360 | except error.Abort as e: | |
1359 | raise error.Abort(_('bundle at %s is corrupted:\n%s') % |
|
1361 | raise error.Abort(_('bundle at %s is corrupted:\n%s') % | |
1360 | (util.hidepassword(raw_url), str(e))) |
|
1362 | (util.hidepassword(raw_url), str(e))) | |
1361 | assert not inpart.read() |
|
1363 | assert not inpart.read() | |
1362 |
|
1364 | |||
1363 | @parthandler('reply:changegroup', ('return', 'in-reply-to')) |
|
1365 | @parthandler('reply:changegroup', ('return', 'in-reply-to')) | |
1364 | def handlereplychangegroup(op, inpart): |
|
1366 | def handlereplychangegroup(op, inpart): | |
1365 | ret = int(inpart.params['return']) |
|
1367 | ret = int(inpart.params['return']) | |
1366 | replyto = int(inpart.params['in-reply-to']) |
|
1368 | replyto = int(inpart.params['in-reply-to']) | |
1367 | op.records.add('changegroup', {'return': ret}, replyto) |
|
1369 | op.records.add('changegroup', {'return': ret}, replyto) | |
1368 |
|
1370 | |||
1369 | @parthandler('check:heads') |
|
1371 | @parthandler('check:heads') | |
1370 | def handlecheckheads(op, inpart): |
|
1372 | def handlecheckheads(op, inpart): | |
1371 | """check that head of the repo did not change |
|
1373 | """check that head of the repo did not change | |
1372 |
|
1374 | |||
1373 | This is used to detect a push race when using unbundle. |
|
1375 | This is used to detect a push race when using unbundle. | |
1374 | This replaces the "heads" argument of unbundle.""" |
|
1376 | This replaces the "heads" argument of unbundle.""" | |
1375 | h = inpart.read(20) |
|
1377 | h = inpart.read(20) | |
1376 | heads = [] |
|
1378 | heads = [] | |
1377 | while len(h) == 20: |
|
1379 | while len(h) == 20: | |
1378 | heads.append(h) |
|
1380 | heads.append(h) | |
1379 | h = inpart.read(20) |
|
1381 | h = inpart.read(20) | |
1380 | assert not h |
|
1382 | assert not h | |
1381 | # Trigger a transaction so that we are guaranteed to have the lock now. |
|
1383 | # Trigger a transaction so that we are guaranteed to have the lock now. | |
1382 | if op.ui.configbool('experimental', 'bundle2lazylocking'): |
|
1384 | if op.ui.configbool('experimental', 'bundle2lazylocking'): | |
1383 | op.gettransaction() |
|
1385 | op.gettransaction() | |
1384 | if heads != op.repo.heads(): |
|
1386 | if heads != op.repo.heads(): | |
1385 | raise error.PushRaced('repository changed while pushing - ' |
|
1387 | raise error.PushRaced('repository changed while pushing - ' | |
1386 | 'please try again') |
|
1388 | 'please try again') | |
1387 |
|
1389 | |||
1388 | @parthandler('output') |
|
1390 | @parthandler('output') | |
1389 | def handleoutput(op, inpart): |
|
1391 | def handleoutput(op, inpart): | |
1390 | """forward output captured on the server to the client""" |
|
1392 | """forward output captured on the server to the client""" | |
1391 | for line in inpart.read().splitlines(): |
|
1393 | for line in inpart.read().splitlines(): | |
1392 | op.ui.status(('remote: %s\n' % line)) |
|
1394 | op.ui.status(('remote: %s\n' % line)) | |
1393 |
|
1395 | |||
1394 | @parthandler('replycaps') |
|
1396 | @parthandler('replycaps') | |
1395 | def handlereplycaps(op, inpart): |
|
1397 | def handlereplycaps(op, inpart): | |
1396 | """Notify that a reply bundle should be created |
|
1398 | """Notify that a reply bundle should be created | |
1397 |
|
1399 | |||
1398 | The payload contains the capabilities information for the reply""" |
|
1400 | The payload contains the capabilities information for the reply""" | |
1399 | caps = decodecaps(inpart.read()) |
|
1401 | caps = decodecaps(inpart.read()) | |
1400 | if op.reply is None: |
|
1402 | if op.reply is None: | |
1401 | op.reply = bundle20(op.ui, caps) |
|
1403 | op.reply = bundle20(op.ui, caps) | |
1402 |
|
1404 | |||
1403 | @parthandler('error:abort', ('message', 'hint')) |
|
1405 | @parthandler('error:abort', ('message', 'hint')) | |
1404 | def handleerrorabort(op, inpart): |
|
1406 | def handleerrorabort(op, inpart): | |
1405 | """Used to transmit abort error over the wire""" |
|
1407 | """Used to transmit abort error over the wire""" | |
1406 | raise error.Abort(inpart.params['message'], hint=inpart.params.get('hint')) |
|
1408 | raise error.Abort(inpart.params['message'], hint=inpart.params.get('hint')) | |
1407 |
|
1409 | |||
1408 | @parthandler('error:pushkey', ('namespace', 'key', 'new', 'old', 'ret', |
|
1410 | @parthandler('error:pushkey', ('namespace', 'key', 'new', 'old', 'ret', | |
1409 | 'in-reply-to')) |
|
1411 | 'in-reply-to')) | |
1410 | def handleerrorpushkey(op, inpart): |
|
1412 | def handleerrorpushkey(op, inpart): | |
1411 | """Used to transmit failure of a mandatory pushkey over the wire""" |
|
1413 | """Used to transmit failure of a mandatory pushkey over the wire""" | |
1412 | kwargs = {} |
|
1414 | kwargs = {} | |
1413 | for name in ('namespace', 'key', 'new', 'old', 'ret'): |
|
1415 | for name in ('namespace', 'key', 'new', 'old', 'ret'): | |
1414 | value = inpart.params.get(name) |
|
1416 | value = inpart.params.get(name) | |
1415 | if value is not None: |
|
1417 | if value is not None: | |
1416 | kwargs[name] = value |
|
1418 | kwargs[name] = value | |
1417 | raise error.PushkeyFailed(inpart.params['in-reply-to'], **kwargs) |
|
1419 | raise error.PushkeyFailed(inpart.params['in-reply-to'], **kwargs) | |
1418 |
|
1420 | |||
1419 | @parthandler('error:unsupportedcontent', ('parttype', 'params')) |
|
1421 | @parthandler('error:unsupportedcontent', ('parttype', 'params')) | |
1420 | def handleerrorunsupportedcontent(op, inpart): |
|
1422 | def handleerrorunsupportedcontent(op, inpart): | |
1421 | """Used to transmit unknown content error over the wire""" |
|
1423 | """Used to transmit unknown content error over the wire""" | |
1422 | kwargs = {} |
|
1424 | kwargs = {} | |
1423 | parttype = inpart.params.get('parttype') |
|
1425 | parttype = inpart.params.get('parttype') | |
1424 | if parttype is not None: |
|
1426 | if parttype is not None: | |
1425 | kwargs['parttype'] = parttype |
|
1427 | kwargs['parttype'] = parttype | |
1426 | params = inpart.params.get('params') |
|
1428 | params = inpart.params.get('params') | |
1427 | if params is not None: |
|
1429 | if params is not None: | |
1428 | kwargs['params'] = params.split('\0') |
|
1430 | kwargs['params'] = params.split('\0') | |
1429 |
|
1431 | |||
1430 | raise error.BundleUnknownFeatureError(**kwargs) |
|
1432 | raise error.BundleUnknownFeatureError(**kwargs) | |
1431 |
|
1433 | |||
1432 | @parthandler('error:pushraced', ('message',)) |
|
1434 | @parthandler('error:pushraced', ('message',)) | |
1433 | def handleerrorpushraced(op, inpart): |
|
1435 | def handleerrorpushraced(op, inpart): | |
1434 | """Used to transmit push race error over the wire""" |
|
1436 | """Used to transmit push race error over the wire""" | |
1435 | raise error.ResponseError(_('push failed:'), inpart.params['message']) |
|
1437 | raise error.ResponseError(_('push failed:'), inpart.params['message']) | |
1436 |
|
1438 | |||
1437 | @parthandler('listkeys', ('namespace',)) |
|
1439 | @parthandler('listkeys', ('namespace',)) | |
1438 | def handlelistkeys(op, inpart): |
|
1440 | def handlelistkeys(op, inpart): | |
1439 | """retrieve pushkey namespace content stored in a bundle2""" |
|
1441 | """retrieve pushkey namespace content stored in a bundle2""" | |
1440 | namespace = inpart.params['namespace'] |
|
1442 | namespace = inpart.params['namespace'] | |
1441 | r = pushkey.decodekeys(inpart.read()) |
|
1443 | r = pushkey.decodekeys(inpart.read()) | |
1442 | op.records.add('listkeys', (namespace, r)) |
|
1444 | op.records.add('listkeys', (namespace, r)) | |
1443 |
|
1445 | |||
1444 | @parthandler('pushkey', ('namespace', 'key', 'old', 'new')) |
|
1446 | @parthandler('pushkey', ('namespace', 'key', 'old', 'new')) | |
1445 | def handlepushkey(op, inpart): |
|
1447 | def handlepushkey(op, inpart): | |
1446 | """process a pushkey request""" |
|
1448 | """process a pushkey request""" | |
1447 | dec = pushkey.decode |
|
1449 | dec = pushkey.decode | |
1448 | namespace = dec(inpart.params['namespace']) |
|
1450 | namespace = dec(inpart.params['namespace']) | |
1449 | key = dec(inpart.params['key']) |
|
1451 | key = dec(inpart.params['key']) | |
1450 | old = dec(inpart.params['old']) |
|
1452 | old = dec(inpart.params['old']) | |
1451 | new = dec(inpart.params['new']) |
|
1453 | new = dec(inpart.params['new']) | |
1452 | # Grab the transaction to ensure that we have the lock before performing the |
|
1454 | # Grab the transaction to ensure that we have the lock before performing the | |
1453 | # pushkey. |
|
1455 | # pushkey. | |
1454 | if op.ui.configbool('experimental', 'bundle2lazylocking'): |
|
1456 | if op.ui.configbool('experimental', 'bundle2lazylocking'): | |
1455 | op.gettransaction() |
|
1457 | op.gettransaction() | |
1456 | ret = op.repo.pushkey(namespace, key, old, new) |
|
1458 | ret = op.repo.pushkey(namespace, key, old, new) | |
1457 | record = {'namespace': namespace, |
|
1459 | record = {'namespace': namespace, | |
1458 | 'key': key, |
|
1460 | 'key': key, | |
1459 | 'old': old, |
|
1461 | 'old': old, | |
1460 | 'new': new} |
|
1462 | 'new': new} | |
1461 | op.records.add('pushkey', record) |
|
1463 | op.records.add('pushkey', record) | |
1462 | if op.reply is not None: |
|
1464 | if op.reply is not None: | |
1463 | rpart = op.reply.newpart('reply:pushkey') |
|
1465 | rpart = op.reply.newpart('reply:pushkey') | |
1464 | rpart.addparam('in-reply-to', str(inpart.id), mandatory=False) |
|
1466 | rpart.addparam('in-reply-to', str(inpart.id), mandatory=False) | |
1465 | rpart.addparam('return', '%i' % ret, mandatory=False) |
|
1467 | rpart.addparam('return', '%i' % ret, mandatory=False) | |
1466 | if inpart.mandatory and not ret: |
|
1468 | if inpart.mandatory and not ret: | |
1467 | kwargs = {} |
|
1469 | kwargs = {} | |
1468 | for key in ('namespace', 'key', 'new', 'old', 'ret'): |
|
1470 | for key in ('namespace', 'key', 'new', 'old', 'ret'): | |
1469 | if key in inpart.params: |
|
1471 | if key in inpart.params: | |
1470 | kwargs[key] = inpart.params[key] |
|
1472 | kwargs[key] = inpart.params[key] | |
1471 | raise error.PushkeyFailed(partid=str(inpart.id), **kwargs) |
|
1473 | raise error.PushkeyFailed(partid=str(inpart.id), **kwargs) | |
1472 |
|
1474 | |||
1473 | @parthandler('reply:pushkey', ('return', 'in-reply-to')) |
|
1475 | @parthandler('reply:pushkey', ('return', 'in-reply-to')) | |
1474 | def handlepushkeyreply(op, inpart): |
|
1476 | def handlepushkeyreply(op, inpart): | |
1475 | """retrieve the result of a pushkey request""" |
|
1477 | """retrieve the result of a pushkey request""" | |
1476 | ret = int(inpart.params['return']) |
|
1478 | ret = int(inpart.params['return']) | |
1477 | partid = int(inpart.params['in-reply-to']) |
|
1479 | partid = int(inpart.params['in-reply-to']) | |
1478 | op.records.add('pushkey', {'return': ret}, partid) |
|
1480 | op.records.add('pushkey', {'return': ret}, partid) | |
1479 |
|
1481 | |||
1480 | @parthandler('obsmarkers') |
|
1482 | @parthandler('obsmarkers') | |
1481 | def handleobsmarker(op, inpart): |
|
1483 | def handleobsmarker(op, inpart): | |
1482 | """add a stream of obsmarkers to the repo""" |
|
1484 | """add a stream of obsmarkers to the repo""" | |
1483 | tr = op.gettransaction() |
|
1485 | tr = op.gettransaction() | |
1484 | markerdata = inpart.read() |
|
1486 | markerdata = inpart.read() | |
1485 | if op.ui.config('experimental', 'obsmarkers-exchange-debug', False): |
|
1487 | if op.ui.config('experimental', 'obsmarkers-exchange-debug', False): | |
1486 | op.ui.write(('obsmarker-exchange: %i bytes received\n') |
|
1488 | op.ui.write(('obsmarker-exchange: %i bytes received\n') | |
1487 | % len(markerdata)) |
|
1489 | % len(markerdata)) | |
1488 | # The mergemarkers call will crash if marker creation is not enabled. |
|
1490 | # The mergemarkers call will crash if marker creation is not enabled. | |
1489 | # we want to avoid this if the part is advisory. |
|
1491 | # we want to avoid this if the part is advisory. | |
1490 | if not inpart.mandatory and op.repo.obsstore.readonly: |
|
1492 | if not inpart.mandatory and op.repo.obsstore.readonly: | |
1491 | op.repo.ui.debug('ignoring obsolescence markers, feature not enabled') |
|
1493 | op.repo.ui.debug('ignoring obsolescence markers, feature not enabled') | |
1492 | return |
|
1494 | return | |
1493 | new = op.repo.obsstore.mergemarkers(tr, markerdata) |
|
1495 | new = op.repo.obsstore.mergemarkers(tr, markerdata) | |
1494 | if new: |
|
1496 | if new: | |
1495 | op.repo.ui.status(_('%i new obsolescence markers\n') % new) |
|
1497 | op.repo.ui.status(_('%i new obsolescence markers\n') % new) | |
1496 | op.records.add('obsmarkers', {'new': new}) |
|
1498 | op.records.add('obsmarkers', {'new': new}) | |
1497 | if op.reply is not None: |
|
1499 | if op.reply is not None: | |
1498 | rpart = op.reply.newpart('reply:obsmarkers') |
|
1500 | rpart = op.reply.newpart('reply:obsmarkers') | |
1499 | rpart.addparam('in-reply-to', str(inpart.id), mandatory=False) |
|
1501 | rpart.addparam('in-reply-to', str(inpart.id), mandatory=False) | |
1500 | rpart.addparam('new', '%i' % new, mandatory=False) |
|
1502 | rpart.addparam('new', '%i' % new, mandatory=False) | |
1501 |
|
1503 | |||
1502 |
|
1504 | |||
1503 | @parthandler('reply:obsmarkers', ('new', 'in-reply-to')) |
|
1505 | @parthandler('reply:obsmarkers', ('new', 'in-reply-to')) | |
1504 | def handleobsmarkerreply(op, inpart): |
|
1506 | def handleobsmarkerreply(op, inpart): | |
1505 | """retrieve the result of a pushkey request""" |
|
1507 | """retrieve the result of a pushkey request""" | |
1506 | ret = int(inpart.params['new']) |
|
1508 | ret = int(inpart.params['new']) | |
1507 | partid = int(inpart.params['in-reply-to']) |
|
1509 | partid = int(inpart.params['in-reply-to']) | |
1508 | op.records.add('obsmarkers', {'new': ret}, partid) |
|
1510 | op.records.add('obsmarkers', {'new': ret}, partid) | |
1509 |
|
1511 | |||
1510 | @parthandler('hgtagsfnodes') |
|
1512 | @parthandler('hgtagsfnodes') | |
1511 | def handlehgtagsfnodes(op, inpart): |
|
1513 | def handlehgtagsfnodes(op, inpart): | |
1512 | """Applies .hgtags fnodes cache entries to the local repo. |
|
1514 | """Applies .hgtags fnodes cache entries to the local repo. | |
1513 |
|
1515 | |||
1514 | Payload is pairs of 20 byte changeset nodes and filenodes. |
|
1516 | Payload is pairs of 20 byte changeset nodes and filenodes. | |
1515 | """ |
|
1517 | """ | |
1516 | # Grab the transaction so we ensure that we have the lock at this point. |
|
1518 | # Grab the transaction so we ensure that we have the lock at this point. | |
1517 | if op.ui.configbool('experimental', 'bundle2lazylocking'): |
|
1519 | if op.ui.configbool('experimental', 'bundle2lazylocking'): | |
1518 | op.gettransaction() |
|
1520 | op.gettransaction() | |
1519 | cache = tags.hgtagsfnodescache(op.repo.unfiltered()) |
|
1521 | cache = tags.hgtagsfnodescache(op.repo.unfiltered()) | |
1520 |
|
1522 | |||
1521 | count = 0 |
|
1523 | count = 0 | |
1522 | while True: |
|
1524 | while True: | |
1523 | node = inpart.read(20) |
|
1525 | node = inpart.read(20) | |
1524 | fnode = inpart.read(20) |
|
1526 | fnode = inpart.read(20) | |
1525 | if len(node) < 20 or len(fnode) < 20: |
|
1527 | if len(node) < 20 or len(fnode) < 20: | |
1526 | op.ui.debug('ignoring incomplete received .hgtags fnodes data\n') |
|
1528 | op.ui.debug('ignoring incomplete received .hgtags fnodes data\n') | |
1527 | break |
|
1529 | break | |
1528 | cache.setfnode(node, fnode) |
|
1530 | cache.setfnode(node, fnode) | |
1529 | count += 1 |
|
1531 | count += 1 | |
1530 |
|
1532 | |||
1531 | cache.write() |
|
1533 | cache.write() | |
1532 | op.ui.debug('applied %i hgtags fnodes cache entries\n' % count) |
|
1534 | op.ui.debug('applied %i hgtags fnodes cache entries\n' % count) |
General Comments 0
You need to be logged in to leave comments.
Login now