Show More
@@ -1,1146 +1,1146 b'' | |||||
1 | # obsolete.py - obsolete markers handling |
|
1 | # obsolete.py - obsolete markers handling | |
2 | # |
|
2 | # | |
3 | # Copyright 2012 Pierre-Yves David <pierre-yves.david@ens-lyon.org> |
|
3 | # Copyright 2012 Pierre-Yves David <pierre-yves.david@ens-lyon.org> | |
4 | # Logilab SA <contact@logilab.fr> |
|
4 | # Logilab SA <contact@logilab.fr> | |
5 | # |
|
5 | # | |
6 | # This software may be used and distributed according to the terms of the |
|
6 | # This software may be used and distributed according to the terms of the | |
7 | # GNU General Public License version 2 or any later version. |
|
7 | # GNU General Public License version 2 or any later version. | |
8 |
|
8 | |||
9 | """Obsolete marker handling |
|
9 | """Obsolete marker handling | |
10 |
|
10 | |||
11 | An obsolete marker maps an old changeset to a list of new |
|
11 | An obsolete marker maps an old changeset to a list of new | |
12 | changesets. If the list of new changesets is empty, the old changeset |
|
12 | changesets. If the list of new changesets is empty, the old changeset | |
13 | is said to be "killed". Otherwise, the old changeset is being |
|
13 | is said to be "killed". Otherwise, the old changeset is being | |
14 | "replaced" by the new changesets. |
|
14 | "replaced" by the new changesets. | |
15 |
|
15 | |||
16 | Obsolete markers can be used to record and distribute changeset graph |
|
16 | Obsolete markers can be used to record and distribute changeset graph | |
17 | transformations performed by history rewrite operations, and help |
|
17 | transformations performed by history rewrite operations, and help | |
18 | building new tools to reconcile conflicting rewrite actions. To |
|
18 | building new tools to reconcile conflicting rewrite actions. To | |
19 | facilitate conflict resolution, markers include various annotations |
|
19 | facilitate conflict resolution, markers include various annotations | |
20 | besides old and news changeset identifiers, such as creation date or |
|
20 | besides old and news changeset identifiers, such as creation date or | |
21 | author name. |
|
21 | author name. | |
22 |
|
22 | |||
23 | The old obsoleted changeset is called a "precursor" and possible |
|
23 | The old obsoleted changeset is called a "precursor" and possible | |
24 | replacements are called "successors". Markers that used changeset X as |
|
24 | replacements are called "successors". Markers that used changeset X as | |
25 | a precursor are called "successor markers of X" because they hold |
|
25 | a precursor are called "successor markers of X" because they hold | |
26 | information about the successors of X. Markers that use changeset Y as |
|
26 | information about the successors of X. Markers that use changeset Y as | |
27 | a successors are call "precursor markers of Y" because they hold |
|
27 | a successors are call "precursor markers of Y" because they hold | |
28 | information about the precursors of Y. |
|
28 | information about the precursors of Y. | |
29 |
|
29 | |||
30 | Examples: |
|
30 | Examples: | |
31 |
|
31 | |||
32 | - When changeset A is replaced by changeset A', one marker is stored: |
|
32 | - When changeset A is replaced by changeset A', one marker is stored: | |
33 |
|
33 | |||
34 | (A, (A',)) |
|
34 | (A, (A',)) | |
35 |
|
35 | |||
36 | - When changesets A and B are folded into a new changeset C, two markers are |
|
36 | - When changesets A and B are folded into a new changeset C, two markers are | |
37 | stored: |
|
37 | stored: | |
38 |
|
38 | |||
39 | (A, (C,)) and (B, (C,)) |
|
39 | (A, (C,)) and (B, (C,)) | |
40 |
|
40 | |||
41 | - When changeset A is simply "pruned" from the graph, a marker is created: |
|
41 | - When changeset A is simply "pruned" from the graph, a marker is created: | |
42 |
|
42 | |||
43 | (A, ()) |
|
43 | (A, ()) | |
44 |
|
44 | |||
45 | - When changeset A is split into B and C, a single marker are used: |
|
45 | - When changeset A is split into B and C, a single marker are used: | |
46 |
|
46 | |||
47 | (A, (C, C)) |
|
47 | (A, (C, C)) | |
48 |
|
48 | |||
49 | We use a single marker to distinguish the "split" case from the "divergence" |
|
49 | We use a single marker to distinguish the "split" case from the "divergence" | |
50 | case. If two independent operations rewrite the same changeset A in to A' and |
|
50 | case. If two independent operations rewrite the same changeset A in to A' and | |
51 | A'', we have an error case: divergent rewriting. We can detect it because |
|
51 | A'', we have an error case: divergent rewriting. We can detect it because | |
52 | two markers will be created independently: |
|
52 | two markers will be created independently: | |
53 |
|
53 | |||
54 | (A, (B,)) and (A, (C,)) |
|
54 | (A, (B,)) and (A, (C,)) | |
55 |
|
55 | |||
56 | Format |
|
56 | Format | |
57 | ------ |
|
57 | ------ | |
58 |
|
58 | |||
59 | Markers are stored in an append-only file stored in |
|
59 | Markers are stored in an append-only file stored in | |
60 | '.hg/store/obsstore'. |
|
60 | '.hg/store/obsstore'. | |
61 |
|
61 | |||
62 | The file starts with a version header: |
|
62 | The file starts with a version header: | |
63 |
|
63 | |||
64 | - 1 unsigned byte: version number, starting at zero. |
|
64 | - 1 unsigned byte: version number, starting at zero. | |
65 |
|
65 | |||
66 | The header is followed by the markers. Marker format depend of the version. See |
|
66 | The header is followed by the markers. Marker format depend of the version. See | |
67 | comment associated with each format for details. |
|
67 | comment associated with each format for details. | |
68 |
|
68 | |||
69 | """ |
|
69 | """ | |
70 | import struct |
|
70 | import struct | |
71 | import util, base85, node |
|
71 | import util, base85, node | |
72 | import phases |
|
72 | import phases | |
73 | from i18n import _ |
|
73 | from i18n import _ | |
74 |
|
74 | |||
75 | _pack = struct.pack |
|
75 | _pack = struct.pack | |
76 | _unpack = struct.unpack |
|
76 | _unpack = struct.unpack | |
77 |
|
77 | |||
78 | _SEEK_END = 2 # os.SEEK_END was introduced in Python 2.5 |
|
78 | _SEEK_END = 2 # os.SEEK_END was introduced in Python 2.5 | |
79 |
|
79 | |||
80 | # the obsolete feature is not mature enough to be enabled by default. |
|
80 | # the obsolete feature is not mature enough to be enabled by default. | |
81 | # you have to rely on third party extension extension to enable this. |
|
81 | # you have to rely on third party extension extension to enable this. | |
82 | _enabled = False |
|
82 | _enabled = False | |
83 |
|
83 | |||
84 | ### obsolescence marker flag |
|
84 | ### obsolescence marker flag | |
85 |
|
85 | |||
86 | ## bumpedfix flag |
|
86 | ## bumpedfix flag | |
87 | # |
|
87 | # | |
88 | # When a changeset A' succeed to a changeset A which became public, we call A' |
|
88 | # When a changeset A' succeed to a changeset A which became public, we call A' | |
89 | # "bumped" because it's a successors of a public changesets |
|
89 | # "bumped" because it's a successors of a public changesets | |
90 | # |
|
90 | # | |
91 | # o A' (bumped) |
|
91 | # o A' (bumped) | |
92 | # |`: |
|
92 | # |`: | |
93 | # | o A |
|
93 | # | o A | |
94 | # |/ |
|
94 | # |/ | |
95 | # o Z |
|
95 | # o Z | |
96 | # |
|
96 | # | |
97 | # The way to solve this situation is to create a new changeset Ad as children |
|
97 | # The way to solve this situation is to create a new changeset Ad as children | |
98 | # of A. This changeset have the same content than A'. So the diff from A to A' |
|
98 | # of A. This changeset have the same content than A'. So the diff from A to A' | |
99 | # is the same than the diff from A to Ad. Ad is marked as a successors of A' |
|
99 | # is the same than the diff from A to Ad. Ad is marked as a successors of A' | |
100 | # |
|
100 | # | |
101 | # o Ad |
|
101 | # o Ad | |
102 | # |`: |
|
102 | # |`: | |
103 | # | x A' |
|
103 | # | x A' | |
104 | # |'| |
|
104 | # |'| | |
105 | # o | A |
|
105 | # o | A | |
106 | # |/ |
|
106 | # |/ | |
107 | # o Z |
|
107 | # o Z | |
108 | # |
|
108 | # | |
109 | # But by transitivity Ad is also a successors of A. To avoid having Ad marked |
|
109 | # But by transitivity Ad is also a successors of A. To avoid having Ad marked | |
110 | # as bumped too, we add the `bumpedfix` flag to the marker. <A', (Ad,)>. |
|
110 | # as bumped too, we add the `bumpedfix` flag to the marker. <A', (Ad,)>. | |
111 | # This flag mean that the successors express the changes between the public and |
|
111 | # This flag mean that the successors express the changes between the public and | |
112 | # bumped version and fix the situation, breaking the transitivity of |
|
112 | # bumped version and fix the situation, breaking the transitivity of | |
113 | # "bumped" here. |
|
113 | # "bumped" here. | |
114 | bumpedfix = 1 |
|
114 | bumpedfix = 1 | |
115 | usingsha256 = 2 |
|
115 | usingsha256 = 2 | |
116 |
|
116 | |||
117 | ## Parsing and writing of version "0" |
|
117 | ## Parsing and writing of version "0" | |
118 | # |
|
118 | # | |
119 | # The header is followed by the markers. Each marker is made of: |
|
119 | # The header is followed by the markers. Each marker is made of: | |
120 | # |
|
120 | # | |
121 | # - 1 uint8 : number of new changesets "N", can be zero. |
|
121 | # - 1 uint8 : number of new changesets "N", can be zero. | |
122 | # |
|
122 | # | |
123 | # - 1 uint32: metadata size "M" in bytes. |
|
123 | # - 1 uint32: metadata size "M" in bytes. | |
124 | # |
|
124 | # | |
125 | # - 1 byte: a bit field. It is reserved for flags used in common |
|
125 | # - 1 byte: a bit field. It is reserved for flags used in common | |
126 | # obsolete marker operations, to avoid repeated decoding of metadata |
|
126 | # obsolete marker operations, to avoid repeated decoding of metadata | |
127 | # entries. |
|
127 | # entries. | |
128 | # |
|
128 | # | |
129 | # - 20 bytes: obsoleted changeset identifier. |
|
129 | # - 20 bytes: obsoleted changeset identifier. | |
130 | # |
|
130 | # | |
131 | # - N*20 bytes: new changesets identifiers. |
|
131 | # - N*20 bytes: new changesets identifiers. | |
132 | # |
|
132 | # | |
133 | # - M bytes: metadata as a sequence of nul-terminated strings. Each |
|
133 | # - M bytes: metadata as a sequence of nul-terminated strings. Each | |
134 | # string contains a key and a value, separated by a colon ':', without |
|
134 | # string contains a key and a value, separated by a colon ':', without | |
135 | # additional encoding. Keys cannot contain '\0' or ':' and values |
|
135 | # additional encoding. Keys cannot contain '\0' or ':' and values | |
136 | # cannot contain '\0'. |
|
136 | # cannot contain '\0'. | |
137 | _fm0version = 0 |
|
137 | _fm0version = 0 | |
138 | _fm0fixed = '>BIB20s' |
|
138 | _fm0fixed = '>BIB20s' | |
139 | _fm0node = '20s' |
|
139 | _fm0node = '20s' | |
140 | _fm0fsize = struct.calcsize(_fm0fixed) |
|
140 | _fm0fsize = struct.calcsize(_fm0fixed) | |
141 | _fm0fnodesize = struct.calcsize(_fm0node) |
|
141 | _fm0fnodesize = struct.calcsize(_fm0node) | |
142 |
|
142 | |||
143 | def _fm0readmarkers(data, off=0): |
|
143 | def _fm0readmarkers(data, off=0): | |
144 | # Loop on markers |
|
144 | # Loop on markers | |
145 | l = len(data) |
|
145 | l = len(data) | |
146 | while off + _fm0fsize <= l: |
|
146 | while off + _fm0fsize <= l: | |
147 | # read fixed part |
|
147 | # read fixed part | |
148 | cur = data[off:off + _fm0fsize] |
|
148 | cur = data[off:off + _fm0fsize] | |
149 | off += _fm0fsize |
|
149 | off += _fm0fsize | |
150 | numsuc, mdsize, flags, pre = _unpack(_fm0fixed, cur) |
|
150 | numsuc, mdsize, flags, pre = _unpack(_fm0fixed, cur) | |
151 | # read replacement |
|
151 | # read replacement | |
152 | sucs = () |
|
152 | sucs = () | |
153 | if numsuc: |
|
153 | if numsuc: | |
154 | s = (_fm0fnodesize * numsuc) |
|
154 | s = (_fm0fnodesize * numsuc) | |
155 | cur = data[off:off + s] |
|
155 | cur = data[off:off + s] | |
156 | sucs = _unpack(_fm0node * numsuc, cur) |
|
156 | sucs = _unpack(_fm0node * numsuc, cur) | |
157 | off += s |
|
157 | off += s | |
158 | # read metadata |
|
158 | # read metadata | |
159 | # (metadata will be decoded on demand) |
|
159 | # (metadata will be decoded on demand) | |
160 | metadata = data[off:off + mdsize] |
|
160 | metadata = data[off:off + mdsize] | |
161 | if len(metadata) != mdsize: |
|
161 | if len(metadata) != mdsize: | |
162 | raise util.Abort(_('parsing obsolete marker: metadata is too ' |
|
162 | raise util.Abort(_('parsing obsolete marker: metadata is too ' | |
163 | 'short, %d bytes expected, got %d') |
|
163 | 'short, %d bytes expected, got %d') | |
164 | % (mdsize, len(metadata))) |
|
164 | % (mdsize, len(metadata))) | |
165 | off += mdsize |
|
165 | off += mdsize | |
166 | metadata = _fm0decodemeta(metadata) |
|
166 | metadata = _fm0decodemeta(metadata) | |
167 | try: |
|
167 | try: | |
168 | when, offset = metadata.pop('date', '0 0').split(' ') |
|
168 | when, offset = metadata.pop('date', '0 0').split(' ') | |
169 | date = float(when), int(offset) |
|
169 | date = float(when), int(offset) | |
170 | except ValueError: |
|
170 | except ValueError: | |
171 | date = (0., 0) |
|
171 | date = (0., 0) | |
172 | parents = None |
|
172 | parents = None | |
173 | if 'p2' in metadata: |
|
173 | if 'p2' in metadata: | |
174 | parents = (metadata.pop('p1', None), metadata.pop('p2', None)) |
|
174 | parents = (metadata.pop('p1', None), metadata.pop('p2', None)) | |
175 | elif 'p1' in metadata: |
|
175 | elif 'p1' in metadata: | |
176 | parents = (metadata.pop('p1', None),) |
|
176 | parents = (metadata.pop('p1', None),) | |
177 | elif 'p0' in metadata: |
|
177 | elif 'p0' in metadata: | |
178 | parents = () |
|
178 | parents = () | |
179 | if parents is not None: |
|
179 | if parents is not None: | |
180 | try: |
|
180 | try: | |
181 | parents = tuple(node.bin(p) for p in parents) |
|
181 | parents = tuple(node.bin(p) for p in parents) | |
182 | # if parent content is not a nodeid, drop the data |
|
182 | # if parent content is not a nodeid, drop the data | |
183 | for p in parents: |
|
183 | for p in parents: | |
184 | if len(p) != 20: |
|
184 | if len(p) != 20: | |
185 | parents = None |
|
185 | parents = None | |
186 | break |
|
186 | break | |
187 | except TypeError: |
|
187 | except TypeError: | |
188 | # if content cannot be translated to nodeid drop the data. |
|
188 | # if content cannot be translated to nodeid drop the data. | |
189 | parents = None |
|
189 | parents = None | |
190 |
|
190 | |||
191 | metadata = tuple(sorted(metadata.iteritems())) |
|
191 | metadata = tuple(sorted(metadata.iteritems())) | |
192 |
|
192 | |||
193 | yield (pre, sucs, flags, metadata, date, parents) |
|
193 | yield (pre, sucs, flags, metadata, date, parents) | |
194 |
|
194 | |||
195 | def _fm0encodeonemarker(marker): |
|
195 | def _fm0encodeonemarker(marker): | |
196 | pre, sucs, flags, metadata, date, parents = marker |
|
196 | pre, sucs, flags, metadata, date, parents = marker | |
197 | if flags & usingsha256: |
|
197 | if flags & usingsha256: | |
198 | raise util.Abort(_('cannot handle sha256 with old obsstore format')) |
|
198 | raise util.Abort(_('cannot handle sha256 with old obsstore format')) | |
199 | metadata = dict(metadata) |
|
199 | metadata = dict(metadata) | |
200 | metadata['date'] = '%d %i' % date |
|
200 | metadata['date'] = '%d %i' % date | |
201 | if parents is not None: |
|
201 | if parents is not None: | |
202 | if not parents: |
|
202 | if not parents: | |
203 | # mark that we explicitly recorded no parents |
|
203 | # mark that we explicitly recorded no parents | |
204 | metadata['p0'] = '' |
|
204 | metadata['p0'] = '' | |
205 | for i, p in enumerate(parents): |
|
205 | for i, p in enumerate(parents): | |
206 | metadata['p%i' % (i + 1)] = node.hex(p) |
|
206 | metadata['p%i' % (i + 1)] = node.hex(p) | |
207 | metadata = _fm0encodemeta(metadata) |
|
207 | metadata = _fm0encodemeta(metadata) | |
208 | numsuc = len(sucs) |
|
208 | numsuc = len(sucs) | |
209 | format = _fm0fixed + (_fm0node * numsuc) |
|
209 | format = _fm0fixed + (_fm0node * numsuc) | |
210 | data = [numsuc, len(metadata), flags, pre] |
|
210 | data = [numsuc, len(metadata), flags, pre] | |
211 | data.extend(sucs) |
|
211 | data.extend(sucs) | |
212 | return _pack(format, *data) + metadata |
|
212 | return _pack(format, *data) + metadata | |
213 |
|
213 | |||
214 | def _fm0encodemeta(meta): |
|
214 | def _fm0encodemeta(meta): | |
215 | """Return encoded metadata string to string mapping. |
|
215 | """Return encoded metadata string to string mapping. | |
216 |
|
216 | |||
217 | Assume no ':' in key and no '\0' in both key and value.""" |
|
217 | Assume no ':' in key and no '\0' in both key and value.""" | |
218 | for key, value in meta.iteritems(): |
|
218 | for key, value in meta.iteritems(): | |
219 | if ':' in key or '\0' in key: |
|
219 | if ':' in key or '\0' in key: | |
220 | raise ValueError("':' and '\0' are forbidden in metadata key'") |
|
220 | raise ValueError("':' and '\0' are forbidden in metadata key'") | |
221 | if '\0' in value: |
|
221 | if '\0' in value: | |
222 | raise ValueError("':' is forbidden in metadata value'") |
|
222 | raise ValueError("':' is forbidden in metadata value'") | |
223 | return '\0'.join(['%s:%s' % (k, meta[k]) for k in sorted(meta)]) |
|
223 | return '\0'.join(['%s:%s' % (k, meta[k]) for k in sorted(meta)]) | |
224 |
|
224 | |||
225 | def _fm0decodemeta(data): |
|
225 | def _fm0decodemeta(data): | |
226 | """Return string to string dictionary from encoded version.""" |
|
226 | """Return string to string dictionary from encoded version.""" | |
227 | d = {} |
|
227 | d = {} | |
228 | for l in data.split('\0'): |
|
228 | for l in data.split('\0'): | |
229 | if l: |
|
229 | if l: | |
230 | key, value = l.split(':') |
|
230 | key, value = l.split(':') | |
231 | d[key] = value |
|
231 | d[key] = value | |
232 | return d |
|
232 | return d | |
233 |
|
233 | |||
234 | ## Parsing and writing of version "1" |
|
234 | ## Parsing and writing of version "1" | |
235 | # |
|
235 | # | |
236 | # The header is followed by the markers. Each marker is made of: |
|
236 | # The header is followed by the markers. Each marker is made of: | |
237 | # |
|
237 | # | |
238 | # - uint32: total size of the marker (including this field) |
|
238 | # - uint32: total size of the marker (including this field) | |
239 | # |
|
239 | # | |
240 | # - float64: date in seconds since epoch |
|
240 | # - float64: date in seconds since epoch | |
241 | # |
|
241 | # | |
242 | # - int16: timezone offset in minutes |
|
242 | # - int16: timezone offset in minutes | |
243 | # |
|
243 | # | |
244 | # - uint16: a bit field. It is reserved for flags used in common |
|
244 | # - uint16: a bit field. It is reserved for flags used in common | |
245 | # obsolete marker operations, to avoid repeated decoding of metadata |
|
245 | # obsolete marker operations, to avoid repeated decoding of metadata | |
246 | # entries. |
|
246 | # entries. | |
247 | # |
|
247 | # | |
248 | # - uint8: number of successors "N", can be zero. |
|
248 | # - uint8: number of successors "N", can be zero. | |
249 | # |
|
249 | # | |
250 | # - uint8: number of parents "P", can be zero. |
|
250 | # - uint8: number of parents "P", can be zero. | |
251 | # |
|
251 | # | |
252 | # 0: parents data stored but no parent, |
|
252 | # 0: parents data stored but no parent, | |
253 | # 1: one parent stored, |
|
253 | # 1: one parent stored, | |
254 | # 2: two parents stored, |
|
254 | # 2: two parents stored, | |
255 | # 3: no parent data stored |
|
255 | # 3: no parent data stored | |
256 | # |
|
256 | # | |
257 | # - uint8: number of metadata entries M |
|
257 | # - uint8: number of metadata entries M | |
258 | # |
|
258 | # | |
259 | # - 20 or 32 bytes: precursor changeset identifier. |
|
259 | # - 20 or 32 bytes: precursor changeset identifier. | |
260 | # |
|
260 | # | |
261 | # - N*(20 or 32) bytes: successors changesets identifiers. |
|
261 | # - N*(20 or 32) bytes: successors changesets identifiers. | |
262 | # |
|
262 | # | |
263 | # - P*(20 or 32) bytes: parents of the precursors changesets. |
|
263 | # - P*(20 or 32) bytes: parents of the precursors changesets. | |
264 | # |
|
264 | # | |
265 | # - M*(uint8, uint8): size of all metadata entries (key and value) |
|
265 | # - M*(uint8, uint8): size of all metadata entries (key and value) | |
266 | # |
|
266 | # | |
267 | # - remaining bytes: the metadata, each (key, value) pair after the other. |
|
267 | # - remaining bytes: the metadata, each (key, value) pair after the other. | |
268 | _fm1version = 1 |
|
268 | _fm1version = 1 | |
269 | _fm1fixed = '>IdhHBBB20s' |
|
269 | _fm1fixed = '>IdhHBBB20s' | |
270 | _fm1nodesha1 = '20s' |
|
270 | _fm1nodesha1 = '20s' | |
271 | _fm1nodesha256 = '32s' |
|
271 | _fm1nodesha256 = '32s' | |
272 | _fm1fsize = struct.calcsize(_fm1fixed) |
|
272 | _fm1fsize = struct.calcsize(_fm1fixed) | |
273 | _fm1parentnone = 3 |
|
273 | _fm1parentnone = 3 | |
274 | _fm1parentshift = 14 |
|
274 | _fm1parentshift = 14 | |
275 | _fm1parentmask = (_fm1parentnone << _fm1parentshift) |
|
275 | _fm1parentmask = (_fm1parentnone << _fm1parentshift) | |
276 | _fm1metapair = 'BB' |
|
276 | _fm1metapair = 'BB' | |
277 | _fm1metapairsize = struct.calcsize('BB') |
|
277 | _fm1metapairsize = struct.calcsize('BB') | |
278 |
|
278 | |||
279 | def _fm1readmarkers(data, off=0): |
|
279 | def _fm1readmarkers(data, off=0): | |
280 | # Loop on markers |
|
280 | # Loop on markers | |
281 | l = len(data) |
|
281 | l = len(data) | |
282 | while off + _fm1fsize <= l: |
|
282 | while off + _fm1fsize <= l: | |
283 | # read fixed part |
|
283 | # read fixed part | |
284 | cur = data[off:off + _fm1fsize] |
|
284 | cur = data[off:off + _fm1fsize] | |
285 | off += _fm1fsize |
|
285 | off += _fm1fsize | |
286 | fixeddata = _unpack(_fm1fixed, cur) |
|
286 | fixeddata = _unpack(_fm1fixed, cur) | |
287 | ttsize, seconds, tz, flags, numsuc, numpar, nummeta, prec = fixeddata |
|
287 | ttsize, seconds, tz, flags, numsuc, numpar, nummeta, prec = fixeddata | |
288 | # extract the number of parents information |
|
288 | # extract the number of parents information | |
289 | if numpar == _fm1parentnone: |
|
289 | if numpar == _fm1parentnone: | |
290 | numpar = None |
|
290 | numpar = None | |
291 | # build the date tuple (upgrade tz minutes to seconds) |
|
291 | # build the date tuple (upgrade tz minutes to seconds) | |
292 | date = (seconds, tz * 60) |
|
292 | date = (seconds, tz * 60) | |
293 | _fm1node = _fm1nodesha1 |
|
293 | _fm1node = _fm1nodesha1 | |
294 | if flags & usingsha256: |
|
294 | if flags & usingsha256: | |
295 | _fm1node = _fm1nodesha256 |
|
295 | _fm1node = _fm1nodesha256 | |
296 | fnodesize = struct.calcsize(_fm1node) |
|
296 | fnodesize = struct.calcsize(_fm1node) | |
297 | # read replacement |
|
297 | # read replacement | |
298 | sucs = () |
|
298 | sucs = () | |
299 | if numsuc: |
|
299 | if numsuc: | |
300 | s = (fnodesize * numsuc) |
|
300 | s = (fnodesize * numsuc) | |
301 | cur = data[off:off + s] |
|
301 | cur = data[off:off + s] | |
302 | sucs = _unpack(_fm1node * numsuc, cur) |
|
302 | sucs = _unpack(_fm1node * numsuc, cur) | |
303 | off += s |
|
303 | off += s | |
304 | # read parents |
|
304 | # read parents | |
305 | if numpar is None: |
|
305 | if numpar is None: | |
306 | parents = None |
|
306 | parents = None | |
307 | elif numpar == 0: |
|
307 | elif numpar == 0: | |
308 | parents = () |
|
308 | parents = () | |
309 | elif numpar: # neither None nor zero |
|
309 | elif numpar: # neither None nor zero | |
310 | s = (fnodesize * numpar) |
|
310 | s = (fnodesize * numpar) | |
311 | cur = data[off:off + s] |
|
311 | cur = data[off:off + s] | |
312 | parents = _unpack(_fm1node * numpar, cur) |
|
312 | parents = _unpack(_fm1node * numpar, cur) | |
313 | off += s |
|
313 | off += s | |
314 | # read metadata |
|
314 | # read metadata | |
315 | metaformat = '>' + (_fm1metapair * nummeta) |
|
315 | metaformat = '>' + (_fm1metapair * nummeta) | |
316 | s = _fm1metapairsize * nummeta |
|
316 | s = _fm1metapairsize * nummeta | |
317 | metapairsize = _unpack(metaformat, data[off:off + s]) |
|
317 | metapairsize = _unpack(metaformat, data[off:off + s]) | |
318 | off += s |
|
318 | off += s | |
319 | metadata = [] |
|
319 | metadata = [] | |
320 | for idx in xrange(0, len(metapairsize), 2): |
|
320 | for idx in xrange(0, len(metapairsize), 2): | |
321 | sk = metapairsize[idx] |
|
321 | sk = metapairsize[idx] | |
322 | sv = metapairsize[idx + 1] |
|
322 | sv = metapairsize[idx + 1] | |
323 | key = data[off:off + sk] |
|
323 | key = data[off:off + sk] | |
324 | value = data[off + sk:off + sk + sv] |
|
324 | value = data[off + sk:off + sk + sv] | |
325 | assert len(key) == sk |
|
325 | assert len(key) == sk | |
326 | assert len(value) == sv |
|
326 | assert len(value) == sv | |
327 | metadata.append((key, value)) |
|
327 | metadata.append((key, value)) | |
328 | off += sk + sv |
|
328 | off += sk + sv | |
329 | metadata = tuple(metadata) |
|
329 | metadata = tuple(metadata) | |
330 |
|
330 | |||
331 | yield (prec, sucs, flags, metadata, date, parents) |
|
331 | yield (prec, sucs, flags, metadata, date, parents) | |
332 |
|
332 | |||
333 | def _fm1encodeonemarker(marker): |
|
333 | def _fm1encodeonemarker(marker): | |
334 | pre, sucs, flags, metadata, date, parents = marker |
|
334 | pre, sucs, flags, metadata, date, parents = marker | |
335 | # determine node size |
|
335 | # determine node size | |
336 | _fm1node = _fm1nodesha1 |
|
336 | _fm1node = _fm1nodesha1 | |
337 | if flags & usingsha256: |
|
337 | if flags & usingsha256: | |
338 | _fm1node = _fm1nodesha256 |
|
338 | _fm1node = _fm1nodesha256 | |
339 | numsuc = len(sucs) |
|
339 | numsuc = len(sucs) | |
340 | numextranodes = numsuc |
|
340 | numextranodes = numsuc | |
341 | if parents is None: |
|
341 | if parents is None: | |
342 | numpar = _fm1parentnone |
|
342 | numpar = _fm1parentnone | |
343 | else: |
|
343 | else: | |
344 | numpar = len(parents) |
|
344 | numpar = len(parents) | |
345 | numextranodes += numpar |
|
345 | numextranodes += numpar | |
346 | formatnodes = _fm1node * numextranodes |
|
346 | formatnodes = _fm1node * numextranodes | |
347 | formatmeta = _fm1metapair * len(metadata) |
|
347 | formatmeta = _fm1metapair * len(metadata) | |
348 | format = _fm1fixed + formatnodes + formatmeta |
|
348 | format = _fm1fixed + formatnodes + formatmeta | |
349 | # tz is stored in minutes so we divide by 60 |
|
349 | # tz is stored in minutes so we divide by 60 | |
350 | tz = date[1]//60 |
|
350 | tz = date[1]//60 | |
351 | data = [None, date[0], tz, flags, numsuc, numpar, len(metadata), pre] |
|
351 | data = [None, date[0], tz, flags, numsuc, numpar, len(metadata), pre] | |
352 | data.extend(sucs) |
|
352 | data.extend(sucs) | |
353 | if parents is not None: |
|
353 | if parents is not None: | |
354 | data.extend(parents) |
|
354 | data.extend(parents) | |
355 | totalsize = struct.calcsize(format) |
|
355 | totalsize = struct.calcsize(format) | |
356 | for key, value in metadata: |
|
356 | for key, value in metadata: | |
357 | lk = len(key) |
|
357 | lk = len(key) | |
358 | lv = len(value) |
|
358 | lv = len(value) | |
359 | data.append(lk) |
|
359 | data.append(lk) | |
360 | data.append(lv) |
|
360 | data.append(lv) | |
361 | totalsize += lk + lv |
|
361 | totalsize += lk + lv | |
362 | data[0] = totalsize |
|
362 | data[0] = totalsize | |
363 | data = [_pack(format, *data)] |
|
363 | data = [_pack(format, *data)] | |
364 | for key, value in metadata: |
|
364 | for key, value in metadata: | |
365 | data.append(key) |
|
365 | data.append(key) | |
366 | data.append(value) |
|
366 | data.append(value) | |
367 | return ''.join(data) |
|
367 | return ''.join(data) | |
368 |
|
368 | |||
369 | # mapping to read/write various marker formats |
|
369 | # mapping to read/write various marker formats | |
370 | # <version> -> (decoder, encoder) |
|
370 | # <version> -> (decoder, encoder) | |
371 | formats = {_fm0version: (_fm0readmarkers, _fm0encodeonemarker), |
|
371 | formats = {_fm0version: (_fm0readmarkers, _fm0encodeonemarker), | |
372 | _fm1version: (_fm1readmarkers, _fm1encodeonemarker)} |
|
372 | _fm1version: (_fm1readmarkers, _fm1encodeonemarker)} | |
373 |
|
373 | |||
374 | def _readmarkers(data): |
|
374 | def _readmarkers(data): | |
375 | """Read and enumerate markers from raw data""" |
|
375 | """Read and enumerate markers from raw data""" | |
376 | off = 0 |
|
376 | off = 0 | |
377 | diskversion = _unpack('>B', data[off:off + 1])[0] |
|
377 | diskversion = _unpack('>B', data[off:off + 1])[0] | |
378 | off += 1 |
|
378 | off += 1 | |
379 | if diskversion not in formats: |
|
379 | if diskversion not in formats: | |
380 | raise util.Abort(_('parsing obsolete marker: unknown version %r') |
|
380 | raise util.Abort(_('parsing obsolete marker: unknown version %r') | |
381 | % diskversion) |
|
381 | % diskversion) | |
382 | return diskversion, formats[diskversion][0](data, off) |
|
382 | return diskversion, formats[diskversion][0](data, off) | |
383 |
|
383 | |||
384 | def encodemarkers(markers, addheader=False, version=_fm0version): |
|
384 | def encodemarkers(markers, addheader=False, version=_fm0version): | |
385 | # Kept separate from flushmarkers(), it will be reused for |
|
385 | # Kept separate from flushmarkers(), it will be reused for | |
386 | # markers exchange. |
|
386 | # markers exchange. | |
387 | encodeone = formats[version][1] |
|
387 | encodeone = formats[version][1] | |
388 | if addheader: |
|
388 | if addheader: | |
389 | yield _pack('>B', version) |
|
389 | yield _pack('>B', version) | |
390 | for marker in markers: |
|
390 | for marker in markers: | |
391 | yield encodeone(marker) |
|
391 | yield encodeone(marker) | |
392 |
|
392 | |||
393 |
|
393 | |||
394 | class marker(object): |
|
394 | class marker(object): | |
395 | """Wrap obsolete marker raw data""" |
|
395 | """Wrap obsolete marker raw data""" | |
396 |
|
396 | |||
397 | def __init__(self, repo, data): |
|
397 | def __init__(self, repo, data): | |
398 | # the repo argument will be used to create changectx in later version |
|
398 | # the repo argument will be used to create changectx in later version | |
399 | self._repo = repo |
|
399 | self._repo = repo | |
400 | self._data = data |
|
400 | self._data = data | |
401 | self._decodedmeta = None |
|
401 | self._decodedmeta = None | |
402 |
|
402 | |||
403 | def __hash__(self): |
|
403 | def __hash__(self): | |
404 | return hash(self._data) |
|
404 | return hash(self._data) | |
405 |
|
405 | |||
406 | def __eq__(self, other): |
|
406 | def __eq__(self, other): | |
407 | if type(other) != type(self): |
|
407 | if type(other) != type(self): | |
408 | return False |
|
408 | return False | |
409 | return self._data == other._data |
|
409 | return self._data == other._data | |
410 |
|
410 | |||
411 | def precnode(self): |
|
411 | def precnode(self): | |
412 | """Precursor changeset node identifier""" |
|
412 | """Precursor changeset node identifier""" | |
413 | return self._data[0] |
|
413 | return self._data[0] | |
414 |
|
414 | |||
415 | def succnodes(self): |
|
415 | def succnodes(self): | |
416 | """List of successor changesets node identifiers""" |
|
416 | """List of successor changesets node identifiers""" | |
417 | return self._data[1] |
|
417 | return self._data[1] | |
418 |
|
418 | |||
419 | def parentnodes(self): |
|
419 | def parentnodes(self): | |
420 | """Parents of the precursors (None if not recorded)""" |
|
420 | """Parents of the precursors (None if not recorded)""" | |
421 | return self._data[5] |
|
421 | return self._data[5] | |
422 |
|
422 | |||
423 | def metadata(self): |
|
423 | def metadata(self): | |
424 | """Decoded metadata dictionary""" |
|
424 | """Decoded metadata dictionary""" | |
425 | return dict(self._data[3]) |
|
425 | return dict(self._data[3]) | |
426 |
|
426 | |||
427 | def date(self): |
|
427 | def date(self): | |
428 | """Creation date as (unixtime, offset)""" |
|
428 | """Creation date as (unixtime, offset)""" | |
429 | return self._data[4] |
|
429 | return self._data[4] | |
430 |
|
430 | |||
431 | def flags(self): |
|
431 | def flags(self): | |
432 | """The flags field of the marker""" |
|
432 | """The flags field of the marker""" | |
433 | return self._data[2] |
|
433 | return self._data[2] | |
434 |
|
434 | |||
435 | class obsstore(object): |
|
435 | class obsstore(object): | |
436 | """Store obsolete markers |
|
436 | """Store obsolete markers | |
437 |
|
437 | |||
438 | Markers can be accessed with two mappings: |
|
438 | Markers can be accessed with two mappings: | |
439 | - precursors[x] -> set(markers on precursors edges of x) |
|
439 | - precursors[x] -> set(markers on precursors edges of x) | |
440 | - successors[x] -> set(markers on successors edges of x) |
|
440 | - successors[x] -> set(markers on successors edges of x) | |
441 | - children[x] -> set(markers on precursors edges of children(x) |
|
441 | - children[x] -> set(markers on precursors edges of children(x) | |
442 | """ |
|
442 | """ | |
443 |
|
443 | |||
444 | fields = ('prec', 'succs', 'flag', 'meta', 'date', 'parents') |
|
444 | fields = ('prec', 'succs', 'flag', 'meta', 'date', 'parents') | |
445 | # prec: nodeid, precursor changesets |
|
445 | # prec: nodeid, precursor changesets | |
446 | # succs: tuple of nodeid, successor changesets (0-N length) |
|
446 | # succs: tuple of nodeid, successor changesets (0-N length) | |
447 | # flag: integer, flag field carrying modifier for the markers (see doc) |
|
447 | # flag: integer, flag field carrying modifier for the markers (see doc) | |
448 | # meta: binary blob, encoded metadata dictionary |
|
448 | # meta: binary blob, encoded metadata dictionary | |
449 | # date: (float, int) tuple, date of marker creation |
|
449 | # date: (float, int) tuple, date of marker creation | |
450 | # parents: (tuple of nodeid) or None, parents of precursors |
|
450 | # parents: (tuple of nodeid) or None, parents of precursors | |
451 | # None is used when no data has been recorded |
|
451 | # None is used when no data has been recorded | |
452 |
|
452 | |||
453 |
def __init__(self, sopener, defaultformat=_fm |
|
453 | def __init__(self, sopener, defaultformat=_fm1version): | |
454 | # caches for various obsolescence related cache |
|
454 | # caches for various obsolescence related cache | |
455 | self.caches = {} |
|
455 | self.caches = {} | |
456 | self._all = [] |
|
456 | self._all = [] | |
457 | self.precursors = {} |
|
457 | self.precursors = {} | |
458 | self.successors = {} |
|
458 | self.successors = {} | |
459 | self.children = {} |
|
459 | self.children = {} | |
460 | self.sopener = sopener |
|
460 | self.sopener = sopener | |
461 | data = sopener.tryread('obsstore') |
|
461 | data = sopener.tryread('obsstore') | |
462 | self._version = defaultformat |
|
462 | self._version = defaultformat | |
463 | if data: |
|
463 | if data: | |
464 | self._version, markers = _readmarkers(data) |
|
464 | self._version, markers = _readmarkers(data) | |
465 | self._load(markers) |
|
465 | self._load(markers) | |
466 |
|
466 | |||
467 | def __iter__(self): |
|
467 | def __iter__(self): | |
468 | return iter(self._all) |
|
468 | return iter(self._all) | |
469 |
|
469 | |||
470 | def __len__(self): |
|
470 | def __len__(self): | |
471 | return len(self._all) |
|
471 | return len(self._all) | |
472 |
|
472 | |||
473 | def __nonzero__(self): |
|
473 | def __nonzero__(self): | |
474 | return bool(self._all) |
|
474 | return bool(self._all) | |
475 |
|
475 | |||
476 | def create(self, transaction, prec, succs=(), flag=0, parents=None, |
|
476 | def create(self, transaction, prec, succs=(), flag=0, parents=None, | |
477 | date=None, metadata=None): |
|
477 | date=None, metadata=None): | |
478 | """obsolete: add a new obsolete marker |
|
478 | """obsolete: add a new obsolete marker | |
479 |
|
479 | |||
480 | * ensuring it is hashable |
|
480 | * ensuring it is hashable | |
481 | * check mandatory metadata |
|
481 | * check mandatory metadata | |
482 | * encode metadata |
|
482 | * encode metadata | |
483 |
|
483 | |||
484 | If you are a human writing code creating marker you want to use the |
|
484 | If you are a human writing code creating marker you want to use the | |
485 | `createmarkers` function in this module instead. |
|
485 | `createmarkers` function in this module instead. | |
486 |
|
486 | |||
487 | return True if a new marker have been added, False if the markers |
|
487 | return True if a new marker have been added, False if the markers | |
488 | already existed (no op). |
|
488 | already existed (no op). | |
489 | """ |
|
489 | """ | |
490 | if metadata is None: |
|
490 | if metadata is None: | |
491 | metadata = {} |
|
491 | metadata = {} | |
492 | if date is None: |
|
492 | if date is None: | |
493 | if 'date' in metadata: |
|
493 | if 'date' in metadata: | |
494 | # as a courtesy for out-of-tree extensions |
|
494 | # as a courtesy for out-of-tree extensions | |
495 | date = util.parsedate(metadata.pop('date')) |
|
495 | date = util.parsedate(metadata.pop('date')) | |
496 | else: |
|
496 | else: | |
497 | date = util.makedate() |
|
497 | date = util.makedate() | |
498 | if len(prec) != 20: |
|
498 | if len(prec) != 20: | |
499 | raise ValueError(prec) |
|
499 | raise ValueError(prec) | |
500 | for succ in succs: |
|
500 | for succ in succs: | |
501 | if len(succ) != 20: |
|
501 | if len(succ) != 20: | |
502 | raise ValueError(succ) |
|
502 | raise ValueError(succ) | |
503 | if prec in succs: |
|
503 | if prec in succs: | |
504 | raise ValueError(_('in-marker cycle with %s') % node.hex(prec)) |
|
504 | raise ValueError(_('in-marker cycle with %s') % node.hex(prec)) | |
505 |
|
505 | |||
506 | metadata = tuple(sorted(metadata.iteritems())) |
|
506 | metadata = tuple(sorted(metadata.iteritems())) | |
507 |
|
507 | |||
508 | marker = (str(prec), tuple(succs), int(flag), metadata, date, parents) |
|
508 | marker = (str(prec), tuple(succs), int(flag), metadata, date, parents) | |
509 | return bool(self.add(transaction, [marker])) |
|
509 | return bool(self.add(transaction, [marker])) | |
510 |
|
510 | |||
511 | def add(self, transaction, markers): |
|
511 | def add(self, transaction, markers): | |
512 | """Add new markers to the store |
|
512 | """Add new markers to the store | |
513 |
|
513 | |||
514 | Take care of filtering duplicate. |
|
514 | Take care of filtering duplicate. | |
515 | Return the number of new marker.""" |
|
515 | Return the number of new marker.""" | |
516 | if not _enabled: |
|
516 | if not _enabled: | |
517 | raise util.Abort('obsolete feature is not enabled on this repo') |
|
517 | raise util.Abort('obsolete feature is not enabled on this repo') | |
518 | known = set(self._all) |
|
518 | known = set(self._all) | |
519 | new = [] |
|
519 | new = [] | |
520 | for m in markers: |
|
520 | for m in markers: | |
521 | if m not in known: |
|
521 | if m not in known: | |
522 | known.add(m) |
|
522 | known.add(m) | |
523 | new.append(m) |
|
523 | new.append(m) | |
524 | if new: |
|
524 | if new: | |
525 | f = self.sopener('obsstore', 'ab') |
|
525 | f = self.sopener('obsstore', 'ab') | |
526 | try: |
|
526 | try: | |
527 | # Whether the file's current position is at the begin or at |
|
527 | # Whether the file's current position is at the begin or at | |
528 | # the end after opening a file for appending is implementation |
|
528 | # the end after opening a file for appending is implementation | |
529 | # defined. So we must seek to the end before calling tell(), |
|
529 | # defined. So we must seek to the end before calling tell(), | |
530 | # or we may get a zero offset for non-zero sized files on |
|
530 | # or we may get a zero offset for non-zero sized files on | |
531 | # some platforms (issue3543). |
|
531 | # some platforms (issue3543). | |
532 | f.seek(0, _SEEK_END) |
|
532 | f.seek(0, _SEEK_END) | |
533 | offset = f.tell() |
|
533 | offset = f.tell() | |
534 | transaction.add('obsstore', offset) |
|
534 | transaction.add('obsstore', offset) | |
535 | # offset == 0: new file - add the version header |
|
535 | # offset == 0: new file - add the version header | |
536 | for bytes in encodemarkers(new, offset == 0, self._version): |
|
536 | for bytes in encodemarkers(new, offset == 0, self._version): | |
537 | f.write(bytes) |
|
537 | f.write(bytes) | |
538 | finally: |
|
538 | finally: | |
539 | # XXX: f.close() == filecache invalidation == obsstore rebuilt. |
|
539 | # XXX: f.close() == filecache invalidation == obsstore rebuilt. | |
540 | # call 'filecacheentry.refresh()' here |
|
540 | # call 'filecacheentry.refresh()' here | |
541 | f.close() |
|
541 | f.close() | |
542 | self._load(new) |
|
542 | self._load(new) | |
543 | # new marker *may* have changed several set. invalidate the cache. |
|
543 | # new marker *may* have changed several set. invalidate the cache. | |
544 | self.caches.clear() |
|
544 | self.caches.clear() | |
545 | # records the number of new markers for the transaction hooks |
|
545 | # records the number of new markers for the transaction hooks | |
546 | previous = int(transaction.hookargs.get('new_obsmarkers', '0')) |
|
546 | previous = int(transaction.hookargs.get('new_obsmarkers', '0')) | |
547 | transaction.hookargs['new_obsmarkers'] = str(previous + len(new)) |
|
547 | transaction.hookargs['new_obsmarkers'] = str(previous + len(new)) | |
548 | return len(new) |
|
548 | return len(new) | |
549 |
|
549 | |||
550 | def mergemarkers(self, transaction, data): |
|
550 | def mergemarkers(self, transaction, data): | |
551 | """merge a binary stream of markers inside the obsstore |
|
551 | """merge a binary stream of markers inside the obsstore | |
552 |
|
552 | |||
553 | Returns the number of new markers added.""" |
|
553 | Returns the number of new markers added.""" | |
554 | version, markers = _readmarkers(data) |
|
554 | version, markers = _readmarkers(data) | |
555 | return self.add(transaction, markers) |
|
555 | return self.add(transaction, markers) | |
556 |
|
556 | |||
557 | def _load(self, markers): |
|
557 | def _load(self, markers): | |
558 | for mark in markers: |
|
558 | for mark in markers: | |
559 | self._all.append(mark) |
|
559 | self._all.append(mark) | |
560 | pre, sucs = mark[:2] |
|
560 | pre, sucs = mark[:2] | |
561 | self.successors.setdefault(pre, set()).add(mark) |
|
561 | self.successors.setdefault(pre, set()).add(mark) | |
562 | for suc in sucs: |
|
562 | for suc in sucs: | |
563 | self.precursors.setdefault(suc, set()).add(mark) |
|
563 | self.precursors.setdefault(suc, set()).add(mark) | |
564 | parents = mark[5] |
|
564 | parents = mark[5] | |
565 | if parents is not None: |
|
565 | if parents is not None: | |
566 | for p in parents: |
|
566 | for p in parents: | |
567 | self.children.setdefault(p, set()).add(mark) |
|
567 | self.children.setdefault(p, set()).add(mark) | |
568 | if node.nullid in self.precursors: |
|
568 | if node.nullid in self.precursors: | |
569 | raise util.Abort(_('bad obsolescence marker detected: ' |
|
569 | raise util.Abort(_('bad obsolescence marker detected: ' | |
570 | 'invalid successors nullid')) |
|
570 | 'invalid successors nullid')) | |
571 | def relevantmarkers(self, nodes): |
|
571 | def relevantmarkers(self, nodes): | |
572 | """return a set of all obsolescence markers relevant to a set of nodes. |
|
572 | """return a set of all obsolescence markers relevant to a set of nodes. | |
573 |
|
573 | |||
574 | "relevant" to a set of nodes mean: |
|
574 | "relevant" to a set of nodes mean: | |
575 |
|
575 | |||
576 | - marker that use this changeset as successor |
|
576 | - marker that use this changeset as successor | |
577 | - prune marker of direct children on this changeset |
|
577 | - prune marker of direct children on this changeset | |
578 | - recursive application of the two rules on precursors of these markers |
|
578 | - recursive application of the two rules on precursors of these markers | |
579 |
|
579 | |||
580 | It is a set so you cannot rely on order.""" |
|
580 | It is a set so you cannot rely on order.""" | |
581 |
|
581 | |||
582 | pendingnodes = set(nodes) |
|
582 | pendingnodes = set(nodes) | |
583 | seenmarkers = set() |
|
583 | seenmarkers = set() | |
584 | seennodes = set(pendingnodes) |
|
584 | seennodes = set(pendingnodes) | |
585 | precursorsmarkers = self.precursors |
|
585 | precursorsmarkers = self.precursors | |
586 | children = self.children |
|
586 | children = self.children | |
587 | while pendingnodes: |
|
587 | while pendingnodes: | |
588 | direct = set() |
|
588 | direct = set() | |
589 | for current in pendingnodes: |
|
589 | for current in pendingnodes: | |
590 | direct.update(precursorsmarkers.get(current, ())) |
|
590 | direct.update(precursorsmarkers.get(current, ())) | |
591 | pruned = [m for m in children.get(current, ()) if not m[1]] |
|
591 | pruned = [m for m in children.get(current, ()) if not m[1]] | |
592 | direct.update(pruned) |
|
592 | direct.update(pruned) | |
593 | direct -= seenmarkers |
|
593 | direct -= seenmarkers | |
594 | pendingnodes = set([m[0] for m in direct]) |
|
594 | pendingnodes = set([m[0] for m in direct]) | |
595 | seenmarkers |= direct |
|
595 | seenmarkers |= direct | |
596 | pendingnodes -= seennodes |
|
596 | pendingnodes -= seennodes | |
597 | seennodes |= pendingnodes |
|
597 | seennodes |= pendingnodes | |
598 | return seenmarkers |
|
598 | return seenmarkers | |
599 |
|
599 | |||
600 | def commonversion(versions): |
|
600 | def commonversion(versions): | |
601 | """Return the newest version listed in both versions and our local formats. |
|
601 | """Return the newest version listed in both versions and our local formats. | |
602 |
|
602 | |||
603 | Returns None if no common version exists. |
|
603 | Returns None if no common version exists. | |
604 | """ |
|
604 | """ | |
605 | versions.sort(reverse=True) |
|
605 | versions.sort(reverse=True) | |
606 | # search for highest version known on both side |
|
606 | # search for highest version known on both side | |
607 | for v in versions: |
|
607 | for v in versions: | |
608 | if v in formats: |
|
608 | if v in formats: | |
609 | return v |
|
609 | return v | |
610 | return None |
|
610 | return None | |
611 |
|
611 | |||
612 | # arbitrary picked to fit into 8K limit from HTTP server |
|
612 | # arbitrary picked to fit into 8K limit from HTTP server | |
613 | # you have to take in account: |
|
613 | # you have to take in account: | |
614 | # - the version header |
|
614 | # - the version header | |
615 | # - the base85 encoding |
|
615 | # - the base85 encoding | |
616 | _maxpayload = 5300 |
|
616 | _maxpayload = 5300 | |
617 |
|
617 | |||
618 | def _pushkeyescape(markers): |
|
618 | def _pushkeyescape(markers): | |
619 | """encode markers into a dict suitable for pushkey exchange |
|
619 | """encode markers into a dict suitable for pushkey exchange | |
620 |
|
620 | |||
621 | - binary data is base85 encoded |
|
621 | - binary data is base85 encoded | |
622 | - split in chunks smaller than 5300 bytes""" |
|
622 | - split in chunks smaller than 5300 bytes""" | |
623 | keys = {} |
|
623 | keys = {} | |
624 | parts = [] |
|
624 | parts = [] | |
625 | currentlen = _maxpayload * 2 # ensure we create a new part |
|
625 | currentlen = _maxpayload * 2 # ensure we create a new part | |
626 | for marker in markers: |
|
626 | for marker in markers: | |
627 | nextdata = _fm0encodeonemarker(marker) |
|
627 | nextdata = _fm0encodeonemarker(marker) | |
628 | if (len(nextdata) + currentlen > _maxpayload): |
|
628 | if (len(nextdata) + currentlen > _maxpayload): | |
629 | currentpart = [] |
|
629 | currentpart = [] | |
630 | currentlen = 0 |
|
630 | currentlen = 0 | |
631 | parts.append(currentpart) |
|
631 | parts.append(currentpart) | |
632 | currentpart.append(nextdata) |
|
632 | currentpart.append(nextdata) | |
633 | currentlen += len(nextdata) |
|
633 | currentlen += len(nextdata) | |
634 | for idx, part in enumerate(reversed(parts)): |
|
634 | for idx, part in enumerate(reversed(parts)): | |
635 | data = ''.join([_pack('>B', _fm0version)] + part) |
|
635 | data = ''.join([_pack('>B', _fm0version)] + part) | |
636 | keys['dump%i' % idx] = base85.b85encode(data) |
|
636 | keys['dump%i' % idx] = base85.b85encode(data) | |
637 | return keys |
|
637 | return keys | |
638 |
|
638 | |||
639 | def listmarkers(repo): |
|
639 | def listmarkers(repo): | |
640 | """List markers over pushkey""" |
|
640 | """List markers over pushkey""" | |
641 | if not repo.obsstore: |
|
641 | if not repo.obsstore: | |
642 | return {} |
|
642 | return {} | |
643 | return _pushkeyescape(repo.obsstore) |
|
643 | return _pushkeyescape(repo.obsstore) | |
644 |
|
644 | |||
645 | def pushmarker(repo, key, old, new): |
|
645 | def pushmarker(repo, key, old, new): | |
646 | """Push markers over pushkey""" |
|
646 | """Push markers over pushkey""" | |
647 | if not key.startswith('dump'): |
|
647 | if not key.startswith('dump'): | |
648 | repo.ui.warn(_('unknown key: %r') % key) |
|
648 | repo.ui.warn(_('unknown key: %r') % key) | |
649 | return 0 |
|
649 | return 0 | |
650 | if old: |
|
650 | if old: | |
651 | repo.ui.warn(_('unexpected old value for %r') % key) |
|
651 | repo.ui.warn(_('unexpected old value for %r') % key) | |
652 | return 0 |
|
652 | return 0 | |
653 | data = base85.b85decode(new) |
|
653 | data = base85.b85decode(new) | |
654 | lock = repo.lock() |
|
654 | lock = repo.lock() | |
655 | try: |
|
655 | try: | |
656 | tr = repo.transaction('pushkey: obsolete markers') |
|
656 | tr = repo.transaction('pushkey: obsolete markers') | |
657 | try: |
|
657 | try: | |
658 | repo.obsstore.mergemarkers(tr, data) |
|
658 | repo.obsstore.mergemarkers(tr, data) | |
659 | tr.close() |
|
659 | tr.close() | |
660 | return 1 |
|
660 | return 1 | |
661 | finally: |
|
661 | finally: | |
662 | tr.release() |
|
662 | tr.release() | |
663 | finally: |
|
663 | finally: | |
664 | lock.release() |
|
664 | lock.release() | |
665 |
|
665 | |||
666 | def getmarkers(repo, nodes=None): |
|
666 | def getmarkers(repo, nodes=None): | |
667 | """returns markers known in a repository |
|
667 | """returns markers known in a repository | |
668 |
|
668 | |||
669 | If <nodes> is specified, only markers "relevant" to those nodes are are |
|
669 | If <nodes> is specified, only markers "relevant" to those nodes are are | |
670 | returned""" |
|
670 | returned""" | |
671 | if nodes is None: |
|
671 | if nodes is None: | |
672 | rawmarkers = repo.obsstore |
|
672 | rawmarkers = repo.obsstore | |
673 | else: |
|
673 | else: | |
674 | rawmarkers = repo.obsstore.relevantmarkers(nodes) |
|
674 | rawmarkers = repo.obsstore.relevantmarkers(nodes) | |
675 |
|
675 | |||
676 | for markerdata in rawmarkers: |
|
676 | for markerdata in rawmarkers: | |
677 | yield marker(repo, markerdata) |
|
677 | yield marker(repo, markerdata) | |
678 |
|
678 | |||
679 | def relevantmarkers(repo, node): |
|
679 | def relevantmarkers(repo, node): | |
680 | """all obsolete markers relevant to some revision""" |
|
680 | """all obsolete markers relevant to some revision""" | |
681 | for markerdata in repo.obsstore.relevantmarkers(node): |
|
681 | for markerdata in repo.obsstore.relevantmarkers(node): | |
682 | yield marker(repo, markerdata) |
|
682 | yield marker(repo, markerdata) | |
683 |
|
683 | |||
684 |
|
684 | |||
685 | def precursormarkers(ctx): |
|
685 | def precursormarkers(ctx): | |
686 | """obsolete marker marking this changeset as a successors""" |
|
686 | """obsolete marker marking this changeset as a successors""" | |
687 | for data in ctx._repo.obsstore.precursors.get(ctx.node(), ()): |
|
687 | for data in ctx._repo.obsstore.precursors.get(ctx.node(), ()): | |
688 | yield marker(ctx._repo, data) |
|
688 | yield marker(ctx._repo, data) | |
689 |
|
689 | |||
690 | def successormarkers(ctx): |
|
690 | def successormarkers(ctx): | |
691 | """obsolete marker making this changeset obsolete""" |
|
691 | """obsolete marker making this changeset obsolete""" | |
692 | for data in ctx._repo.obsstore.successors.get(ctx.node(), ()): |
|
692 | for data in ctx._repo.obsstore.successors.get(ctx.node(), ()): | |
693 | yield marker(ctx._repo, data) |
|
693 | yield marker(ctx._repo, data) | |
694 |
|
694 | |||
695 | def allsuccessors(obsstore, nodes, ignoreflags=0): |
|
695 | def allsuccessors(obsstore, nodes, ignoreflags=0): | |
696 | """Yield node for every successor of <nodes>. |
|
696 | """Yield node for every successor of <nodes>. | |
697 |
|
697 | |||
698 | Some successors may be unknown locally. |
|
698 | Some successors may be unknown locally. | |
699 |
|
699 | |||
700 | This is a linear yield unsuited to detecting split changesets. It includes |
|
700 | This is a linear yield unsuited to detecting split changesets. It includes | |
701 | initial nodes too.""" |
|
701 | initial nodes too.""" | |
702 | remaining = set(nodes) |
|
702 | remaining = set(nodes) | |
703 | seen = set(remaining) |
|
703 | seen = set(remaining) | |
704 | while remaining: |
|
704 | while remaining: | |
705 | current = remaining.pop() |
|
705 | current = remaining.pop() | |
706 | yield current |
|
706 | yield current | |
707 | for mark in obsstore.successors.get(current, ()): |
|
707 | for mark in obsstore.successors.get(current, ()): | |
708 | # ignore marker flagged with specified flag |
|
708 | # ignore marker flagged with specified flag | |
709 | if mark[2] & ignoreflags: |
|
709 | if mark[2] & ignoreflags: | |
710 | continue |
|
710 | continue | |
711 | for suc in mark[1]: |
|
711 | for suc in mark[1]: | |
712 | if suc not in seen: |
|
712 | if suc not in seen: | |
713 | seen.add(suc) |
|
713 | seen.add(suc) | |
714 | remaining.add(suc) |
|
714 | remaining.add(suc) | |
715 |
|
715 | |||
716 | def allprecursors(obsstore, nodes, ignoreflags=0): |
|
716 | def allprecursors(obsstore, nodes, ignoreflags=0): | |
717 | """Yield node for every precursors of <nodes>. |
|
717 | """Yield node for every precursors of <nodes>. | |
718 |
|
718 | |||
719 | Some precursors may be unknown locally. |
|
719 | Some precursors may be unknown locally. | |
720 |
|
720 | |||
721 | This is a linear yield unsuited to detecting folded changesets. It includes |
|
721 | This is a linear yield unsuited to detecting folded changesets. It includes | |
722 | initial nodes too.""" |
|
722 | initial nodes too.""" | |
723 |
|
723 | |||
724 | remaining = set(nodes) |
|
724 | remaining = set(nodes) | |
725 | seen = set(remaining) |
|
725 | seen = set(remaining) | |
726 | while remaining: |
|
726 | while remaining: | |
727 | current = remaining.pop() |
|
727 | current = remaining.pop() | |
728 | yield current |
|
728 | yield current | |
729 | for mark in obsstore.precursors.get(current, ()): |
|
729 | for mark in obsstore.precursors.get(current, ()): | |
730 | # ignore marker flagged with specified flag |
|
730 | # ignore marker flagged with specified flag | |
731 | if mark[2] & ignoreflags: |
|
731 | if mark[2] & ignoreflags: | |
732 | continue |
|
732 | continue | |
733 | suc = mark[0] |
|
733 | suc = mark[0] | |
734 | if suc not in seen: |
|
734 | if suc not in seen: | |
735 | seen.add(suc) |
|
735 | seen.add(suc) | |
736 | remaining.add(suc) |
|
736 | remaining.add(suc) | |
737 |
|
737 | |||
738 | def foreground(repo, nodes): |
|
738 | def foreground(repo, nodes): | |
739 | """return all nodes in the "foreground" of other node |
|
739 | """return all nodes in the "foreground" of other node | |
740 |
|
740 | |||
741 | The foreground of a revision is anything reachable using parent -> children |
|
741 | The foreground of a revision is anything reachable using parent -> children | |
742 | or precursor -> successor relation. It is very similar to "descendant" but |
|
742 | or precursor -> successor relation. It is very similar to "descendant" but | |
743 | augmented with obsolescence information. |
|
743 | augmented with obsolescence information. | |
744 |
|
744 | |||
745 | Beware that possible obsolescence cycle may result if complex situation. |
|
745 | Beware that possible obsolescence cycle may result if complex situation. | |
746 | """ |
|
746 | """ | |
747 | repo = repo.unfiltered() |
|
747 | repo = repo.unfiltered() | |
748 | foreground = set(repo.set('%ln::', nodes)) |
|
748 | foreground = set(repo.set('%ln::', nodes)) | |
749 | if repo.obsstore: |
|
749 | if repo.obsstore: | |
750 | # We only need this complicated logic if there is obsolescence |
|
750 | # We only need this complicated logic if there is obsolescence | |
751 | # XXX will probably deserve an optimised revset. |
|
751 | # XXX will probably deserve an optimised revset. | |
752 | nm = repo.changelog.nodemap |
|
752 | nm = repo.changelog.nodemap | |
753 | plen = -1 |
|
753 | plen = -1 | |
754 | # compute the whole set of successors or descendants |
|
754 | # compute the whole set of successors or descendants | |
755 | while len(foreground) != plen: |
|
755 | while len(foreground) != plen: | |
756 | plen = len(foreground) |
|
756 | plen = len(foreground) | |
757 | succs = set(c.node() for c in foreground) |
|
757 | succs = set(c.node() for c in foreground) | |
758 | mutable = [c.node() for c in foreground if c.mutable()] |
|
758 | mutable = [c.node() for c in foreground if c.mutable()] | |
759 | succs.update(allsuccessors(repo.obsstore, mutable)) |
|
759 | succs.update(allsuccessors(repo.obsstore, mutable)) | |
760 | known = (n for n in succs if n in nm) |
|
760 | known = (n for n in succs if n in nm) | |
761 | foreground = set(repo.set('%ln::', known)) |
|
761 | foreground = set(repo.set('%ln::', known)) | |
762 | return set(c.node() for c in foreground) |
|
762 | return set(c.node() for c in foreground) | |
763 |
|
763 | |||
764 |
|
764 | |||
765 | def successorssets(repo, initialnode, cache=None): |
|
765 | def successorssets(repo, initialnode, cache=None): | |
766 | """Return all set of successors of initial nodes |
|
766 | """Return all set of successors of initial nodes | |
767 |
|
767 | |||
768 | The successors set of a changeset A are a group of revisions that succeed |
|
768 | The successors set of a changeset A are a group of revisions that succeed | |
769 | A. It succeeds A as a consistent whole, each revision being only a partial |
|
769 | A. It succeeds A as a consistent whole, each revision being only a partial | |
770 | replacement. The successors set contains non-obsolete changesets only. |
|
770 | replacement. The successors set contains non-obsolete changesets only. | |
771 |
|
771 | |||
772 | This function returns the full list of successor sets which is why it |
|
772 | This function returns the full list of successor sets which is why it | |
773 | returns a list of tuples and not just a single tuple. Each tuple is a valid |
|
773 | returns a list of tuples and not just a single tuple. Each tuple is a valid | |
774 | successors set. Not that (A,) may be a valid successors set for changeset A |
|
774 | successors set. Not that (A,) may be a valid successors set for changeset A | |
775 | (see below). |
|
775 | (see below). | |
776 |
|
776 | |||
777 | In most cases, a changeset A will have a single element (e.g. the changeset |
|
777 | In most cases, a changeset A will have a single element (e.g. the changeset | |
778 | A is replaced by A') in its successors set. Though, it is also common for a |
|
778 | A is replaced by A') in its successors set. Though, it is also common for a | |
779 | changeset A to have no elements in its successor set (e.g. the changeset |
|
779 | changeset A to have no elements in its successor set (e.g. the changeset | |
780 | has been pruned). Therefore, the returned list of successors sets will be |
|
780 | has been pruned). Therefore, the returned list of successors sets will be | |
781 | [(A',)] or [], respectively. |
|
781 | [(A',)] or [], respectively. | |
782 |
|
782 | |||
783 | When a changeset A is split into A' and B', however, it will result in a |
|
783 | When a changeset A is split into A' and B', however, it will result in a | |
784 | successors set containing more than a single element, i.e. [(A',B')]. |
|
784 | successors set containing more than a single element, i.e. [(A',B')]. | |
785 | Divergent changesets will result in multiple successors sets, i.e. [(A',), |
|
785 | Divergent changesets will result in multiple successors sets, i.e. [(A',), | |
786 | (A'')]. |
|
786 | (A'')]. | |
787 |
|
787 | |||
788 | If a changeset A is not obsolete, then it will conceptually have no |
|
788 | If a changeset A is not obsolete, then it will conceptually have no | |
789 | successors set. To distinguish this from a pruned changeset, the successor |
|
789 | successors set. To distinguish this from a pruned changeset, the successor | |
790 | set will only contain itself, i.e. [(A,)]. |
|
790 | set will only contain itself, i.e. [(A,)]. | |
791 |
|
791 | |||
792 | Finally, successors unknown locally are considered to be pruned (obsoleted |
|
792 | Finally, successors unknown locally are considered to be pruned (obsoleted | |
793 | without any successors). |
|
793 | without any successors). | |
794 |
|
794 | |||
795 | The optional `cache` parameter is a dictionary that may contain precomputed |
|
795 | The optional `cache` parameter is a dictionary that may contain precomputed | |
796 | successors sets. It is meant to reuse the computation of a previous call to |
|
796 | successors sets. It is meant to reuse the computation of a previous call to | |
797 | `successorssets` when multiple calls are made at the same time. The cache |
|
797 | `successorssets` when multiple calls are made at the same time. The cache | |
798 | dictionary is updated in place. The caller is responsible for its live |
|
798 | dictionary is updated in place. The caller is responsible for its live | |
799 | spawn. Code that makes multiple calls to `successorssets` *must* use this |
|
799 | spawn. Code that makes multiple calls to `successorssets` *must* use this | |
800 | cache mechanism or suffer terrible performances. |
|
800 | cache mechanism or suffer terrible performances. | |
801 |
|
801 | |||
802 | """ |
|
802 | """ | |
803 |
|
803 | |||
804 | succmarkers = repo.obsstore.successors |
|
804 | succmarkers = repo.obsstore.successors | |
805 |
|
805 | |||
806 | # Stack of nodes we search successors sets for |
|
806 | # Stack of nodes we search successors sets for | |
807 | toproceed = [initialnode] |
|
807 | toproceed = [initialnode] | |
808 | # set version of above list for fast loop detection |
|
808 | # set version of above list for fast loop detection | |
809 | # element added to "toproceed" must be added here |
|
809 | # element added to "toproceed" must be added here | |
810 | stackedset = set(toproceed) |
|
810 | stackedset = set(toproceed) | |
811 | if cache is None: |
|
811 | if cache is None: | |
812 | cache = {} |
|
812 | cache = {} | |
813 |
|
813 | |||
814 | # This while loop is the flattened version of a recursive search for |
|
814 | # This while loop is the flattened version of a recursive search for | |
815 | # successors sets |
|
815 | # successors sets | |
816 | # |
|
816 | # | |
817 | # def successorssets(x): |
|
817 | # def successorssets(x): | |
818 | # successors = directsuccessors(x) |
|
818 | # successors = directsuccessors(x) | |
819 | # ss = [[]] |
|
819 | # ss = [[]] | |
820 | # for succ in directsuccessors(x): |
|
820 | # for succ in directsuccessors(x): | |
821 | # # product as in itertools cartesian product |
|
821 | # # product as in itertools cartesian product | |
822 | # ss = product(ss, successorssets(succ)) |
|
822 | # ss = product(ss, successorssets(succ)) | |
823 | # return ss |
|
823 | # return ss | |
824 | # |
|
824 | # | |
825 | # But we can not use plain recursive calls here: |
|
825 | # But we can not use plain recursive calls here: | |
826 | # - that would blow the python call stack |
|
826 | # - that would blow the python call stack | |
827 | # - obsolescence markers may have cycles, we need to handle them. |
|
827 | # - obsolescence markers may have cycles, we need to handle them. | |
828 | # |
|
828 | # | |
829 | # The `toproceed` list act as our call stack. Every node we search |
|
829 | # The `toproceed` list act as our call stack. Every node we search | |
830 | # successors set for are stacked there. |
|
830 | # successors set for are stacked there. | |
831 | # |
|
831 | # | |
832 | # The `stackedset` is set version of this stack used to check if a node is |
|
832 | # The `stackedset` is set version of this stack used to check if a node is | |
833 | # already stacked. This check is used to detect cycles and prevent infinite |
|
833 | # already stacked. This check is used to detect cycles and prevent infinite | |
834 | # loop. |
|
834 | # loop. | |
835 | # |
|
835 | # | |
836 | # successors set of all nodes are stored in the `cache` dictionary. |
|
836 | # successors set of all nodes are stored in the `cache` dictionary. | |
837 | # |
|
837 | # | |
838 | # After this while loop ends we use the cache to return the successors sets |
|
838 | # After this while loop ends we use the cache to return the successors sets | |
839 | # for the node requested by the caller. |
|
839 | # for the node requested by the caller. | |
840 | while toproceed: |
|
840 | while toproceed: | |
841 | # Every iteration tries to compute the successors sets of the topmost |
|
841 | # Every iteration tries to compute the successors sets of the topmost | |
842 | # node of the stack: CURRENT. |
|
842 | # node of the stack: CURRENT. | |
843 | # |
|
843 | # | |
844 | # There are four possible outcomes: |
|
844 | # There are four possible outcomes: | |
845 | # |
|
845 | # | |
846 | # 1) We already know the successors sets of CURRENT: |
|
846 | # 1) We already know the successors sets of CURRENT: | |
847 | # -> mission accomplished, pop it from the stack. |
|
847 | # -> mission accomplished, pop it from the stack. | |
848 | # 2) Node is not obsolete: |
|
848 | # 2) Node is not obsolete: | |
849 | # -> the node is its own successors sets. Add it to the cache. |
|
849 | # -> the node is its own successors sets. Add it to the cache. | |
850 | # 3) We do not know successors set of direct successors of CURRENT: |
|
850 | # 3) We do not know successors set of direct successors of CURRENT: | |
851 | # -> We add those successors to the stack. |
|
851 | # -> We add those successors to the stack. | |
852 | # 4) We know successors sets of all direct successors of CURRENT: |
|
852 | # 4) We know successors sets of all direct successors of CURRENT: | |
853 | # -> We can compute CURRENT successors set and add it to the |
|
853 | # -> We can compute CURRENT successors set and add it to the | |
854 | # cache. |
|
854 | # cache. | |
855 | # |
|
855 | # | |
856 | current = toproceed[-1] |
|
856 | current = toproceed[-1] | |
857 | if current in cache: |
|
857 | if current in cache: | |
858 | # case (1): We already know the successors sets |
|
858 | # case (1): We already know the successors sets | |
859 | stackedset.remove(toproceed.pop()) |
|
859 | stackedset.remove(toproceed.pop()) | |
860 | elif current not in succmarkers: |
|
860 | elif current not in succmarkers: | |
861 | # case (2): The node is not obsolete. |
|
861 | # case (2): The node is not obsolete. | |
862 | if current in repo: |
|
862 | if current in repo: | |
863 | # We have a valid last successors. |
|
863 | # We have a valid last successors. | |
864 | cache[current] = [(current,)] |
|
864 | cache[current] = [(current,)] | |
865 | else: |
|
865 | else: | |
866 | # Final obsolete version is unknown locally. |
|
866 | # Final obsolete version is unknown locally. | |
867 | # Do not count that as a valid successors |
|
867 | # Do not count that as a valid successors | |
868 | cache[current] = [] |
|
868 | cache[current] = [] | |
869 | else: |
|
869 | else: | |
870 | # cases (3) and (4) |
|
870 | # cases (3) and (4) | |
871 | # |
|
871 | # | |
872 | # We proceed in two phases. Phase 1 aims to distinguish case (3) |
|
872 | # We proceed in two phases. Phase 1 aims to distinguish case (3) | |
873 | # from case (4): |
|
873 | # from case (4): | |
874 | # |
|
874 | # | |
875 | # For each direct successors of CURRENT, we check whether its |
|
875 | # For each direct successors of CURRENT, we check whether its | |
876 | # successors sets are known. If they are not, we stack the |
|
876 | # successors sets are known. If they are not, we stack the | |
877 | # unknown node and proceed to the next iteration of the while |
|
877 | # unknown node and proceed to the next iteration of the while | |
878 | # loop. (case 3) |
|
878 | # loop. (case 3) | |
879 | # |
|
879 | # | |
880 | # During this step, we may detect obsolescence cycles: a node |
|
880 | # During this step, we may detect obsolescence cycles: a node | |
881 | # with unknown successors sets but already in the call stack. |
|
881 | # with unknown successors sets but already in the call stack. | |
882 | # In such a situation, we arbitrary set the successors sets of |
|
882 | # In such a situation, we arbitrary set the successors sets of | |
883 | # the node to nothing (node pruned) to break the cycle. |
|
883 | # the node to nothing (node pruned) to break the cycle. | |
884 | # |
|
884 | # | |
885 | # If no break was encountered we proceed to phase 2. |
|
885 | # If no break was encountered we proceed to phase 2. | |
886 | # |
|
886 | # | |
887 | # Phase 2 computes successors sets of CURRENT (case 4); see details |
|
887 | # Phase 2 computes successors sets of CURRENT (case 4); see details | |
888 | # in phase 2 itself. |
|
888 | # in phase 2 itself. | |
889 | # |
|
889 | # | |
890 | # Note the two levels of iteration in each phase. |
|
890 | # Note the two levels of iteration in each phase. | |
891 | # - The first one handles obsolescence markers using CURRENT as |
|
891 | # - The first one handles obsolescence markers using CURRENT as | |
892 | # precursor (successors markers of CURRENT). |
|
892 | # precursor (successors markers of CURRENT). | |
893 | # |
|
893 | # | |
894 | # Having multiple entry here means divergence. |
|
894 | # Having multiple entry here means divergence. | |
895 | # |
|
895 | # | |
896 | # - The second one handles successors defined in each marker. |
|
896 | # - The second one handles successors defined in each marker. | |
897 | # |
|
897 | # | |
898 | # Having none means pruned node, multiple successors means split, |
|
898 | # Having none means pruned node, multiple successors means split, | |
899 | # single successors are standard replacement. |
|
899 | # single successors are standard replacement. | |
900 | # |
|
900 | # | |
901 | for mark in sorted(succmarkers[current]): |
|
901 | for mark in sorted(succmarkers[current]): | |
902 | for suc in mark[1]: |
|
902 | for suc in mark[1]: | |
903 | if suc not in cache: |
|
903 | if suc not in cache: | |
904 | if suc in stackedset: |
|
904 | if suc in stackedset: | |
905 | # cycle breaking |
|
905 | # cycle breaking | |
906 | cache[suc] = [] |
|
906 | cache[suc] = [] | |
907 | else: |
|
907 | else: | |
908 | # case (3) If we have not computed successors sets |
|
908 | # case (3) If we have not computed successors sets | |
909 | # of one of those successors we add it to the |
|
909 | # of one of those successors we add it to the | |
910 | # `toproceed` stack and stop all work for this |
|
910 | # `toproceed` stack and stop all work for this | |
911 | # iteration. |
|
911 | # iteration. | |
912 | toproceed.append(suc) |
|
912 | toproceed.append(suc) | |
913 | stackedset.add(suc) |
|
913 | stackedset.add(suc) | |
914 | break |
|
914 | break | |
915 | else: |
|
915 | else: | |
916 | continue |
|
916 | continue | |
917 | break |
|
917 | break | |
918 | else: |
|
918 | else: | |
919 | # case (4): we know all successors sets of all direct |
|
919 | # case (4): we know all successors sets of all direct | |
920 | # successors |
|
920 | # successors | |
921 | # |
|
921 | # | |
922 | # Successors set contributed by each marker depends on the |
|
922 | # Successors set contributed by each marker depends on the | |
923 | # successors sets of all its "successors" node. |
|
923 | # successors sets of all its "successors" node. | |
924 | # |
|
924 | # | |
925 | # Each different marker is a divergence in the obsolescence |
|
925 | # Each different marker is a divergence in the obsolescence | |
926 | # history. It contributes successors sets distinct from other |
|
926 | # history. It contributes successors sets distinct from other | |
927 | # markers. |
|
927 | # markers. | |
928 | # |
|
928 | # | |
929 | # Within a marker, a successor may have divergent successors |
|
929 | # Within a marker, a successor may have divergent successors | |
930 | # sets. In such a case, the marker will contribute multiple |
|
930 | # sets. In such a case, the marker will contribute multiple | |
931 | # divergent successors sets. If multiple successors have |
|
931 | # divergent successors sets. If multiple successors have | |
932 | # divergent successors sets, a Cartesian product is used. |
|
932 | # divergent successors sets, a Cartesian product is used. | |
933 | # |
|
933 | # | |
934 | # At the end we post-process successors sets to remove |
|
934 | # At the end we post-process successors sets to remove | |
935 | # duplicated entry and successors set that are strict subset of |
|
935 | # duplicated entry and successors set that are strict subset of | |
936 | # another one. |
|
936 | # another one. | |
937 | succssets = [] |
|
937 | succssets = [] | |
938 | for mark in sorted(succmarkers[current]): |
|
938 | for mark in sorted(succmarkers[current]): | |
939 | # successors sets contributed by this marker |
|
939 | # successors sets contributed by this marker | |
940 | markss = [[]] |
|
940 | markss = [[]] | |
941 | for suc in mark[1]: |
|
941 | for suc in mark[1]: | |
942 | # cardinal product with previous successors |
|
942 | # cardinal product with previous successors | |
943 | productresult = [] |
|
943 | productresult = [] | |
944 | for prefix in markss: |
|
944 | for prefix in markss: | |
945 | for suffix in cache[suc]: |
|
945 | for suffix in cache[suc]: | |
946 | newss = list(prefix) |
|
946 | newss = list(prefix) | |
947 | for part in suffix: |
|
947 | for part in suffix: | |
948 | # do not duplicated entry in successors set |
|
948 | # do not duplicated entry in successors set | |
949 | # first entry wins. |
|
949 | # first entry wins. | |
950 | if part not in newss: |
|
950 | if part not in newss: | |
951 | newss.append(part) |
|
951 | newss.append(part) | |
952 | productresult.append(newss) |
|
952 | productresult.append(newss) | |
953 | markss = productresult |
|
953 | markss = productresult | |
954 | succssets.extend(markss) |
|
954 | succssets.extend(markss) | |
955 | # remove duplicated and subset |
|
955 | # remove duplicated and subset | |
956 | seen = [] |
|
956 | seen = [] | |
957 | final = [] |
|
957 | final = [] | |
958 | candidate = sorted(((set(s), s) for s in succssets if s), |
|
958 | candidate = sorted(((set(s), s) for s in succssets if s), | |
959 | key=lambda x: len(x[1]), reverse=True) |
|
959 | key=lambda x: len(x[1]), reverse=True) | |
960 | for setversion, listversion in candidate: |
|
960 | for setversion, listversion in candidate: | |
961 | for seenset in seen: |
|
961 | for seenset in seen: | |
962 | if setversion.issubset(seenset): |
|
962 | if setversion.issubset(seenset): | |
963 | break |
|
963 | break | |
964 | else: |
|
964 | else: | |
965 | final.append(listversion) |
|
965 | final.append(listversion) | |
966 | seen.append(setversion) |
|
966 | seen.append(setversion) | |
967 | final.reverse() # put small successors set first |
|
967 | final.reverse() # put small successors set first | |
968 | cache[current] = final |
|
968 | cache[current] = final | |
969 | return cache[initialnode] |
|
969 | return cache[initialnode] | |
970 |
|
970 | |||
971 | def _knownrevs(repo, nodes): |
|
971 | def _knownrevs(repo, nodes): | |
972 | """yield revision numbers of known nodes passed in parameters |
|
972 | """yield revision numbers of known nodes passed in parameters | |
973 |
|
973 | |||
974 | Unknown revisions are silently ignored.""" |
|
974 | Unknown revisions are silently ignored.""" | |
975 | torev = repo.changelog.nodemap.get |
|
975 | torev = repo.changelog.nodemap.get | |
976 | for n in nodes: |
|
976 | for n in nodes: | |
977 | rev = torev(n) |
|
977 | rev = torev(n) | |
978 | if rev is not None: |
|
978 | if rev is not None: | |
979 | yield rev |
|
979 | yield rev | |
980 |
|
980 | |||
981 | # mapping of 'set-name' -> <function to compute this set> |
|
981 | # mapping of 'set-name' -> <function to compute this set> | |
982 | cachefuncs = {} |
|
982 | cachefuncs = {} | |
983 | def cachefor(name): |
|
983 | def cachefor(name): | |
984 | """Decorator to register a function as computing the cache for a set""" |
|
984 | """Decorator to register a function as computing the cache for a set""" | |
985 | def decorator(func): |
|
985 | def decorator(func): | |
986 | assert name not in cachefuncs |
|
986 | assert name not in cachefuncs | |
987 | cachefuncs[name] = func |
|
987 | cachefuncs[name] = func | |
988 | return func |
|
988 | return func | |
989 | return decorator |
|
989 | return decorator | |
990 |
|
990 | |||
991 | def getrevs(repo, name): |
|
991 | def getrevs(repo, name): | |
992 | """Return the set of revision that belong to the <name> set |
|
992 | """Return the set of revision that belong to the <name> set | |
993 |
|
993 | |||
994 | Such access may compute the set and cache it for future use""" |
|
994 | Such access may compute the set and cache it for future use""" | |
995 | repo = repo.unfiltered() |
|
995 | repo = repo.unfiltered() | |
996 | if not repo.obsstore: |
|
996 | if not repo.obsstore: | |
997 | return frozenset() |
|
997 | return frozenset() | |
998 | if name not in repo.obsstore.caches: |
|
998 | if name not in repo.obsstore.caches: | |
999 | repo.obsstore.caches[name] = cachefuncs[name](repo) |
|
999 | repo.obsstore.caches[name] = cachefuncs[name](repo) | |
1000 | return repo.obsstore.caches[name] |
|
1000 | return repo.obsstore.caches[name] | |
1001 |
|
1001 | |||
1002 | # To be simple we need to invalidate obsolescence cache when: |
|
1002 | # To be simple we need to invalidate obsolescence cache when: | |
1003 | # |
|
1003 | # | |
1004 | # - new changeset is added: |
|
1004 | # - new changeset is added: | |
1005 | # - public phase is changed |
|
1005 | # - public phase is changed | |
1006 | # - obsolescence marker are added |
|
1006 | # - obsolescence marker are added | |
1007 | # - strip is used a repo |
|
1007 | # - strip is used a repo | |
1008 | def clearobscaches(repo): |
|
1008 | def clearobscaches(repo): | |
1009 | """Remove all obsolescence related cache from a repo |
|
1009 | """Remove all obsolescence related cache from a repo | |
1010 |
|
1010 | |||
1011 | This remove all cache in obsstore is the obsstore already exist on the |
|
1011 | This remove all cache in obsstore is the obsstore already exist on the | |
1012 | repo. |
|
1012 | repo. | |
1013 |
|
1013 | |||
1014 | (We could be smarter here given the exact event that trigger the cache |
|
1014 | (We could be smarter here given the exact event that trigger the cache | |
1015 | clearing)""" |
|
1015 | clearing)""" | |
1016 | # only clear cache is there is obsstore data in this repo |
|
1016 | # only clear cache is there is obsstore data in this repo | |
1017 | if 'obsstore' in repo._filecache: |
|
1017 | if 'obsstore' in repo._filecache: | |
1018 | repo.obsstore.caches.clear() |
|
1018 | repo.obsstore.caches.clear() | |
1019 |
|
1019 | |||
1020 | @cachefor('obsolete') |
|
1020 | @cachefor('obsolete') | |
1021 | def _computeobsoleteset(repo): |
|
1021 | def _computeobsoleteset(repo): | |
1022 | """the set of obsolete revisions""" |
|
1022 | """the set of obsolete revisions""" | |
1023 | obs = set() |
|
1023 | obs = set() | |
1024 | getrev = repo.changelog.nodemap.get |
|
1024 | getrev = repo.changelog.nodemap.get | |
1025 | getphase = repo._phasecache.phase |
|
1025 | getphase = repo._phasecache.phase | |
1026 | for n in repo.obsstore.successors: |
|
1026 | for n in repo.obsstore.successors: | |
1027 | rev = getrev(n) |
|
1027 | rev = getrev(n) | |
1028 | if rev is not None and getphase(repo, rev): |
|
1028 | if rev is not None and getphase(repo, rev): | |
1029 | obs.add(rev) |
|
1029 | obs.add(rev) | |
1030 | return obs |
|
1030 | return obs | |
1031 |
|
1031 | |||
1032 | @cachefor('unstable') |
|
1032 | @cachefor('unstable') | |
1033 | def _computeunstableset(repo): |
|
1033 | def _computeunstableset(repo): | |
1034 | """the set of non obsolete revisions with obsolete parents""" |
|
1034 | """the set of non obsolete revisions with obsolete parents""" | |
1035 | # revset is not efficient enough here |
|
1035 | # revset is not efficient enough here | |
1036 | # we do (obsolete()::) - obsolete() by hand |
|
1036 | # we do (obsolete()::) - obsolete() by hand | |
1037 | obs = getrevs(repo, 'obsolete') |
|
1037 | obs = getrevs(repo, 'obsolete') | |
1038 | if not obs: |
|
1038 | if not obs: | |
1039 | return set() |
|
1039 | return set() | |
1040 | cl = repo.changelog |
|
1040 | cl = repo.changelog | |
1041 | return set(r for r in cl.descendants(obs) if r not in obs) |
|
1041 | return set(r for r in cl.descendants(obs) if r not in obs) | |
1042 |
|
1042 | |||
1043 | @cachefor('suspended') |
|
1043 | @cachefor('suspended') | |
1044 | def _computesuspendedset(repo): |
|
1044 | def _computesuspendedset(repo): | |
1045 | """the set of obsolete parents with non obsolete descendants""" |
|
1045 | """the set of obsolete parents with non obsolete descendants""" | |
1046 | suspended = repo.changelog.ancestors(getrevs(repo, 'unstable')) |
|
1046 | suspended = repo.changelog.ancestors(getrevs(repo, 'unstable')) | |
1047 | return set(r for r in getrevs(repo, 'obsolete') if r in suspended) |
|
1047 | return set(r for r in getrevs(repo, 'obsolete') if r in suspended) | |
1048 |
|
1048 | |||
1049 | @cachefor('extinct') |
|
1049 | @cachefor('extinct') | |
1050 | def _computeextinctset(repo): |
|
1050 | def _computeextinctset(repo): | |
1051 | """the set of obsolete parents without non obsolete descendants""" |
|
1051 | """the set of obsolete parents without non obsolete descendants""" | |
1052 | return getrevs(repo, 'obsolete') - getrevs(repo, 'suspended') |
|
1052 | return getrevs(repo, 'obsolete') - getrevs(repo, 'suspended') | |
1053 |
|
1053 | |||
1054 |
|
1054 | |||
1055 | @cachefor('bumped') |
|
1055 | @cachefor('bumped') | |
1056 | def _computebumpedset(repo): |
|
1056 | def _computebumpedset(repo): | |
1057 | """the set of revs trying to obsolete public revisions""" |
|
1057 | """the set of revs trying to obsolete public revisions""" | |
1058 | bumped = set() |
|
1058 | bumped = set() | |
1059 | # util function (avoid attribute lookup in the loop) |
|
1059 | # util function (avoid attribute lookup in the loop) | |
1060 | phase = repo._phasecache.phase # would be faster to grab the full list |
|
1060 | phase = repo._phasecache.phase # would be faster to grab the full list | |
1061 | public = phases.public |
|
1061 | public = phases.public | |
1062 | cl = repo.changelog |
|
1062 | cl = repo.changelog | |
1063 | torev = cl.nodemap.get |
|
1063 | torev = cl.nodemap.get | |
1064 | obs = getrevs(repo, 'obsolete') |
|
1064 | obs = getrevs(repo, 'obsolete') | |
1065 | for rev in repo: |
|
1065 | for rev in repo: | |
1066 | # We only evaluate mutable, non-obsolete revision |
|
1066 | # We only evaluate mutable, non-obsolete revision | |
1067 | if (public < phase(repo, rev)) and (rev not in obs): |
|
1067 | if (public < phase(repo, rev)) and (rev not in obs): | |
1068 | node = cl.node(rev) |
|
1068 | node = cl.node(rev) | |
1069 | # (future) A cache of precursors may worth if split is very common |
|
1069 | # (future) A cache of precursors may worth if split is very common | |
1070 | for pnode in allprecursors(repo.obsstore, [node], |
|
1070 | for pnode in allprecursors(repo.obsstore, [node], | |
1071 | ignoreflags=bumpedfix): |
|
1071 | ignoreflags=bumpedfix): | |
1072 | prev = torev(pnode) # unfiltered! but so is phasecache |
|
1072 | prev = torev(pnode) # unfiltered! but so is phasecache | |
1073 | if (prev is not None) and (phase(repo, prev) <= public): |
|
1073 | if (prev is not None) and (phase(repo, prev) <= public): | |
1074 | # we have a public precursors |
|
1074 | # we have a public precursors | |
1075 | bumped.add(rev) |
|
1075 | bumped.add(rev) | |
1076 | break # Next draft! |
|
1076 | break # Next draft! | |
1077 | return bumped |
|
1077 | return bumped | |
1078 |
|
1078 | |||
1079 | @cachefor('divergent') |
|
1079 | @cachefor('divergent') | |
1080 | def _computedivergentset(repo): |
|
1080 | def _computedivergentset(repo): | |
1081 | """the set of rev that compete to be the final successors of some revision. |
|
1081 | """the set of rev that compete to be the final successors of some revision. | |
1082 | """ |
|
1082 | """ | |
1083 | divergent = set() |
|
1083 | divergent = set() | |
1084 | obsstore = repo.obsstore |
|
1084 | obsstore = repo.obsstore | |
1085 | newermap = {} |
|
1085 | newermap = {} | |
1086 | for ctx in repo.set('(not public()) - obsolete()'): |
|
1086 | for ctx in repo.set('(not public()) - obsolete()'): | |
1087 | mark = obsstore.precursors.get(ctx.node(), ()) |
|
1087 | mark = obsstore.precursors.get(ctx.node(), ()) | |
1088 | toprocess = set(mark) |
|
1088 | toprocess = set(mark) | |
1089 | while toprocess: |
|
1089 | while toprocess: | |
1090 | prec = toprocess.pop()[0] |
|
1090 | prec = toprocess.pop()[0] | |
1091 | if prec not in newermap: |
|
1091 | if prec not in newermap: | |
1092 | successorssets(repo, prec, newermap) |
|
1092 | successorssets(repo, prec, newermap) | |
1093 | newer = [n for n in newermap[prec] if n] |
|
1093 | newer = [n for n in newermap[prec] if n] | |
1094 | if len(newer) > 1: |
|
1094 | if len(newer) > 1: | |
1095 | divergent.add(ctx.rev()) |
|
1095 | divergent.add(ctx.rev()) | |
1096 | break |
|
1096 | break | |
1097 | toprocess.update(obsstore.precursors.get(prec, ())) |
|
1097 | toprocess.update(obsstore.precursors.get(prec, ())) | |
1098 | return divergent |
|
1098 | return divergent | |
1099 |
|
1099 | |||
1100 |
|
1100 | |||
1101 | def createmarkers(repo, relations, flag=0, date=None, metadata=None): |
|
1101 | def createmarkers(repo, relations, flag=0, date=None, metadata=None): | |
1102 | """Add obsolete markers between changesets in a repo |
|
1102 | """Add obsolete markers between changesets in a repo | |
1103 |
|
1103 | |||
1104 | <relations> must be an iterable of (<old>, (<new>, ...)[,{metadata}]) |
|
1104 | <relations> must be an iterable of (<old>, (<new>, ...)[,{metadata}]) | |
1105 | tuple. `old` and `news` are changectx. metadata is an optional dictionary |
|
1105 | tuple. `old` and `news` are changectx. metadata is an optional dictionary | |
1106 | containing metadata for this marker only. It is merged with the global |
|
1106 | containing metadata for this marker only. It is merged with the global | |
1107 | metadata specified through the `metadata` argument of this function, |
|
1107 | metadata specified through the `metadata` argument of this function, | |
1108 |
|
1108 | |||
1109 | Trying to obsolete a public changeset will raise an exception. |
|
1109 | Trying to obsolete a public changeset will raise an exception. | |
1110 |
|
1110 | |||
1111 | Current user and date are used except if specified otherwise in the |
|
1111 | Current user and date are used except if specified otherwise in the | |
1112 | metadata attribute. |
|
1112 | metadata attribute. | |
1113 |
|
1113 | |||
1114 | This function operates within a transaction of its own, but does |
|
1114 | This function operates within a transaction of its own, but does | |
1115 | not take any lock on the repo. |
|
1115 | not take any lock on the repo. | |
1116 | """ |
|
1116 | """ | |
1117 | # prepare metadata |
|
1117 | # prepare metadata | |
1118 | if metadata is None: |
|
1118 | if metadata is None: | |
1119 | metadata = {} |
|
1119 | metadata = {} | |
1120 | if 'user' not in metadata: |
|
1120 | if 'user' not in metadata: | |
1121 | metadata['user'] = repo.ui.username() |
|
1121 | metadata['user'] = repo.ui.username() | |
1122 | tr = repo.transaction('add-obsolescence-marker') |
|
1122 | tr = repo.transaction('add-obsolescence-marker') | |
1123 | try: |
|
1123 | try: | |
1124 | for rel in relations: |
|
1124 | for rel in relations: | |
1125 | prec = rel[0] |
|
1125 | prec = rel[0] | |
1126 | sucs = rel[1] |
|
1126 | sucs = rel[1] | |
1127 | localmetadata = metadata.copy() |
|
1127 | localmetadata = metadata.copy() | |
1128 | if 2 < len(rel): |
|
1128 | if 2 < len(rel): | |
1129 | localmetadata.update(rel[2]) |
|
1129 | localmetadata.update(rel[2]) | |
1130 |
|
1130 | |||
1131 | if not prec.mutable(): |
|
1131 | if not prec.mutable(): | |
1132 | raise util.Abort("cannot obsolete immutable changeset: %s" |
|
1132 | raise util.Abort("cannot obsolete immutable changeset: %s" | |
1133 | % prec) |
|
1133 | % prec) | |
1134 | nprec = prec.node() |
|
1134 | nprec = prec.node() | |
1135 | nsucs = tuple(s.node() for s in sucs) |
|
1135 | nsucs = tuple(s.node() for s in sucs) | |
1136 | npare = None |
|
1136 | npare = None | |
1137 | if not nsucs: |
|
1137 | if not nsucs: | |
1138 | npare = tuple(p.node() for p in prec.parents()) |
|
1138 | npare = tuple(p.node() for p in prec.parents()) | |
1139 | if nprec in nsucs: |
|
1139 | if nprec in nsucs: | |
1140 | raise util.Abort("changeset %s cannot obsolete itself" % prec) |
|
1140 | raise util.Abort("changeset %s cannot obsolete itself" % prec) | |
1141 | repo.obsstore.create(tr, nprec, nsucs, flag, parents=npare, |
|
1141 | repo.obsstore.create(tr, nprec, nsucs, flag, parents=npare, | |
1142 | date=date, metadata=localmetadata) |
|
1142 | date=date, metadata=localmetadata) | |
1143 | repo.filteredrevcache.clear() |
|
1143 | repo.filteredrevcache.clear() | |
1144 | tr.close() |
|
1144 | tr.close() | |
1145 | finally: |
|
1145 | finally: | |
1146 | tr.release() |
|
1146 | tr.release() |
General Comments 0
You need to be logged in to leave comments.
Login now