Show More
@@ -1,1224 +1,1224 b'' | |||||
1 | # coding: utf8 |
|
1 | # coding: utf8 | |
2 | # copies.py - copy detection for Mercurial |
|
2 | # copies.py - copy detection for Mercurial | |
3 | # |
|
3 | # | |
4 | # Copyright 2008 Matt Mackall <mpm@selenic.com> |
|
4 | # Copyright 2008 Matt Mackall <mpm@selenic.com> | |
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 | from __future__ import absolute_import |
|
9 | from __future__ import absolute_import | |
10 |
|
10 | |||
11 | import collections |
|
11 | import collections | |
12 | import os |
|
12 | import os | |
13 |
|
13 | |||
14 | from .i18n import _ |
|
14 | from .i18n import _ | |
15 | from .node import ( |
|
15 | from .node import ( | |
16 | nullid, |
|
16 | nullid, | |
17 | nullrev, |
|
17 | nullrev, | |
18 | ) |
|
18 | ) | |
19 |
|
19 | |||
20 | from . import ( |
|
20 | from . import ( | |
21 | match as matchmod, |
|
21 | match as matchmod, | |
22 | pathutil, |
|
22 | pathutil, | |
23 | policy, |
|
23 | policy, | |
24 | pycompat, |
|
24 | pycompat, | |
25 | util, |
|
25 | util, | |
26 | ) |
|
26 | ) | |
27 |
|
27 | |||
28 |
|
28 | |||
29 | from .utils import stringutil |
|
29 | from .utils import stringutil | |
30 |
|
30 | |||
31 | from .revlogutils import ( |
|
31 | from .revlogutils import ( | |
32 | flagutil, |
|
32 | flagutil, | |
33 | sidedata as sidedatamod, |
|
33 | sidedata as sidedatamod, | |
34 | ) |
|
34 | ) | |
35 |
|
35 | |||
36 | rustmod = policy.importrust("copy_tracing") |
|
36 | rustmod = policy.importrust("copy_tracing") | |
37 |
|
37 | |||
38 |
|
38 | |||
39 | def _filter(src, dst, t): |
|
39 | def _filter(src, dst, t): | |
40 | """filters out invalid copies after chaining""" |
|
40 | """filters out invalid copies after chaining""" | |
41 |
|
41 | |||
42 | # When _chain()'ing copies in 'a' (from 'src' via some other commit 'mid') |
|
42 | # When _chain()'ing copies in 'a' (from 'src' via some other commit 'mid') | |
43 | # with copies in 'b' (from 'mid' to 'dst'), we can get the different cases |
|
43 | # with copies in 'b' (from 'mid' to 'dst'), we can get the different cases | |
44 | # in the following table (not including trivial cases). For example, case 2 |
|
44 | # in the following table (not including trivial cases). For example, case 2 | |
45 | # is where a file existed in 'src' and remained under that name in 'mid' and |
|
45 | # is where a file existed in 'src' and remained under that name in 'mid' and | |
46 | # then was renamed between 'mid' and 'dst'. |
|
46 | # then was renamed between 'mid' and 'dst'. | |
47 | # |
|
47 | # | |
48 | # case src mid dst result |
|
48 | # case src mid dst result | |
49 | # 1 x y - - |
|
49 | # 1 x y - - | |
50 | # 2 x y y x->y |
|
50 | # 2 x y y x->y | |
51 | # 3 x y x - |
|
51 | # 3 x y x - | |
52 | # 4 x y z x->z |
|
52 | # 4 x y z x->z | |
53 | # 5 - x y - |
|
53 | # 5 - x y - | |
54 | # 6 x x y x->y |
|
54 | # 6 x x y x->y | |
55 | # |
|
55 | # | |
56 | # _chain() takes care of chaining the copies in 'a' and 'b', but it |
|
56 | # _chain() takes care of chaining the copies in 'a' and 'b', but it | |
57 | # cannot tell the difference between cases 1 and 2, between 3 and 4, or |
|
57 | # cannot tell the difference between cases 1 and 2, between 3 and 4, or | |
58 | # between 5 and 6, so it includes all cases in its result. |
|
58 | # between 5 and 6, so it includes all cases in its result. | |
59 | # Cases 1, 3, and 5 are then removed by _filter(). |
|
59 | # Cases 1, 3, and 5 are then removed by _filter(). | |
60 |
|
60 | |||
61 | for k, v in list(t.items()): |
|
61 | for k, v in list(t.items()): | |
62 | # remove copies from files that didn't exist |
|
62 | # remove copies from files that didn't exist | |
63 | if v not in src: |
|
63 | if v not in src: | |
64 | del t[k] |
|
64 | del t[k] | |
65 | # remove criss-crossed copies |
|
65 | # remove criss-crossed copies | |
66 | elif k in src and v in dst: |
|
66 | elif k in src and v in dst: | |
67 | del t[k] |
|
67 | del t[k] | |
68 | # remove copies to files that were then removed |
|
68 | # remove copies to files that were then removed | |
69 | elif k not in dst: |
|
69 | elif k not in dst: | |
70 | del t[k] |
|
70 | del t[k] | |
71 |
|
71 | |||
72 |
|
72 | |||
73 | def _chain(prefix, suffix): |
|
73 | def _chain(prefix, suffix): | |
74 | """chain two sets of copies 'prefix' and 'suffix'""" |
|
74 | """chain two sets of copies 'prefix' and 'suffix'""" | |
75 | result = prefix.copy() |
|
75 | result = prefix.copy() | |
76 | for key, value in pycompat.iteritems(suffix): |
|
76 | for key, value in pycompat.iteritems(suffix): | |
77 | result[key] = prefix.get(value, value) |
|
77 | result[key] = prefix.get(value, value) | |
78 | return result |
|
78 | return result | |
79 |
|
79 | |||
80 |
|
80 | |||
81 | def _tracefile(fctx, am, basemf): |
|
81 | def _tracefile(fctx, am, basemf): | |
82 | """return file context that is the ancestor of fctx present in ancestor |
|
82 | """return file context that is the ancestor of fctx present in ancestor | |
83 | manifest am |
|
83 | manifest am | |
84 |
|
84 | |||
85 | Note: we used to try and stop after a given limit, however checking if that |
|
85 | Note: we used to try and stop after a given limit, however checking if that | |
86 | limit is reached turned out to be very expensive. we are better off |
|
86 | limit is reached turned out to be very expensive. we are better off | |
87 | disabling that feature.""" |
|
87 | disabling that feature.""" | |
88 |
|
88 | |||
89 | for f in fctx.ancestors(): |
|
89 | for f in fctx.ancestors(): | |
90 | path = f.path() |
|
90 | path = f.path() | |
91 | if am.get(path, None) == f.filenode(): |
|
91 | if am.get(path, None) == f.filenode(): | |
92 | return path |
|
92 | return path | |
93 | if basemf and basemf.get(path, None) == f.filenode(): |
|
93 | if basemf and basemf.get(path, None) == f.filenode(): | |
94 | return path |
|
94 | return path | |
95 |
|
95 | |||
96 |
|
96 | |||
97 | def _dirstatecopies(repo, match=None): |
|
97 | def _dirstatecopies(repo, match=None): | |
98 | ds = repo.dirstate |
|
98 | ds = repo.dirstate | |
99 | c = ds.copies().copy() |
|
99 | c = ds.copies().copy() | |
100 | for k in list(c): |
|
100 | for k in list(c): | |
101 | if ds[k] not in b'anm' or (match and not match(k)): |
|
101 | if ds[k] not in b'anm' or (match and not match(k)): | |
102 | del c[k] |
|
102 | del c[k] | |
103 | return c |
|
103 | return c | |
104 |
|
104 | |||
105 |
|
105 | |||
106 | def _computeforwardmissing(a, b, match=None): |
|
106 | def _computeforwardmissing(a, b, match=None): | |
107 | """Computes which files are in b but not a. |
|
107 | """Computes which files are in b but not a. | |
108 | This is its own function so extensions can easily wrap this call to see what |
|
108 | This is its own function so extensions can easily wrap this call to see what | |
109 | files _forwardcopies is about to process. |
|
109 | files _forwardcopies is about to process. | |
110 | """ |
|
110 | """ | |
111 | ma = a.manifest() |
|
111 | ma = a.manifest() | |
112 | mb = b.manifest() |
|
112 | mb = b.manifest() | |
113 | return mb.filesnotin(ma, match=match) |
|
113 | return mb.filesnotin(ma, match=match) | |
114 |
|
114 | |||
115 |
|
115 | |||
116 | def usechangesetcentricalgo(repo): |
|
116 | def usechangesetcentricalgo(repo): | |
117 | """Checks if we should use changeset-centric copy algorithms""" |
|
117 | """Checks if we should use changeset-centric copy algorithms""" | |
118 | if repo.filecopiesmode == b'changeset-sidedata': |
|
118 | if repo.filecopiesmode == b'changeset-sidedata': | |
119 | return True |
|
119 | return True | |
120 | readfrom = repo.ui.config(b'experimental', b'copies.read-from') |
|
120 | readfrom = repo.ui.config(b'experimental', b'copies.read-from') | |
121 | changesetsource = (b'changeset-only', b'compatibility') |
|
121 | changesetsource = (b'changeset-only', b'compatibility') | |
122 | return readfrom in changesetsource |
|
122 | return readfrom in changesetsource | |
123 |
|
123 | |||
124 |
|
124 | |||
125 | def _committedforwardcopies(a, b, base, match): |
|
125 | def _committedforwardcopies(a, b, base, match): | |
126 | """Like _forwardcopies(), but b.rev() cannot be None (working copy)""" |
|
126 | """Like _forwardcopies(), but b.rev() cannot be None (working copy)""" | |
127 | # files might have to be traced back to the fctx parent of the last |
|
127 | # files might have to be traced back to the fctx parent of the last | |
128 | # one-side-only changeset, but not further back than that |
|
128 | # one-side-only changeset, but not further back than that | |
129 | repo = a._repo |
|
129 | repo = a._repo | |
130 |
|
130 | |||
131 | if usechangesetcentricalgo(repo): |
|
131 | if usechangesetcentricalgo(repo): | |
132 | return _changesetforwardcopies(a, b, match) |
|
132 | return _changesetforwardcopies(a, b, match) | |
133 |
|
133 | |||
134 | debug = repo.ui.debugflag and repo.ui.configbool(b'devel', b'debug.copies') |
|
134 | debug = repo.ui.debugflag and repo.ui.configbool(b'devel', b'debug.copies') | |
135 | dbg = repo.ui.debug |
|
135 | dbg = repo.ui.debug | |
136 | if debug: |
|
136 | if debug: | |
137 | dbg(b'debug.copies: looking into rename from %s to %s\n' % (a, b)) |
|
137 | dbg(b'debug.copies: looking into rename from %s to %s\n' % (a, b)) | |
138 | am = a.manifest() |
|
138 | am = a.manifest() | |
139 | basemf = None if base is None else base.manifest() |
|
139 | basemf = None if base is None else base.manifest() | |
140 |
|
140 | |||
141 | # find where new files came from |
|
141 | # find where new files came from | |
142 | # we currently don't try to find where old files went, too expensive |
|
142 | # we currently don't try to find where old files went, too expensive | |
143 | # this means we can miss a case like 'hg rm b; hg cp a b' |
|
143 | # this means we can miss a case like 'hg rm b; hg cp a b' | |
144 | cm = {} |
|
144 | cm = {} | |
145 |
|
145 | |||
146 | # Computing the forward missing is quite expensive on large manifests, since |
|
146 | # Computing the forward missing is quite expensive on large manifests, since | |
147 | # it compares the entire manifests. We can optimize it in the common use |
|
147 | # it compares the entire manifests. We can optimize it in the common use | |
148 | # case of computing what copies are in a commit versus its parent (like |
|
148 | # case of computing what copies are in a commit versus its parent (like | |
149 | # during a rebase or histedit). Note, we exclude merge commits from this |
|
149 | # during a rebase or histedit). Note, we exclude merge commits from this | |
150 | # optimization, since the ctx.files() for a merge commit is not correct for |
|
150 | # optimization, since the ctx.files() for a merge commit is not correct for | |
151 | # this comparison. |
|
151 | # this comparison. | |
152 | forwardmissingmatch = match |
|
152 | forwardmissingmatch = match | |
153 | if b.p1() == a and b.p2().node() == nullid: |
|
153 | if b.p1() == a and b.p2().node() == nullid: | |
154 | filesmatcher = matchmod.exact(b.files()) |
|
154 | filesmatcher = matchmod.exact(b.files()) | |
155 | forwardmissingmatch = matchmod.intersectmatchers(match, filesmatcher) |
|
155 | forwardmissingmatch = matchmod.intersectmatchers(match, filesmatcher) | |
156 | missing = _computeforwardmissing(a, b, match=forwardmissingmatch) |
|
156 | missing = _computeforwardmissing(a, b, match=forwardmissingmatch) | |
157 |
|
157 | |||
158 | ancestrycontext = a._repo.changelog.ancestors([b.rev()], inclusive=True) |
|
158 | ancestrycontext = a._repo.changelog.ancestors([b.rev()], inclusive=True) | |
159 |
|
159 | |||
160 | if debug: |
|
160 | if debug: | |
161 | dbg(b'debug.copies: missing files to search: %d\n' % len(missing)) |
|
161 | dbg(b'debug.copies: missing files to search: %d\n' % len(missing)) | |
162 |
|
162 | |||
163 | for f in sorted(missing): |
|
163 | for f in sorted(missing): | |
164 | if debug: |
|
164 | if debug: | |
165 | dbg(b'debug.copies: tracing file: %s\n' % f) |
|
165 | dbg(b'debug.copies: tracing file: %s\n' % f) | |
166 | fctx = b[f] |
|
166 | fctx = b[f] | |
167 | fctx._ancestrycontext = ancestrycontext |
|
167 | fctx._ancestrycontext = ancestrycontext | |
168 |
|
168 | |||
169 | if debug: |
|
169 | if debug: | |
170 | start = util.timer() |
|
170 | start = util.timer() | |
171 | opath = _tracefile(fctx, am, basemf) |
|
171 | opath = _tracefile(fctx, am, basemf) | |
172 | if opath: |
|
172 | if opath: | |
173 | if debug: |
|
173 | if debug: | |
174 | dbg(b'debug.copies: rename of: %s\n' % opath) |
|
174 | dbg(b'debug.copies: rename of: %s\n' % opath) | |
175 | cm[f] = opath |
|
175 | cm[f] = opath | |
176 | if debug: |
|
176 | if debug: | |
177 | dbg( |
|
177 | dbg( | |
178 | b'debug.copies: time: %f seconds\n' |
|
178 | b'debug.copies: time: %f seconds\n' | |
179 | % (util.timer() - start) |
|
179 | % (util.timer() - start) | |
180 | ) |
|
180 | ) | |
181 | return cm |
|
181 | return cm | |
182 |
|
182 | |||
183 |
|
183 | |||
184 | def _revinfo_getter(repo, match): |
|
184 | def _revinfo_getter(repo, match): | |
185 | """returns a function that returns the following data given a <rev>" |
|
185 | """returns a function that returns the following data given a <rev>" | |
186 |
|
186 | |||
187 | * p1: revision number of first parent |
|
187 | * p1: revision number of first parent | |
188 | * p2: revision number of first parent |
|
188 | * p2: revision number of first parent | |
189 | * changes: a ChangingFiles object |
|
189 | * changes: a ChangingFiles object | |
190 | """ |
|
190 | """ | |
191 | cl = repo.changelog |
|
191 | cl = repo.changelog | |
192 | parents = cl.parentrevs |
|
192 | parents = cl.parentrevs | |
193 | flags = cl.flags |
|
193 | flags = cl.flags | |
194 |
|
194 | |||
195 | HASCOPIESINFO = flagutil.REVIDX_HASCOPIESINFO |
|
195 | HASCOPIESINFO = flagutil.REVIDX_HASCOPIESINFO | |
196 |
|
196 | |||
197 | changelogrevision = cl.changelogrevision |
|
197 | changelogrevision = cl.changelogrevision | |
198 |
|
198 | |||
199 | if rustmod is not None: |
|
199 | if rustmod is not None: | |
200 |
|
200 | |||
201 | def revinfo(rev): |
|
201 | def revinfo(rev): | |
202 | p1, p2 = parents(rev) |
|
202 | p1, p2 = parents(rev) | |
203 | if flags(rev) & HASCOPIESINFO: |
|
203 | if flags(rev) & HASCOPIESINFO: | |
204 | raw = changelogrevision(rev)._sidedata.get(sidedatamod.SD_FILES) |
|
204 | raw = changelogrevision(rev)._sidedata.get(sidedatamod.SD_FILES) | |
205 | else: |
|
205 | else: | |
206 | raw = None |
|
206 | raw = None | |
207 | return (p1, p2, raw) |
|
207 | return (p1, p2, raw) | |
208 |
|
208 | |||
209 | else: |
|
209 | else: | |
210 |
|
210 | |||
211 | def revinfo(rev): |
|
211 | def revinfo(rev): | |
212 | p1, p2 = parents(rev) |
|
212 | p1, p2 = parents(rev) | |
213 | if flags(rev) & HASCOPIESINFO: |
|
213 | if flags(rev) & HASCOPIESINFO: | |
214 | changes = changelogrevision(rev).changes |
|
214 | changes = changelogrevision(rev).changes | |
215 | else: |
|
215 | else: | |
216 | changes = None |
|
216 | changes = None | |
217 | return (p1, p2, changes) |
|
217 | return (p1, p2, changes) | |
218 |
|
218 | |||
219 | return revinfo |
|
219 | return revinfo | |
220 |
|
220 | |||
221 |
|
221 | |||
222 | def cached_is_ancestor(is_ancestor): |
|
222 | def cached_is_ancestor(is_ancestor): | |
223 | """return a cached version of is_ancestor""" |
|
223 | """return a cached version of is_ancestor""" | |
224 | cache = {} |
|
224 | cache = {} | |
225 |
|
225 | |||
226 | def _is_ancestor(anc, desc): |
|
226 | def _is_ancestor(anc, desc): | |
227 | if anc > desc: |
|
227 | if anc > desc: | |
228 | return False |
|
228 | return False | |
229 | elif anc == desc: |
|
229 | elif anc == desc: | |
230 | return True |
|
230 | return True | |
231 | key = (anc, desc) |
|
231 | key = (anc, desc) | |
232 | ret = cache.get(key) |
|
232 | ret = cache.get(key) | |
233 | if ret is None: |
|
233 | if ret is None: | |
234 | ret = cache[key] = is_ancestor(anc, desc) |
|
234 | ret = cache[key] = is_ancestor(anc, desc) | |
235 | return ret |
|
235 | return ret | |
236 |
|
236 | |||
237 | return _is_ancestor |
|
237 | return _is_ancestor | |
238 |
|
238 | |||
239 |
|
239 | |||
240 | def _changesetforwardcopies(a, b, match): |
|
240 | def _changesetforwardcopies(a, b, match): | |
241 | if a.rev() in (nullrev, b.rev()): |
|
241 | if a.rev() in (nullrev, b.rev()): | |
242 | return {} |
|
242 | return {} | |
243 |
|
243 | |||
244 | repo = a.repo().unfiltered() |
|
244 | repo = a.repo().unfiltered() | |
245 | children = {} |
|
245 | children = {} | |
246 |
|
246 | |||
247 | cl = repo.changelog |
|
247 | cl = repo.changelog | |
248 | isancestor = cl.isancestorrev |
|
248 | isancestor = cl.isancestorrev | |
249 |
|
249 | |||
250 | # To track rename from "A" to B, we need to gather all parent β children |
|
250 | # To track rename from "A" to B, we need to gather all parent β children | |
251 | # edges that are contains in `::B` but not in `::A`. |
|
251 | # edges that are contains in `::B` but not in `::A`. | |
252 | # |
|
252 | # | |
253 | # |
|
253 | # | |
254 | # To do so, we need to gather all revisions exclusiveΒΉ to "B" (ieΒΉ: `::b - |
|
254 | # To do so, we need to gather all revisions exclusiveΒΉ to "B" (ieΒΉ: `::b - | |
255 | # ::a`) and also all the "roots point", ie the parents of the exclusive set |
|
255 | # ::a`) and also all the "roots point", ie the parents of the exclusive set | |
256 | # that belong to ::a. These are exactly all the revisions needed to express |
|
256 | # that belong to ::a. These are exactly all the revisions needed to express | |
257 | # the parent β children we need to combine. |
|
257 | # the parent β children we need to combine. | |
258 | # |
|
258 | # | |
259 | # [1] actually, we need to gather all the edges within `(::a)::b`, ie: |
|
259 | # [1] actually, we need to gather all the edges within `(::a)::b`, ie: | |
260 | # excluding paths that leads to roots that are not ancestors of `a`. We |
|
260 | # excluding paths that leads to roots that are not ancestors of `a`. We | |
261 | # keep this out of the explanation because it is hard enough without this special case.. |
|
261 | # keep this out of the explanation because it is hard enough without this special case.. | |
262 |
|
262 | |||
263 | parents = cl._uncheckedparentrevs |
|
263 | parents = cl._uncheckedparentrevs | |
264 | graph_roots = (nullrev, nullrev) |
|
264 | graph_roots = (nullrev, nullrev) | |
265 |
|
265 | |||
266 | ancestors = cl.ancestors([a.rev()], inclusive=True) |
|
266 | ancestors = cl.ancestors([a.rev()], inclusive=True) | |
267 | revs = cl.findmissingrevs(common=[a.rev()], heads=[b.rev()]) |
|
267 | revs = cl.findmissingrevs(common=[a.rev()], heads=[b.rev()]) | |
268 | roots = set() |
|
268 | roots = set() | |
269 | has_graph_roots = False |
|
269 | has_graph_roots = False | |
270 |
|
270 | |||
271 | # iterate over `only(B, A)` |
|
271 | # iterate over `only(B, A)` | |
272 | for r in revs: |
|
272 | for r in revs: | |
273 | ps = parents(r) |
|
273 | ps = parents(r) | |
274 | if ps == graph_roots: |
|
274 | if ps == graph_roots: | |
275 | has_graph_roots = True |
|
275 | has_graph_roots = True | |
276 | else: |
|
276 | else: | |
277 | p1, p2 = ps |
|
277 | p1, p2 = ps | |
278 |
|
278 | |||
279 | # find all the "root points" (see larger comment above) |
|
279 | # find all the "root points" (see larger comment above) | |
280 | if p1 != nullrev and p1 in ancestors: |
|
280 | if p1 != nullrev and p1 in ancestors: | |
281 | roots.add(p1) |
|
281 | roots.add(p1) | |
282 | if p2 != nullrev and p2 in ancestors: |
|
282 | if p2 != nullrev and p2 in ancestors: | |
283 | roots.add(p2) |
|
283 | roots.add(p2) | |
284 | if not roots: |
|
284 | if not roots: | |
285 | # no common revision to track copies from |
|
285 | # no common revision to track copies from | |
286 | return {} |
|
286 | return {} | |
287 | if has_graph_roots: |
|
287 | if has_graph_roots: | |
288 | # this deal with the special case mentionned in the [1] footnotes. We |
|
288 | # this deal with the special case mentionned in the [1] footnotes. We | |
289 | # must filter out revisions that leads to non-common graphroots. |
|
289 | # must filter out revisions that leads to non-common graphroots. | |
290 | roots = list(roots) |
|
290 | roots = list(roots) | |
291 | m = min(roots) |
|
291 | m = min(roots) | |
292 | h = [b.rev()] |
|
292 | h = [b.rev()] | |
293 | roots_to_head = cl.reachableroots(m, h, roots, includepath=True) |
|
293 | roots_to_head = cl.reachableroots(m, h, roots, includepath=True) | |
294 | roots_to_head = set(roots_to_head) |
|
294 | roots_to_head = set(roots_to_head) | |
295 | revs = [r for r in revs if r in roots_to_head] |
|
295 | revs = [r for r in revs if r in roots_to_head] | |
296 |
|
296 | |||
297 | if repo.filecopiesmode == b'changeset-sidedata': |
|
297 | if repo.filecopiesmode == b'changeset-sidedata': | |
298 | # When using side-data, we will process the edges "from" the children. |
|
298 | # When using side-data, we will process the edges "from" the children. | |
299 | # We iterate over the childre, gathering previous collected data for |
|
299 | # We iterate over the childre, gathering previous collected data for | |
300 | # the parents. Do know when the parents data is no longer necessary, we |
|
300 | # the parents. Do know when the parents data is no longer necessary, we | |
301 | # keep a counter of how many children each revision has. |
|
301 | # keep a counter of how many children each revision has. | |
302 | # |
|
302 | # | |
303 | # An interresting property of `children_count` is that it only contains |
|
303 | # An interresting property of `children_count` is that it only contains | |
304 | # revision that will be relevant for a edge of the graph. So if a |
|
304 | # revision that will be relevant for a edge of the graph. So if a | |
305 | # children has parent not in `children_count`, that edges should not be |
|
305 | # children has parent not in `children_count`, that edges should not be | |
306 | # processed. |
|
306 | # processed. | |
307 | children_count = dict((r, 0) for r in roots) |
|
307 | children_count = dict((r, 0) for r in roots) | |
308 | for r in revs: |
|
308 | for r in revs: | |
309 | for p in cl.parentrevs(r): |
|
309 | for p in cl.parentrevs(r): | |
310 | if p == nullrev: |
|
310 | if p == nullrev: | |
311 | continue |
|
311 | continue | |
312 | children_count[r] = 0 |
|
312 | children_count[r] = 0 | |
313 | if p in children_count: |
|
313 | if p in children_count: | |
314 | children_count[p] += 1 |
|
314 | children_count[p] += 1 | |
315 | revinfo = _revinfo_getter(repo, match) |
|
315 | revinfo = _revinfo_getter(repo, match) | |
316 | return _combine_changeset_copies( |
|
316 | return _combine_changeset_copies( | |
317 | revs, children_count, b.rev(), revinfo, match, isancestor |
|
317 | revs, children_count, b.rev(), revinfo, match, isancestor | |
318 | ) |
|
318 | ) | |
319 | else: |
|
319 | else: | |
320 | # When not using side-data, we will process the edges "from" the parent. |
|
320 | # When not using side-data, we will process the edges "from" the parent. | |
321 | # so we need a full mapping of the parent -> children relation. |
|
321 | # so we need a full mapping of the parent -> children relation. | |
322 | children = dict((r, []) for r in roots) |
|
322 | children = dict((r, []) for r in roots) | |
323 | for r in revs: |
|
323 | for r in revs: | |
324 | for p in cl.parentrevs(r): |
|
324 | for p in cl.parentrevs(r): | |
325 | if p == nullrev: |
|
325 | if p == nullrev: | |
326 | continue |
|
326 | continue | |
327 | children[r] = [] |
|
327 | children[r] = [] | |
328 | if p in children: |
|
328 | if p in children: | |
329 | children[p].append(r) |
|
329 | children[p].append(r) | |
330 | x = revs.pop() |
|
330 | x = revs.pop() | |
331 | assert x == b.rev() |
|
331 | assert x == b.rev() | |
332 | revs.extend(roots) |
|
332 | revs.extend(roots) | |
333 | revs.sort() |
|
333 | revs.sort() | |
334 |
|
334 | |||
335 | revinfo = _revinfo_getter_extra(repo) |
|
335 | revinfo = _revinfo_getter_extra(repo) | |
336 | return _combine_changeset_copies_extra( |
|
336 | return _combine_changeset_copies_extra( | |
337 | revs, children, b.rev(), revinfo, match, isancestor |
|
337 | revs, children, b.rev(), revinfo, match, isancestor | |
338 | ) |
|
338 | ) | |
339 |
|
339 | |||
340 |
|
340 | |||
341 | def _combine_changeset_copies( |
|
341 | def _combine_changeset_copies( | |
342 | revs, children_count, targetrev, revinfo, match, isancestor |
|
342 | revs, children_count, targetrev, revinfo, match, isancestor | |
343 | ): |
|
343 | ): | |
344 | """combine the copies information for each item of iterrevs |
|
344 | """combine the copies information for each item of iterrevs | |
345 |
|
345 | |||
346 | revs: sorted iterable of revision to visit |
|
346 | revs: sorted iterable of revision to visit | |
347 | children_count: a {parent: <number-of-relevant-children>} mapping. |
|
347 | children_count: a {parent: <number-of-relevant-children>} mapping. | |
348 | targetrev: the final copies destination revision (not in iterrevs) |
|
348 | targetrev: the final copies destination revision (not in iterrevs) | |
349 | revinfo(rev): a function that return (p1, p2, p1copies, p2copies, removed) |
|
349 | revinfo(rev): a function that return (p1, p2, p1copies, p2copies, removed) | |
350 | match: a matcher |
|
350 | match: a matcher | |
351 |
|
351 | |||
352 | It returns the aggregated copies information for `targetrev`. |
|
352 | It returns the aggregated copies information for `targetrev`. | |
353 | """ |
|
353 | """ | |
354 |
|
354 | |||
355 | alwaysmatch = match.always() |
|
355 | alwaysmatch = match.always() | |
356 |
|
356 | |||
357 | if rustmod is not None: |
|
357 | if rustmod is not None: | |
358 | final_copies = rustmod.combine_changeset_copies( |
|
358 | final_copies = rustmod.combine_changeset_copies( | |
359 | list(revs), children_count, targetrev, revinfo, isancestor |
|
359 | list(revs), children_count, targetrev, revinfo, isancestor | |
360 | ) |
|
360 | ) | |
361 | else: |
|
361 | else: | |
362 | isancestor = cached_is_ancestor(isancestor) |
|
362 | isancestor = cached_is_ancestor(isancestor) | |
363 |
|
363 | |||
364 | all_copies = {} |
|
364 | all_copies = {} | |
365 | # iterate over all the "children" side of copy tracing "edge" |
|
365 | # iterate over all the "children" side of copy tracing "edge" | |
366 | for current_rev in revs: |
|
366 | for current_rev in revs: | |
367 | p1, p2, changes = revinfo(current_rev) |
|
367 | p1, p2, changes = revinfo(current_rev) | |
368 | current_copies = None |
|
368 | current_copies = None | |
369 | # iterate over all parents to chain the existing data with the |
|
369 | # iterate over all parents to chain the existing data with the | |
370 | # data from the parent β child edge. |
|
370 | # data from the parent β child edge. | |
371 | for parent, parent_rev in ((1, p1), (2, p2)): |
|
371 | for parent, parent_rev in ((1, p1), (2, p2)): | |
372 | if parent_rev == nullrev: |
|
372 | if parent_rev == nullrev: | |
373 | continue |
|
373 | continue | |
374 | remaining_children = children_count.get(parent_rev) |
|
374 | remaining_children = children_count.get(parent_rev) | |
375 | if remaining_children is None: |
|
375 | if remaining_children is None: | |
376 | continue |
|
376 | continue | |
377 | remaining_children -= 1 |
|
377 | remaining_children -= 1 | |
378 | children_count[parent_rev] = remaining_children |
|
378 | children_count[parent_rev] = remaining_children | |
379 | if remaining_children: |
|
379 | if remaining_children: | |
380 | copies = all_copies.get(parent_rev, None) |
|
380 | copies = all_copies.get(parent_rev, None) | |
381 | else: |
|
381 | else: | |
382 | copies = all_copies.pop(parent_rev, None) |
|
382 | copies = all_copies.pop(parent_rev, None) | |
383 |
|
383 | |||
384 | if copies is None: |
|
384 | if copies is None: | |
385 | # this is a root |
|
385 | # this is a root | |
386 | copies = {} |
|
386 | copies = {} | |
387 |
|
387 | |||
388 | newcopies = copies |
|
388 | newcopies = copies | |
389 | # chain the data in the edge with the existing data |
|
389 | # chain the data in the edge with the existing data | |
390 | if changes is not None: |
|
390 | if changes is not None: | |
391 | childcopies = {} |
|
391 | childcopies = {} | |
392 | if parent == 1: |
|
392 | if parent == 1: | |
393 | childcopies = changes.copied_from_p1 |
|
393 | childcopies = changes.copied_from_p1 | |
394 | elif parent == 2: |
|
394 | elif parent == 2: | |
395 | childcopies = changes.copied_from_p2 |
|
395 | childcopies = changes.copied_from_p2 | |
396 |
|
396 | |||
397 | if childcopies: |
|
397 | if childcopies: | |
398 | newcopies = copies.copy() |
|
398 | newcopies = copies.copy() | |
399 | for dest, source in pycompat.iteritems(childcopies): |
|
399 | for dest, source in pycompat.iteritems(childcopies): | |
400 | prev = copies.get(source) |
|
400 | prev = copies.get(source) | |
401 | if prev is not None and prev[1] is not None: |
|
401 | if prev is not None and prev[1] is not None: | |
402 | source = prev[1] |
|
402 | source = prev[1] | |
403 | newcopies[dest] = (current_rev, source) |
|
403 | newcopies[dest] = (current_rev, source) | |
404 | assert newcopies is not copies |
|
404 | assert newcopies is not copies | |
405 | if changes.removed: |
|
405 | if changes.removed: | |
406 | if newcopies is copies: |
|
406 | if newcopies is copies: | |
407 | newcopies = copies.copy() |
|
407 | newcopies = copies.copy() | |
408 | for f in changes.removed: |
|
408 | for f in changes.removed: | |
409 | if f in newcopies: |
|
409 | if f in newcopies: | |
410 | if newcopies is copies: |
|
410 | if newcopies is copies: | |
411 | # copy on write to avoid affecting potential other |
|
411 | # copy on write to avoid affecting potential other | |
412 | # branches. when there are no other branches, this |
|
412 | # branches. when there are no other branches, this | |
413 | # could be avoided. |
|
413 | # could be avoided. | |
414 | newcopies = copies.copy() |
|
414 | newcopies = copies.copy() | |
415 | newcopies[f] = (current_rev, None) |
|
415 | newcopies[f] = (current_rev, None) | |
416 | # check potential need to combine the data from another parent (for |
|
416 | # check potential need to combine the data from another parent (for | |
417 | # that child). See comment below for details. |
|
417 | # that child). See comment below for details. | |
418 | if current_copies is None: |
|
418 | if current_copies is None: | |
419 | current_copies = newcopies |
|
419 | current_copies = newcopies | |
420 | elif current_copies is newcopies: |
|
420 | elif current_copies is newcopies: | |
421 | # nothing to merge: |
|
421 | # nothing to merge: | |
422 | pass |
|
422 | pass | |
423 | else: |
|
423 | else: | |
424 | # we are the second parent to work on c, we need to merge our |
|
424 | # we are the second parent to work on c, we need to merge our | |
425 | # work with the other. |
|
425 | # work with the other. | |
426 | # |
|
426 | # | |
427 | # In case of conflict, parent 1 take precedence over parent 2. |
|
427 | # In case of conflict, parent 1 take precedence over parent 2. | |
428 | # This is an arbitrary choice made anew when implementing |
|
428 | # This is an arbitrary choice made anew when implementing | |
429 | # changeset based copies. It was made without regards with |
|
429 | # changeset based copies. It was made without regards with | |
430 | # potential filelog related behavior. |
|
430 | # potential filelog related behavior. | |
431 | assert parent == 2 |
|
431 | assert parent == 2 | |
432 | current_copies = _merge_copies_dict( |
|
432 | current_copies = _merge_copies_dict( | |
433 | newcopies, current_copies, isancestor, changes |
|
433 | newcopies, current_copies, isancestor, changes | |
434 | ) |
|
434 | ) | |
435 | all_copies[current_rev] = current_copies |
|
435 | all_copies[current_rev] = current_copies | |
436 |
|
436 | |||
437 | # filter out internal details and return a {dest: source mapping} |
|
437 | # filter out internal details and return a {dest: source mapping} | |
438 | final_copies = {} |
|
438 | final_copies = {} | |
439 | for dest, (tt, source) in all_copies[targetrev].items(): |
|
439 | for dest, (tt, source) in all_copies[targetrev].items(): | |
440 | if source is not None: |
|
440 | if source is not None: | |
441 | final_copies[dest] = source |
|
441 | final_copies[dest] = source | |
442 | if not alwaysmatch: |
|
442 | if not alwaysmatch: | |
443 | for filename in list(final_copies.keys()): |
|
443 | for filename in list(final_copies.keys()): | |
444 | if not match(filename): |
|
444 | if not match(filename): | |
445 | del final_copies[filename] |
|
445 | del final_copies[filename] | |
446 | return final_copies |
|
446 | return final_copies | |
447 |
|
447 | |||
448 |
|
448 | |||
449 | # constant to decide which side to pick with _merge_copies_dict |
|
449 | # constant to decide which side to pick with _merge_copies_dict | |
450 | PICK_MINOR = 0 |
|
450 | PICK_MINOR = 0 | |
451 | PICK_MAJOR = 1 |
|
451 | PICK_MAJOR = 1 | |
452 | PICK_EITHER = 2 |
|
452 | PICK_EITHER = 2 | |
453 |
|
453 | |||
454 |
|
454 | |||
455 | def _merge_copies_dict(minor, major, isancestor, changes): |
|
455 | def _merge_copies_dict(minor, major, isancestor, changes): | |
456 | """merge two copies-mapping together, minor and major |
|
456 | """merge two copies-mapping together, minor and major | |
457 |
|
457 | |||
458 | In case of conflict, value from "major" will be picked. |
|
458 | In case of conflict, value from "major" will be picked. | |
459 |
|
459 | |||
460 | - `isancestors(low_rev, high_rev)`: callable return True if `low_rev` is an |
|
460 | - `isancestors(low_rev, high_rev)`: callable return True if `low_rev` is an | |
461 | ancestors of `high_rev`, |
|
461 | ancestors of `high_rev`, | |
462 |
|
462 | |||
463 | - `ismerged(path)`: callable return True if `path` have been merged in the |
|
463 | - `ismerged(path)`: callable return True if `path` have been merged in the | |
464 | current revision, |
|
464 | current revision, | |
465 |
|
465 | |||
466 | return the resulting dict (in practice, the "minor" object, updated) |
|
466 | return the resulting dict (in practice, the "minor" object, updated) | |
467 | """ |
|
467 | """ | |
468 | for dest, value in major.items(): |
|
468 | for dest, value in major.items(): | |
469 | other = minor.get(dest) |
|
469 | other = minor.get(dest) | |
470 | if other is None: |
|
470 | if other is None: | |
471 | minor[dest] = value |
|
471 | minor[dest] = value | |
472 | else: |
|
472 | else: | |
473 | pick = _compare_values(changes, isancestor, dest, other, value) |
|
473 | pick = _compare_values(changes, isancestor, dest, other, value) | |
474 | if pick == PICK_MAJOR: |
|
474 | if pick == PICK_MAJOR: | |
475 | minor[dest] = value |
|
475 | minor[dest] = value | |
476 | return minor |
|
476 | return minor | |
477 |
|
477 | |||
478 |
|
478 | |||
479 |
def _compare_values(changes, isancestor, dest, o |
|
479 | def _compare_values(changes, isancestor, dest, minor, major): | |
480 | """compare two value within a _merge_copies_dict loop iteration""" |
|
480 | """compare two value within a _merge_copies_dict loop iteration""" | |
481 |
|
|
481 | major_tt = major[0] | |
482 |
o |
|
482 | minor_tt = minor[0] | |
483 |
|
483 | |||
484 |
if |
|
484 | if major[1] == minor[1]: | |
485 | return PICK_EITHER |
|
485 | return PICK_EITHER | |
486 | # content from "major" wins, unless it is older |
|
486 | # content from "major" wins, unless it is older | |
487 | # than the branch point or there is a merge |
|
487 | # than the branch point or there is a merge | |
488 |
if |
|
488 | if major_tt == minor_tt: | |
489 | return PICK_MAJOR |
|
489 | return PICK_MAJOR | |
490 |
elif changes is not None and |
|
490 | elif changes is not None and major[1] is None and dest in changes.salvaged: | |
491 | return PICK_MINOR |
|
491 | return PICK_MINOR | |
492 |
elif changes is not None and o |
|
492 | elif changes is not None and minor[1] is None and dest in changes.salvaged: | |
493 | return PICK_MAJOR |
|
493 | return PICK_MAJOR | |
494 | elif changes is not None and dest in changes.merged: |
|
494 | elif changes is not None and dest in changes.merged: | |
495 | return PICK_MAJOR |
|
495 | return PICK_MAJOR | |
496 |
elif not isancestor( |
|
496 | elif not isancestor(major_tt, minor_tt): | |
497 |
if |
|
497 | if major[1] is not None: | |
498 | return PICK_MAJOR |
|
498 | return PICK_MAJOR | |
499 |
elif isancestor(o |
|
499 | elif isancestor(minor_tt, major_tt): | |
500 | return PICK_MAJOR |
|
500 | return PICK_MAJOR | |
501 | return PICK_MINOR |
|
501 | return PICK_MINOR | |
502 |
|
502 | |||
503 |
|
503 | |||
504 | def _revinfo_getter_extra(repo): |
|
504 | def _revinfo_getter_extra(repo): | |
505 | """return a function that return multiple data given a <rev>"i |
|
505 | """return a function that return multiple data given a <rev>"i | |
506 |
|
506 | |||
507 | * p1: revision number of first parent |
|
507 | * p1: revision number of first parent | |
508 | * p2: revision number of first parent |
|
508 | * p2: revision number of first parent | |
509 | * p1copies: mapping of copies from p1 |
|
509 | * p1copies: mapping of copies from p1 | |
510 | * p2copies: mapping of copies from p2 |
|
510 | * p2copies: mapping of copies from p2 | |
511 | * removed: a list of removed files |
|
511 | * removed: a list of removed files | |
512 | * ismerged: a callback to know if file was merged in that revision |
|
512 | * ismerged: a callback to know if file was merged in that revision | |
513 | """ |
|
513 | """ | |
514 | cl = repo.changelog |
|
514 | cl = repo.changelog | |
515 | parents = cl.parentrevs |
|
515 | parents = cl.parentrevs | |
516 |
|
516 | |||
517 | def get_ismerged(rev): |
|
517 | def get_ismerged(rev): | |
518 | ctx = repo[rev] |
|
518 | ctx = repo[rev] | |
519 |
|
519 | |||
520 | def ismerged(path): |
|
520 | def ismerged(path): | |
521 | if path not in ctx.files(): |
|
521 | if path not in ctx.files(): | |
522 | return False |
|
522 | return False | |
523 | fctx = ctx[path] |
|
523 | fctx = ctx[path] | |
524 | parents = fctx._filelog.parents(fctx._filenode) |
|
524 | parents = fctx._filelog.parents(fctx._filenode) | |
525 | nb_parents = 0 |
|
525 | nb_parents = 0 | |
526 | for n in parents: |
|
526 | for n in parents: | |
527 | if n != nullid: |
|
527 | if n != nullid: | |
528 | nb_parents += 1 |
|
528 | nb_parents += 1 | |
529 | return nb_parents >= 2 |
|
529 | return nb_parents >= 2 | |
530 |
|
530 | |||
531 | return ismerged |
|
531 | return ismerged | |
532 |
|
532 | |||
533 | def revinfo(rev): |
|
533 | def revinfo(rev): | |
534 | p1, p2 = parents(rev) |
|
534 | p1, p2 = parents(rev) | |
535 | ctx = repo[rev] |
|
535 | ctx = repo[rev] | |
536 | p1copies, p2copies = ctx._copies |
|
536 | p1copies, p2copies = ctx._copies | |
537 | removed = ctx.filesremoved() |
|
537 | removed = ctx.filesremoved() | |
538 | return p1, p2, p1copies, p2copies, removed, get_ismerged(rev) |
|
538 | return p1, p2, p1copies, p2copies, removed, get_ismerged(rev) | |
539 |
|
539 | |||
540 | return revinfo |
|
540 | return revinfo | |
541 |
|
541 | |||
542 |
|
542 | |||
543 | def _combine_changeset_copies_extra( |
|
543 | def _combine_changeset_copies_extra( | |
544 | revs, children, targetrev, revinfo, match, isancestor |
|
544 | revs, children, targetrev, revinfo, match, isancestor | |
545 | ): |
|
545 | ): | |
546 | """version of `_combine_changeset_copies` that works with the Google |
|
546 | """version of `_combine_changeset_copies` that works with the Google | |
547 | specific "extra" based storage for copy information""" |
|
547 | specific "extra" based storage for copy information""" | |
548 | all_copies = {} |
|
548 | all_copies = {} | |
549 | alwaysmatch = match.always() |
|
549 | alwaysmatch = match.always() | |
550 | for r in revs: |
|
550 | for r in revs: | |
551 | copies = all_copies.pop(r, None) |
|
551 | copies = all_copies.pop(r, None) | |
552 | if copies is None: |
|
552 | if copies is None: | |
553 | # this is a root |
|
553 | # this is a root | |
554 | copies = {} |
|
554 | copies = {} | |
555 | for i, c in enumerate(children[r]): |
|
555 | for i, c in enumerate(children[r]): | |
556 | p1, p2, p1copies, p2copies, removed, ismerged = revinfo(c) |
|
556 | p1, p2, p1copies, p2copies, removed, ismerged = revinfo(c) | |
557 | if r == p1: |
|
557 | if r == p1: | |
558 | parent = 1 |
|
558 | parent = 1 | |
559 | childcopies = p1copies |
|
559 | childcopies = p1copies | |
560 | else: |
|
560 | else: | |
561 | assert r == p2 |
|
561 | assert r == p2 | |
562 | parent = 2 |
|
562 | parent = 2 | |
563 | childcopies = p2copies |
|
563 | childcopies = p2copies | |
564 | if not alwaysmatch: |
|
564 | if not alwaysmatch: | |
565 | childcopies = { |
|
565 | childcopies = { | |
566 | dst: src for dst, src in childcopies.items() if match(dst) |
|
566 | dst: src for dst, src in childcopies.items() if match(dst) | |
567 | } |
|
567 | } | |
568 | newcopies = copies |
|
568 | newcopies = copies | |
569 | if childcopies: |
|
569 | if childcopies: | |
570 | newcopies = copies.copy() |
|
570 | newcopies = copies.copy() | |
571 | for dest, source in pycompat.iteritems(childcopies): |
|
571 | for dest, source in pycompat.iteritems(childcopies): | |
572 | prev = copies.get(source) |
|
572 | prev = copies.get(source) | |
573 | if prev is not None and prev[1] is not None: |
|
573 | if prev is not None and prev[1] is not None: | |
574 | source = prev[1] |
|
574 | source = prev[1] | |
575 | newcopies[dest] = (c, source) |
|
575 | newcopies[dest] = (c, source) | |
576 | assert newcopies is not copies |
|
576 | assert newcopies is not copies | |
577 | for f in removed: |
|
577 | for f in removed: | |
578 | if f in newcopies: |
|
578 | if f in newcopies: | |
579 | if newcopies is copies: |
|
579 | if newcopies is copies: | |
580 | # copy on write to avoid affecting potential other |
|
580 | # copy on write to avoid affecting potential other | |
581 | # branches. when there are no other branches, this |
|
581 | # branches. when there are no other branches, this | |
582 | # could be avoided. |
|
582 | # could be avoided. | |
583 | newcopies = copies.copy() |
|
583 | newcopies = copies.copy() | |
584 | newcopies[f] = (c, None) |
|
584 | newcopies[f] = (c, None) | |
585 | othercopies = all_copies.get(c) |
|
585 | othercopies = all_copies.get(c) | |
586 | if othercopies is None: |
|
586 | if othercopies is None: | |
587 | all_copies[c] = newcopies |
|
587 | all_copies[c] = newcopies | |
588 | else: |
|
588 | else: | |
589 | # we are the second parent to work on c, we need to merge our |
|
589 | # we are the second parent to work on c, we need to merge our | |
590 | # work with the other. |
|
590 | # work with the other. | |
591 | # |
|
591 | # | |
592 | # In case of conflict, parent 1 take precedence over parent 2. |
|
592 | # In case of conflict, parent 1 take precedence over parent 2. | |
593 | # This is an arbitrary choice made anew when implementing |
|
593 | # This is an arbitrary choice made anew when implementing | |
594 | # changeset based copies. It was made without regards with |
|
594 | # changeset based copies. It was made without regards with | |
595 | # potential filelog related behavior. |
|
595 | # potential filelog related behavior. | |
596 | if parent == 1: |
|
596 | if parent == 1: | |
597 | _merge_copies_dict_extra( |
|
597 | _merge_copies_dict_extra( | |
598 | othercopies, newcopies, isancestor, ismerged |
|
598 | othercopies, newcopies, isancestor, ismerged | |
599 | ) |
|
599 | ) | |
600 | else: |
|
600 | else: | |
601 | _merge_copies_dict_extra( |
|
601 | _merge_copies_dict_extra( | |
602 | newcopies, othercopies, isancestor, ismerged |
|
602 | newcopies, othercopies, isancestor, ismerged | |
603 | ) |
|
603 | ) | |
604 | all_copies[c] = newcopies |
|
604 | all_copies[c] = newcopies | |
605 |
|
605 | |||
606 | final_copies = {} |
|
606 | final_copies = {} | |
607 | for dest, (tt, source) in all_copies[targetrev].items(): |
|
607 | for dest, (tt, source) in all_copies[targetrev].items(): | |
608 | if source is not None: |
|
608 | if source is not None: | |
609 | final_copies[dest] = source |
|
609 | final_copies[dest] = source | |
610 | return final_copies |
|
610 | return final_copies | |
611 |
|
611 | |||
612 |
|
612 | |||
613 | def _merge_copies_dict_extra(minor, major, isancestor, ismerged): |
|
613 | def _merge_copies_dict_extra(minor, major, isancestor, ismerged): | |
614 | """version of `_merge_copies_dict` that works with the Google |
|
614 | """version of `_merge_copies_dict` that works with the Google | |
615 | specific "extra" based storage for copy information""" |
|
615 | specific "extra" based storage for copy information""" | |
616 | for dest, value in major.items(): |
|
616 | for dest, value in major.items(): | |
617 | other = minor.get(dest) |
|
617 | other = minor.get(dest) | |
618 | if other is None: |
|
618 | if other is None: | |
619 | minor[dest] = value |
|
619 | minor[dest] = value | |
620 | else: |
|
620 | else: | |
621 | new_tt = value[0] |
|
621 | new_tt = value[0] | |
622 | other_tt = other[0] |
|
622 | other_tt = other[0] | |
623 | if value[1] == other[1]: |
|
623 | if value[1] == other[1]: | |
624 | continue |
|
624 | continue | |
625 | # content from "major" wins, unless it is older |
|
625 | # content from "major" wins, unless it is older | |
626 | # than the branch point or there is a merge |
|
626 | # than the branch point or there is a merge | |
627 | if ( |
|
627 | if ( | |
628 | new_tt == other_tt |
|
628 | new_tt == other_tt | |
629 | or not isancestor(new_tt, other_tt) |
|
629 | or not isancestor(new_tt, other_tt) | |
630 | or ismerged(dest) |
|
630 | or ismerged(dest) | |
631 | ): |
|
631 | ): | |
632 | minor[dest] = value |
|
632 | minor[dest] = value | |
633 |
|
633 | |||
634 |
|
634 | |||
635 | def _forwardcopies(a, b, base=None, match=None): |
|
635 | def _forwardcopies(a, b, base=None, match=None): | |
636 | """find {dst@b: src@a} copy mapping where a is an ancestor of b""" |
|
636 | """find {dst@b: src@a} copy mapping where a is an ancestor of b""" | |
637 |
|
637 | |||
638 | if base is None: |
|
638 | if base is None: | |
639 | base = a |
|
639 | base = a | |
640 | match = a.repo().narrowmatch(match) |
|
640 | match = a.repo().narrowmatch(match) | |
641 | # check for working copy |
|
641 | # check for working copy | |
642 | if b.rev() is None: |
|
642 | if b.rev() is None: | |
643 | cm = _committedforwardcopies(a, b.p1(), base, match) |
|
643 | cm = _committedforwardcopies(a, b.p1(), base, match) | |
644 | # combine copies from dirstate if necessary |
|
644 | # combine copies from dirstate if necessary | |
645 | copies = _chain(cm, _dirstatecopies(b._repo, match)) |
|
645 | copies = _chain(cm, _dirstatecopies(b._repo, match)) | |
646 | else: |
|
646 | else: | |
647 | copies = _committedforwardcopies(a, b, base, match) |
|
647 | copies = _committedforwardcopies(a, b, base, match) | |
648 | return copies |
|
648 | return copies | |
649 |
|
649 | |||
650 |
|
650 | |||
651 | def _backwardrenames(a, b, match): |
|
651 | def _backwardrenames(a, b, match): | |
652 | if a._repo.ui.config(b'experimental', b'copytrace') == b'off': |
|
652 | if a._repo.ui.config(b'experimental', b'copytrace') == b'off': | |
653 | return {} |
|
653 | return {} | |
654 |
|
654 | |||
655 | # Even though we're not taking copies into account, 1:n rename situations |
|
655 | # Even though we're not taking copies into account, 1:n rename situations | |
656 | # can still exist (e.g. hg cp a b; hg mv a c). In those cases we |
|
656 | # can still exist (e.g. hg cp a b; hg mv a c). In those cases we | |
657 | # arbitrarily pick one of the renames. |
|
657 | # arbitrarily pick one of the renames. | |
658 | # We don't want to pass in "match" here, since that would filter |
|
658 | # We don't want to pass in "match" here, since that would filter | |
659 | # the destination by it. Since we're reversing the copies, we want |
|
659 | # the destination by it. Since we're reversing the copies, we want | |
660 | # to filter the source instead. |
|
660 | # to filter the source instead. | |
661 | f = _forwardcopies(b, a) |
|
661 | f = _forwardcopies(b, a) | |
662 | r = {} |
|
662 | r = {} | |
663 | for k, v in sorted(pycompat.iteritems(f)): |
|
663 | for k, v in sorted(pycompat.iteritems(f)): | |
664 | if match and not match(v): |
|
664 | if match and not match(v): | |
665 | continue |
|
665 | continue | |
666 | # remove copies |
|
666 | # remove copies | |
667 | if v in a: |
|
667 | if v in a: | |
668 | continue |
|
668 | continue | |
669 | r[v] = k |
|
669 | r[v] = k | |
670 | return r |
|
670 | return r | |
671 |
|
671 | |||
672 |
|
672 | |||
673 | def pathcopies(x, y, match=None): |
|
673 | def pathcopies(x, y, match=None): | |
674 | """find {dst@y: src@x} copy mapping for directed compare""" |
|
674 | """find {dst@y: src@x} copy mapping for directed compare""" | |
675 | repo = x._repo |
|
675 | repo = x._repo | |
676 | debug = repo.ui.debugflag and repo.ui.configbool(b'devel', b'debug.copies') |
|
676 | debug = repo.ui.debugflag and repo.ui.configbool(b'devel', b'debug.copies') | |
677 | if debug: |
|
677 | if debug: | |
678 | repo.ui.debug( |
|
678 | repo.ui.debug( | |
679 | b'debug.copies: searching copies from %s to %s\n' % (x, y) |
|
679 | b'debug.copies: searching copies from %s to %s\n' % (x, y) | |
680 | ) |
|
680 | ) | |
681 | if x == y or not x or not y: |
|
681 | if x == y or not x or not y: | |
682 | return {} |
|
682 | return {} | |
683 | if y.rev() is None and x == y.p1(): |
|
683 | if y.rev() is None and x == y.p1(): | |
684 | if debug: |
|
684 | if debug: | |
685 | repo.ui.debug(b'debug.copies: search mode: dirstate\n') |
|
685 | repo.ui.debug(b'debug.copies: search mode: dirstate\n') | |
686 | # short-circuit to avoid issues with merge states |
|
686 | # short-circuit to avoid issues with merge states | |
687 | return _dirstatecopies(repo, match) |
|
687 | return _dirstatecopies(repo, match) | |
688 | a = y.ancestor(x) |
|
688 | a = y.ancestor(x) | |
689 | if a == x: |
|
689 | if a == x: | |
690 | if debug: |
|
690 | if debug: | |
691 | repo.ui.debug(b'debug.copies: search mode: forward\n') |
|
691 | repo.ui.debug(b'debug.copies: search mode: forward\n') | |
692 | copies = _forwardcopies(x, y, match=match) |
|
692 | copies = _forwardcopies(x, y, match=match) | |
693 | elif a == y: |
|
693 | elif a == y: | |
694 | if debug: |
|
694 | if debug: | |
695 | repo.ui.debug(b'debug.copies: search mode: backward\n') |
|
695 | repo.ui.debug(b'debug.copies: search mode: backward\n') | |
696 | copies = _backwardrenames(x, y, match=match) |
|
696 | copies = _backwardrenames(x, y, match=match) | |
697 | else: |
|
697 | else: | |
698 | if debug: |
|
698 | if debug: | |
699 | repo.ui.debug(b'debug.copies: search mode: combined\n') |
|
699 | repo.ui.debug(b'debug.copies: search mode: combined\n') | |
700 | base = None |
|
700 | base = None | |
701 | if a.rev() != nullrev: |
|
701 | if a.rev() != nullrev: | |
702 | base = x |
|
702 | base = x | |
703 | copies = _chain( |
|
703 | copies = _chain( | |
704 | _backwardrenames(x, a, match=match), |
|
704 | _backwardrenames(x, a, match=match), | |
705 | _forwardcopies(a, y, base, match=match), |
|
705 | _forwardcopies(a, y, base, match=match), | |
706 | ) |
|
706 | ) | |
707 | _filter(x, y, copies) |
|
707 | _filter(x, y, copies) | |
708 | return copies |
|
708 | return copies | |
709 |
|
709 | |||
710 |
|
710 | |||
711 | def mergecopies(repo, c1, c2, base): |
|
711 | def mergecopies(repo, c1, c2, base): | |
712 | """ |
|
712 | """ | |
713 | Finds moves and copies between context c1 and c2 that are relevant for |
|
713 | Finds moves and copies between context c1 and c2 that are relevant for | |
714 | merging. 'base' will be used as the merge base. |
|
714 | merging. 'base' will be used as the merge base. | |
715 |
|
715 | |||
716 | Copytracing is used in commands like rebase, merge, unshelve, etc to merge |
|
716 | Copytracing is used in commands like rebase, merge, unshelve, etc to merge | |
717 | files that were moved/ copied in one merge parent and modified in another. |
|
717 | files that were moved/ copied in one merge parent and modified in another. | |
718 | For example: |
|
718 | For example: | |
719 |
|
719 | |||
720 | o ---> 4 another commit |
|
720 | o ---> 4 another commit | |
721 | | |
|
721 | | | |
722 | | o ---> 3 commit that modifies a.txt |
|
722 | | o ---> 3 commit that modifies a.txt | |
723 | | / |
|
723 | | / | |
724 | o / ---> 2 commit that moves a.txt to b.txt |
|
724 | o / ---> 2 commit that moves a.txt to b.txt | |
725 | |/ |
|
725 | |/ | |
726 | o ---> 1 merge base |
|
726 | o ---> 1 merge base | |
727 |
|
727 | |||
728 | If we try to rebase revision 3 on revision 4, since there is no a.txt in |
|
728 | If we try to rebase revision 3 on revision 4, since there is no a.txt in | |
729 | revision 4, and if user have copytrace disabled, we prints the following |
|
729 | revision 4, and if user have copytrace disabled, we prints the following | |
730 | message: |
|
730 | message: | |
731 |
|
731 | |||
732 | ```other changed <file> which local deleted``` |
|
732 | ```other changed <file> which local deleted``` | |
733 |
|
733 | |||
734 | Returns a tuple where: |
|
734 | Returns a tuple where: | |
735 |
|
735 | |||
736 | "branch_copies" an instance of branch_copies. |
|
736 | "branch_copies" an instance of branch_copies. | |
737 |
|
737 | |||
738 | "diverge" is a mapping of source name -> list of destination names |
|
738 | "diverge" is a mapping of source name -> list of destination names | |
739 | for divergent renames. |
|
739 | for divergent renames. | |
740 |
|
740 | |||
741 | This function calls different copytracing algorithms based on config. |
|
741 | This function calls different copytracing algorithms based on config. | |
742 | """ |
|
742 | """ | |
743 | # avoid silly behavior for update from empty dir |
|
743 | # avoid silly behavior for update from empty dir | |
744 | if not c1 or not c2 or c1 == c2: |
|
744 | if not c1 or not c2 or c1 == c2: | |
745 | return branch_copies(), branch_copies(), {} |
|
745 | return branch_copies(), branch_copies(), {} | |
746 |
|
746 | |||
747 | narrowmatch = c1.repo().narrowmatch() |
|
747 | narrowmatch = c1.repo().narrowmatch() | |
748 |
|
748 | |||
749 | # avoid silly behavior for parent -> working dir |
|
749 | # avoid silly behavior for parent -> working dir | |
750 | if c2.node() is None and c1.node() == repo.dirstate.p1(): |
|
750 | if c2.node() is None and c1.node() == repo.dirstate.p1(): | |
751 | return ( |
|
751 | return ( | |
752 | branch_copies(_dirstatecopies(repo, narrowmatch)), |
|
752 | branch_copies(_dirstatecopies(repo, narrowmatch)), | |
753 | branch_copies(), |
|
753 | branch_copies(), | |
754 | {}, |
|
754 | {}, | |
755 | ) |
|
755 | ) | |
756 |
|
756 | |||
757 | copytracing = repo.ui.config(b'experimental', b'copytrace') |
|
757 | copytracing = repo.ui.config(b'experimental', b'copytrace') | |
758 | if stringutil.parsebool(copytracing) is False: |
|
758 | if stringutil.parsebool(copytracing) is False: | |
759 | # stringutil.parsebool() returns None when it is unable to parse the |
|
759 | # stringutil.parsebool() returns None when it is unable to parse the | |
760 | # value, so we should rely on making sure copytracing is on such cases |
|
760 | # value, so we should rely on making sure copytracing is on such cases | |
761 | return branch_copies(), branch_copies(), {} |
|
761 | return branch_copies(), branch_copies(), {} | |
762 |
|
762 | |||
763 | if usechangesetcentricalgo(repo): |
|
763 | if usechangesetcentricalgo(repo): | |
764 | # The heuristics don't make sense when we need changeset-centric algos |
|
764 | # The heuristics don't make sense when we need changeset-centric algos | |
765 | return _fullcopytracing(repo, c1, c2, base) |
|
765 | return _fullcopytracing(repo, c1, c2, base) | |
766 |
|
766 | |||
767 | # Copy trace disabling is explicitly below the node == p1 logic above |
|
767 | # Copy trace disabling is explicitly below the node == p1 logic above | |
768 | # because the logic above is required for a simple copy to be kept across a |
|
768 | # because the logic above is required for a simple copy to be kept across a | |
769 | # rebase. |
|
769 | # rebase. | |
770 | if copytracing == b'heuristics': |
|
770 | if copytracing == b'heuristics': | |
771 | # Do full copytracing if only non-public revisions are involved as |
|
771 | # Do full copytracing if only non-public revisions are involved as | |
772 | # that will be fast enough and will also cover the copies which could |
|
772 | # that will be fast enough and will also cover the copies which could | |
773 | # be missed by heuristics |
|
773 | # be missed by heuristics | |
774 | if _isfullcopytraceable(repo, c1, base): |
|
774 | if _isfullcopytraceable(repo, c1, base): | |
775 | return _fullcopytracing(repo, c1, c2, base) |
|
775 | return _fullcopytracing(repo, c1, c2, base) | |
776 | return _heuristicscopytracing(repo, c1, c2, base) |
|
776 | return _heuristicscopytracing(repo, c1, c2, base) | |
777 | else: |
|
777 | else: | |
778 | return _fullcopytracing(repo, c1, c2, base) |
|
778 | return _fullcopytracing(repo, c1, c2, base) | |
779 |
|
779 | |||
780 |
|
780 | |||
781 | def _isfullcopytraceable(repo, c1, base): |
|
781 | def _isfullcopytraceable(repo, c1, base): | |
782 | """Checks that if base, source and destination are all no-public branches, |
|
782 | """Checks that if base, source and destination are all no-public branches, | |
783 | if yes let's use the full copytrace algorithm for increased capabilities |
|
783 | if yes let's use the full copytrace algorithm for increased capabilities | |
784 | since it will be fast enough. |
|
784 | since it will be fast enough. | |
785 |
|
785 | |||
786 | `experimental.copytrace.sourcecommitlimit` can be used to set a limit for |
|
786 | `experimental.copytrace.sourcecommitlimit` can be used to set a limit for | |
787 | number of changesets from c1 to base such that if number of changesets are |
|
787 | number of changesets from c1 to base such that if number of changesets are | |
788 | more than the limit, full copytracing algorithm won't be used. |
|
788 | more than the limit, full copytracing algorithm won't be used. | |
789 | """ |
|
789 | """ | |
790 | if c1.rev() is None: |
|
790 | if c1.rev() is None: | |
791 | c1 = c1.p1() |
|
791 | c1 = c1.p1() | |
792 | if c1.mutable() and base.mutable(): |
|
792 | if c1.mutable() and base.mutable(): | |
793 | sourcecommitlimit = repo.ui.configint( |
|
793 | sourcecommitlimit = repo.ui.configint( | |
794 | b'experimental', b'copytrace.sourcecommitlimit' |
|
794 | b'experimental', b'copytrace.sourcecommitlimit' | |
795 | ) |
|
795 | ) | |
796 | commits = len(repo.revs(b'%d::%d', base.rev(), c1.rev())) |
|
796 | commits = len(repo.revs(b'%d::%d', base.rev(), c1.rev())) | |
797 | return commits < sourcecommitlimit |
|
797 | return commits < sourcecommitlimit | |
798 | return False |
|
798 | return False | |
799 |
|
799 | |||
800 |
|
800 | |||
801 | def _checksinglesidecopies( |
|
801 | def _checksinglesidecopies( | |
802 | src, dsts1, m1, m2, mb, c2, base, copy, renamedelete |
|
802 | src, dsts1, m1, m2, mb, c2, base, copy, renamedelete | |
803 | ): |
|
803 | ): | |
804 | if src not in m2: |
|
804 | if src not in m2: | |
805 | # deleted on side 2 |
|
805 | # deleted on side 2 | |
806 | if src not in m1: |
|
806 | if src not in m1: | |
807 | # renamed on side 1, deleted on side 2 |
|
807 | # renamed on side 1, deleted on side 2 | |
808 | renamedelete[src] = dsts1 |
|
808 | renamedelete[src] = dsts1 | |
809 | elif src not in mb: |
|
809 | elif src not in mb: | |
810 | # Work around the "short-circuit to avoid issues with merge states" |
|
810 | # Work around the "short-circuit to avoid issues with merge states" | |
811 | # thing in pathcopies(): pathcopies(x, y) can return a copy where the |
|
811 | # thing in pathcopies(): pathcopies(x, y) can return a copy where the | |
812 | # destination doesn't exist in y. |
|
812 | # destination doesn't exist in y. | |
813 | pass |
|
813 | pass | |
814 | elif mb[src] != m2[src] and not _related(c2[src], base[src]): |
|
814 | elif mb[src] != m2[src] and not _related(c2[src], base[src]): | |
815 | return |
|
815 | return | |
816 | elif mb[src] != m2[src] or mb.flags(src) != m2.flags(src): |
|
816 | elif mb[src] != m2[src] or mb.flags(src) != m2.flags(src): | |
817 | # modified on side 2 |
|
817 | # modified on side 2 | |
818 | for dst in dsts1: |
|
818 | for dst in dsts1: | |
819 | copy[dst] = src |
|
819 | copy[dst] = src | |
820 |
|
820 | |||
821 |
|
821 | |||
822 | class branch_copies(object): |
|
822 | class branch_copies(object): | |
823 | """Information about copies made on one side of a merge/graft. |
|
823 | """Information about copies made on one side of a merge/graft. | |
824 |
|
824 | |||
825 | "copy" is a mapping from destination name -> source name, |
|
825 | "copy" is a mapping from destination name -> source name, | |
826 | where source is in c1 and destination is in c2 or vice-versa. |
|
826 | where source is in c1 and destination is in c2 or vice-versa. | |
827 |
|
827 | |||
828 | "movewithdir" is a mapping from source name -> destination name, |
|
828 | "movewithdir" is a mapping from source name -> destination name, | |
829 | where the file at source present in one context but not the other |
|
829 | where the file at source present in one context but not the other | |
830 | needs to be moved to destination by the merge process, because the |
|
830 | needs to be moved to destination by the merge process, because the | |
831 | other context moved the directory it is in. |
|
831 | other context moved the directory it is in. | |
832 |
|
832 | |||
833 | "renamedelete" is a mapping of source name -> list of destination |
|
833 | "renamedelete" is a mapping of source name -> list of destination | |
834 | names for files deleted in c1 that were renamed in c2 or vice-versa. |
|
834 | names for files deleted in c1 that were renamed in c2 or vice-versa. | |
835 |
|
835 | |||
836 | "dirmove" is a mapping of detected source dir -> destination dir renames. |
|
836 | "dirmove" is a mapping of detected source dir -> destination dir renames. | |
837 | This is needed for handling changes to new files previously grafted into |
|
837 | This is needed for handling changes to new files previously grafted into | |
838 | renamed directories. |
|
838 | renamed directories. | |
839 | """ |
|
839 | """ | |
840 |
|
840 | |||
841 | def __init__( |
|
841 | def __init__( | |
842 | self, copy=None, renamedelete=None, dirmove=None, movewithdir=None |
|
842 | self, copy=None, renamedelete=None, dirmove=None, movewithdir=None | |
843 | ): |
|
843 | ): | |
844 | self.copy = {} if copy is None else copy |
|
844 | self.copy = {} if copy is None else copy | |
845 | self.renamedelete = {} if renamedelete is None else renamedelete |
|
845 | self.renamedelete = {} if renamedelete is None else renamedelete | |
846 | self.dirmove = {} if dirmove is None else dirmove |
|
846 | self.dirmove = {} if dirmove is None else dirmove | |
847 | self.movewithdir = {} if movewithdir is None else movewithdir |
|
847 | self.movewithdir = {} if movewithdir is None else movewithdir | |
848 |
|
848 | |||
849 | def __repr__(self): |
|
849 | def __repr__(self): | |
850 | return '<branch_copies\n copy=%r\n renamedelete=%r\n dirmove=%r\n movewithdir=%r\n>' % ( |
|
850 | return '<branch_copies\n copy=%r\n renamedelete=%r\n dirmove=%r\n movewithdir=%r\n>' % ( | |
851 | self.copy, |
|
851 | self.copy, | |
852 | self.renamedelete, |
|
852 | self.renamedelete, | |
853 | self.dirmove, |
|
853 | self.dirmove, | |
854 | self.movewithdir, |
|
854 | self.movewithdir, | |
855 | ) |
|
855 | ) | |
856 |
|
856 | |||
857 |
|
857 | |||
858 | def _fullcopytracing(repo, c1, c2, base): |
|
858 | def _fullcopytracing(repo, c1, c2, base): | |
859 | """The full copytracing algorithm which finds all the new files that were |
|
859 | """The full copytracing algorithm which finds all the new files that were | |
860 | added from merge base up to the top commit and for each file it checks if |
|
860 | added from merge base up to the top commit and for each file it checks if | |
861 | this file was copied from another file. |
|
861 | this file was copied from another file. | |
862 |
|
862 | |||
863 | This is pretty slow when a lot of changesets are involved but will track all |
|
863 | This is pretty slow when a lot of changesets are involved but will track all | |
864 | the copies. |
|
864 | the copies. | |
865 | """ |
|
865 | """ | |
866 | m1 = c1.manifest() |
|
866 | m1 = c1.manifest() | |
867 | m2 = c2.manifest() |
|
867 | m2 = c2.manifest() | |
868 | mb = base.manifest() |
|
868 | mb = base.manifest() | |
869 |
|
869 | |||
870 | copies1 = pathcopies(base, c1) |
|
870 | copies1 = pathcopies(base, c1) | |
871 | copies2 = pathcopies(base, c2) |
|
871 | copies2 = pathcopies(base, c2) | |
872 |
|
872 | |||
873 | if not (copies1 or copies2): |
|
873 | if not (copies1 or copies2): | |
874 | return branch_copies(), branch_copies(), {} |
|
874 | return branch_copies(), branch_copies(), {} | |
875 |
|
875 | |||
876 | inversecopies1 = {} |
|
876 | inversecopies1 = {} | |
877 | inversecopies2 = {} |
|
877 | inversecopies2 = {} | |
878 | for dst, src in copies1.items(): |
|
878 | for dst, src in copies1.items(): | |
879 | inversecopies1.setdefault(src, []).append(dst) |
|
879 | inversecopies1.setdefault(src, []).append(dst) | |
880 | for dst, src in copies2.items(): |
|
880 | for dst, src in copies2.items(): | |
881 | inversecopies2.setdefault(src, []).append(dst) |
|
881 | inversecopies2.setdefault(src, []).append(dst) | |
882 |
|
882 | |||
883 | copy1 = {} |
|
883 | copy1 = {} | |
884 | copy2 = {} |
|
884 | copy2 = {} | |
885 | diverge = {} |
|
885 | diverge = {} | |
886 | renamedelete1 = {} |
|
886 | renamedelete1 = {} | |
887 | renamedelete2 = {} |
|
887 | renamedelete2 = {} | |
888 | allsources = set(inversecopies1) | set(inversecopies2) |
|
888 | allsources = set(inversecopies1) | set(inversecopies2) | |
889 | for src in allsources: |
|
889 | for src in allsources: | |
890 | dsts1 = inversecopies1.get(src) |
|
890 | dsts1 = inversecopies1.get(src) | |
891 | dsts2 = inversecopies2.get(src) |
|
891 | dsts2 = inversecopies2.get(src) | |
892 | if dsts1 and dsts2: |
|
892 | if dsts1 and dsts2: | |
893 | # copied/renamed on both sides |
|
893 | # copied/renamed on both sides | |
894 | if src not in m1 and src not in m2: |
|
894 | if src not in m1 and src not in m2: | |
895 | # renamed on both sides |
|
895 | # renamed on both sides | |
896 | dsts1 = set(dsts1) |
|
896 | dsts1 = set(dsts1) | |
897 | dsts2 = set(dsts2) |
|
897 | dsts2 = set(dsts2) | |
898 | # If there's some overlap in the rename destinations, we |
|
898 | # If there's some overlap in the rename destinations, we | |
899 | # consider it not divergent. For example, if side 1 copies 'a' |
|
899 | # consider it not divergent. For example, if side 1 copies 'a' | |
900 | # to 'b' and 'c' and deletes 'a', and side 2 copies 'a' to 'c' |
|
900 | # to 'b' and 'c' and deletes 'a', and side 2 copies 'a' to 'c' | |
901 | # and 'd' and deletes 'a'. |
|
901 | # and 'd' and deletes 'a'. | |
902 | if dsts1 & dsts2: |
|
902 | if dsts1 & dsts2: | |
903 | for dst in dsts1 & dsts2: |
|
903 | for dst in dsts1 & dsts2: | |
904 | copy1[dst] = src |
|
904 | copy1[dst] = src | |
905 | copy2[dst] = src |
|
905 | copy2[dst] = src | |
906 | else: |
|
906 | else: | |
907 | diverge[src] = sorted(dsts1 | dsts2) |
|
907 | diverge[src] = sorted(dsts1 | dsts2) | |
908 | elif src in m1 and src in m2: |
|
908 | elif src in m1 and src in m2: | |
909 | # copied on both sides |
|
909 | # copied on both sides | |
910 | dsts1 = set(dsts1) |
|
910 | dsts1 = set(dsts1) | |
911 | dsts2 = set(dsts2) |
|
911 | dsts2 = set(dsts2) | |
912 | for dst in dsts1 & dsts2: |
|
912 | for dst in dsts1 & dsts2: | |
913 | copy1[dst] = src |
|
913 | copy1[dst] = src | |
914 | copy2[dst] = src |
|
914 | copy2[dst] = src | |
915 | # TODO: Handle cases where it was renamed on one side and copied |
|
915 | # TODO: Handle cases where it was renamed on one side and copied | |
916 | # on the other side |
|
916 | # on the other side | |
917 | elif dsts1: |
|
917 | elif dsts1: | |
918 | # copied/renamed only on side 1 |
|
918 | # copied/renamed only on side 1 | |
919 | _checksinglesidecopies( |
|
919 | _checksinglesidecopies( | |
920 | src, dsts1, m1, m2, mb, c2, base, copy1, renamedelete1 |
|
920 | src, dsts1, m1, m2, mb, c2, base, copy1, renamedelete1 | |
921 | ) |
|
921 | ) | |
922 | elif dsts2: |
|
922 | elif dsts2: | |
923 | # copied/renamed only on side 2 |
|
923 | # copied/renamed only on side 2 | |
924 | _checksinglesidecopies( |
|
924 | _checksinglesidecopies( | |
925 | src, dsts2, m2, m1, mb, c1, base, copy2, renamedelete2 |
|
925 | src, dsts2, m2, m1, mb, c1, base, copy2, renamedelete2 | |
926 | ) |
|
926 | ) | |
927 |
|
927 | |||
928 | # find interesting file sets from manifests |
|
928 | # find interesting file sets from manifests | |
929 | cache = [] |
|
929 | cache = [] | |
930 |
|
930 | |||
931 | def _get_addedfiles(idx): |
|
931 | def _get_addedfiles(idx): | |
932 | if not cache: |
|
932 | if not cache: | |
933 | addedinm1 = m1.filesnotin(mb, repo.narrowmatch()) |
|
933 | addedinm1 = m1.filesnotin(mb, repo.narrowmatch()) | |
934 | addedinm2 = m2.filesnotin(mb, repo.narrowmatch()) |
|
934 | addedinm2 = m2.filesnotin(mb, repo.narrowmatch()) | |
935 | u1 = sorted(addedinm1 - addedinm2) |
|
935 | u1 = sorted(addedinm1 - addedinm2) | |
936 | u2 = sorted(addedinm2 - addedinm1) |
|
936 | u2 = sorted(addedinm2 - addedinm1) | |
937 | cache.extend((u1, u2)) |
|
937 | cache.extend((u1, u2)) | |
938 | return cache[idx] |
|
938 | return cache[idx] | |
939 |
|
939 | |||
940 | u1fn = lambda: _get_addedfiles(0) |
|
940 | u1fn = lambda: _get_addedfiles(0) | |
941 | u2fn = lambda: _get_addedfiles(1) |
|
941 | u2fn = lambda: _get_addedfiles(1) | |
942 | if repo.ui.debugflag: |
|
942 | if repo.ui.debugflag: | |
943 | u1 = u1fn() |
|
943 | u1 = u1fn() | |
944 | u2 = u2fn() |
|
944 | u2 = u2fn() | |
945 |
|
945 | |||
946 | header = b" unmatched files in %s" |
|
946 | header = b" unmatched files in %s" | |
947 | if u1: |
|
947 | if u1: | |
948 | repo.ui.debug( |
|
948 | repo.ui.debug( | |
949 | b"%s:\n %s\n" % (header % b'local', b"\n ".join(u1)) |
|
949 | b"%s:\n %s\n" % (header % b'local', b"\n ".join(u1)) | |
950 | ) |
|
950 | ) | |
951 | if u2: |
|
951 | if u2: | |
952 | repo.ui.debug( |
|
952 | repo.ui.debug( | |
953 | b"%s:\n %s\n" % (header % b'other', b"\n ".join(u2)) |
|
953 | b"%s:\n %s\n" % (header % b'other', b"\n ".join(u2)) | |
954 | ) |
|
954 | ) | |
955 |
|
955 | |||
956 | renamedeleteset = set() |
|
956 | renamedeleteset = set() | |
957 | divergeset = set() |
|
957 | divergeset = set() | |
958 | for dsts in diverge.values(): |
|
958 | for dsts in diverge.values(): | |
959 | divergeset.update(dsts) |
|
959 | divergeset.update(dsts) | |
960 | for dsts in renamedelete1.values(): |
|
960 | for dsts in renamedelete1.values(): | |
961 | renamedeleteset.update(dsts) |
|
961 | renamedeleteset.update(dsts) | |
962 | for dsts in renamedelete2.values(): |
|
962 | for dsts in renamedelete2.values(): | |
963 | renamedeleteset.update(dsts) |
|
963 | renamedeleteset.update(dsts) | |
964 |
|
964 | |||
965 | repo.ui.debug( |
|
965 | repo.ui.debug( | |
966 | b" all copies found (* = to merge, ! = divergent, " |
|
966 | b" all copies found (* = to merge, ! = divergent, " | |
967 | b"% = renamed and deleted):\n" |
|
967 | b"% = renamed and deleted):\n" | |
968 | ) |
|
968 | ) | |
969 | for side, copies in ((b"local", copies1), (b"remote", copies2)): |
|
969 | for side, copies in ((b"local", copies1), (b"remote", copies2)): | |
970 | if not copies: |
|
970 | if not copies: | |
971 | continue |
|
971 | continue | |
972 | repo.ui.debug(b" on %s side:\n" % side) |
|
972 | repo.ui.debug(b" on %s side:\n" % side) | |
973 | for f in sorted(copies): |
|
973 | for f in sorted(copies): | |
974 | note = b"" |
|
974 | note = b"" | |
975 | if f in copy1 or f in copy2: |
|
975 | if f in copy1 or f in copy2: | |
976 | note += b"*" |
|
976 | note += b"*" | |
977 | if f in divergeset: |
|
977 | if f in divergeset: | |
978 | note += b"!" |
|
978 | note += b"!" | |
979 | if f in renamedeleteset: |
|
979 | if f in renamedeleteset: | |
980 | note += b"%" |
|
980 | note += b"%" | |
981 | repo.ui.debug( |
|
981 | repo.ui.debug( | |
982 | b" src: '%s' -> dst: '%s' %s\n" % (copies[f], f, note) |
|
982 | b" src: '%s' -> dst: '%s' %s\n" % (copies[f], f, note) | |
983 | ) |
|
983 | ) | |
984 | del renamedeleteset |
|
984 | del renamedeleteset | |
985 | del divergeset |
|
985 | del divergeset | |
986 |
|
986 | |||
987 | repo.ui.debug(b" checking for directory renames\n") |
|
987 | repo.ui.debug(b" checking for directory renames\n") | |
988 |
|
988 | |||
989 | dirmove1, movewithdir2 = _dir_renames(repo, c1, copy1, copies1, u2fn) |
|
989 | dirmove1, movewithdir2 = _dir_renames(repo, c1, copy1, copies1, u2fn) | |
990 | dirmove2, movewithdir1 = _dir_renames(repo, c2, copy2, copies2, u1fn) |
|
990 | dirmove2, movewithdir1 = _dir_renames(repo, c2, copy2, copies2, u1fn) | |
991 |
|
991 | |||
992 | branch_copies1 = branch_copies(copy1, renamedelete1, dirmove1, movewithdir1) |
|
992 | branch_copies1 = branch_copies(copy1, renamedelete1, dirmove1, movewithdir1) | |
993 | branch_copies2 = branch_copies(copy2, renamedelete2, dirmove2, movewithdir2) |
|
993 | branch_copies2 = branch_copies(copy2, renamedelete2, dirmove2, movewithdir2) | |
994 |
|
994 | |||
995 | return branch_copies1, branch_copies2, diverge |
|
995 | return branch_copies1, branch_copies2, diverge | |
996 |
|
996 | |||
997 |
|
997 | |||
998 | def _dir_renames(repo, ctx, copy, fullcopy, addedfilesfn): |
|
998 | def _dir_renames(repo, ctx, copy, fullcopy, addedfilesfn): | |
999 | """Finds moved directories and files that should move with them. |
|
999 | """Finds moved directories and files that should move with them. | |
1000 |
|
1000 | |||
1001 | ctx: the context for one of the sides |
|
1001 | ctx: the context for one of the sides | |
1002 | copy: files copied on the same side (as ctx) |
|
1002 | copy: files copied on the same side (as ctx) | |
1003 | fullcopy: files copied on the same side (as ctx), including those that |
|
1003 | fullcopy: files copied on the same side (as ctx), including those that | |
1004 | merge.manifestmerge() won't care about |
|
1004 | merge.manifestmerge() won't care about | |
1005 | addedfilesfn: function returning added files on the other side (compared to |
|
1005 | addedfilesfn: function returning added files on the other side (compared to | |
1006 | ctx) |
|
1006 | ctx) | |
1007 | """ |
|
1007 | """ | |
1008 | # generate a directory move map |
|
1008 | # generate a directory move map | |
1009 | invalid = set() |
|
1009 | invalid = set() | |
1010 | dirmove = {} |
|
1010 | dirmove = {} | |
1011 |
|
1011 | |||
1012 | # examine each file copy for a potential directory move, which is |
|
1012 | # examine each file copy for a potential directory move, which is | |
1013 | # when all the files in a directory are moved to a new directory |
|
1013 | # when all the files in a directory are moved to a new directory | |
1014 | for dst, src in pycompat.iteritems(fullcopy): |
|
1014 | for dst, src in pycompat.iteritems(fullcopy): | |
1015 | dsrc, ddst = pathutil.dirname(src), pathutil.dirname(dst) |
|
1015 | dsrc, ddst = pathutil.dirname(src), pathutil.dirname(dst) | |
1016 | if dsrc in invalid: |
|
1016 | if dsrc in invalid: | |
1017 | # already seen to be uninteresting |
|
1017 | # already seen to be uninteresting | |
1018 | continue |
|
1018 | continue | |
1019 | elif ctx.hasdir(dsrc) and ctx.hasdir(ddst): |
|
1019 | elif ctx.hasdir(dsrc) and ctx.hasdir(ddst): | |
1020 | # directory wasn't entirely moved locally |
|
1020 | # directory wasn't entirely moved locally | |
1021 | invalid.add(dsrc) |
|
1021 | invalid.add(dsrc) | |
1022 | elif dsrc in dirmove and dirmove[dsrc] != ddst: |
|
1022 | elif dsrc in dirmove and dirmove[dsrc] != ddst: | |
1023 | # files from the same directory moved to two different places |
|
1023 | # files from the same directory moved to two different places | |
1024 | invalid.add(dsrc) |
|
1024 | invalid.add(dsrc) | |
1025 | else: |
|
1025 | else: | |
1026 | # looks good so far |
|
1026 | # looks good so far | |
1027 | dirmove[dsrc] = ddst |
|
1027 | dirmove[dsrc] = ddst | |
1028 |
|
1028 | |||
1029 | for i in invalid: |
|
1029 | for i in invalid: | |
1030 | if i in dirmove: |
|
1030 | if i in dirmove: | |
1031 | del dirmove[i] |
|
1031 | del dirmove[i] | |
1032 | del invalid |
|
1032 | del invalid | |
1033 |
|
1033 | |||
1034 | if not dirmove: |
|
1034 | if not dirmove: | |
1035 | return {}, {} |
|
1035 | return {}, {} | |
1036 |
|
1036 | |||
1037 | dirmove = {k + b"/": v + b"/" for k, v in pycompat.iteritems(dirmove)} |
|
1037 | dirmove = {k + b"/": v + b"/" for k, v in pycompat.iteritems(dirmove)} | |
1038 |
|
1038 | |||
1039 | for d in dirmove: |
|
1039 | for d in dirmove: | |
1040 | repo.ui.debug( |
|
1040 | repo.ui.debug( | |
1041 | b" discovered dir src: '%s' -> dst: '%s'\n" % (d, dirmove[d]) |
|
1041 | b" discovered dir src: '%s' -> dst: '%s'\n" % (d, dirmove[d]) | |
1042 | ) |
|
1042 | ) | |
1043 |
|
1043 | |||
1044 | movewithdir = {} |
|
1044 | movewithdir = {} | |
1045 | # check unaccounted nonoverlapping files against directory moves |
|
1045 | # check unaccounted nonoverlapping files against directory moves | |
1046 | for f in addedfilesfn(): |
|
1046 | for f in addedfilesfn(): | |
1047 | if f not in fullcopy: |
|
1047 | if f not in fullcopy: | |
1048 | for d in dirmove: |
|
1048 | for d in dirmove: | |
1049 | if f.startswith(d): |
|
1049 | if f.startswith(d): | |
1050 | # new file added in a directory that was moved, move it |
|
1050 | # new file added in a directory that was moved, move it | |
1051 | df = dirmove[d] + f[len(d) :] |
|
1051 | df = dirmove[d] + f[len(d) :] | |
1052 | if df not in copy: |
|
1052 | if df not in copy: | |
1053 | movewithdir[f] = df |
|
1053 | movewithdir[f] = df | |
1054 | repo.ui.debug( |
|
1054 | repo.ui.debug( | |
1055 | b" pending file src: '%s' -> dst: '%s'\n" |
|
1055 | b" pending file src: '%s' -> dst: '%s'\n" | |
1056 | % (f, df) |
|
1056 | % (f, df) | |
1057 | ) |
|
1057 | ) | |
1058 | break |
|
1058 | break | |
1059 |
|
1059 | |||
1060 | return dirmove, movewithdir |
|
1060 | return dirmove, movewithdir | |
1061 |
|
1061 | |||
1062 |
|
1062 | |||
1063 | def _heuristicscopytracing(repo, c1, c2, base): |
|
1063 | def _heuristicscopytracing(repo, c1, c2, base): | |
1064 | """Fast copytracing using filename heuristics |
|
1064 | """Fast copytracing using filename heuristics | |
1065 |
|
1065 | |||
1066 | Assumes that moves or renames are of following two types: |
|
1066 | Assumes that moves or renames are of following two types: | |
1067 |
|
1067 | |||
1068 | 1) Inside a directory only (same directory name but different filenames) |
|
1068 | 1) Inside a directory only (same directory name but different filenames) | |
1069 | 2) Move from one directory to another |
|
1069 | 2) Move from one directory to another | |
1070 | (same filenames but different directory names) |
|
1070 | (same filenames but different directory names) | |
1071 |
|
1071 | |||
1072 | Works only when there are no merge commits in the "source branch". |
|
1072 | Works only when there are no merge commits in the "source branch". | |
1073 | Source branch is commits from base up to c2 not including base. |
|
1073 | Source branch is commits from base up to c2 not including base. | |
1074 |
|
1074 | |||
1075 | If merge is involved it fallbacks to _fullcopytracing(). |
|
1075 | If merge is involved it fallbacks to _fullcopytracing(). | |
1076 |
|
1076 | |||
1077 | Can be used by setting the following config: |
|
1077 | Can be used by setting the following config: | |
1078 |
|
1078 | |||
1079 | [experimental] |
|
1079 | [experimental] | |
1080 | copytrace = heuristics |
|
1080 | copytrace = heuristics | |
1081 |
|
1081 | |||
1082 | In some cases the copy/move candidates found by heuristics can be very large |
|
1082 | In some cases the copy/move candidates found by heuristics can be very large | |
1083 | in number and that will make the algorithm slow. The number of possible |
|
1083 | in number and that will make the algorithm slow. The number of possible | |
1084 | candidates to check can be limited by using the config |
|
1084 | candidates to check can be limited by using the config | |
1085 | `experimental.copytrace.movecandidateslimit` which defaults to 100. |
|
1085 | `experimental.copytrace.movecandidateslimit` which defaults to 100. | |
1086 | """ |
|
1086 | """ | |
1087 |
|
1087 | |||
1088 | if c1.rev() is None: |
|
1088 | if c1.rev() is None: | |
1089 | c1 = c1.p1() |
|
1089 | c1 = c1.p1() | |
1090 | if c2.rev() is None: |
|
1090 | if c2.rev() is None: | |
1091 | c2 = c2.p1() |
|
1091 | c2 = c2.p1() | |
1092 |
|
1092 | |||
1093 | changedfiles = set() |
|
1093 | changedfiles = set() | |
1094 | m1 = c1.manifest() |
|
1094 | m1 = c1.manifest() | |
1095 | if not repo.revs(b'%d::%d', base.rev(), c2.rev()): |
|
1095 | if not repo.revs(b'%d::%d', base.rev(), c2.rev()): | |
1096 | # If base is not in c2 branch, we switch to fullcopytracing |
|
1096 | # If base is not in c2 branch, we switch to fullcopytracing | |
1097 | repo.ui.debug( |
|
1097 | repo.ui.debug( | |
1098 | b"switching to full copytracing as base is not " |
|
1098 | b"switching to full copytracing as base is not " | |
1099 | b"an ancestor of c2\n" |
|
1099 | b"an ancestor of c2\n" | |
1100 | ) |
|
1100 | ) | |
1101 | return _fullcopytracing(repo, c1, c2, base) |
|
1101 | return _fullcopytracing(repo, c1, c2, base) | |
1102 |
|
1102 | |||
1103 | ctx = c2 |
|
1103 | ctx = c2 | |
1104 | while ctx != base: |
|
1104 | while ctx != base: | |
1105 | if len(ctx.parents()) == 2: |
|
1105 | if len(ctx.parents()) == 2: | |
1106 | # To keep things simple let's not handle merges |
|
1106 | # To keep things simple let's not handle merges | |
1107 | repo.ui.debug(b"switching to full copytracing because of merges\n") |
|
1107 | repo.ui.debug(b"switching to full copytracing because of merges\n") | |
1108 | return _fullcopytracing(repo, c1, c2, base) |
|
1108 | return _fullcopytracing(repo, c1, c2, base) | |
1109 | changedfiles.update(ctx.files()) |
|
1109 | changedfiles.update(ctx.files()) | |
1110 | ctx = ctx.p1() |
|
1110 | ctx = ctx.p1() | |
1111 |
|
1111 | |||
1112 | copies2 = {} |
|
1112 | copies2 = {} | |
1113 | cp = _forwardcopies(base, c2) |
|
1113 | cp = _forwardcopies(base, c2) | |
1114 | for dst, src in pycompat.iteritems(cp): |
|
1114 | for dst, src in pycompat.iteritems(cp): | |
1115 | if src in m1: |
|
1115 | if src in m1: | |
1116 | copies2[dst] = src |
|
1116 | copies2[dst] = src | |
1117 |
|
1117 | |||
1118 | # file is missing if it isn't present in the destination, but is present in |
|
1118 | # file is missing if it isn't present in the destination, but is present in | |
1119 | # the base and present in the source. |
|
1119 | # the base and present in the source. | |
1120 | # Presence in the base is important to exclude added files, presence in the |
|
1120 | # Presence in the base is important to exclude added files, presence in the | |
1121 | # source is important to exclude removed files. |
|
1121 | # source is important to exclude removed files. | |
1122 | filt = lambda f: f not in m1 and f in base and f in c2 |
|
1122 | filt = lambda f: f not in m1 and f in base and f in c2 | |
1123 | missingfiles = [f for f in changedfiles if filt(f)] |
|
1123 | missingfiles = [f for f in changedfiles if filt(f)] | |
1124 |
|
1124 | |||
1125 | copies1 = {} |
|
1125 | copies1 = {} | |
1126 | if missingfiles: |
|
1126 | if missingfiles: | |
1127 | basenametofilename = collections.defaultdict(list) |
|
1127 | basenametofilename = collections.defaultdict(list) | |
1128 | dirnametofilename = collections.defaultdict(list) |
|
1128 | dirnametofilename = collections.defaultdict(list) | |
1129 |
|
1129 | |||
1130 | for f in m1.filesnotin(base.manifest()): |
|
1130 | for f in m1.filesnotin(base.manifest()): | |
1131 | basename = os.path.basename(f) |
|
1131 | basename = os.path.basename(f) | |
1132 | dirname = os.path.dirname(f) |
|
1132 | dirname = os.path.dirname(f) | |
1133 | basenametofilename[basename].append(f) |
|
1133 | basenametofilename[basename].append(f) | |
1134 | dirnametofilename[dirname].append(f) |
|
1134 | dirnametofilename[dirname].append(f) | |
1135 |
|
1135 | |||
1136 | for f in missingfiles: |
|
1136 | for f in missingfiles: | |
1137 | basename = os.path.basename(f) |
|
1137 | basename = os.path.basename(f) | |
1138 | dirname = os.path.dirname(f) |
|
1138 | dirname = os.path.dirname(f) | |
1139 | samebasename = basenametofilename[basename] |
|
1139 | samebasename = basenametofilename[basename] | |
1140 | samedirname = dirnametofilename[dirname] |
|
1140 | samedirname = dirnametofilename[dirname] | |
1141 | movecandidates = samebasename + samedirname |
|
1141 | movecandidates = samebasename + samedirname | |
1142 | # f is guaranteed to be present in c2, that's why |
|
1142 | # f is guaranteed to be present in c2, that's why | |
1143 | # c2.filectx(f) won't fail |
|
1143 | # c2.filectx(f) won't fail | |
1144 | f2 = c2.filectx(f) |
|
1144 | f2 = c2.filectx(f) | |
1145 | # we can have a lot of candidates which can slow down the heuristics |
|
1145 | # we can have a lot of candidates which can slow down the heuristics | |
1146 | # config value to limit the number of candidates moves to check |
|
1146 | # config value to limit the number of candidates moves to check | |
1147 | maxcandidates = repo.ui.configint( |
|
1147 | maxcandidates = repo.ui.configint( | |
1148 | b'experimental', b'copytrace.movecandidateslimit' |
|
1148 | b'experimental', b'copytrace.movecandidateslimit' | |
1149 | ) |
|
1149 | ) | |
1150 |
|
1150 | |||
1151 | if len(movecandidates) > maxcandidates: |
|
1151 | if len(movecandidates) > maxcandidates: | |
1152 | repo.ui.status( |
|
1152 | repo.ui.status( | |
1153 | _( |
|
1153 | _( | |
1154 | b"skipping copytracing for '%s', more " |
|
1154 | b"skipping copytracing for '%s', more " | |
1155 | b"candidates than the limit: %d\n" |
|
1155 | b"candidates than the limit: %d\n" | |
1156 | ) |
|
1156 | ) | |
1157 | % (f, len(movecandidates)) |
|
1157 | % (f, len(movecandidates)) | |
1158 | ) |
|
1158 | ) | |
1159 | continue |
|
1159 | continue | |
1160 |
|
1160 | |||
1161 | for candidate in movecandidates: |
|
1161 | for candidate in movecandidates: | |
1162 | f1 = c1.filectx(candidate) |
|
1162 | f1 = c1.filectx(candidate) | |
1163 | if _related(f1, f2): |
|
1163 | if _related(f1, f2): | |
1164 | # if there are a few related copies then we'll merge |
|
1164 | # if there are a few related copies then we'll merge | |
1165 | # changes into all of them. This matches the behaviour |
|
1165 | # changes into all of them. This matches the behaviour | |
1166 | # of upstream copytracing |
|
1166 | # of upstream copytracing | |
1167 | copies1[candidate] = f |
|
1167 | copies1[candidate] = f | |
1168 |
|
1168 | |||
1169 | return branch_copies(copies1), branch_copies(copies2), {} |
|
1169 | return branch_copies(copies1), branch_copies(copies2), {} | |
1170 |
|
1170 | |||
1171 |
|
1171 | |||
1172 | def _related(f1, f2): |
|
1172 | def _related(f1, f2): | |
1173 | """return True if f1 and f2 filectx have a common ancestor |
|
1173 | """return True if f1 and f2 filectx have a common ancestor | |
1174 |
|
1174 | |||
1175 | Walk back to common ancestor to see if the two files originate |
|
1175 | Walk back to common ancestor to see if the two files originate | |
1176 | from the same file. Since workingfilectx's rev() is None it messes |
|
1176 | from the same file. Since workingfilectx's rev() is None it messes | |
1177 | up the integer comparison logic, hence the pre-step check for |
|
1177 | up the integer comparison logic, hence the pre-step check for | |
1178 | None (f1 and f2 can only be workingfilectx's initially). |
|
1178 | None (f1 and f2 can only be workingfilectx's initially). | |
1179 | """ |
|
1179 | """ | |
1180 |
|
1180 | |||
1181 | if f1 == f2: |
|
1181 | if f1 == f2: | |
1182 | return True # a match |
|
1182 | return True # a match | |
1183 |
|
1183 | |||
1184 | g1, g2 = f1.ancestors(), f2.ancestors() |
|
1184 | g1, g2 = f1.ancestors(), f2.ancestors() | |
1185 | try: |
|
1185 | try: | |
1186 | f1r, f2r = f1.linkrev(), f2.linkrev() |
|
1186 | f1r, f2r = f1.linkrev(), f2.linkrev() | |
1187 |
|
1187 | |||
1188 | if f1r is None: |
|
1188 | if f1r is None: | |
1189 | f1 = next(g1) |
|
1189 | f1 = next(g1) | |
1190 | if f2r is None: |
|
1190 | if f2r is None: | |
1191 | f2 = next(g2) |
|
1191 | f2 = next(g2) | |
1192 |
|
1192 | |||
1193 | while True: |
|
1193 | while True: | |
1194 | f1r, f2r = f1.linkrev(), f2.linkrev() |
|
1194 | f1r, f2r = f1.linkrev(), f2.linkrev() | |
1195 | if f1r > f2r: |
|
1195 | if f1r > f2r: | |
1196 | f1 = next(g1) |
|
1196 | f1 = next(g1) | |
1197 | elif f2r > f1r: |
|
1197 | elif f2r > f1r: | |
1198 | f2 = next(g2) |
|
1198 | f2 = next(g2) | |
1199 | else: # f1 and f2 point to files in the same linkrev |
|
1199 | else: # f1 and f2 point to files in the same linkrev | |
1200 | return f1 == f2 # true if they point to the same file |
|
1200 | return f1 == f2 # true if they point to the same file | |
1201 | except StopIteration: |
|
1201 | except StopIteration: | |
1202 | return False |
|
1202 | return False | |
1203 |
|
1203 | |||
1204 |
|
1204 | |||
1205 | def graftcopies(wctx, ctx, base): |
|
1205 | def graftcopies(wctx, ctx, base): | |
1206 | """reproduce copies between base and ctx in the wctx |
|
1206 | """reproduce copies between base and ctx in the wctx | |
1207 |
|
1207 | |||
1208 | Unlike mergecopies(), this function will only consider copies between base |
|
1208 | Unlike mergecopies(), this function will only consider copies between base | |
1209 | and ctx; it will ignore copies between base and wctx. Also unlike |
|
1209 | and ctx; it will ignore copies between base and wctx. Also unlike | |
1210 | mergecopies(), this function will apply copies to the working copy (instead |
|
1210 | mergecopies(), this function will apply copies to the working copy (instead | |
1211 | of just returning information about the copies). That makes it cheaper |
|
1211 | of just returning information about the copies). That makes it cheaper | |
1212 | (especially in the common case of base==ctx.p1()) and useful also when |
|
1212 | (especially in the common case of base==ctx.p1()) and useful also when | |
1213 | experimental.copytrace=off. |
|
1213 | experimental.copytrace=off. | |
1214 |
|
1214 | |||
1215 | merge.update() will have already marked most copies, but it will only |
|
1215 | merge.update() will have already marked most copies, but it will only | |
1216 | mark copies if it thinks the source files are related (see |
|
1216 | mark copies if it thinks the source files are related (see | |
1217 | merge._related()). It will also not mark copies if the file wasn't modified |
|
1217 | merge._related()). It will also not mark copies if the file wasn't modified | |
1218 | on the local side. This function adds the copies that were "missed" |
|
1218 | on the local side. This function adds the copies that were "missed" | |
1219 | by merge.update(). |
|
1219 | by merge.update(). | |
1220 | """ |
|
1220 | """ | |
1221 | new_copies = pathcopies(base, ctx) |
|
1221 | new_copies = pathcopies(base, ctx) | |
1222 | _filter(wctx.p1(), wctx, new_copies) |
|
1222 | _filter(wctx.p1(), wctx, new_copies) | |
1223 | for dst, src in pycompat.iteritems(new_copies): |
|
1223 | for dst, src in pycompat.iteritems(new_copies): | |
1224 | wctx[dst].markcopied(src) |
|
1224 | wctx[dst].markcopied(src) |
General Comments 0
You need to be logged in to leave comments.
Login now