bookmarks: resolve divergent bookmark when moving across a branch This patch resolves a single divergent bookmark if a divergent bookmark exists in the target revision and it current bookmark is not an ancestor of the target revision, else it would already be handled by the previous patch in this series. Test coverage is added.

bookmarks: resolve divergent bookmarks when moving active bookmark forward This patch resolves divergent bookmarks between the current active bookmark MARK and the new destination. This situation can arise when pulling new changesets, abandoning your current changesets actively bookmarked with MARK via strip, and then doing a bare update. The non-divergent but active bookmark MARK is then moved to a common ancestor of the new changesets and the abandoned changesets. Test coverage is added.

bookmarks: resolve divergent bookmarks when fowarding bookmark to descendant This patch is a follow-up to 56dd55da2f7d that resolves divergent bookmarks between the to-be-forwarded bookmark MARK and the new descendant. This situation can happen when pulling new changesets, updating to the divergent bookmark, abandoning the previous changesets with strip, and then moving MARK to MARK@N. Test coverage is added.

tags: update tag type only if tag node is updated (issue3911) Before this patch, tag type information is always updated, even if tag previously read in has higher priority than one newly read in. This causes that the tag type is displayed as "local", even if global tag overwrites existing local one successfully. This patch updates tag type only if tag node is updated. This patch tests overwriting local tags below: - visible one (normal case) - already removed one (recorded as null)

match: fix root calculation for combining regexps with simple paths The fall-back root for walking is the repo root, not no root. The "roots" do however also end up in m.files() which is used in various ways, for instance to indicate whether matches are exact. The change could thus have other impacts.

icasefs: enhance test to prevent regression by changes in the future This patch adds steps to test case-folding collision detection on the revision tree built up in the order different from the one reproducing issue3452. This may prevent regression by changes around "copy detection" and/or "case-folding collision detection" logic in the future.

icasefs: rewrite case-folding collision detection (issue3452) Before this patch, case-folding collision detection uses "copies.pathcopies()" before "manifestmerge()", and is not aware of renaming in some cases. For example, in the case of issue3452, "copies.pathcopies()" can't detect renaming, if the file is renamed at the revision before common ancestor of merging. So, "hg merge" is aborted unexpectedly on case insensitive filesystem. This patch fully rewrites case-folding collision detection, and relocate it into "manifestmerge()". New implementation uses list of actions held in "actions" and "prompts" to build provisional merged manifest up. Provisional merged manifest should be correct, if actions required to build merge result up in working directory are listed up in "actions" and "prompts" correctly. This patch checks case-folding collision still before prompting for merge, to avoid aborting after some interactions with users. So, this assumes that user would choose not "deleted" but "changed". This patch also changes existing abort message, because sorting before collision detection changes order of checked files.

icasefs: refactor "test-casecollision-merge.t" to reuse in succeeding fixes This patch refactors "test-casecollision-merge.t" to increase reusability in succeeding patches. It is confirmed that changed test also can detect issue3370.

icasefs: ignore removed files at building "dirstate._foldmap" up on icasefs Before this patch, all files in dirstate are used to build "_foldmap" up on case insensitive filesystem regardless of their statuses. For example, when dirstate contains both removed file 'a' and added file 'A', "_foldmap" may be updated finally by removed file 'a'. This causes unexpected status information for added file 'A' at "hg status" invocation. This patch ignores removed files at building "_foldmap" up on case insensitive filessytem. This patch doesn't add any test, because this issue is difficult to reproduce intentionally: it depends on iteration order of "dirstate._map".

i18n-pt_BR: synchronized with 3f5e75c22585

merge with i18n

i18n-ja: synchronized with f01a351db791

dispatch: add support for --option=value to _earlygetopt This fixes a very confusing error message: $ hg --config=pager.enabled=off st abort: option --config may not be abbreviated!

dispatch: add doctests for _earlygetopt

push: make locking of source optional (issue3684) Having the permission to lock the source repo on push is now optional. When the repo cannot be locked, phase are not changed locally. A status message is issue when some actual phase movement are skipped: cannot lock source repo, skipping local public phase update A debug message with the exact reason of the locking failure is issued in all case.

push: factorise phase movement in a simple closure Having all phases movement centralised will help to handle special case when the local repo can not be locked as describe in issue 3684.

merge: increase safety of parallel updating/removing on icasefs "merge.applyupdates()" sorts "actions" in removal first order, and "workeractions" derived from it should be also sorted. If each actions in "workeractions" are executed in serial, this sorting ensures that merging/updating process is collision free, because updating the file in target context is always executed after removing the existing file which causes case-folding collision against the former. In the other hand, if each actions are executed in parallel, updating on a worker process may be executed before removing on another worker process, because "worker.partition()" partitions list of actions regardless of type of each actions. This patch divides "workeractions" into removing and updating, and executes the former first. This patch still scans "actions"/"workeractions" some times for ease of patch review, even though large list may cost much in this way. (total cost should be as same as before) This also changes some tests, because dividing "workeractions" affects progress indication.

hgweb: handle filtered "0" rev in navigation Before this changeset, navigation generation crashed if revision "0" was filtered. We introduce a `_first` methods on revision navigation that return the lowest unfiltered element and use it in two place were the "0" changeset was explicitly referenced. Test case are introduced.

hgweb: fix empty navigation detection For some obscure reason, changelog.node(0) returns nullid if changelog is empty. this break empty navigation detection. We fix this code by using the length of the changelog. Using the length have some issue with revision filtering but this is a small step in the right direction. Proper fix comes in later changeset.

tests: AIX can't handle negative date in test-dirstate.t test-dirstate.t fails on AIX in the absurd date test. AIX touch errors on any date prior to 1970. AIX mktime() gives an error on such dates, so the problem is deeper than touch and attempts to work around touch in Python failed. Give up. Add an AIX test to hghave and skip the absurd date test on AIX.

win32: use explicit path to "python.exe" only if it exists Before this patch, "hg.bat" for Windows environment always uses "%~dp0..\python" as explicit path to "python.exe". This path may not be valid in some cases. For example, on the environment using "virtualenv" python package, both "python.exe" and "hg.bat" are placed in the same directory. In this case, "python.exe" should be found on PATH, because virtualenv activation script puts "python.exe" on the PATH. This patch uses explicit path to "python.exe" only if it exists, and expects that "python.exe" can be found on PATH otherwise.

test-mq-strip.t: add a test for strip --keep with clean working dir This helped uncover a bug in a patchset I've been writing.

largefiles: use repo.wwrite for writing standins (issue3909)

largefiles: drop repo wrapping detection After 257afe5489d4 I see: $ hg id -q largefiles: repo method 'commit' appears to have already been wrapped by another extension: largefiles may behave incorrectly largefiles: repo method 'push' appears to have already been wrapped by another extension: largefiles may behave incorrectly be207d9b7e4b The warning is bad: * The message gives no hint what the problem is and how it can be resolved. The message is useless. * Largefiles do have its share of problems, but I don't think I ever have seen a problem where this warning would have helped. The 'may' in the warning seems like an exaggeration of the risk. Having largefiles enabled in combination with for instance mq, hggit and hgsubversion causes a warning (depending on the configuration order) but do not cause problems. Extensions might of course be incompatible, but they can be that in many other ways. The check and the message are incorrect. It would thus be better to remove the check and the warning completely. Before 257afe5489d4 the check always failed. That change made the check work more like intended ... but the intention was wrong. This change will thus also back that change out.

config: discard "%unset" values defined in the other files read in previously Before this patch, "%unset" can't unset values defined in the other files read in previously, even though online help document says that it can. It can unset only values defined in the same configuration file. For example, the value defined in "~/.hgrc" can't be unset by "%unset" in ".hg/hgrc" of the repository. This patch records "%unset"-ed values in "config.parse()", and discards corresponding values in "config.update()".

tests: rename from test-config-case.t to test-config.t for centralization Before this patch, there is no test script testing configuration handling generally. "test-config-case.t" seems to be specific for testing case sensitive configuration. This patch renames from "test-config-case.t" to "test-config.t" for centralization of tests around configuration handling.

i18n: show the non-ASCII password prompt text correctly Before this patch, the prompt text for asking password is directly passed to "getpass.getpass()" of Python standard library. In "getpass.getpass()" implementation on Windows environment, the prompt text is split into byte sequence and "msvcrt.putch()" is applied on each bytes in it. This splitting causes non-ASCII prompt text to be broken. This patch shows the prompt text for asking password on "ui.getpass()" side, and invokes "getpass.getpass()" with empty prompt text. This prevents non-ASCII prompt text from being broken in "getpass.getpass()" implementation. This patch also sets "ui.prompt" label to prompt text to follow "ui.prompt()" style.

tests: make sed usage in test-unionrepo.t cross-platform Usage of the 'i' command proves tricky. I tried to write a check-code rule, but failed.

check-code: fix sed 'i' command rule newline matching The regular expression was meant to match cases where an 'i' command was not followed by precisely a '\' and then a newline; it failed to match the newline, so cases with a '\' but no newline would erroneously pass.

blackbox: don't run permission tests on non-unix systems The windows and vfat test runs were failing due to read/write permissions not working the same on those systems. On vfat, permissions can't be changed at all, and on windows it seems the chmod emulation doesn't remove read permissions. We could theoretically get the 'cannot write to blacklog.log' test to pass on windows but there's no #if condition to let us exclude vfat only. Verified that test-blackbox passes on windows now.

check-code: expand sed rule to include more offenders Expands the rule added in 5e4491c114b2 to include cases where the address is a line number instead of a regular expression, and fixes an instance of this pattern in test-unionrepo.t.

check-code: add a rule against a GNU sed-ism BSD sed requires the 'i' command to be followed with a backslash and a newline, like so: $ sed -e '/^@/i\ > other' We've encountered this problem before, e.g. in test-mq.t (900767dfa80d). This change adds a check-code rule and fixes two instances of the problem in test-record.t.

hgweb: make help verbose again (issue3899) Due to regression introduced in f5db3092790f, help in hgweb was rendered in non-verbose form (issue3899)

merge with i18n

i18n-pt_BR: synchronized with 64ea454e7d76

css: remove repeated property 'margin' property was repeated for the same selector

css: fixed font-family There is no 'sans' font-family, replaced with 'sans-serif'

color: add a test with extension loaded and ui.formatted=False

color: set _colormode to None when mode is unset (issue3895) Previously, colorui assumed that it would only be called when mode wasn't None. 7ae12ce87594 changed that, so now colorui needs to care about whether it should colorize output.

color: turn colorui functions into forwards when color is None colorui will be set to None as necessary in an upcoming patch.

largefiles: fix typos in documentation

translations: change label integer error to not specify the kind of label The current error message used the kind (bookmark, branch, tag) in the message. Unfortunately this isn't easily translatable since some languages give different genders to these words. The fix is to not specify it at all, since it should be implicit based on the command the user just ran. Relevant discussions around this area: http://selenic.com/pipermail/mercurial-devel/2012-October/045567.html http://selenic.com/pipermail/mercurial-devel/2012-October/045600.html

Added signature for changeset 292cd385856d

Added tag 2.6-rc for changeset 292cd385856d

merge default int stable for 2.6 code freeze

blackbox: automatically rotate log files If enabled, log rotation prevents the amount of space used by the blackbox log from growing without bound. This becomes important in cases where there are a lot of busy repositories managed by humans and automation on many machines. In large deployments, we cannot reasonably track all the repos where blackbox logs need to be managed, so it is safer to have blackbox manage its own logs than to move responsibility to an external tool such as logrotate. This change adds two configuration keys: * blackbox.maxsize is the maximum allowable size of the current log * blackbox.maxfiles is the number of log files to maintain

ui: add a configbytes method, for space configuration This accepts a floating point number, followed by optional whitespace, followed by an optional one- or two-letter unit specifier (for bytes, kilobytes, megabytes, or gigabytes).

mq: do not inherit settings form base repo in mqrepo (Fixes issue2358) This is done by using repo.baseui when creating the queue repo.

color: set the ui class in reposetup, before a repo is created Previously only the class of repo.ui was set. The class was set in the wrapper of dispatch._runcommand. But this is called after the repo is created. Only the class of repo.ui is set because repo.baseui already exists and is not accessible. We now set the class in uisetup because this is called before repo creation. This is also the place suggested in the wiki for modifying classes.

parsers: remove warning: format ‘%ld’ expects argument of type ‘long int’ gcc 4.6.3 on 12.04 Ubuntu machine emits warnings: mercurial/parsers.c: In function ‘find_deepest’: mercurial/parsers.c:1288:9: warning: format ‘%ld’ expects argument of type ‘long int’, but argument 3 has type ‘Py_ssize_t’ [-Wformat] mercurial/parsers.c:1288:9: warning: format ‘%ld’ expects argument of type ‘long int’, but argument 4 has type ‘Py_ssize_t’ [-Wformat]

annotate: discard refcount of discarded annotation for memory efficiency Before this patch, refcount (managed in "needed") of the annotation result is kept as 1, even if corresponding annotation result is discarded from "hist", because it isn't decreased and discarded. In the history tree including merging revision, the most recent common ancestor of merged revisions is scanned twice. Refcount of such ancestor never becomes 0, because refcount is started from 1 at the second scanning. This prevents annotation results of merging revision in "hist" from being discarded, and decreases memory efficiency. This patch discards refcount of the annotation result, when the corresponding annotation is discarded from "hist".

revlog: fix a regression with null revision Introduced in the patch which fixes issue3497 Part of that patch was erroneously submitted and it shouldn't be in the code

rebase: check no-op before checking phase (issue3891) This could lead to user confusion, because phases aren't really involved at all when attempting to perform a no-op rebase.

template: call runtemplate on the src argument to sub This allows using a template keyword on calls to the sub function and brings the function inline with most other semantics of the other template functions.

run-tests: introduce --random for running tests in random error This is convenient when investigating failures one by one to avoid running the same fixed tests first every time.

largefiles: wlock in status before lfdirstate.write()

transplant: use set for prune lookup

obsolete: extract obsolescence marker pulling into a dedicated function Having a dedicated function will allow us to experiment with other exchange strategies in an extension. As we have no solid clues about how to do it right, being able to experiment is vital. Some transaction tricks are necessary for pull. But nothing too scary.

obsolete: extract obsolescence marker pushing into a dedicated function Having a dedicated function will allows us to experiment with other exchange strategies in an extension. As we have no solid clues about how to do it right, being able to experiment is vital. I intended a more ambitious extraction of push logic, but we are far too advanced in the release cycle for it.

blackbox: fix a case of name capture

pathencode: eliminate comma at end of enum list to avoid pedantic warning

smtp: use 465 as default port for SMTPS Before this patch, port 25 (wellknown port of SMTP) is used as default port, even if "[smtp] tls" is configured as "smtps". This patch uses port 465 (wellknown port of SMTPS) as default port, if "[smtp] tls" is configured as "smtps".

subrepo: fix test breakage introduced in 2b34d004e644 The test didn't break when I originally committed it (git version 1.7.12.4 (Apple Git-37)), and still doesn't on that machine, but the output changes elsewhere with "Warning: you are leaving N commits behind..." This change sets a ref on the commit we're updating the subrepo _away_ from to quash the warning.

histedit: more precise user message when changeset is missing Now that we explicitly detect duplicated changesets, we can explicitly detect missing ones. We cover the same cases as before, some others and we offer a better error message in all cases.

histedit: protect against duplicated entries Before this change one would issue rules with duplicated entries. For this to happen some other changeset had to be missing to maintain the rules length.

histedit: track short hash instead of changectx object This simplifies set usage and allow us to add a missing revision check in a later commit.

histedit: ensure rules return short hex at all time Handling homogeneous data is both simpler and safer.

histedit-test: test unknown command We had no tested for this case.

histedit-test: test unknown changeset We had no tested for this case.

histedit-test: test malformed line We had not tested for this case.

histedit-test: test that extra revisions are detected We had not tested for this case.

histedit-test: test that missing revisions are detected We had not tested for this case.

histedit: handle multiple spaces between action and hash (issue3893) There is some clue that the previous code intended to handle that but it was actually not the case. As a result action could apply to the empty string '' changeset, leading to the use the current working directory parent in some operations.

http2: sane readline It turns out that it pays off to read more than a byte at a time with a select in between :)

http2: make read use pushchunk/popchunk, eschew itertools The itertools approach was showing up high in the profile output.

http2: track available data as it changes instead of recomputing it With a large amount of available data, this computation can become costly.

histedit-test: rename histedit-revspec into histedit-argument We have no testing for argument. This small file seems perfect once but needs a rename.

histedit-test: remove unnecessary graphlog extension Graphlog is now in core.

histedit-test: drop skip logic for older mercurial version We are shipped in core now.

hgweb: refactor checks for granting and revoking user permissions Provides an entry point for extensions implementing more complex authorization schemes. Original patch proposed by Markus Zapke-Gründemann.

check-code: add check for lock release order

parsers: fix variable declaration position issue

zsh_completion: complete all files for `commit/backout --logfile` Up until now commit/backout `--logfile` option was completed only with *.txt-files. This requirement is a bit too strong (and is nowhere to be stated) so it's better to complete with all files. Stolen from zsh.git repo, c5fce37b changeset by Oliver Kiddle.

transplant: improve documentation

transplant: clarify what --branch do - it has nothing to do with branches --branch specifies heads and has nothing to do with named branches, and it only deals with topological branches to the same extent as all other DAG operations do.

import: inline checkexact function We have a sngle call now, no need to make it a function.

import: factor out checkexact The two calls are strictly identical. We can simply move it after the if/else clause.

amend: fix unlocking order - first lock then wlock

codingstyle: remove trailing spaces in various text files Better do it once than see random changes in diffs later.

tests: run check-code on Python files without .py extension

histedit: move outgoing processing to its own function Every piece of code extracted from the main command is a win. We simplify changeset determination in the process. Parent ceases being a list before becoming a node. We how have a root variable containing a node all the time.

histedit: move all arguments checks to the beginning of the command This changeset move all checks and raises related to arguments validation to the top of the file. This gathers all the logic in one place and clarifies the code doing actual work. This paves the way for splitting this gigantic function in separated functions. A `goal` variable is introduced in the process. It holds the action to be done by this invocation (new, continue or abort). An invalid invocation is found in the process (the new code is a bit stricter).

histedit-test: generalise --commands "-" usage This is simpler than temporary file version. There some minor test changes since commit messages are no longer modifed. There is still some tests using --commands with a real file.

histedit: allow "-" as a command file When "-" is specified as a command file, we read rules from stdin. Alter a single test to demo the behavior, but most of them could benefit from this. There is minor change in test output resulting from the fact we no longer change log messages.

histedit: properly handle --continue on empty fold When all changes from the fold have been dropped, the --continue code was confused. This changeset handles this case. The test for this case existed but was broken.

histedit-test: replace obscure python script by a plain file One of the tests uses a python script to edit the histedit commands on the fly. It was hard to read and likely to break. This is replaced by a plain command file. Doing so made me discover that the python script is incorrect. The "+6" regexp never matched because the revision number is wrong. This error is kept in the replacement for now. Fixing the histedit command triggers a traceback in histedit. A later changeset will fix the error and restore the intended test.

test: improve documentation of some histedit tests Each test section is clearly separated from the others and gains some helpful documentation.

test: use a lighter log style in histedit test The default log displays a log of information that is useless for histedit tests. Having a cleaner log help readability of those tests a lot. This does not change anything to the test semantic.

subrepo: add regression test for issue3870

subrepo: clone of git sub-repository creates incorrect git branch (issue3870) Mercurial handles git subrepos by incorrect way. If the mercurial repo has a git sub-repo and somebody started a new branch in the subrepo and push it into git, the next one who will clone the whole repo will get incorrect branch name in the git subrepo.

dispatch: print 'abort:' when a pre-command hook fails (BC) This also changes the exit code from whatever the hook returned to 255. This brings it in line with all the other hooks that abort.

largefiles: don't hash all largefiles when initializing a lfdirstate The largefiles will be hashed on demand if necessary ... and sometimes it isn't necessary.

largefiles: use filechunkiter for iterating largefile when serving getlfile The default file iterator is line based and will give odd chunk sizes - often very short and relatively expensive.

largefiles: stat all largefiles in one batch before downloading This avoids a lot of expensive roundtrips to remote repositories ... but might be slightly slower for local operations. This will also change some aborts on missing files to warnings. That will in some situations make it possible to continue working on a repository with missing largefiles.

largefiles: 'put' should store 'source' file in under 'hash', also in localstore

largefiles: getlfile must hit end of HTTP chunked streams to reuse connections We did read the exactly the right number of bytes from the response body. But if the response came in chunked encoding then that meant that the HTTP layer still hadn't read the last 0-sized chunk and expected the app layer to read more data from the stream. The app layer was however happy and sent another request which had to be sent on another HTTP connection while the old one was lingering until some other event closed the connection. Adding an extra read where we expect to hit the end of file makes the HTTP connection ready for reuse. This thus plugs a real socket leak. To distinguish HTTP from SSH we look at self's class, just like it is done in putlfile.

largefiles: drop limitreader, use filechunkiter limit filechunkiter.close was a noop.

largefiles: move protocol conversion into getlfile and make it an iterable Avoid the intermediate limitreader and filechunkiter between getlfile and copyandhash - return the right protocol and put the complexity where it better can be managed.

largefiles: don't close the fd passed to store._getfile

largefiles: remove blecch from lfutil.copyandhash - don't close the passed fd

largefiles: drop lfutil.blockstream - use filechunkiter like everybody else The old chunk size is kept - just to avoid changing it.

largefiles: refactoring - use findfile in localstore._getfile

largefiles: refactoring - return hex from _getfile and copyandhash

largefiles: refactoring - create destination dir in lfutil.link

summary: clear "commonincoming" also if branches are different Before this patch, "commonincoming" calculated by "discovery.findcommonincoming()" is cleared, only if "default" URL without branch part (tail "#branch" of URL) differs from "default-push" URL without branch part. But common revisions in "commonincoming" calculated for a branch doesn't include ones for another branch, even if URLs without branch part are same. The result of "discovery.findcommonoutgoing()" invocation with such "commonincoming" becomes incorrect in some cases. This patch clears "commonincoming", also if branch part of "default" differs from one of "default-push". To avoid redundant looking up: - "ui.expandpath('default')" and "ui.expandpath('default-push', 'default')" are not compared directly, even though they contain branch information, because they are not yet normalized by "hg.parseurl()": tail "/" of path, for example - "commonincoming" is not cleared, if branch isn't specified in "default" URL, because such "commonincoming" contains common revisions for all branches This patch also tests "different path, same branch" pattern to check careless degrading around comparison between source and destination.

summary: make "incoming" information sensitive to branch in URL (issue3830) Before this patch, "incoming" information of "hg summary --remote" is not sensitive to the branch specified in the URL of the destination repository, even though "hg pull"/"hg incoming" are so. Invocation of "discovery.findcommonincoming()" without "heads" argument treats revisions on branches other than the one specified in the URL as incoming ones unexpectedly. This patch looks head revisions, which are already detected by "hg.addbranchrevs()" from URL, up against "other" repository, and invokes "discovery.findcommonincoming()" with list of them as "heads" to limit calculation of incoming revisions.

histedit: make "hg histedit" sensitive to branch in URL Before this patch, "hg histedit" are not sensitive to the branch specified in the URL of the destination repository, even though "hg push"/"hg outgoing" are so: Invocation of "discovery.findcommonoutgoing()" without "onlyheads" argument treats revisions on branches other than the one specified in the URL as outgoing ones unexpectedly. This patch specifies list of head revisions, which are already detected by "hg.addbranchrevs()" from URL and looked up against local repository, as "onlyheads" to "discovery.findcommonoutgoing()" to limit calculation of outgoing revisions.

summary: make "outgoing" information sensitive to branch in URL (issue3829) Before this patch, "outgoing" information of "hg summary --remote" is not sensitive to the branch specified in the URL of the destination repository, even though "hg push"/"hg outgoing" are so: Invocation of "discovery.findcommonoutgoing()" without "onlyheads" argument treats revisions on branches other than the one specified in the URL as outgoing ones unexpectedly. This patch looks head revisions, which are already detected by "hg.addbranchrevs()" from URL, up against local repository, and invokes "discovery.findcommonoutgoing()" with list of them as "onlyheads" to limit calculation of outgoing revisions.

annotate: increase refcount of each revisions correctly (issue3841) Before this patch, refcount (managed in "needed") of parents of each revisions in "visit" is increased, only when parent is not annotated yet (examined by "p not in hist"). But this causes less refcount of the revision like "A" in the tree below ("A" is assumed as the second parent of "C"): A --- B --- C \ / \-----/ Steps of annotation for "C" in this case are shown below: 1. for "C" 1.1 increase refcount of "B" 1.2 increase refcount of "A" (=> 1) 1.3 defer annotation for "C" 2. for "A" 2.1 annotate for "A" (=> put result into "hist[A]") 2.2 clear "pcache[A]" ("pcache[A] = []") 3. for "B" 3.1 not increase refcount of "A", because "A not in hist" is False 3.2 annotate for "B" 3.3 decrease refcount of "A" (=> 0) 3.4 delete "hist[A]", even though "A" is still needed by "C" 3.5 clear "pcache[B]" 4. for "C", again 4.1 not increase refcount of "B", because "B not in hist" is False 4.2 increase refcount of "A" (=> 1) 4.3 defer annotation for "C" 5. for "A", again 5.1 annotate for "A" (=> put result into "hist[A]", again) 5.2 clear "pcache[A]" 6. for "C", once again 6.1 not increase refcount of "B", because "B not in hist" is False 6.2 not increase refcount of "A", because "A not in hist" is False 6.3 annotate for "C" 6.4 decrease refcount of "A", and delete "hist[A]" 6.5 decrease refcount of "B", and delete "hist[B]" 6.6 clear "pcache[C]" At step (5.1), annotation for "A" mis-recognizes that all lines are created at "A", because "pcache[A]" already cleared at step (2.2) prevents from scanning ancestors of "A". So, annotation for "C" or its descendants loses information about "A" or its ancestors. The root cause of this problem is that refcount of "A" is decreased at step (3.3), even though it isn't increased at step (3.1). To increase refcount correctly, this patch increases refcount of each parents of each revisions: - regardless of "p not in hist" or not, and - only once for each revisions in "visit" (by "not pcached") In fact, this problem should occur only on legacy repositories in which a filelog includes the merging between the revision and its ancestor (as the second parent), because: - tree is scanned in depth-first without such merging, revisions in "visit" refer different revisions as parent each other - recent Mercurial doesn't allow such merging changelog and manifest can include such merging someway, but filelogs can't, because "localrepository._filecommit()" converts such merging request to linear history. This patch tests merging cases below: these cases are from filelog of "mercurial/commands.py" in the repository of Mercurial itself. - both parents are same 10 --- 11 --- 12 \_/ filelogrev: changesetid: 10 00ea3613f82c 11 fc4a6e5b5812 12 4f802588cdfb - the second parent is also ancestor of the first one 37 --- 38 --- 39 --- 40 \________/ filelogrev: changesetid: 37 f8d56da6ac8f 38 38919e1c254d 39 d3400605d246 40 f06a4a3b86a7

annotate: reuse already calculated annotation Before this patch, annotation is re-calculated even if it is already calculated. This may cause unexpected annotation, because already cleared "pcache" ("pcache[f] = []") prevents from scanning ancestors. This patch reuses already calculated annotation if it is available. In fact, "reusable" situation should be seen only on legacy repositories in which a filelog include the merging between the revision and its ancestor, because: - tree is scanned in depth-first without such merging, annotation result should be released soon - recent Mercurial doesn't allow such merging changelog and manifest can include such merging someway, but filelogs can't, because "localrepository._filecommit()" converts such merging request to linear history.