Automatic nesting into running transactions in the same repository. This associates a transaction handle with a given repository object, and any additional calls to start new transactions reuse that transaction. For the 2700 patch import run, this brings the system time down from 1m20s to 50s, mostly by skipping backups of the dirstate file. (note, this patch does not change hg import to use the nested transaction, mq is the only user right now)

Fix bash_completion on Solaris TK Soh noticed that the awk that ships with Solaris doesn't allow you to define new functions. According to http://www.shelldorado.com/articles/awkcompat.html it looks like it also doesn't have some other stuff that we were using, like deletion of an array element and gsub. Rewrite the parsing of hg --debug help to work around that. This version doesn't filter aliases that are simple abbreviations for debug commands (not a big problem, since there are none right now).

refactor the dodiff optimization

speed up hg log --patch Changing dodiff to read the manifest/changelog for node1 before calling repo.update allows us to take advantage of the revlog revision cache. Before this patch and my previous "speed up hg log --debug" patch, when using hg log -p to display three revisions (A, B and C), dodiff and repo.changes would end up reading the manifests in this order: B A B A C B C B With both patches, this order becomes: A A B B B B C C (This considers only dodiff and repo.changes. I'm not sure how other parts of hg log enter the picture.) The speed up will depend on the revisions being displayed. (All "before" times already have my previous "speed up hg log --debug" patch applied.) hg repo (tip = 414e81ae971f). hg log -p before after real 0m50.981s 0m45.279s user 0m47.930s 0m42.560s sys 0m2.526s 0m2.523s output size: 6917897 bytes kernel repo (tip = 9d4e135960ed). hg log -p -l64 before after real 2m14.995s 1m45.025s user 2m9.509s 1m33.900s sys 0m3.663s 0m2.942s output size: 31497621 bytes same kernel repo. hg log -p -l64 -r c84c2069592f:0 before after real 1m48.045s 1m0.076s user 1m44.094s 0m58.492s sys 0m2.603s 0m1.103s output size: 197983 bytes c84c2069592f was the tip of a 10 day old kernel repo that I had lying around and was where I first tested this patch. For some weird coincidence it's also a place where the patch makes a huge difference.

speed up hg log --debug hg repo: before: real 0m6.998s user 0m6.952s sys 0m0.046s after: real 0m3.884s user 0m3.839s sys 0m0.042s

on revert, only undele files that are removed in the dirstate

Removed now obsolete min/max check in walkchangerevs().

Replaced fixed window size for walkchangerevs with an increasing one. Window sizes starts at 8 (for good interactiveness) and doubles with each window until it is 512, which seems to be the maximum efficient value.

hgmerge: do not use file with markers in tools merging themselves This time I forgot about a crucial part of the logic change. The patch makes kdiff3 and tkdiff useful again.

Fix hgmerge: mkdir "'/tmp'/hgmerge..." and ask_if_merged() didn't work.

Fixed wrong (copy&paste) usage of tmpl instead of self.t in hgweb.

Fix 'hg serve' not using CSS. hgweb.py (templatepath): Return normalized path [fixes later checks in staticfile()]

only write the dirstate when something changed

Allow serving static files from hgwebdir to fix CSS and favicon.

Cleanup of tabs and coding when sending content type.

Merge with jeffpc's hg-static

[hgweb] Added favicon to standard/gitweb templates

Merge with mercurial/tonfa

change the default timeout to 600 seconds

add a timeout when a lock is held (default 1024 sec) - change the wait keyword from lock.lock to timeout, a negative timeout of means "wait forever" - refactor the two lock functions from localrepo.py - make them use the timeout (default 1024, can be changed with ui.timeout in the config file - update the doc

Don't use mktemp in tests, we're already in a secure temp dir.

Added test cases for repo corruption fixed in 2e0a288ca93e (issue132) (submitted by Alexis S. L. Carvalho <alexis@cecm.usp.br>)

revalidate revlog data after locking the repo (issue132)

merge with matt

merge with crew

add an optional argument to push only the specified revisions (push -r)

Merge with jeffpc

Merge with mpm

[hgweb] Move CSS from headers into static files

[hgweb] Static content serving

Merge

Merge with crew

hgmerge: add meld support

hgmerge: improve "apps not found" message

hgmerge: logic changes Details: - keep automatic merge conflicts in $LOCAL - do not assume that $EDITOR returns proper error code - fix a slight mistake in diff+patch merge success detection (it could succeed if the $EDITOR failed to run)

hgmerge: add and use ask_if_merged function This patch, unlike the earlier ones, doesn't use unportable select builtin.

hgmerge: fix diff+patch detection; cleanups Details: - actually show help message when no tools are found - whitespace cleanup (stupid emacs) - quote more variables - simplify merge/diff3 error handling

Merged RSS feed for tags from Peter van Dijk

add rss link to tags page

implement entriesnotip for tags in hgweb.py ; change entries to entriesnotip in templates/tags-rss.tmpl

tags-rss support; still shows tip

hg log --limit shouldn't count unprinted revisions (found by TK Soh)

Cleanup: Put initialization of br in front of --branches option handling.

Make option --branch vs. --branches more consistent. As you can abbreviate --branches to --branch, old scripts using the long name will still work.

Add option -b/--branch to hg tip (as done for hg parents).

merge with crew

fix typo in debugrebuildstate

Sunpro compiler patch The compiling runs through without warning, but runnig the newly builded hg emmits a message: | ImportError: ld.so.1: python: fatal: relocation error: | file /opt/local/lib/python2.3/site-packages/mercurial/bdiff.so: | symbol cmp: referenced symbol not found Removing the inline infront of cmp corrects this error message.

add options to log man page.

merge with crew.

add -l,--limit to log command.

add 'debugrebuildstate' to rebuild the dirstate from a given revision - added and removed files will be lost while recreating the dirstate - modifications are not lost

use repo.lock when cloning via copy, use lock.LockException when necessary

Raise a different exception when the lock is not available When the filesystem is read-only or if we have some other error, there is no need to wait. Raise a lock.LockUnavailable exception.

merge with crew

localrepo: refactor the locking functions

make test-pull-permission cleanup correctly in case of errors

fix warnings from pychecker (unused variables and shadowing)

bisect: fix for recent changes disallowing "false" merges

transitive copy/rename Add support for transitive copy or rename operations, e.g. hg rename a b ... (possibly edit b) hg rename b c hg commit ... by actually renaming "b" as "c" while recording a copy of "a" to "c" and a removal of "a".

Fixed misleading indentation in mpatch.c

merge with crew.

make server tests use --daemon.

Merge with upstream

Added help string for --daemon-pipefds

stop read end of pipe from staying open forever in child process.

add --daemon option to serve command. for issue 45. code looks odd because it is portable to windows. windows does not have os.fork, so have to spawn and use pipe to tell parent ready instead.

AmbiguousCommand is raised too soon. Right now, hg raises AmbiguousCommand as soon as it finds two commands/aliases that start with the substring it's searching for, even though it may still find a full match later on. This is a bit hard to hit on purpose, because hg checks the list of commands in whatever order is returned by table.keys(), which will change when you add an alias to a command. You should be able to hit it by adding an alias "u" to the "identify" command - not that that makes a lot of sense...

Don't hard-code /tmp in test-ro-message This makes the test succeed when $TMPDIR is set.

Merge with crew

add preoutgoing and outgoing hooks. preoutgoing lets prevent pull over http or ssh. outgoing lets notify after pull.

i forgot to change the test output for test-hook.

tests for new hooks. fix things i found when writing tests.

fix test-help diff options.

merge with crew.

add -p option to tip. for issue 64.

add prechangegroup and pretxnchangegroup hooks. prechangegroup lets you stop push, pull or unbundle before it begins. pretxnchangegroup lets you inspect changegroup before transaction is committed, and roll back if you not like it.

merge with crew.

add -w and -p options to diff. this is for issue 126.

fix names of parent changeset ids in hooks. fix hook part of man page.

prefix hook env var names with HG_. old names are still provided, but doc says they deprecated.

merge with crew

add a -b/--branch option to 'hg parents'

Fix diff against an empty file (issue124) and add a test for this.

catch errors and throw exception with invalid binary patch data

add pretxncommit hook. hook allows check of changeset after create, but before transaction is committed. hook failure rolls transaction back. makes place for local policies like commit message must contain bug id or reviewer signoff. change also adds parent changeset ids to commit hook environment, because is cheap and useful.

add pretag and tag hooks. pretag hook lets hook decide how tags can be named. tag hook is notifier.

hook parameters must be strings.

make hook code nicer. pre hooks can raise exception if fail. tidy up better if exception raised (like KeyboardInterrupt) when running hook program.

hooks run after transactions finish must not affect method results.

Refactor excessive merge detection, add test

Revert convert-repo changes

fix hgrc doc hook descriptions.

hooks: distinguish between locally committed and incoming changes. "commit" hook is now for locally committed changes only. "incoming" is for changes pulled, pushed, or unbundled into local repo.

Allow callers to pass in the dirstate lock in most localrepo.py funcs. This makes it possible to take the lock once and commit a large number of patches, without having to read and write the dirstate for each patch.

revlog.strip should clear the chunkcache The chunkcache isn't valid after a strip, it needs to be cleared out.

Show "(No help text available)" for commands without doc string. Currently this only occurs with "hg help qcommit" from mq extension.

Create default commit message in a more pythonic way.

Removed "HG: manifest hash" line from commit message.

Put the empty line in commit text above the branch merge line.

Run commit message editor in the repo root (like hooks). This makes the hgeditor script work with hg commit -R path/to/repo

add changenav at bottom of page if one reads through the changelog and wants to go to the previous or next page one needs the navigation links at the bottom, not at the top of the page

display revision numbers revision numbers are useful for orientation when browsing through the changelog

fix changenav for maxchanges != 10 for maxchanges = 25 the changelog navigation had a minimum stepsize of 30; try to calculate more useful navigation steps

add --mbox output to patchbomb add option to save generated mails in an mbox file for later processing with formail, instead of sending them directly

hgmerge: various cleanups Details: - put temporary file names into variables - make all temporary file names end with random part - cleanup FileMerge handling - do not use hardlinks in FileMerge change test (breaks on FAT) - try harder to keep file mtime unchanged in case of failed merge

hgmerge: add cleanup functions This patch adds functions cleanup, success and failure. The last two of these should be used instead of exit. Current code was changed to use them. It also moves $HGTMP to the top of the file (it's used in the cleanup function), changes the comment and removes now unneeded trap in the diff+patch merge.

hgmerge: add and use more tool variables This patch adds even more customisable variables with tool paths. Also moves $FILEMERGE where it belongs.

Update copyright notice

Merge with crew

Merge with Peter van Dijk

remove escaping from decriptions on repo index page

Added raw view for file annotate in hgweb.

hgweb/templates: remove unused vars from map-raw annotateline contains html and it is unused filenodelink is unused but should be defined

hgweb/templates: no need to escape the '<' '>' and '&' in raw templates

Removed executable bit from patchbomb extension.

Update version strings and add release notes to Windows specific files

Fall back to asciidoc6 html backend if html4 doesn't work. This reenables building documentation on Debian sarge, which broke after 9346a7fb3fe2.

Reverted name change (mercurial -> Mercurial) from dd5085897010.

merge with crew.

make stuff for macos x binary package. it is now easy to make prebuilt binary for macos x. i have no hosting space. i hope someone else can make binaries.

extension gpg.py: really raise the exception in case of invalid data

Cleanup of spacing in bash_completion

Use user specified path to hg in bash_completion e.g. for multiple installs with different versions or extensions.

fix an exception in gpg.py with multiples sigs for the same cset - silly error, do not overwrite the variable we use for iterating - correct 'hg help sign'

fixes for gpg.py extension - add copyright and license - add i18n - add 'sigs' command behaving like 'hg tags' - change 'role' to 'comment' and output it only if there is one - refactoring

Merge with crew

Merge with upstream

reverting 11d12bd6e1dcd9610fa26e97d25e7ad553e8ffa5 passing local everywhere violate the layering

cleanup of revlog.group when repository is local revlog.group cached every chunk from the revlog, the behaviour was needed to minimize the roundtrip with old-http. The patch export the information that the repository is local or not from the repository object down to the revlog. Then it uses the workaround for old-http only if the repository is non-local. The memory used server side when pulling goes down to less than 30Mo maximum whereas without the patch more than 160Mo was used when cloning the linux kernel repository. The time used by cloning is roughly the same (although some caching could be implemented if needed): before 110.25user 20.90system 2:52.00elapsed 76%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+708707minor)pagefaults 0swaps after 112.85user 22.98system 2:50.66elapsed 79%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+862862minor)pagefaults 0swaps

Added missing space in output of gpg.py

merge with crew

abort when using 'update -m' and this is not a merge add more testcases for updating with local changes

merge with self.

add documentation for email command.

forgot to add import statement for _.

add _ to several strings

turn patchbomb script into an extension module. command name is now 'hg email'.

[hgweb] Added rename information to gitweb templates & misc cleanups Fixed up file log template's title Removed double entry for filelogentry in gitweb map file

verify: notice extra data in indices

Added tag 0.8 for changeset 3a56574f329a368d645853e0f9e09472aee62349

Merge with crew

Safer version of FileMerge merge

return a non-zero exit code when a merge failed

Added missing quote in warning text to tests, too.

Added missing quote in warning text.

Fix revlog signature check for empty revlogs

Add revlog version signature check

doc: clone and outgoing point to pull for valid path docs

deprecate the rawcommit command

convert-repo: use commit rather than rawcommit

convert-repo: automatically create empty map file

deprecate the use of 'hg tag NAME [REV]' add some tests and some doc

hgweb: show copy/rename links in file history

unbreak templates/summary-gitweb.tmpl

Fix comment syntax

changed template escape filter, added urlescape filter; audited all templates for insertion bugs; added note to manifest.py about newlines in filenames

hgweb: fix breakage on manifest subdirs from path cleaning

Process working files through encode filters in dodiff()

Use of opendiff as merge program on MacOS X Hello, If one has the developer tools installed on MacOS X there's another alternative for the merge program. Invocation could be done somehow like in the attached patch for hgmerge.

Clean up paths passed to hgweb (spotted by Peter van Dijk)

add removed files to the changelog file list - this should allow better detection of removed file when walking in the history (like hg log) it doesn't help for the fast path of hg log where we only look at the filelog - users of the changelog file list shouldn't assume anymore that the file still exist (anyway it won't be found in the manifest like in 5ecf05541e11) - fix the tests (some hashes changed)

Remove duplicate bunidiff code from hgdiff, importing from mdiff.py instead

bash_completion: ignore hg help error messages They are usually not useful when you're completing stuff and end up cluttering the screen.

bash_completion: always use single quotes with compgen -W This avoids a bug in bash 2.05a

bash_completion: be more careful about whitespaces - use awk to parse the output of hg help. - print one completion candidate per line - print the debug commands after regular commands (this eases the shell side of the parsing) - don't print aliases that are simple abbreviations (e.g. up/update, id/identify)

Added hint to hg-ssh that you can use shell pattern matching.

_hg_status improvements in bash_completion: - Use "hg status -n" option instead of cut command. - Sort status flags accoring to "hg status --help" - Show deleted files instead of "hg remove"d for "hg remove"

bash_completion: small updates and fixes - $result in _hg_commands is not an array - use "hg status ." - update "hg cat" completion - let bash itself generate the default completion

Add new bdiff based unidiff generation.

contrib: add Chris Mason's stand-alone diff tool This uses Mercurial's diff algorithm to generate unidiffs like the traditional diff tool.

Fix rcpath for hgwebdir case (sys.argv is empty)

copy & rename can ignore removed source files As mercurial now distinguishes deleted files from removed files, we can skip removed source files while copying or renaming.

copy/rename '.' or '..' correctly

Merge with Eric

Merge with crew

This fixes a bug that Chris Mason found. As for a test case, I can't think of one. It's a very weird case. Basically, if there is a file listed as changed in the changelog entry, but not showing up in any of the associated manifest entries, hg would abort when trying to create a changeset. Now it just decides the file must not have any versions relevant to the changeset.

Copy manifest map before modifying it (see issue86)

[hgweb] Prevent search engine bots from indexing the entire repository

fix for hgweb.filterfiles some variable weren't converted, fix a possible out of range error

make the order of the arguments for filterfiles consistent fix a bug where the order was reversed

correct the relative path when walking from a subdir - yield the correct relative path when walking from a subdir (regardless the walk is repo-wide or not) - have a consistent output from all commands: filename are absolute when no patterns are specified and relative otherwise.

Added test for relative paths and all status flags for 'hg status'

Added test for the currently broken addremove with the expected output.

fix locate broken since 63799b01985c - fix the keyword arguments for walk in commands.locate - implement 'hg locate -r <rev>' - add a test for 'hg locate'

Restored old behaviour for 'hg co' and 'hg co -C' for removed files.

Adjusted all but one test output for the new behaviour of localrepo.changes()

Distinguish removed and deleted files. Tests are not fixed yet. hg status will now show "R filename" for "hg rm"ed files and "! filename" for files which were deleted manually. Manually deleted files are considered unmodified.

Renamed c, a, d, u to modified, added, removed, unknown for users of changes()

Make localrepo.changes() internally distinguish between removed and deleted.

Cleaned up localrepo.changes()

Cleanup of indentation, spacing, newlines, strings and line length

hgk.py is not a script (no '#!' line needed)

Added archives links to gitweb template

Adjusted test output to match new hgignore error messages.

make invalid pattern message not confusing.

if hgignore contains errors, print message that is not confusing.

eliminate backtrace when piping output on windows. this fixes issue 54.

document revert command more accurately. this fixes issue 87.

add '_' to several strings

- duplicate the parent link logic to show child links - don't show trivial parent or children links

merge with mainline

add --plain option to patchbomb. this omits the hg patch headers. it makes a patchbomb email look like a regular email with a patch.

make patchbomb work with recent changes to export

strictly adher to 80 chars per line

Error handling for unbundle, fix option line wrapping

add --update support to 'hg unbundle'

remove the gpg stuff from hgeditor (superseded by the signing extension) remove the gpg stuff from hgeditor, update the comments to make it clear it can be used to show a diff while commiting

cleanup of revlog.group when repository is local revlog.group cached every chunk from the revlog, the behaviour was needed to minimize the roundtrip with old-http. We now cache the revlog data ~4MB at a time. The memory used server side when pulling goes down to 35Mo maximum whereas without the patch more than 160Mo was used when cloning the linux kernel repository. The time used by cloning is higher mainly because of the check in revlog.revision. before 110.25user 20.90system 2:52.00elapsed 76%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+708707minor)pagefaults 0swaps after 117.56user 18.86system 2:50.43elapsed 80%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+748366minor)pagefaults 0swaps

remove a duplicate check in commit / fix a bug when merging files followed by revert

fix 'hg tag <tagname> <revision> - fix a collision with the option keyword rev and the option name - update the tests

Re-enable the renamed check fastpath

Merge with crew

if a filename contains spaces, patch adds quote around it

gpg signing extension for hg the principle is almost the same as how tags work: .hgsigs stores signatures, localsigs stores local signatures the format of the signatures is: nodeid sigversion base64_detached_sig sigversion 0 signs simply the nodeid (maybe we would like to sign other things in the future). you can associate fingerprints with roles in hgrc like: [gpg] fingerprint_of_a_key_without_spaces = release fingerprint_of_a_key_without_spaces = contributor, testing the key used for signing can be specified on the command line or via hgrc (key =) thanks to Eric Hopper for testing and bugfixing

do not copy atime and mtime in util.copyfiles it is unsupported on some fs (vfat) and we don't need to copy it one clone anyway

reloads manifest and changelog after recovering

fix a bug in hg export --switch-parent

Better error message (without /.hg appended) when repository is not found. This makes the error for missing default or default-push more readable.

Less annoying directory completion (see http://bugs.debian.org/343458) The current bash completion script is quite painful in conjuntion with deep directory trees because it adds a space after each successful directory completion. Eg. "hg clone /ho<tab>" is completed to "hg clone /home " when what you really want is "hg clone /home/" (assuming the complete path to the repository looks like /home/foo/hg...). That's because the 'complete' command does not know about the type of completion it receives from the _hg shell function. When only a single completion is returned, it assumes completion is complete and tells readline to add a trailing space. This behaviour is usually wanted, but not in the case of directory completion. I've attached a patch that circumvents this problem by only returning successful completions for directories that contain a .hg subdirectory. If no repositories are found, no completions are returned either, and bash falls back to ordinary (filename) completion. I find this behaviour a lot less annoying than the current one. Alternative: Use option nospace for the 'complete' command and let _hg itself take care of adding a trailing space where appropriate. That's a far more intrusive change, though.

Merge with upstream.

delete unintended print statement

merge with crew

make mercurial look in more places for config files. now it searches <install dir>/etc/mercurial, /etc/mercurial, and user hgrc. this allows site-wide configuration to be shared over automounted nfs partition, instead of chenging on every system. option of having local configuration on every system remains. old code for searching /etc/mercurial/hgrc.d never worked, this code is tested and works.

fix the cat command - improve localrepo.walk when passed a node - make the differents walk commands in commands.py accept a node - change commands.cat to walk over a revision - add a test

abort when merging two heads and repository has local changes

Merge http://sh0n.net/jeffpc/repos/hg-gitweb/

Remove tabs, and trailing whitespace from hgweb.py

[hgweb] Remove reference to a non-existing file

[hgweb] Fixed up bad link

File log and some FIXMEs

[hgweb] gitweb style: File annotate converted, file revision made more like the deafault style Namespace polution prevention (manifest vs. tagmanifest - shouldn't be a problem yet) File annotate updated File revision now has commit info

[hgweb] More gitweb style related fixes Fixed manifest link on tags page Fixed changeset link on manifest page

File revision style fixed

[hgweb] Initial import of the "gitweb" style

add a --switch-parent option to export against the other parent

hg log -p should show file deletions

quote the filename when calling HGEDITOR

make all commands be repo-wide by default - refactor code in commands.py so that: if no pattern is given, the command will be repo-wide else it will work on the working dir - update the doc - fix the tests

make command defaults works for aliases