dirstate: make parents() faster. This allows localrepository.status() to avoid reading all but the first 40 bytes of the dirstate in the common case of running "hg status".

localrepository.status: only acquire wlock if actually needed. This speeds up the common case of not needing to update the dirstate, and avoids the need to reload and parse the dirstate "just in case".

util._matcher: speed up regexp matching. In 4babaa52badf, Benoit made a change that substantially slows matching when a big .hgignore file is in play, because it calls into the regexp matching engine potentially hundreds of times per file to be matched. I've partly rolled back his change, so that we only call into the matcher once per file, but preserved the ability to report a meaningful error message if there's a syntax error in the regexp.

Merge with crew-stable

Work around python 2.5.1 tarfile regression

Merge with crew-stable

Avoid float rounding errors when checking http protocol version.

Work around problem where 'del os.environ["FOO"]' doesn't update environment. test-hook failed since changeset 949516072388 because of this on Solaris 8.

Use python instead of shell printf with \x for test-simplemerge-cmd Needed e.g. to be compatible with Solaris.

Merge with crew.

change tests to use simplemerge by default

polish the simplemerge command; add a test

actually port simplemerge to hg - use bdiff instead of patiencediff; this is a larger change, since bdiff works on 2 multi-line strings, while patiencediff works on 2 lists; - rename the main class from Merge3 to Merge3Text and add a Merge3 class that derives from Merge3Text. This new Merge3 class has the same interface from the original class, so that the tests still work; - Merge3 uses util.binary to detect binary data and raises util.Abort instead of a specific exception; - don't use the @decorator syntax, to keep python2.3 compatibility; - the test uses unittest, which likes to print how long it took to run. This obviously doesn't play too well with hg's test suite, so we override time.time to fool unittest; - one test has a different (but still valid) output because of the different diff algorithm used; - the TestCase class used by bzr has some extras to help debugging. test-merge3.py used 2 of them: - log method to log some data - assertEqualDiff method to ease viewing diffs of diffs We add a dummy log method and use regular assertEquals instead of assertEqualDiff. - make simplemerge executable and add "#!/usr/bin/env python" header

Import 3-way merge code from bzr merge3.py is imported as contrib/simplemerge test_merge3.py is imported as tests/test-simplemerge.py

add mdiff.get_matching_blocks

Merge with stable

Merge with crew

merge: fix quadratic behavior in find-copies

Merge with crew-stable.

test-command-template: improve test coverage

command line templates: add formatnode filter This will print a full hash node with --debug and a short one otherwise. Use it with some default templates and in map-cmdline.default to print the parents. This should fix issue538.

fix 'hg <not-log> -v --template foo' with revisions without copies

Merge with stable

hgweb: fix rfind bug in PATH_INFO handling

Merge with crew

Merge with stable

Merge with crew

hgweb: make PATH_INFO handling slightly more robust

Merge with crew

Merge with Matt

Merge with stable

Add graphlog extension

Merge with crew-stable

mq: fix qtop when some patches beneath are guarded

mq: test qtop with holes in the patch sequence.

Merge with crew-stable

minor typo fix in templater's docstring

Merge with crew-stable.

Add test for issue 529 - "mq aborts when merging patch deleting files".

mq: don't abort when merging a patch that removes files

atomicfile: don't copy the original file if it'll be truncated

use atomictemp files to write the dirstate

use atomictemp files to write branch.cache

util.opener: if requested, use atomicfile even if the file doesn't exist Right now, surprisingly enough, if you request an atomic file but the file still doesn't exist, you get a regular file. AFAICS, the only time this happens is during the initial creation of the dirstate.

cache os.umask even on windows

pull umask-checking up

Merge with crew-stable.

help: log --keyword is case-insensitive

hgweb: expand keyword search to full list of files

log: expand keyword search to full list of files

run-tests.py: use coverage.py with "#!/usr/bin/env python" tests

run-tests.py: small cleanup

run-tests.py: use coverage.py with *.py tests

run-tests.py: tell coverage.py to ignore errors Otherwise there'll be some IOErrors when it tries to open python files created during the execution of e.g. test-hook.

test-bad-pull: try to avoid timing-related failures

Try to find diffstat in PATH before calling it At least on Solaris, /bin/sh will print a "command not found" message even if we redirect stderr.

Add a pointer to "hg -v help" to the bottom of hg help {,cmd} output This should make it easier to discover global options. As a bonus, they are no longer displayed by something like a hg -v help dates

revlog.py: fix/tweak read ahead code in lazyparser

Merge with Matt

Merge with stable