ui: assumetrusted -> trust

ui: _get_cdata -> _data

config: getsource -> source

ui: simplify fixconfig

ui: fix-up and warn about deprecated %%

bugzilla: fix section reading

templater: use new config parser This gives us the ability to use includes and continuations

config: add section filter to read move section filtering out of ui.readconfig kill section filtering in config.update

config: deal with spaces at end of line more carefully

hgweb: kill parentui references

ui: kill most users of parentui name and arg, replace with .copy()

ui: replace parentui mechanism with repo.baseui

add cmdutil.remoteui remoteui sorts out the issues of getting ssh config options from the local repo into the remote one while not copying other options like hooks.

ui: less links to parentui

config: add some helper methods

config: split source data out into separate map

config: add %unset name support

config: allow including other config files

showconfig: show source file and line with --debug

util: kill configparser wrapper

hgweb: use config.config

clone: config escaping no longer needed

Merge with stable

Merge with crew-stable

transplant: forbid transplant to nonempty repositories with no working directory. Partially undoes d3fb413667e5.

ui.isatty() accesses sys.stdin, causing mod_wsgi to throw IOError.

hg: allow hg.parseurl(url, None) In many places hg.parseurl is called with a url and "opts.get('rev')", suggesting the second, optional argument can be None. Because opts['rev'] usually defaults to [] this never happens in practice. However, extensions don't necessarily behave the same, but do copy this pattern. Also, include wider hg.parseurl tests, beyond a demonstration of the problem.

transplant: remove the restriction that the destination be nonempty. Test fixed by Patrick Mezard <pmezard@gmail.com>

Fix subversion convert not detecting empty changesets. Need to check for the empty as list as well as None.

cvsps: fix crash when log message refers to non-existent branch (issue1615).

cvsps: add test for merge detection (issue1615). Currently only testing convert.cvsps.mergefrom; this script should also probably test convert.cvsps.mergeto.

convert/cvs: add an option to disable remote log caching Useful for debugging or testing when using --tmpdir.

tests: removed redundant "-u test" from test scripts The tests are executed with a .hgrc file which adds "-u test" by default.

tests: removed redundant "-d '0 0'" from test scripts The tests are executed with a .hgrc file which adds "-d '0 0'" by default.

Merge with crew-stable

convert/bzr: make it work with filemaps (issue1631) The bzr converter maintains a child -> parents mapping and drop entries whenever a child is read. It does not work with filemaps, getchangedfiles() may be called more than once when filtered files belong to merge revisions. getchanges() still works that way but it is not clear whether a similar issue can arise when interacting with merges.

verify: combine sets instead of concatenating lists

rebase, revlog: use set(x) instead of set(x.keys()) The latter is both unnecessary and slower.

monoblue: Fix indentation and other styling in the annotation view.

run-tests: cosmetics

patchbomb: add user agent header to mails

help: document stripdir template filter

templatefilters: add new stripdir filter Adds a new template filter for removing directory levels from a string. Examples: {foo|stripdir} -> 'foo' {foo/bar|stripdir} -> 'foo' {foo/bar/more|stripdir} -> 'foo/bar' {foo/bar/more|stripdir|stripdir} -> 'foo'

context: use Python 2.4 decorator syntax

localrepo: use sets in findcommonincoming

strutil: removed rsplit

notify: turned a set-like dict into a real set

revlog: let nodestotag be a set instead of a list

replace set-like dictionaries with real sets Many of the dictionaries created by dict.fromkeys were emulating sets. These can now be replaced with real sets.

util: use built-in set instead of util.unique

util: use built-in set and frozenset This drops Python 2.3 compatibility.

Merge with crew-stable

convert/bzr: fix symlink handling (issue1626)

win32text: be more careful about rejecting violating changesets We now try to walk changesets in reverse order from newest to oldest, so that if we see a file multiple times, we treat the newest version as canonical. This should prevent us from rejecting a changegroup that contains an unacceptable commit followed later by a commit that fixes the problem.

keepalive: attempt to fix issue1003 This is a reimport of the relevant piece of the upstream urlgrabber, which appears to be more correct.

test-archive: silence stupid messages from GNU tar Recent versions of GNU tar have apparently decided they're old enough that it's ok for them to prattle on senselessly about things no one cares about without anyone objecting. We object; apply duct tape.