color: wrap qrecord The qrecord command is registered in uisetup phase, so we do our wrapping in extsetup phase.

record: use uisetup instead of extsetup to register qrecord New commands should be registered in uisetup so that other extensions have a change of wrapping them in extsetup.

merge: supply base node to merge tools in the environment Merge tools will be able to exploit this to correctly merge backouts. This won't work fully, though, until issue 1327 is solved, since the node information is not necessarily correct.

test-bisect: normalize shell error message Different shells on vary in their error messages. Here we're satisfied if the shell mentions './foobar' at all.

run-tests: add "debug" mode: don't capture child output, just show it. In debug mode, we: - don't try to diff expected vs actual output - disallow --interactive - disable timeouts (warn if user tried to supply one) - don't try to parse hghave output, since we don't have the output - don't try to save output to .err file

run-tests: make --tmpdir option more useful. - instead of creating HGTMP inside tmpdir, now HGTMP is tmpdir (thus, fail if tmpdir already exists) - passing --tmpdir automatically turns on --keep-tmpdir

hg.1: kill spurious backslash I believe the backslash prevented asciidoc from automatically turning "(C)" into a real copyright symbol. This replacement is not done in reST in the first place.

help: markup email addresses as inline literals Otherwise we get useless links to user@example.org in the HTML version of the man pages.

doc/Makefile: add help/*.txt as a dependency

doc: add some internal links to man pages

hgrc.5: combined short paragraphs into one Short and precise sentences are good, but we can have more than one such sentence per paragraph.

doc: use reST note directive The 'NOTE: bla bla' syntax was for asciidoc and is still present in a couple of docstrings. The docstrings will be converted to reST format when minirst knows how to handle it.

hgrc.5: capitalize True and False consistently

commands: slightly better help for --traceback

hgrc.5: document ui.traceback setting

fix tests output changed by 8269fe2d48f6

byterange: backport fix from upstream

hgweb: send proper error messages to the client Fixes a bug in protocol which caused an exception during exception handling in some cases on Windows. Also makes sure the server error message is correctly propagated to the client, instead of being thrown away.

transaction: always remove empty journal on abort When transactions without entries were aborted, the journal (of size 0) was not unlinked, which prevents subsequent operations until hg recover is run on the repository. We also make sure the journal is unlinked when committing, even if the provided hook doesn't do so.

test-bisect: fix shell-dependent output

change 'journal already exists' to 'abandoned transaction found' Journal already exists is a pretty internal piece of information, which doesn't necessarily mean much to people who are not familiar with the code. The new text is a more well-known concept.

verify: report existence of journal

bisect: no need to save the state if it wasn't changed

record: get the right position (account for skipped files)

log --limit: break after a limited number of csets (broken by f3d60543924f)

transaction: more specific exceptions, os.unlink can raise OSError

lock: catch specific exceptions

patch: use the public ctx API instead of the internals

patch: don't use mutable object as default argument

patch: initialize all attributes of the hunk class

patch: simplify logic

lock: use '==' instead of 'is' for integer equality ('is' may not work)

pychecker: remove unused local variables

remove unused imports

changelog: do not use a mutable default value

bundlerepo: keep the bundlerevlog interface in sync with revlog

localrepo/branchcache: remove lbranchmap(), convert users to use utf-8 names We don't need a "local-charset" aware branchmap() function, we can convert the names when needed during the output.

localrepo/branchcache: kill unused localrepo.branchcache The UTF-8 version of the branchcache is the master, always up-to-date, we can rename it to localrepo._branchcache.

localrepo/branchcache: rebuild the "charset-aware" branchcache when needed It's simpler and the two cache weren't always in sync otherwise: localrepo.branchcache would not be updated after calling branchmap()

churn: fix --progress (broken by f3d60543924f)

transfer branchmap branch names over the wire in utf-8

churn: fix changeset count (broken by 9b127e888640)

churn: ability to display added/removed lines separately

cmdutil: fix bug in finddate() implementation

merge changes from mpm

cmdutil: update finddate() walkchangerevs() now returns contexts, does not take a cache func.

walkchangerevs: drop ui arg

walkchangerevs: reset cache between windows

log: tidy up some filter tests

walkchangerevs: move 'add' to callback Now walkchangerevs is a simple iterator over contexts

hgweb: added test case for extension loading phases (issue1824) this checks the order of module loading phases of hgweb. `4) reposetup' lines are duplicated because hgweb calls hg.repository() twice, one by __init__, another by refresh.

extensions: changed to call extsetup() from extensions.loadall() previously uisetup() was invoked by extensions.loadall(), but extsetup() was by _dispatch(). there's no need to split them because we have nothing to do between uisetup() and extsetup(). this fixes issue1824 indirectly.

hgrc.5.txt: mention hgrc categories with higher precedence first Reordering the FILES section accordingly. The previous ordering of categories might have been nice from the viewpoint of a site admin doing an initial install, but presenting a higher-precedence-first ordering is more relevant and natural for the average end user, since he will most likely resort to editing rc files in the order of their precedence, overriding whatever "sane" defaults are coming from more general files. Note that this patch does not change the texts, it just moves them. So, whatever bugs, grammar errors, or typos may have been in the texts before this patch: they are still there. On purpose. Because this patch here does not want to reword texts while moving them.

hook: only redirect stdout if it and stderr are valid files When using hgwebdir with WSGI via the IIS ISAPI-WSGI extension, both stdout and stderr filenos are set to -2, which makes the os.dup call in hook.py fail.

verify: filter the candidate list for broken linkrevs

Merge with crew