Merge with crew-stable

test-convert-git: Use ~ instead of ^ Solaris sh interprets ^ as pipe symbol and the test thus failed in strange ways. git can however use of HEAD~ instead of HEAD^.

transplant: Add trailing LF in tmp file for filtering Transplant filtering used a tmp file ending with the (stripped) commit message and thus no final LF. Text files not ending with LF is usually not used on Unix and is thus wrong ;-) The missing LF had bad consequences because Solaris sed chokes on unterminated input lines, so echo -n foo|sed 's,o,0,g' doesn't output anything, and the filter used in tests/test-transplant thus stripped the last (and only) line in the message away on solaris.

Merge with crew-stable

convert: fix history topology when using hg.tagsbranch When using hg.tagsbranch, all conversions fail at transcribing the source repo topology with conversion scenarios interleaving: - initial repo conversion, - then upstream tag - then later incremental conversion The problem is caused by the usage of the last source revid as source revid for the tagnode in the revision map. On a later incremental conversion, when resolving some child commit of the true last source revid, convert will parent the nodes to the tag node. If using tagsbranch you get a bad topology where all incremental commit nodes end up in the tag branch. Analysis and test by Edouard Gomez <ed.gomez@free.fr>

Merge with crew-stable

merge: exclude common ancestor from --preview (issue1818)

inotify: _inotify.c: bugfix: remove_watch has to return PyNone on success We are not currently using that API function in inotify, hence the reason for the "silent" bug. But returning NULL here causes the interpreter to crash. Let's keep code clean for reusers :) The whole "bail" logic was unneeded here.

parsers.c: parse_manifest: fixing refcount of flags When flags was DECREF'ed, scope was referencing to the outer variable, outside of the block. It was in fact always NULL: the real Python object was never decref'ed.

test-highlight: remove redundant test Changeset 799373ff2554 is testing how a EUC-JP encoded Japanese file is treated in different encodings, so the ISO-8859-1 encoded German file is no longer needed.

Merge with crew-stable

highlight: fixes garbled text in non-UTF-8 environment This patch treats all files inside repository as encoded by locale's encoding when pygmentize. We can assume that most files are written in locale's encoding, but current implementation treats them as UTF-8. So there's no way to specify the encoding of files. Current implementation, db7557359636 (issue1341): 1. Convert original `text`, which is treated as UTF-8, to locale's encoding. `encoding.tolocal()` is the method to convert from internal UTF-8 to local. If original `text` is not UTF-8, e.g. Japanese EUC-JP, some characters become garbled here. 2. pygmentize, with no UnicodeDecodeError. This patch: 1. Convert original `text`, which is treated as locale's encoding, to unicode. Pygments prefers unicode object than raw str. [1]_ If original `text` is not encoded by locale's encoding, some characters become garbled here. 2. pygmentize, also with no UnicodeDecodeError :) 3. Convert unicode back to raw str, which is encoded by locale's. .. [1] http://pygments.org/docs/unicode/

Make distinct lookup error for localrepo.lookup This allows clone/share to correctly distinguish lookup errors from corruption errors and catch only the former.

doc: add TOC to hg.1.html and hgrc.5.html

log: prevent negative date range from displaying entire log (issue1805)

manifest/revlog: do not let the revlog cache mutable objects If a buffer of an mutable object is passed to revlog.addrevision(), the revlog will happily store it in its cache. Later when the revlog reuses the cached entry, if the manifest modified the object in-between, all kind of bugs appears. We fix it by: - passing immutable objects to addrevision() if they are already available - only storing the text in the cache if it's of str type Then we can remove the conversion of the cache entry to str() during retrieval. That was probably just there hiding the bug for the common cases but not really fixing it.

doc/README: simplify instructions

doc: use our own rst2man.py script (issue1746) The rst2man tool has not yet been part of an official Docutils release, and it is not present in most distributions. This poses a problem for people who want to install Mercurial from source, or who want to create a Mercurial package for such a distribution -- how to specify the build-dependencies? By including the rst2man.py script with Mercurial people only need a normal Docutils installation in order to install Mercurial.

util, minirst: do not crash with COLUMNS=0

manifestdict: remove unnecessary dictionary copy No need to copy the dict, dict.__init__() will do that for us. It was responsible for a non-negligeable waste of time during a qpush of an -mm queue on the kernel repo.

manifest.add(): cleanup worklist construction and iteration

manifest: simplify cache handling, use a unique cache

manifest.add(): simplify with iterators and generator expressions

color: do all setup in uisetup Now that extensions are loaded in phases, we are guaranteed to always find mq in uisetup with extensions.find and we know that the mq commands are not yet copied into commands.table.

extensions: pass ui arguement to extsetup if accepted

extensions: load and configure extensions in well-defined phases Extensions are now loaded with a call-graph like this: dispatch._dispatch extensions.loadall extensions.load # add foo module to extensions._extensions extensions.load # add bar module to extensions._extensions foo.uisetup(ui) bar.uisetup(ui) foo.extsetup() bar.extsetup() commands.table.update(foo.cmdtable) commands.table.update(bar.cmdtable) hg.repository foo.reposetup(ui, repo) bar.reposetup(ui, repo) The uisetup calls could easily be moved out to dispatch._dispatch, but have been kept in extensions.loadall since at least TortoiseHg calls extensions.loadall and expects it to call uisetup. The extensions.load function called uisetup. It now has an unused ui argument which has been kept for backwards compatibility.

highlight: move code from module top-level into extsetup

run-tests: standardize on --foo instead of -f/--foo We write -f/--foo in Mercurial help texts (docstrings), but I think it looks strange in error messages.

run-tests: use parser.error for cmdline option errors

doc/README: rst2man is now in Docutils proper

doc/Makefile: detect the right name for rst2man and rst2html docutils uses the .py extension on the commands, and so do their installer. Distribution packages might strip the .py, but the official name should work too.

Merge with crew-stable

test-keyword: update output to match 5d49fdef6fd0

hgweb: show diff header line in raw diffs

doc/Makefile: Don't show Makefile comments in output The comments makes sense when editing Makefile, but not when watching its output where it looks too much like error messages.

doc/Makefile: detect rst2man errors Make couldn't detect rst2man errors when it wasn't the last command, so empty man page files would be installed.

test-hooks: Don't hide the value of HG_PENDING Hiding the value of HG_PENDING made it less obvious what was going on in case of test failure

ui: fix NameError in ui.progress due to unit/units typo

util: quicker fspath, do not lower names when the length is different

test-casefolding: fix wrong case on hfs on linux on linux hfs keeps the old case of the file stored force it by renameing to another name and back

hghave: check for case insensitive filesystem in current dir check was done on TEMP

run-tests: expand --tmpdir and create it if needed

patch: remove the unused, broken reverse() function

patch: readgitpatch: remove unused variable 'src'

use 'x in dict' instead of 'dict.has_key(x)' "in" is faster, and has_key will be removed in py3k

churn: use genexps now that we dropped 2.3 compatibility

merge with crew-stable

churn: issue833 was reintroduced in 9bc46d069a76, correct it and add a test