fix util.canonpath on windows. old test for absolute path was not portable. use os.path api instead.

make appendfile simpler so it does not break with revlogng on windows. it used to cache open files. this made revlogng break because it wants to rename files when splitting .i into .i/.d, but cannot rename or unlink open files on windows. new code is bit slower, but safe on linux and windows. proper fix for too many open/close of changelog/manifest belongs in different place. can get 10% speed improvement back.

Exit test-hup only after the background process has died. Otherwise the open file might let the tests fail on NFS.

merge with mainline.

mq: do not fail if directory to create exists

merge revlogng with mpm tip

Fix transaction handling during interleaved file split.

merge 0.8.1 with revlogng

Additional appendfile fixes for interleaved data/index files The appendfile code was not passing default version info to the changelog or manifest classes, and so they were always being created as version 0. revlog.checkinlinesize had to be corrected to seek to the end of the index file when no index file was passed (only clone does this)

Speedup revlog.ancestors for the linear case revlog.ancestors can be expensive on big repos. This cuts down the overall time for hg update by ~19% by short cutting revlog.ancestors when one of the revisions is reachable from another.

Reduce index memory usage by storing the bare string instead of tuples Storing the tuple returned by struct.unpack significantly increases the memory required to store the entire index in ram. This patch uses struct.unpack on demand instead.

New lazy index code for revlogs. This tunes for large repositories. It does not read the whole index file in one big chunk, but tries to buffer reads in more reasonable chunks instead. Search speeds are improved in two ways. When trying to find a specific sha hash, it searches from the end of the file backward. More recent entries are more likely to be relevant, especially the tip. Also, this can load only the mapping of nodes to revlog index number. Loading the map uses less cpu (no struct.unpack) and much less memory than loading both the map and the index. This cuts down the time for hg tip on the 80,000 changeset kernel repo from 1.8s to 3.69s. Most commands the pull a single rev out of a big index get roughly the same benefit. Commands that read the whole index are not slower.

Fill in the uncompressed size during revlog.addgroup This uses code from Matt to calculate the size change that would result from applying a delta to keep an accurate running total of the text size during revlog.addgroup

Fix inlined revlogs to seek to eof after opening "a+"

Create an atomic opener that does not automatically rename on close The revlog.checkinlinesize() uses an atomic opener to replace the index file after converting it from inline to traditional .i and .d files. If this operation is interrupted, the atomic file class can overwrite a valid file with a partially written one. This patch introduces an atomic opener that does not automatically replace the destination file with the tempfile. This way an interrupted checkinlinesize() call turns into a noop.

Make the appendfile class inline-data index friendly The appendfile class needs a few changes to make it work with interleaved index files. It needs to support the tell() method, opening in a+ mode, and it needs to delay the checkinlinesize call until after the append file is written. Given that open(file, "a+") doesn't always seek to the end of the file, this adds seek operations to appendfile that understand whence args

Fix bundle repos to use an index tuple consistent with revlogng The bundle repo code was adding a field to the index tuple, which confused the revlogng changes. This creates a new dict instead to maintain the extra bundle info.

Implement data inlined with the index file This patch allows you to optionally inline data bytes with the revlog index file. It saves considerable space and checkout time by reducing the number of inodes, wasted partial blocks and system calls. To use the inline data add this to your .hgrc [revlog] # inline data only works with revlogng format=1 # inline is the only valid flag right now. flags=inline

Implement revlogng. revlogng results in smaller indexes, can address larger data files, and supports flags and version numbers. By default the original revlog format is used. To use the new format, use the following .hgrc field: [revlog] # format choices are 0 (classic revlog format) and 1 revlogng format=1

import: use gpatch if present on system. patch is broken on solaris. fixes issue 205. add new useful function, util.find_in_path.

merge with crew.

merge with crew.

let run-tests run optional code coverage tests. to enable, set env var TEST_COVERAGE to non-empty string. run-tests will print report after running. after coverage run, file tests/.coverage will contain coverage data. use tests/coverage.py to get more info, like annotated source code.

fix bugs in coverage.py.

add coverage.py module to tests directory. written by ned batchelder and gareth rees.

fix update when a locally added file match the target revision the dirstate entry for the added file wasn't updated. fix issue207

util.unlink should only catch OSError.

benoit asked for comment to make avoid of recursive repo clearer.

merge with crew.

support nested repositories. mercurial will not recurse into a subdirectory that contains a .hg directory. it is treated as a separate repository.

Don't use -u option for diff on SunOS, too. This makes the tests work without using GNU diff on Solaris 8.

run-tests: Use appropriate (that is: none) diff options on HP-UX.

Use the new format of setting tags in view extension (Debian bug #361897)

Never exit directly from commands.dispatch(), but pass return code to caller. Usually the caller is commands.run(). Some extensions still use sys.exit(), this is catched, too. Fixed wrong return statement in commands.recover() yielding a zero exit code.

Polished exception handling when importing extensions: - Warning about failed import is printed on one line. - For this warning the [extensions] section key is used, because the value doesn't have to be set. - Whan a --traceback is requested, exit immediately to not mix up problems when importing an extension with other problems.

Added doc/gendoc.py to MANIFEST.in, so docs can be regenerated on demand.

fix exception handling on windows. move win32 code into util_win32.py.

Include tests/md5sum.py in manifest.

Added tag 0.8.1 for changeset 6a03cff2b0f5d30281e6addefe96b993582f2eac

Try to use /usr/bin/test or /bin/test to work around missing features. The builtin "test" of Solaris 8 /bin/sh doesn't understand -nt.

util.copyfiles: only switch to copy if hardlink raises IOError or OSError. before this, interrupting clone would give "file a same as file b" error because caught KeyboardInterrupt.

Make completion for debugindex<tab><tab> show debugindexdot, too. The special handling for commands with names that are substrings of other commands (e.g. with st and strip) wasn't used with debug commands before.

Set correct exception for another possible malloc error in mpatch.c

Register qversion as a non repository related command

Show remote ssh noise only with --debug and increase the limit to 500 lines. Messages can already be sent to the user over stderr, if one really wants this.

document the possible race in "hg undo"

Don't allow --style and --template for hg update and hg merge. The only way that changesets will be shown with these command is in an error message when trying to update to an ambiguous branch name.

Minor fix for revert: No need to make a backup when forgetting added files. echo foo > foo && hg add foo && hg revert foo creates foo.orig, though foo is not changed or deleted.

small changes to revert command. fix bug 93: work with files that are in target manifest but not dirstate.

bash_completion: completion for commands provided by extensions Make the bash_completion function call _hg_cmd_$cmd to generate completion candidates for $cmd if that function exists. Add basic completion functions for: - mq: - qpop - qpush - qdelete - qsave - qcommit - strip - hbisect: - bisect - patchbomb: - email - gpg: - sign

Don't enter an endless loop if remote hg doesn't answer, show remote noise. The endless loop is not only triggered if the remote shell is too noisy, but even if the local ssh command dies due to an error.

bash_completion: small optimization Right now we always call "hg help $cmd" to get the canonical name of $cmd (i.e. to go from "co" to "update"). This patch optimistically assumes that $cmd is already the canonical form and tries to generate completions for it. If that fails, it falls back to canonicalizing $cmd and trying again. This means that: - if a command or alias is explicitly handled by the _hg_command_specific function, things get somewhat faster - as long as the canonical $cmd is handled by _hg_command_specific, all its aliases and abbreviations are also handled.

Tell the user the real "hg update -C" command, not just "-C" when aborting.

Merged with crew.

Don't abort when backup files already exist. Don't allow alternative names. Rationale: - When the user wants to revert, he shouldn't be stopped from doing this just because some old backups will be overwritten. - To not clobber important files by accident, alternative names for backup files were disabled. As the backup target now has a fixed name, the user doesn't have to be informed about the backup copy (unless --verbose)

Make 'hg tags -q' only list tag names without revision numbers and hashes, and change bash_completion to use this.

add --options to debugcomplete and change bash_completion to use it make debugcomplete print one item per line (this is not needed for the bash_completion script, but should be easier to use in other scripts)

Group changes done by the same developer on the same with --style=changelog Changeset and tags are appended to the change message for non-quiet and non-verbose output, so grouping works. Fixes last bit of issue110.

Don't print filenames in braces for changelog style.

Web site and wiki are now the same.