hgweb.unbundle: call req.httphdr only after the last possible call to bail

Work around a urllib2 bug in Python < 2.4.2 When urllib2 base64-encodes the password needed for the Proxy-authorization header, it forgets to remove the trailing "\n". Later, a "\r\n" sequence is appended to every header, as required by the standard. Some proxies interpret the resulting "\n\r\n" sequence in the same way as "\r\n\r\n": end of headers. This usually doesn't cause trouble for this request, but when the proxy tries to read the next one, it thinks the request starts with some garbage and returns a "400 - Bad Request" error.

Fix confusing message from hg revert (issue332) % mkdir sub1 % touch sub1/file1 % hg add sub1/ adding sub1/file1 % hg rev sub1 sub1: No such file in rev e4c586763258 forgetting sub1/file1

Merge with stable

Merge with crew

revlog: attempt to gracefully handle some interleaved damage

merge with crew-stable

Merge with hg-stable

Merge with mpm

Merge with -stable

tags: add --remove

tags: remove unused variables

Make the tags algorithm smarter

Refactor tags code to prepare for improving the algorithm

merge with crew-stable

Only hg repositories override [paths], not simple directories (fixes issue510)

git-send-email compatibility: stop reading changelog after ^---$

mq: add qparent tag (first parent of qbase)

Merge with -stable, fix small test failure

Remove dummy branch hack from tests

Move branch read/write to dirstate where it belongs

Minor default branch cleanups

Move branches.cache to branch.cache Keeps old clients from conflicting with new caches, makes features logic unnecessary.

"default" is the default branch name

Make [PATCH] removal slightly more robust

Remove undo log after mq operations that rollback would break

Refresh file size after partial qrefresh. This makes a subsequent hg revert; hg status work.

Clarify update help text for update without an explicit revision

Improve hg branch -f help text

Add test for branch shadowing

branch: require --force to shadow existing branches

Strip [PATCH...] from message subject when importing patches

Merge a bunch of matcher and locate fixes.

Optimize return value of util._matcher for hgignore case

Optimize return value of util._matcher for common command line case This will trigger every time somebody runs something like "hg diff" or "hg status" without any arguments. The important part here is returning util.always as the match function, which is a much simpler (and faster) function than the usual return value, and allows other code to just skip the filtering if it knows all files will match.

remove unused "head" hack from util._matcher

locate: exit(1) if we didn't print any file

change locate to use relglobs by default This makes its default behaviour useful again (issue108), and changes it search the entire repository by default (instead of just the cwd), just like all other commands. It also hides issue204 by default, but you'll still see the same behaviour if you give it a relpath: pattern.

localrepo.walk: if we're walking a specific revision, sort the files This is more consistent with our behaviour while walking the working directory.

dirstate.statwalk: explicitly test for ignored directories This removes a hack where we appended '/' to a dirname so that: - it would not appear on the "dc" dict - it would always be matched by the match function This was a contorted way of checking if the directory was matched by some hgignore pattern, and it would still fail with some uses of --include/--exclude patterns. Things would still work fine if we removed the check altogether and just appended things to "work" directly, but then we would end up walking ignored directories too, which could be quite a bit of work. This allows further simplification of the match function returned by util._matcher, and fixes walking the working directory with a --include pattern that matches only the end of a name.

util._matcher: unify pattern normalization This should fix issue347. It also highlights one issue with the directory walking code when you have an --include pattern that matches the end of a filename. This is fixed by the next patch.

make the output of test-walk more readable

util.*matcher: change default "names" argument names=['.'] means "include (recursively) only files from the current subdir"; the function then did a hack to walk the whole tree. Clean that up. This also fixes a problem where "--include ." works in a subdir, but not on the tree root.

util._matcher: fix handling of path: patterns

util._matcher: fix handling of relglob: patterns

A 'glob:foo?bar' pattern determines a root - the tree root

Leave normalization of patterns to util._matcher Passing [] to util.cmdmatcher accidentally fixes walking of files with "\n" in the name.

util._matcher: update comments This should match the code better. Except for bugs...

remove obsolete code from patch.diff repo.status already does this filtering. If the caller supplies a set of changes, it's his responsibility to filter out what he doesn't want.

small globprefix fix

Merge with crew-stable

Test update to branch tip

Make update with no node jump to the tip of the current branch

fix qrefresh'ing an empty patch This is not superefficient (the commit will have to walk the whole tree), but it works.

statwalk: don't put self.root in the files list That list is supposed to contain internal, normalized names ("/"-separated, relative to the repo root), not OS-native paths.

fix qclone to a remote repo This fixes issue467. -stable still needs something like revision 55578a8d7e84 - mq: allow push if -r is given explicitly. In the meantime, hg qpop -a; hg qclone . ssh://... is a workaround.

Save an absolute path in repo.path This should keep things working if the process changes directories after creating a repo object.

merge with crew-stable

Add tests for .hg/branches.cache feature list.

Add a features list to branches.cache to detect caches of old hg versions. The leading space in the written file makes sure that the feature list never can match an existing version, even if the first feature can be read as hex. Additionally old hg versions display the id with --debug, too.

Store empty (default) branch in branch cache, too. Operations on this branch don't work otherwise. Reading branches.cache had to be adjusted to allow an empty label. Adjusted reading of the cache tip for symmetry, no functional change here.

small fixes for the parent patch - format.usestore is a boolean option - python wart: ("revlogv1") is a string, not a tuple - only create a dummy changelog if we're using a store - add a test

Merge with crew

Allow disabling store format to work with absurdly long filenames

hgweb: use the given revision in the name of the archive If you ask for an archive in hgweb by tagname the directory in the archive should include the tagname not the changeset-id.

incoming/outgoing: print the name of the other repo

localrepo.status: use manifest.find on "lookup" files

localrepo.status: fcmp gets a getnode function instead of the manifest

fix manifest.find

merge with crew-stable

Fix sed expression in test-notify to be more portable.

Fix help text for hg status -i

purge.py: fix invocation of statwalk

Delete the README for purge, putting the useful informations in comments

Delete the Purge class, refactoring Purge.purge() in dopurge()

Do not use 'self' in the purge() method Avoid any reference to 'self' in Purge.purge() to allow its refactoring in a simple function.

Use nested functions instead of object methods

Merge _remove_file() and _remove_dir() in a single function

Use "f in dirstate" instead of "dirstate.state(f) == '?'"

Reduce the absolute/relative path conversion in the purge extension

Make the purge extension use the statwalk walker from the dirstate object

Yield directories in dirstate.statwalk()

Add a test for the purge extension

merge with crew-stable

Don't truncate tag and branch names after 30 chars (introduced 73860ffbe798) with test added for a long (local) tag name.

patchbomb: Strip more than one trailing dot (and spaces between them)

patchbomb: Allow to specify subject of single-patch-series (issue475)

test-abort-checkin: use a hook to abort the commit This should make it more portable.

test-bad-extension: avoid printing the file name This should make it simpler on windows

make test-acl a bit more digestable for windows

Create the parent directory when checking out symlinks.

Fix addremove for symlinks to non-existent files

Avoid some false positives for addremove -s The original code uses the similary score 1 - len(diff(after, before)) / len(after) The diff can at most be the size of the 'before' file, so any small 'before' file would be considered very similar. Removing an empty file would cause all files added in the same revision to be considered copies of the removed file. This changes the metric to bytes_overlap(before, after) / len(before + after) i.e. the actual percentage of bytes shared between the two files.

Merge with crew-stable

mq: abort cleanly when invalid patch name is given to qguard

Close keepalive connections to fix server traceback

don't return uninitialized memory from bdiff.blocks() bdiff.blocks() returns a dummy match at the end of both files; the length of that chunk is never set, so it will sometimes contain random heap garbage. There are apparently workarounds for this elsewhere: # bdiff sometimes gives huge matches past eof, this check eats them,

web/server: disable address reuse option for BaseHTTPServer on windows On windows, a socket with the SO_REUSEADDR option set is able to bind to any port, even if there's already an active socket listening on it. test-http: check server address cannot be reused.

stdout raises EINVAL when flush() is called on a closed pipe under win32. Maybe the exception should be caught and translated at raise location instead (sshserver.py).

localrepo: stream_in may raise ZeroDivisionError with nul float elapsed argument.

streamclone: stream_out tried to decodefilename non-normalized paths

Add _xmlplus to the demandimport blacklist. (HGCIA was failing to load.)

Suppress <stdout> before hg export -v (regression from previous patch).

Let patch.export handle any fp-like object

merge with crew-stable

Unified #! paths for python scripts and removed them for test modules.

Use the mercurial i18n infrastructure in the purge extension

Print relative names in the purge extension

Merge with crew

Merge with stable

merge: fix renaming of subdirectories under renamed directories

Break core of repo.tag into dirstate/hook-free repo._tag for convert-repo

Use only integer part of mtime in workingfilectx.date(), fixes test-context.py Similar to what is done in dirstate.status() (changeset 882e703eaa94) and has to be removed when Mercurial switches to subsecond granularity.

Merge with crew

Merge with crew

Merge with crew

convert-repo: use .git/objects/ rather than .git/HEAD to detect git repos

Don't keep trailing white space in cvs tags

Change a bit cvsps arguments - Force updating the cache - Use internal cvs implementation, reducing overhead - Use quiet mode

Fixed workingfilectx.date() (found by Thomas Waldmann) with test.

merge with crew-stable

test-notify: adapt to sed shipped with MacOS MacOS X' sed does not have the \? repetition operator.

git binary patches: don't print the header for identical files

git binary patches: use hashes to detect identical files

add test for diffing identical binary files

Don't lie that "binary file has changed" Without -a option to "hg diff", mdiff.unidiff reported that "Binary file foo has changed" without even trying to compare things. Now it computes MD5 of old and new files, compares them and makes the conclusion.

add some tests for qpush/qpop error codes

Modify qpush/qpop idempotent operations to return success Modify qpush/qpop such that idempotent instances of the operations return success. That is, repeated qpush -a, qpop -a, qpush patch-name or qpop patch-name will return success. The end goal of each of these operations is to reach a particular state. Whether or not the patches were already applied does not affect that state or operation status. Likewise, be careful to retain the error status on a qpush/qpop without arguments when the end/beginning of the series has been reached.

mq: propagate the return error of pop

util_win32.py: fix user_rcpath Fixes issue504.

Combined the two os_rcpath methods into a single one near rcpath in mercurial/util.py The os_rcpath method delegates to system_rcpath and user_rcpath from the different platform implementations.

merge with crew-stable

hgweb: catch util.Abort raised by addchangegroup Right now, if a pretxnchangegroup hook fails, we send some HTML error message to the client and the transaction is not rolled back (issue499). Catching util.Abort allows us to send a decent message to the client and for some reason makes the rollback complete. This patch is not perfect since it doesn't fix the reason why the transaction wasn't rolled back (maybe some circular references?). Also, the transaction is aborted only after we've sent the response back to the client and the "transaction aborted" message ends up in the logs of the web server.

notify: don't try to fix addresses if notify.domain is not set

mail.py: don't try to use TLS if python doesn't have SSL support This should hide the traceback from issue501.

git patches: handle renames of binary files

hgweb/server.py: use hg.repository to create a repo object

qclone: don't call reposetup manually

extdiff: open files in binary mode Problem noticed by Darrell Gallion.

Make extdiff work at root directory

Fix util.shellquote on windows.

Fix accessing the repo through a symlink.

Fix hg showconfig traceback with values that aren't strings

hgweb: allow static files to be served directly by the HTTP server

Unified *_rcpath so the interface is similar across operating systems Changed os_rcpath to combine system_rcpath and user_rcpath. Changed system_rcpath and user_rcpath to both return a list of paths to add to the combined rcpath for both Windows NT and other platforms.

convert-repo converts symlinks from git

merge with crew-stable

hgwebdir: try to get web.style and web.motd from the ui.config system This finishes fixing issue253. As a bonus, web.style and web.motd settings from ~/.hgrc will be used for the hgwebdir index page.

Pass a ui from create_server to hgwebdir and a repo from hgwebdir to hgweb This allows repo pages to respect hg serve --webdir-conf <file> --style=gitweb (part of issue253). Since we're creating a ui object anyway, use it as the parentui of the ui objects created for every repo entry. This has the unintended side-effect that --name=foo on the command line will set the name of all repos. If one of the repos being served has a .hg/hgrc owned by a user that is not trusted, hg will now print the "Not trusting file..." warning when reading it. This is consistent with the behaviour from a hg serve from inside the repo.

hg serve: call setconfig on the parentui It makes sense to do this on the "global" ui object, since command line options should affect the whole process. This should fix hg serve --style=gitweb inside a repo (part of issue253).

Prevent type exception on concatenation if diffstat returns None. This will most often occur if diffstat is not installed in the target platform, though may also happen in other cases where diffstat fails to execute. Signed-off-by: Sean Dague <sean@dague.net>

Switch CGI stdout to binary on windows Problem diagnosed by Andrei Vermel.

Fix dirstate fail at drive root on Windows

Try to pass repo.ui to reposetup hooks The ui object we received in this function may belong to another repo, which could be confusing from the hook point of view. Trying to use the ui object from the newly created repo should avoid this confusion.

BeOS compatibility support

transplant: remote bundle source was not closed before deleting the fetched bundle. It fails under win32 because "normal" opened files cannot be removed.

mq: qinit -c creates a repo even after a regular qinit

Handle functions as the value of a hooks.<name> config variable This should make it easier for extensions to add a hook when they're loaded.

allow values that aren't strings in util.configparser

Merge with stable

Fix wrong module reference in copyfile exception

setup.py: reload __version__.py after writing it. write_version loads __version__.py before writing it (see 7eb694a1c1af), so we have to explicitly reload it to use the correct version. Problem diagnosed by Christian Ebert.

fix strip'ping the second parent of a merge

install reposetup hook right after loading the extension

Open bundle files in binary mode

convert-repo: handle packed git tags

lazyindex: handle __delitem__ in loadblock

commit: catch IOError filecommit can still raise IOError if we're on a filesystem without support for symlinks.

Merge with crew-stable

filecommit: don't forget the local parent on a merge with a local rename

Fallback to ascii if getpreferredencoding raises an exception Fixes issue478.

unbundle: don't use urllib if it's a local file

docopy: deal with globs on windows in a better way

Explicitly expand globs on Windows

mtime can be -1 after a merge and cause tracebacks on win32. Output "unset" instead.

Adjust documentation for 78a0dd93db0b (empty username to force specifying it)

hgwebdir: class hgwebdir should also accept a configparser instance

transplant: test non-local source

transplant: update to current writebundle API

mq: Mention usage of hg add/remove/copy/rename in qrefresh help text.

add pserver support to convert_repo

gitweb: prevent wrapping of "changeset | manifest" links

merge with stable

Abort on empty username so specifying a username can be forced. This behaviour was introduced by c3345b0f2fcd and broken by 98f2507c5551. Added test for this.

Copied files sources were not shown by status -C under Win32. repo.dirstate.copied() takes a normalized paths while local paths were passed.

add test for 540d1059c802

merge with upstream

hgk: remove embedded nulls in descriptions

mq: allow push if -r is given explicitly

Don't use ints in HTTP headers

Enforce unixish style for all generated patch names. User supplied names are left unchanged (including those loaded from series).

Make test-transplant test pull case

Update transplant's merge.update call

Minor test-transplant cleanup

merge with jsmith

use DESTDIR both places its needed

honor DESTDIR env variable for docs' make install

install man pages by default into $(PREFIX)/share/man not $(PREFIX)/man since the FHS puts them there

merge with -stable

bundlerepo: it was meant to be revdiff() instead of chunk()

patchbomb: fix timezone offset in message date header Use mercurial.util instead of time to set and format dates.

fix for digest auth when using keepalive.py The problem was with python > 2.3 which stores part of the headers in unredirected_hdrs. Furthermore, we simplify the code to use httplib directly. fix issue473

Subclass file with a __len__ method instead of setting Content-length This is necessary for the next patch to work with python2.3 Refactor the subclassing of send()

hgweb: add raw template to hgwebdir

Flush stderr after write. stderr can be buffered under win32 when redirected to a file, including stdout.

Merge branchname changes in localrepo.commit.

Use local encoding for "HG: branch" line in commit editor.

Add branch name to editor text.

When committing, record the changeset in the branchcache.

Allow adding symlinks that don't point to files

catch AttributeError in util.checklink There's no symlink function in the os module on windows. Fixes issue468

sync with -stable

Handle exceptions in do_hgweb: Send "Internal Server Error", log traceback

various doc fixes - update can no longer be used to merge - short summary start with lowercase - add merge to short help list - fix tests

catch socket.error fix issue394

fix handling of multiple Content-type headers This can happen if an error happens while sending application/mercurial-0.1 content. The error page will be sent resulting (for at least lighttpd) in the following headers: Content-type: application/mercurial-0.1 Content-type: text/html which result in req.proto = 'application/mercurial-0.1, text/html' fix issue344

add [diff] section to hgrc documentation

inst.reason isn't alway in the form (errno, strerror) urllib2.urlopen("foobar://foo") is an example where inst.reason is a string fix issue383

fix hg help <ext> for extension that do not define any command solve issue462

mq: fix strip failing to unlink a bundle os.unlink doesn't know how to deal with urls

symlinks: minimal support for symlinks in merge/update This will get angry, confused, and sad if you try to merge a symlink with a regular file.

symlinks: add flags param to wwrite Now wwrite can set symlink and exec for files

replace filehandle version of wwrite with wwritedata

unify encode/decode filter routines

symlinks: add _link checking function

symlinks: use is_link wherever is_exec is used

symlinks: don't complain about symlinks

symlinks: add a linkfunc

symlinks: add basic symlink functions to util.py

symlinks: check whether a filesystem supports symlinks

exec: remove last flag from is_exec

exec: add execfunc to simplify exec flag support on non-exec filesystems

pull umask-checking up

exec: checkexec checks whether filesystem supports exec flags

expanded bash_completion for transplant extension

Yield entries from [paths], too, when completing -R/--repository.

Corrected synopsis for transplant.

Improved error message for extensions overriding commands (with test): - Old text "module foo overrides bar" was misread as overriding module bar. - Only print a single warning line per extension instead of one for each command.

sync with stable

Don't use -f for rm in tests where not needed. Drop /bin/ from /bin/rm.