tests: test that the pid returned by `hg serve` looks reasonable This failed on windows before win32.spawndetached has been fixed. The process name was "cmd.exe" and not "hg.exe" or "python.exe".

win32: spawndetached returns pid of detached process and not of cmd.exe win32.spawndetached starts the detached process by `cmd.exe` (or COMSPEC). The pid it returned was the one of cmd.exe and not the one of the detached process. When this pid is used to kill the process, the detached process is not killed, but only cmd.exe. With this patch the pid of the detached process is written to the pid file. Killing the process works as expected. The pid is only evaluated on writing the pid file. It is unnecessary to search the pid when it is not needed. And more important, it probably does not yet exist right after the cmd.exe process was started. When the pid is written to the file, waiting for the start of the detached process has already happened. Use this functionality instead of writing a 2nd wait function. Many tests on windows will not fail anymore, all those with the first failing line "abort: child process failed to start". (The processes still hanging around from previous test runs have to be killed first. They still block a tcp port.) A good test for the functionality of this patch is test-treediscovery.t, because it starts and kills `hg serve -d` several times.

revset: minor changes adding baseset to revsets Changed bits of code to work with baseset implementations.

tests: rewrite path in test-shelve.t for not being mangled on msys msys (on windows) converets '-R bundle:.XX/XX' to '-R bundle:;.XX/XX'. Avoid this by writing '-R bundle://.XX/XX'. This is used more often than the alternative work arounds like '-Rbundle://.XX/XX' or '-R bundle:Y/../.XX/XX'.

hooks: only disable/re-enable demandimport when it's already enabled This fixes an issue introduced in d7c28954d901 where, when disabling demandimport while running hooks, it's inadvertently re-enabled even when it was never enabled in the first place. This doesn't affect normal command line usage of Mercurial; it only matters when Mercurial is run with demandimport intentionally disabled.

merge with stable

convert: secret config option for disabling debugsvnlog Subversion issues involving svn log such as d54467c1a198 can be tricky to debug when it is run in an 'hg debugsvnlog' sub process. Debugging is simpler when convert only uses one process. With this change convert will invoke the svn log directly when setting [convert] svn.debugsvnlog = False This is intentionally not documented.

convert: make subversion revsplit more stable when meeting revisions without @ revsplit would crash for instance if given a subversion string without @ ... and that could somehow happen when playing around with convert.

revset: minor changes adding baseset to revsets Changed bits of code to work with baseset implementations.

revset: added __add__ method to baseset class

revset: added docstring to baseset class

merge: don't overwrite file untracked after remove, abort with 'untracked files' Merge could overwrite untracked files and cause data loss. Instead we now handle the 'local side removed file and has untracked file instead' case as the 'other side added file that local has untracked' case: FILE: untracked file exists abort: untracked files in working directory differ from files in requested revision It could perhaps make sense to create .orig files when overwriting, either instead of aborting or when overwriting anyway because of force ... but for now we stay consistent with similar cases.

shelve: better (and slightly redundant) test coverage for unshelve conflicts

shelve: status messages from unshelve It was hard for the user to know what was going on when unshelving - especially if the user had to resolve conflicts and thus got to see the intermediate states. Seeing that pending changes was gone could scare the user, make him panic, and do stuff that really made him lose data. Merging (both when rebasing and with pending changes) also requires some understanding of where in the process you are and what you are merging. To help the user we now show a couple of status messages (when relevant): temporarily committing pending changes (restore with 'hg unshelve --abort') rebasing shelved changes

shelve: be quiet when unshelve pulls from the shelve bundle unshelve was quite verbose and it was hard for a user to follow what really was going on. It ended up saying 'added 1 changesets' ... but the user just expected and got pending changes and never saw any changeset. The use of bundles is an implementation detail that we don't have to leak here. Pulling is quite verbose, optimized for pulling many changesets from remote repos - that is not the case here. Instead, set the quiet flag when pulling the bundle - not only when temporarily committing pending changes. The 'finally' restore of ui.quiet is moved to the outer try/finally used for locking.

shelve: add 'changes to' prefix to default shelve message The shelved changes _could_ perhaps be amended to the parent changeset but it _is_ not the parent changeset. Using the description from the parent changeset is thus wrong and confusing. Instead, add a 'changes to' prefix.

shelve: mention FILE options in help Shelve do normally take a list of files or patterns to shelve and the command summary should thus show [FILE]... Note: --delete is a bit special and interpret the parameters as a list of shelve names. This change makes that even less obvious from the help. Too bad - we can't please everyone.

shelve: mention walk options in help

shelve: really pass publicancestors to changegroupsubset - not the parents publicancestors returned the parents of the public ancestors ... and changegroupsubset used the parents of these as base for the bundle. That gave bundles with one layer of changesets more than necessary.

shelve: publicancestors do not have to visit nullrev Not visiting nullrev seems more correct and might be a minor optimization.

localrepo: make it clear that changegroupsubset doesn't take bases but roots changegroupsubset will take the parents of the roots to find the bases. Other parts of Mercurial do not expect that, and a result of that is that some bundles contain more changesets than necessary. No real changes here - just renaming a parameter to document what it is.

repoview: use repo.revs() instead of a private revset method Breaks an import cycle.

commands: use ctx.getfileset instead of fileset.getfileset

push: backout 2bc520bd0ce0 due to test failures

mq: fix qimport url check

match: use ctx.getfileset() instead of fileset.getfileset() Resolves an import cycle involving match and merge.

context: add a getfileset() method so fewer things need fileset directly Allows me to break an import cycle in a subsequent change.

fix_bytesmod: fix defects in fix_bytesmod so it produces working code Prior to this change, the fixer did not always add the py3kcompat import to files, and did not manage to blacklist itself.

discovery: improve "note: unsynced remote changes!" warning This note (which actually is a warning) frequently caused confusion. "unsynced" is not a well established user-facing concept in Mercurial and the message was not very specific or helpful. Instead, show a messages like: remote has heads on branch 'default' that are not known locally: 6c0482d977a3 and show it before aborting on "push creates new remote head". This will also give more of a hint in the case where the branch has been closed remotely and 'hg heads' thus not would show any new heads after pulling. A similar (but actually very different) message was addressed in 6b618aa08b6e.

convert: drop unused getheads from sinks

tests: on windows, run command explicitly in sh for working command substitution On windows, os.system runs the command in cmd.exe, which does not know command substitution. To let this work, run the command explicitly in sh. This fixes the failure of this test on windows.

tests: test-module-imports.t works on windows (with backslash path sep) Transform the backslashes to slashes on windows, because they are eaten up by the shell. (Similar as done in test-check-pyflakes.t.) This test will not produce a traceback on windows anymore. But the test still fails because the detected mixed imports differ.

qimport: allow importing URLs

revset: fixed bug where revset returning order was being changed Some revsets were innecesarily turning the subset into a set before iterating over it. This led to returning order changing in some cases.

itersubrepos: move to scmutil to break a direct import cycle

import-checker: handle standard modules with arch in the filename Installations with module names like /usr/lib/python2.7/lib-dynload/bz2.x86_64-linux-gnu.so occurs in the wild. Let's just ignore everything after first '.' when guessing the Python module name.

tests: use ls instead of find, all files are in the same directory In this case find has no advantage compared to ls. Descending into directories is unnecessary, because there are none.

commands: hg debuginstall checks missing templates (issue4151) Missing templates where not reported as a problem, only an empty bracket were shown as indication of no found template directory: $ hg debuginstall *...some lines* checking templates ()... *...some lines* no problems detected Now the problem is reported and extended with some information. The style of the messages is adapted to the other messages of debuginstall. When no templates directories exist, it writes: $ hg debuginstall *...some lines* checking templates ()... no template directories found (templates seem to have been installed incorrectly) *...some lines* 1 problems detected, please check your install! When the template map is not found, it writes: $ hg debuginstall *...some lines* checking templates (/path/to/mercurial/templates)... template 'default' not found (templates seem to have been installed incorrectly) *...some lines* 1 problems detected, please check your install! When the template map is buggy the message is the same as before. The error message is shown before the line "(templates seem ...)". No test is added because testing this failure is complicated. It would require to modify the templates directory of the mercurial installation, or to monkey patch a function (os.listdir or any from mercurial.templater) by a test extension.

contrib/Makefile.python: build local Python from source This provides an easy way to install new or old Python versions on Unixish systems. It is mainly intended for testing Mercurial with different Python versions. Example: $ cd build $ make -f ../contrib/Makefile.python python PYTHONVER=2.4 PREFIX=/tmp/p24 $ /tmp/p24/bin/python -V Python 2.4 Tested on Fedora, Debian and Ubuntu with Python versions from 2.4 to 2.7.6. Most Python versions need some workarounds to compile on Debian-ish systems. The workarounds do not do any significant harm on Fedora.

tests: use small conditional section for tic requirement in test-status-color.t Better test coverage when curses not is available.

import-checker: show stdlib and relative imports separately Make the output more useful for debugging problems.

tests: improved test coverage for HTTP authentication and 401 responses Dump the access log to verify that the client gets the 401s it deserves and handles them correctly. This establishes a baseline for Basic authentication protocol.

tests: ignore http tests that are known wontfix failures on python 2.4 As mentioned on http://bz.selenic.com/show_bug.cgi?id=2739#c17 , b3083042bdda left test-http.t failing on Python < 2.4.3. That has not been noticed because most 2.4 testing has been done with a patched 2.4.2. This makes sure that the tests only are run for Python >= 2.4.3. That makes it possible to verify that everything else works with Python 2.4.

import-checker: fix names of dynamically loaded modules The import checker found standard library modules such as lib-dynload/zlibmodule.so but saw that as a 'zlibmodule' module, not as the 'zlib' module. Debian ships Python with most modules built-in and this incorrect handling of dynamic modules did thus not cause problems on that platform. Fedora ships Python with as many modules as possible loaded dynamically. That made the import checker tests fail with incorrect classification of the following modules: array fcntl grp itertools time zlib. This change makes test-module-imports.t pass on Fedora.

tests: for consistent output flush between writing sterr and stdout Because stdout and stderr are buffered, the order of the output is other way around on some systems.

discovery: cleanup of variable names and comments

localrepo: give a sigh of relief when getting lock after waiting for it A message like this was sometimes shown when pushing: remote: waiting for lock on repository foo held by 'mercurial:20858' That could scare users, making them wonder whether the push actually succeeded. To mitigate that fear, issue an additional "warning" such as: got lock after 2 seconds The return value from lock.lock.lock() was unused - instead we return the delay. This also adds the first test coverage for waiting for locks.

convert: add tagmap option Tests have been updated.