tests: give up and make setsockopt() calls optional in the output I can't figure out what causes these calls to happen or not, and I weary of dealing with them. End the madness for now by marking them all as optional lines of output. Differential Revision: https://phab.mercurial-scm.org/D5762

py3: fix test-remotefilelog-repack.t This is uglier, but more obviously correct in my opinion. I guess Python 3 doesn't hang on to the exception as long, which seems reasonable. Differential Revision: https://phab.mercurial-scm.org/D5781

tests: alter email `From` line to a value that's consistently parsed Python2: >>> email.header.decode_header('=?UTF-8?q?Rapha=C3=ABl=20Hertzog?= <hertzog@debian.org>') [('Rapha\xc3\xabl Hertzog', 'utf-8'), ('<hertzog@debian.org>', None)] Python3: >>> email.header.decode_header('=?UTF-8?q?Rapha=C3=ABl=20Hertzog?= <hertzog@debian.org>') [(b'Rapha\xc3\xabl Hertzog', 'utf-8'), (b' <hertzog@debian.org>', None)] So alter the input to an input that parses to the same result consistently. After skimming the relevant RFC (1342), I'm not sure if what we had was valid, or how I could modify it to be more consistent. Differential Revision: https://phab.mercurial-scm.org/D5769

diff: drop duplicate filter of copies by destination I'm pretty sure we don't need to filter copies by destination, at least since the previous patch. Differential Revision: https://phab.mercurial-scm.org/D5790

diff: use match.intersectmatchers() Differential Revision: https://phab.mercurial-scm.org/D5789

tests: convert ParseError arguments to str on Python 3 Arguments internally are bytes. Printing the exception on Python 3 will include b'' prefixes. This test file uses a .out file, which doesn't support conditional output. The easiest way to get this to pass on Python 3 is to normalize the exception before printing so there are no b'' prefixes. Differential Revision: https://phab.mercurial-scm.org/D5793

tests: quote $PYTHON for py3 support on Windows

py3: byteify the --retest path of run-tests.py

subrepo: clean up lingering bytes/str issues in svn support Path encoding is a little suspect in here, but it's not worse than it was when we were on Python 2 only. Subversion subrepo tests now pass in Python 3. Differential Revision: https://phab.mercurial-scm.org/D5774

tweakdefaults: set ui.relative-paths instead of command.status.relative Differential Revision: https://phab.mercurial-scm.org/D5775

status: introduce higher-level ui.relative-paths The existing commands.status.relative trumps the new config. We need to keep the existing config around for compatibility. However, I don't think we need to introduce similar command-specific options for other commands when they learn to respec ui.relative-paths. Differential Revision: https://phab.mercurial-scm.org/D5746

status: extract helper for producing relative or absolute path for UI I put the helper in scmutil so it can be used by most modules. I would have put it in utils.pathutil, but I think that's supposed to be unaware of repos. Differential Revision: https://phab.mercurial-scm.org/D5745

tags: avoid generating commit for an already deleted tag (issue5752) `$ hg tag --remove <tagname>` will keep on generating commits regardless of checking whether the tag is deleted or not. I added a a fix for `global` tags by aborting if `nullid` is already mapped to the tag that is to be deleted. Differential Revision: https://phab.mercurial-scm.org/D5780

watchman: detect nested mercurial repositories and abort Disable fsmonitor on nested mercurial repositories. Only depth one repositories are detected at the moment. Differential Revision: https://phab.mercurial-scm.org/D5587

watchman: add verbose config knob This new config knob allows to silent watchman log and warning messages when watchman is unavailable. Differential Revision: https://phab.mercurial-scm.org/D5586

extdiff: add --per-file and --confirm options The new options lets the user control how the external program is run. By default, Mercurial passes the 2 snapshot directories as usual, but it can also run the program repeatedly on each file's snapshot pair, and optionally prompt the user each time.

run-tests: sort the skip, failure and error lists in the final output This will help keep the lists consistent, for comparison across runs.

tests: add Python 3 output for test-remotefilelog-gc.t Python 3 raises a slightly different error on invalid paths. Differential Revision: https://phab.mercurial-scm.org/D5773

hg: raise Abort on invalid path Currently, some os.path functions when opening repositories may raise an uncaught TypeError or ValueError if the path is invalid. Let's catch these exceptions and turn them into an Abort for convenience. Differential Revision: https://phab.mercurial-scm.org/D5772

subrepo: bytes/str cleanups on Git support Git subrepo tests now pass on Python 3. Differential Revision: https://phab.mercurial-scm.org/D5768

tests: compare against a bytes in test-lock.py skip-blame: just b'' prefixes Differential Revision: https://phab.mercurial-scm.org/D5771

tests: perform a shallow copy instead of a deep copy Python 3 can't perform a deep copy because it looks like symbols in the locals() namespace are not deep-copyable. For the curious, somehow the deepcopy() is attempting to copy objects attached to the unittest.* functions for the running test! We don't use deepcopy() anywhere in the code base and a shallow object copy should be sufficient to test lock copying. Actually, I'm not sure why we even test this, as I couldn't find copy.copy() being used for lock copying either. Who knows. Differential Revision: https://phab.mercurial-scm.org/D5770

tests: make and use a new `svnurlof.py` helper for constructing svn urls The previous trick of a Python oneliner and some subshells is too hard to make portable, and this lets us consolidate some Windows-specific logic down to a single place. Differential Revision: https://phab.mercurial-scm.org/D5766

server: skip logging of ECONNRESET I believe this was exposed by 5492dc20, because the sending of the 500 would have already failed and prevented this logging. On Python 3, this will be a ConnectionResetError, which is a subtype of OSError, which is why we check for both OSError and socket.error. Bonus: this fixes a race in test-hgweb.t where sometimes the ECONNRESET wouldn't happen, because now we just don't log those errors. Differential Revision: https://phab.mercurial-scm.org/D5764