context: delete unused undelete() Maybe it's been unused since c8e2a5ea7062 (mq: avoid data loss upon qfold + qmv (issue3058), 2011-10-20), maybe not. Differential Revision: https://phab.mercurial-scm.org/D5888

diffordiffstat: avoid looking up contexts twice I'm not worried about performance; this is just simpler. Differential Revision: https://phab.mercurial-scm.org/D5890

fsmonitor: rename new verbose config knob The config knob was introduced in this release cycle under the old extension name, rename it before it is part of a release. Differential Revision: https://phab.mercurial-scm.org/D5880

tests: stabilize test-extdiff.t on Windows The test harness only wanted to drop the (windows !) lines, which also means the actual output matched the (no-windows !) lines. So drop those annotations too. I suspect if the output of the affected tests was actually collected on Windows, things were quoted because TEMPDIR contained a `~` in MSYS. Typically that happens when the username is 9 (8?) or more characters. I explicitly set TEMP and TMP to something short like "C:\temp" to avoid this, because otherwise there are a bunch of failures like this in other tests.

lfs: disable all authentication except Basic for HTTP(S) connections I ran into a problem pushing to an old Apache server- the normal outgoing traffic occurred, the Batch API request and response occurred, and then things suddenly halted. 5 minutes later, a 500 was returned, and the server log had a timeout reading 32K from `self._req.bodyfh` in hgweb.request.sendresponse(). Watching in WireShark, the Batch API got a 401, retried properly, then proceeded to PUT the blob (without authentication headers). This got a 401, but the client never retried with authentication. Worse, the blob was sent over the wire in the failed attempt. This kills digests for both the Batch API and the Transfer API. While in theory we could have the Batch API provide external URLs to a place that supports Basic Authentication, the LFS spec actually calls out using Basic Authentication[1]. It's not clear to me if they've been able to shoehorn in other methods. But let's keep it simple until somebody needs it. If we only had to support python2, we could just not add the handler for digest authentication. However in python3, AbstractBasicAuthHandler raises ValueError if it sees a scheme other than Basic. So we need to intercept all other schemes before it gets to that point. # no-check-commit because of urllib2.OpenerDirector foo_bar calling conventions [1] https://github.com/git-lfs/git-lfs/blob/master/docs/api/authentication.md

locate: respect ui.relative-paths Differential Revision: https://phab.mercurial-scm.org/D5875

revert: respect ui.relative-paths Differential Revision: https://phab.mercurial-scm.org/D5874

scmutil: delete now-unused origpath() (API) It has been replaced by backuppath(). Differential Revision: https://phab.mercurial-scm.org/D5860

subrepo: migrate to scmutil.backuppath() This has a test impact. It seems to me to be for the better. Differential Revision: https://phab.mercurial-scm.org/D5859

resolve: migrate to scmutil.backuppath() Differential Revision: https://phab.mercurial-scm.org/D5858

merge: migrate to scmutil.backuppath() Differential Revision: https://phab.mercurial-scm.org/D5857

filemerge: migrate to scmutil.backuppath() Differential Revision: https://phab.mercurial-scm.org/D5856

mq: migrate to scmutil.backuppath() Differential Revision: https://phab.mercurial-scm.org/D5855

largefiles: migrate to scmutil.backuppath() It seems unnecessary to convert the paths here back to repo-relative paths, but I'll leave that for someone else to clean up. Differential Revision: https://phab.mercurial-scm.org/D5854

revert: migrate to scmutil.backuppath() Differential Revision: https://phab.mercurial-scm.org/D5853

shelve: fix broken backup of conflicting untracked file Differential Revision: https://phab.mercurial-scm.org/D5852

scmutil: introduce a new backuppath() to replace origpath() Unlike most functions in our codebase, origpath() takes a path that is relative to cwd. This commit introduces a replacement for origpath(). The new function takes a path that is relative to the repo root. There is a lot of duplication between the two, but I intend to remove origpath() within the next few commits, so it won't be a maintenance burden. origpath() is also a little weird in that it returns either a a cwd-relative path or an absolute path. It needs to be able to return a path outside the repo, so it makes sense that it can return an absolute path. However, it would be simpler to always return an absolute path. The new function does that. Differential Revision: https://phab.mercurial-scm.org/D5851

tests: demonstrate broken unshelve when backing up untracked file Differential Revision: https://phab.mercurial-scm.org/D5850

subrepo: always show relative path to .orig backup Same as previous commit, but for subrepo. Differential Revision: https://phab.mercurial-scm.org/D5873

mq: always show relative path to .orig backup Same as previous commit, but for mq (I would have folded them, but test-check-commit doesn't like "revert/mq" as a topic and I couldn't think of a better one). Differential Revision: https://phab.mercurial-scm.org/D5872

revert: always show relative path to .orig backup This helps make some future patches easier when I replace origpath() by another function that works with repo-relative paths (origpath() works with cwd-relative paths). Always showing a relative path seems a little more user-friendly and is more consistent between configured ui.origbackuppath and not. OTOH, it's annoying if ui.origbackuppath is far outside the repo. This is just --verbose output, so I don't think it's worth spending much time on (I've already wasted too many hours on it). Differential Revision: https://phab.mercurial-scm.org/D5871

py3: ensure the HTTP password manager returns strings, not bytes The digest handler calls into the password manager on its own, and it apparently expects strings. Perhaps the Basic authentication handler didn't hit this because of its manual password fetch and format in retry_http_basic_auth(). The `pycompat.bytesurl()` on the user and password just above the first url.py diff seems unnecessary, because the password proxy in ui is converting to bytes IIUC.

tests: enable HTTP digest testing I suppose we could spin the client side extension off to a *.py file if it gets more use. I was basically just looking to avoid killing the server and relaunching it just to change authentication schemes, because that doesn't always work on Windows. The test changes capture the problem with py3.

wsgiheaders: make sure __repr__() returns a string When printing `req.headers` on the server side to debug, it complained that '%b' needed to take a string, not bytes. Changing '%s' to '%r' caused it to complain that __repr__ didn't return a string.

tests: add code to handle HTTP digests on the server side It's not hooked up yet. Mostly this was cargoculted and simplified from some python.org code[1]. It's not trying to test the security as much as it is trying to make sure that clients are sending out the right data when challenged. (And they aren't on py3.) [1] http://svn.python.org/projects/sandbox/trunk/digestauth/digestauth.py

run-tests: allow spaces in the --view tool Most tools on Windows are in Program Files, and not necessarily on PATH.

tests: extract the http server authentication extension to a single module We had 4 copy/pastes of this, and no coverage for http digests (which are currently broken on py3).

extdiff: support tools that can be run simultaneously