rebase: don't create merge when continuing rebase interrupted by old hg This fixes the bug described and demonstrated in the previous commit. It does so by practically undoing 8082a77cc3a2 (rebase: remove some redundant setting of dirstate parents, 2020-01-10). Differential Revision: https://phab.mercurial-scm.org/D8356

tests: demonstrate how continuing rebase after upgrade can result in merge If the user starts a rebase with an hg version before 9c9cfecd4600 (rebase: don't use rebased node as dirstate p2 (BC), 2020-01-10) and then runs into conflicts, they will be dropped out to the shell with the rebased node set as the dirstate's second parent. If they then upgrade to a later hg version, it will respect the dirstate's parents and will create a merge commit even if the user was rebasing a non-merge commit. Differential Revision: https://phab.mercurial-scm.org/D8355

manifest: also declare treemanifest as implementing imanifestdict It always has been intended to be, but it had a mismatched flags() method until the previous change. Differential Revision: https://phab.mercurial-scm.org/D8366

manifest: introduce new exception to signal unavailability of fastdelta() I've spent some time reflecting on this, and I think this is the best we can do in this API contract for now. This opens the door to adding treemanifest's implementation to the list of implementers of imanifestdict. Differential Revision: https://phab.mercurial-scm.org/D8365

copies-tests: remove spurious `]` in the template Fixing this typo happily impact all the tests output. Differential Revision: https://phab.mercurial-scm.org/D8258

merge with stable

debugcommands: fix typo in debuguigetpass

dagop: fix subsetparentswalker to set p1/p2 chains at merge revision The previous implementation was wrong because the '1'/'2' key would be appended at a fork revision. Since we traverse the graph from heads, a merge revision is actually a branching point, where the sort key must be generated.

dagop: simplify dict/set reuse condition in subsetparentswalker Prepares for fixing the calculation of p1/p2 sort keys. With this change, there will be one more copying on merge&fork case. I think the copying cost is negligible since we'll have to update each item in the dict on merge/fork.

extensions: refactor function for obtaining disabled extension help The way this worked before was hgext.__index__ was consulted. This file appears to only be present on some Windows distributions. This file contains a dict mapping extension name to its summary line, not its full docstring. The problem with this is that code in the help system was calling this function to resolve help text. If hgext.__index__ was present, only the summary line would be displayed. If not, the full extension help would be printed. This commit changes the function to not use hgext.__index__ such that it always returns the full extension help text. As a result of this change, test-extension.t and test-qrecord.t now pass when run from environments that have an hgext.__index__. Differential Revision: https://phab.mercurial-scm.org/D8344

tests: perform grep manually in test-doctest.py This test has been failing on Windows since 0af56d3ee24c introduced the `hg files` invocation. Specifically, Windows seems to be choking on special characters in the fileset pattern. I believe at least \n and > were causing issues. I attempted various incantations to make the Windows command line parser accept the fileset but couldn't get anything working. I declared bankruptcy and just reimplemented the grepping code in Python. After this change, the test now passes on Windows again. Differential Revision: https://phab.mercurial-scm.org/D8343

tests: prevent printing \r to stdout Like we've done in other recent commits, we need to change sys.stdout on Python 3 to not use os.linesep so output is consistent on Python 3 on Windows. With this change, test-notify.t now passes on Python 3 on Windows! Differential Revision: https://phab.mercurial-scm.org/D8342

tests: force \n newlines when writing to sys.stdout Without this, Python 3 on Windows inserts some \r that aren't present in the input, causing test-http-bad-server.t to fail. After this change, the test passes on Python 3 on Windows! Differential Revision: https://phab.mercurial-scm.org/D8341

dispatch: force \n for newlines on sys.std* streams (BC) The sys.std* streams behave differently on Python 3. On Python 3, these streams are an io.TextIOWrapper that wraps a binary buffer stored on a .buffer attribute. These TextIOWrapper instances normalize \n to os.linesep by default. On Windows, this means that \n is normalized to \r\n. So functions like print() which have an implicit end='\n' will actually emit \r\n for line endings. While most parts of Mercurial go through the ui.write() layer to print output, some code - notably in extensions and hooks - can use print(). If this code was using print() or otherwise writing to sys.std* on Windows, Mercurial would emit \r\n. In reality, pretty much everything on Windows reacts to \n just fine. Mercurial itself doesn't emit \r\n when going through the ui layer. Changing the sys.std* streams to not normalize line endings sounds like a scary change. But I think it is safe. It also makes Mercurial on Python 3 behave similarly to Python 2, which did not perform \r\n normalization in print() by default. .. bc:: sys.{stdout, stderr, stdin} now use \n line endings on Python 3 Differential Revision: https://phab.mercurial-scm.org/D8339

hook: move stdio redirection to context manager The old code was checking stdio redirection in a loop. This didn't make sense. The pattern is better expressed as a context manager IMO, so this commit refactors it to be one. Differential Revision: https://phab.mercurial-scm.org/D8338