v0.46.0

[jennings]

About

jj is a Git-compatible version control system that is both simple and powerful. See
the installation instructions to get started.

Release highlights

• Jujutsu can now colocate workspaces besides the default one by creating Git
  worktrees. Use jj workspace add --[no-]colocate and the setting
  git.colocate to control this.

Breaking changes

• The minimum supported git command version is now 2.42.0, up from 2.41.0.
  jj workspace add uses git worktree add --orphan, which was added in
  2.42.0.

• The minimum supported Rust version (MSRV) is now 1.97.1.

• jj bisect run now runs some consistency checks before proceeding to bisect.
  This helps ensure that the command can tell good and bad revisions apart,
  and that the working copy does go from bad to good over the provided revset.
  Use the new flag --trust-endpoints to disable these checks.

• jj split now opens a single editor session to edit descriptions for the
  split commits.

• jj undo and jj redo now refuse to undo/redo an operation that was
  performed in another workspace. Use --allow-cross-workspace to undo/redo
  it anyway.

• jj workspace list/root no longer omit unreachable paths. All recorded
  paths are now shown, with warnings displayed in jj workspace root.

• The List.get(), .first(), and .last() template functions now return
  Option<T> instead of throwing an error on out-of-bounds access.

New features

• jj workspace add supports --colocate/--no-colocate flags to control
  whether a Git worktree is created alongside the workspace. The default
  colocates when the current workspace is colocated and the git.colocate
  config is true. jj workspace forget removes the corresponding Git
  worktree when one exists.

• jj git colocation status/enable/disable now work on child
  workspaces. status correctly reports colocation state and includes
  the workspace name. enable creates a Git worktree and disable
  removes it, allowing colocation to be toggled after workspace
  creation.

• jj workspace remove removes a workspace and its directory from disk. The
  working-copy state is snapshotted into a commit before removal.

• Added commands jj file edit and jj file delete for editing files in any
  revision without needing to change the working copy.

• jj git push now supports pushing to multiple remotes at the same time.
  This can be configured via git.push set to a string pattern
  or array of string patterns, or with the repeatable --remote flag,
  which also accepts string patterns.

• The default target revisions for jj git push can now be configured via
  revsets.git-push.

• Added the TreeEntry.normal_value() template method and the TreeValue type
  to access resolved tree values, formatted as their full object IDs, including
  Git submodule commit IDs.

• Diff hunk headers now include nearby source symbols for many common
  programming and markup languages.

• fix.tools.<name>.line-range-args (replaces line-range-arg) is an array of
  string template args to pass to the fix tool. This is more flexible in cases
  where you need to pass multiple arguments to the tool, such as separate args
  for the range start and range end.

• jj run now uses the sparse patterns from the workspace it's run from.
  Use the --sparse-patterns option to control this behavior (evaluated
  per each jj run invocation).

• jj util diff <path1> <path2> to compare files on disk.

• Aliases now support setting aliases.<name>.enabled = false, which will
  disable them. This can be used to disable built-in aliases or disable aliases
  in later layers (such as repo config files).

• ui.editor now supports $path and $line substitution variables. Example:
  ui.editor = ["emacs", "+$line", "$path"]

• fill template function now supports an additional named parameter
  break_words, that allows specifying if the template should break words
  longer than width passed in the input to ensure no words overflow the
  specified width.

• The json() template function now supports map literals: json({'key' => value})

• The hunk headers of diff.color-words.conflict = "pair" now include the
  conflict labels of the compared terms.

Fixed bugs

• On Windows, jj no longer hangs when a subprocess needs to prompt the user,
  such as ssh asking for a key passphrase or for confirmation of an unknown
  host key. Subprocesses started from a terminal now inherit its console, rather
  than being given an invisible one by CREATE_NO_WINDOW for the prompt to
  disappear into.
  #6745
  #8547

• On Windows, jj git colocation enable and jj git colocation disable no
  longer fail with "Access is denied (os error 5)" when the Git repository
  contains pack files.
  #8661

• jj undo of jj workspace forget now correctly preserves the workspace's
  recorded path. Previously the path metadata was lost, leaving the workspace
  in a broken state after undo.
  #9991

• at_operation() can now be used with operations that are not ancestors of
  the current operation (e.g. sibling operations created by concurrent
  commands). Previously, evaluating such expressions failed if they resolved
  to commits missing from the current operation's index.

• .gitignore files are now respected even if they aren't materialized in the
  working copy because they are excluded by the sparse patterns. Previously,
  ignored files could become tracked in a sparse working copy.
  #2289

• In-tree ignore files (.gitignore) are no longer read through symlinks,
  matching git behavior. Such files are now silently skipped instead of having
  their symlink target applied. $GIT_DIR/info/exclude and core.excludesFile
  are unaffected and still follow symlinks, as git does.
  #7161

• jj workspace list templates are now labeled with workspace name,
  workspace root, etc.

Contributors

Thanks to the people who made this release happen!

• Aaron Bies (@slerpyyy)
• Austin Seipp (@thoughtpolice)
• Bartok9 (@Bartok9)
• Brice Figureau (@masterzen)
• Bryan O'Sullivan (@bos)
• Caleb White (@calebdw)
• David Rieber (@drieber)
• Farid Zakaria (@fzakaria)
• Gabriel Goller (@kaffarell)
• Gasper Stukelj (@mirkomartn)
• Jakub Stasiak (@jstasiak)
• JamBalaya56562 (@JamBalaya56562)
• Joseph Lou (@josephlou5)
• Karnajeet Gosavi (@kg290)
• LOG (@logarithmone1128)
• Martin von Zweigbergk (@martinvonz)
• Matt Stark (@matts1)
• Mustafa Officewala (@genericusername2709)
• pederbe (@pederbe)
• Philip Metzger (@PhilipMetzger)
• Pro (@twistedfall)
• Remo Senekowitsch (@senekor)
• Sami Hiltunen (@SamiHiltunen)
• sofia (@badp)
• Stephen Jennings (@jennings)
• Vincent Ging Ho Yim (@cenviity)
• xtqqczze (@xtqqczze)
• Yannik Sander (@ysndr)
• Yuya Nishihara (@yuja)

---

This discussion was created from the release v0.46.0.

[jasikpark]

Jujutsu can now colocate workspaces besides the default one by creating Git
worktrees. Use jj workspace add --[no-]colocate and the setting
git.colocate to control this.

yipeeee

[calebdw]

You're welcome 😁

[nevdelap]

Yeah, that is awesome!

[qezz]

Great set of features in this update!

One thing that throws me off is the new behavior of jj split. Previously it would insert 1 header line to each buffer (#6617), which is one keystroke away from deleting, but now it's 7 lines

JJ: Enter or edit commit descriptions after the `JJ: describe` lines.
JJ: Warning:
JJ: - The text you enter will be lost on a syntax error.
JJ: - The syntax of the separator lines may change in the future.
JJ:
JJ: describe 4807d7e9501e -------
JJ: Enter a description for the selected changes.

And I can't naively delete it because both split commits are now in the same buffer. It unfortunately becomes very messy with

[templates]
draft_commit_description = '''
concat(
  builtin_draft_commit_description,
  "\nJJ: ignore-rest\n",
  diff.git(),
)
'''

i.e. the selected change is on the top (say, 100 lines), then the remaining changes are on the bottom, and I need to find them somehow.

There's an new editor hint from #10269, but it seems like it only 'hints' the position of the first commit message, and not the second

Anyway, feels like the only solution to the issue is to teach my editor to jump between commit message slots, and to create a key bind for it.
It doesn't seem there's a way to bring back the old behavior.

Please let me know if my understanding is wrong, and if there's anything that can be done about it

[josephlou5]

Related: #10373