# -*- mode: org -*-
* v4.7.2    2026-11-07

- Tags are now shown in the logs of submodules in the status buffer
  of the super-repository.  Additionally all unpulled tags are shown
  in the headings of these logs.  1c470b54, 5f6c5afa

- When setting the value of a Git variable in a menu, the worktree
  specific value can be set or unset using a prefix argument.  See
  https://github.com/magit/forge/discussions/831.  This is still
  experimental.  2e92cad8

- In the log menu, the ~--authors~ argument now accepts multiple values.
  7851e024

- Improved the default choice when pushing to the upstream, if that
  isn't configured yet.  #5638

- Hi-lock is now instructed to use overlays in Magit buffers so that
  its highlights are not lost when refreshing such buffers.

- Improved the timestamp fallbacks displayed in repository lists for
  repositories that lack release tags.  58a46982

- When committing ~dabbrev-capf~ is now added near the end of
  ~completion-at-point-functions~ to avoid shadowing other functions.
  #5655

- Optimized diff washing by preventing modification hooks from
  triggering and avoiding unnecessary buffer edits.  #5656, #5659

- Added new hook ~magit-post-call-git-functions~.  #5652

- Hunk fontification and refinement are now disabled when
  ~long-lines-threshold~ triggers.  #5647

- ~magit-list-repositories~ can now be instructed (using option
  ~magit-repository-directories~) to include repositories nested inside
  other repositories.  #5657

- ~magit-list-repositories~ can now list different sets of repositories
  depending on ~default-directory~, by setting ~magit-repository-directories~
  in ~.dir-locals.el~.  #5658

Bugfixes:

- Fixed a regression in ~magit-stage-untracked~.  #5635

- ~bug-reference-mode~ did not work in buffers used to write commit
  messages.  fe66e74a

- ~bug-reference-mode~ needed help doing its work in files located
  in the gitdir (which Magit uses when authoring commit messages).
  fe66e74a

- ~magit-branch-delete~ failed to delete remote branches on remotes
  whose name contain a slash.  #5637

- When invoked from a modified file, ~magit-status-here~ no longer
  jumped to the correct location.  #5641

- When visiting a blob from a buffer visiting the respective file in
  the worktree, the position was not restored in the target buffer.

- ~magit-toplevel~ did not work from a subdirectory of the gitdir of
  a submodule, and signaled an error from the gitdir of a regular
  repository, which happens to be a subdirectory of a directory
  named "modules".  #5644

- ~magit-branch~ and ~magit-remote~ were not fully refreshed in an
  edge-case.  #5642

- GnuPG became more verbose in some cases, which we weren't prepared
  to handle, causing failures in properly formatting the output.
  #5643

- After adding a worktree to a bare repository, trying to create
  a commit from the worktree signaled an error.  #5636

- The script used to call back into Emacs from Git hooks failed if the
  path to ~emacsclient~ needs shell quoting.  #5651

- A local file was used to collect standard error of remote processes,
  causing an unnecessary round-trip.  #5654

- In logs dedicated to selecting a commit (such as when starting a
  rebase), the wrong section was highlighted initially.  #5653

- On remote systems and when ~with-editor-emacsclient-executable~ is
  ~nil~, no longer try to call back from Git hooks into Emacs, which
  cannot work.  #5592

- When Git hooks are configured to call back into Emacs, and using
  an ~windows-nt~ Emacs together with Cygwin, members of ~$PATH~ were
  not translated, and the wrong path separator was used.  #5640

- In Emacs 31.1, Tramp changed/extended how it filters out environment
  variables, that should be hidden from processes on the remote.  This
  caused a variable to be filtered out, that must remain set for Magit
  to not drop data when creating a stash.  We now hold the variable as
  is expected from us, and added a safety-net, which asserts that the
  variable indeed remains set.  #5660
...
...
