completions/magick: remove deprecated convert command and dependencies

This is a potential bikeshed project but if fish is being used as a
proper noun, it should either be capitalised all the time or none of the
time, and the latter looks better.

.cirrus.yml

@@ -1,49 +0,0 @@
-env:
-    CIRRUS_CLONE_DEPTH: 100
-    CI: 1
-
-linux_task:
-    matrix:
-        - name: alpine
-          container: &step
-              image: ghcr.io/krobelus/fish-ci/alpine:latest
-              memory: 4GB
-        - name: ubuntu-oldest-supported
-          container:
-              <<: *step
-              image: ghcr.io/krobelus/fish-ci/ubuntu-oldest-supported:latest
-    tests_script:
-        # cirrus at times gives us 32 procs and 2 GB of RAM
-        # Unrestriced parallelism results in OOM
-        - lscpu || true
-        - (cat /proc/meminfo | grep MemTotal) || true
-        - mkdir build && cd build
-        - FISH_TEST_MAX_CONCURRENCY=6 cmake -G Ninja -DCMAKE_BUILD_TYPE=Debug ..
-        - ninja -j 6 fish
-        - ninja fish_run_tests
-    only_if: $CIRRUS_REPO_OWNER == 'fish-shell'
-
-freebsd_task:
-    matrix:
-        - name: FreeBSD Stable
-          freebsd_instance:
-              image: freebsd-15-0-release-amd64-ufs # updatecli.d/cirrus-freebsd.yml
-    tests_script:
-        - pkg update
-        - pkg install -y cmake-core devel/pcre2 devel/ninja gettext git-lite lang/rust misc/py-pexpect
-        # libclang.so is a required build dependency for rust-c++ ffi bridge
-        - pkg install -y llvm
-        # BSDs have the following behavior: root may open or access files even if
-        # the mode bits would otherwise disallow it. For example root may open()
-        # a file with write privileges even if the file has mode 400. This breaks
-        # our tests for e.g. cd and path. So create a new unprivileged user to run tests.
-        - pw user add -n fish-user -s /bin/csh -d /home/fish-user
-        - mkdir -p /home/fish-user
-        - chown -R fish-user /home/fish-user
-        - mkdir build && cd build
-        - chown -R fish-user ..
-        - sudo -u fish-user -s whoami
-        - sudo -u fish-user -s FISH_TEST_MAX_CONCURRENCY=1 cmake -G Ninja -DCMAKE_BUILD_TYPE=Debug ..
-        - sudo -u fish-user -s ninja -j 6 fish
-        - sudo -u fish-user -s ninja fish_run_tests
-    only_if: $CIRRUS_REPO_OWNER == 'fish-shell'

.editorconfig

@@ -21,7 +21,7 @@ indent_size = 4
 [build_tools/release.sh]
 max_line_length = 72
 
-[{Dockerfile,Vagrantfile}]
+[Vagrantfile]
 indent_size = 2
 
 [share/{completions,functions}/**.fish]
@@ -30,5 +30,5 @@ max_line_length = unset
 [{COMMIT_EDITMSG,git-revise-todo,*.jjdescription}]
 max_line_length = 72
 
-[*.yml]
+[*.{toml,yml}]
 indent_size = 2

.github/actions/install-sphinx/action.yml

@@ -13,11 +13,9 @@ runs:
         command -v uv
         command -v uvx
         # Check that pyproject.toml and the lock file are in sync.
-        # TODO Use "uv" to install Python as well.
-        : 'Note that --no-managed-python below would be implied but be explicit'
-        uv='env UV_PYTHON=python uv --no-managed-python'
-        $uv lock --check --exclude-newer="$(awk -F'"' <uv.lock '/^exclude-newer[[:space:]]*=/ {print $2}')"
-        # Install globally.
-        sudo $uv pip install --group=dev --system --break-system-packages
+        uv lock --check --exclude-newer="$(awk -F'"' <uv.lock '/^exclude-newer[[:space:]]*=/ {print $2}')"
+        uv venv ~/.local --allow-existing
+        uv pip install --group=dev
         # Smoke test.
         python -c 'import sphinx; import sphinx_markdown_builder'
+        python3 -c 'import sphinx; import sphinx_markdown_builder'

.github/actions/rust-toolchain/action.yml

@@ -25,7 +25,7 @@ runs:
         set -x
         toolchain=$(
           case "$toolchain_channel" in
-            (stable) echo 1.93 ;; # updatecli.d/rust.yml
+            (stable) echo 1.96 ;; # updatecli.d/rust.yml
             (msrv)   echo 1.85 ;; # updatecli.d/rust.yml
             (*)
               printf >&2 "error: unsupported toolchain channel %s" "$toolchain_channel"

.github/workflows/autolabel_prs.yml

@@ -10,7 +10,7 @@ jobs:
     steps:
     - name: Set label and milestone
       id: set-label-milestone
-      uses: actions/github-script@ed597411d8f924073f98dfc5c65a23a2325f34cd # v8, build_tools/update-dependencies.sh
+      uses: actions/github-script@d746ffe35508b1917358783b479e04febd2b8f71 # v9.0.0, build_tools/update-dependencies.sh
       with:
         script: |
           const completionsLabel = 'completions';

.github/workflows/build_docker_images.yml

@@ -1,64 +0,0 @@
-name: Build Docker test images
-
-on:
-  push:
-    branches:
-      - master
-    paths:
-      - 'docker/**'
-  workflow_dispatch:
-
-concurrency:
-  group: docker-builds
-
-env:
-  REGISTRY: ghcr.io
-  NAMESPACE: fish-ci
-
-jobs:
-  docker-build:
-    if: github.repository_owner == 'fish-shell'
-    permissions:
-      contents: read
-      packages: write
-      attestations: write
-      id-token: write
-
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - os: ubuntu-latest
-            target: alpine
-          - os: ubuntu-latest
-            target: ubuntu-oldest-supported
-    runs-on: ${{ matrix.os }}
-
-    steps:
-      -
-        name: Checkout
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2, build_tools/update-dependencies.sh
-      -
-        name: Login to Container registry
-        uses: docker/login-action@c94ce9fb468520275223c153574b00df6fe4bcc9 # v3.7.0, build_tools/update-dependencies.sh
-        with:
-          registry: ${{ env.REGISTRY }}
-          username: ${{ github.actor }}
-          password: ${{ secrets.GITHUB_TOKEN }}
-      -
-        name: Extract metadata (tags, labels) for Docker
-        id: meta
-        uses: docker/metadata-action@c299e40c65443455700f0fdfc63efafe5b349051 # v5.10.0, build_tools/update-dependencies.sh
-        with:
-          images: ${{ env.REGISTRY }}/${{ github.repository_owner }}/${{ env.NAMESPACE }}/${{ matrix.target }}
-          flavor: |
-            latest=true
-      -
-        name: Build and push
-        uses: docker/build-push-action@263435318d21b8e681c14492fe198d362a7d2c83 # v6.18.0, build_tools/update-dependencies.sh
-        with:
-          context: docker/context
-          push: true
-          file: docker/${{ matrix.target }}.Dockerfile
-          tags: ${{ steps.meta.outputs.tags }}
-          labels: ${{ steps.meta.outputs.labels }}

.github/workflows/development-builds.yml

@@ -0,0 +1,32 @@
+name: Linux development builds
+
+on:
+  push:
+    branches:
+      - buildscript
+
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    environment: linux-development
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2, build_tools/update-dependencies.sh
+      - uses: astral-sh/setup-uv@v7
+      - name: Update package database
+        run: sudo apt-get update
+      - name: Install deps
+        run: sudo apt install debhelper devscripts dpkg-dev
+      - name: Create tarball and source packages
+        run: |
+          version=$(build_tools/git_version_gen.sh --stdout 2>/dev/null)
+          mkdir /tmp/gpg
+          echo "$SIGNING_GPG_KEY" > /tmp/gpg/signing-gpg-key
+          mkdir /tmp/fish-built
+          FISH_ARTEFACT_PATH=/tmp/fish-built ./build_tools/make_tarball.sh
+          FISH_ARTEFACT_PATH=/tmp/fish-built DEB_SIGN_KEYFILE=/tmp/gpg/signing-gpg-key ./build_tools/make_linux_packages.sh $version
+      - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1, build_tools/update-dependencies.sh
+        with:
+          name: linux-source-packages
+          path: |
+            /tmp/fish-built
+            ! /tmp/fish-built/fish-*/* # don't include the unpacked source directory

.github/workflows/lint-dependencies.yml

@@ -17,8 +17,8 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2, build_tools/update-dependencies.sh
-    - uses: EmbarkStudios/cargo-deny-action@44db170f6a7d12a6e90340e9e0fca1f650d34b14 # v2.0.15, build_tools/update-dependencies.sh
+    - uses: EmbarkStudios/cargo-deny-action@8a45e1c7c9a95dfae3276e89a553705e40ae45a2 # v2.0.20, build_tools/update-dependencies.sh
       with:
         command: check licenses
         arguments: --all-features --locked --exclude-dev
-        rust-version: 1.93 # updatecli.d/rust.yml
+        rust-version: 1.96 # updatecli.d/rust.yml

.github/workflows/lint.yml

@@ -16,12 +16,33 @@ jobs:
     - name: install dependencies
       run: pip install ruff
     - name: build fish
-      run: cargo build
+      run: cargo build --bin fish_indent
     - name: check format
-      run: PATH="target/debug:$PATH" build_tools/style.fish --all --check
+      run: PATH="target/debug:$PATH" cargo xtask format --all --check
     - name: check rustfmt
       run: find build.rs crates src -type f -name '*.rs' | xargs rustfmt --check
 
+  shellcheck:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2, build_tools/update-dependencies.sh
+    - uses: ./.github/actions/rust-toolchain@stable
+    - name: Update package database
+      run: sudo apt-get update
+    - name: Install shellcheck
+      run: sudo apt install shellcheck
+    - name: shellcheck
+      run: cargo xtask shellcheck
+
+  po_files_up_to_date:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2, build_tools/update-dependencies.sh
+    - uses: ./.github/actions/rust-toolchain@stable
+    - name: Install deps
+      uses: ./.github/actions/install-dependencies
+    - name: Check PO files
+      run: cargo xtask gettext check
 
   clippy:
     runs-on: ubuntu-latest
@@ -32,6 +53,8 @@ jobs:
             features: ""
           - rust_version: "stable"
             features: "--no-default-features"
+          - rust_version: "stable"
+            features: "--all-features"
           - rust_version: "msrv"
             features: ""
     steps:
@@ -45,6 +68,11 @@ jobs:
     - name: Install deps
       run: |
         sudo apt install gettext
+    - name: Patch Cargo.toml to deny unknown lints
+      run: |
+        if [ "${{ matrix.rust_version }}" = stable ]; then
+          sed -i /^rust.unknown_lints/d Cargo.toml
+        fi
     - name: cargo clippy
       run: cargo clippy --workspace --all-targets ${{ matrix.features }} -- --deny=warnings
 

.github/workflows/lockthreads.yml

@@ -18,7 +18,7 @@ jobs:
       pull-requests: write  # for dessant/lock-threads to lock PRs
     runs-on: ubuntu-latest
     steps:
-      - uses: dessant/lock-threads@f5f995c727ac99a91dec92781a8e34e7c839a65e # v6.0.0, build_tools/update-dependencies.sh
+      - uses: dessant/lock-threads@b2726a6ae6f1e1b06eb0ff28f7e4fb5e4246bbca # v6.0.2, build_tools/update-dependencies.sh
         with:
           github-token: ${{ github.token }}
           issue-inactive-days: '365'

.github/workflows/release.yml

@@ -63,7 +63,7 @@ jobs:
           sed -n 2p "$relnotes" | grep -q '^$'
           sed -i 1,2d "$relnotes"
       - name: Upload tarball artifact
-        uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6.0.0, build_tools/update-dependencies.sh
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1, build_tools/update-dependencies.sh
         with:
           name: source-tarball
           path: |
@@ -104,7 +104,7 @@ jobs:
             tar -cazf fish-$(git describe)-linux-$arch.tar.xz \
               -C target/$arch-unknown-linux-musl/release fish
           done
-      - uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6.0.0, build_tools/update-dependencies.sh
+      - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1, build_tools/update-dependencies.sh
         with:
           name: Static builds for Linux
           path: fish-${{ inputs.version }}-linux-*.tar.xz
@@ -123,14 +123,14 @@ jobs:
           # Workaround for https://github.com/actions/checkout/issues/882
           ref: ${{ inputs.version }}
       - name: Download all artifacts
-        uses: actions/download-artifact@37930b1c2abaa49bbe596cd826c3c89aef350131 # v7.0.0, build_tools/update-dependencies.sh
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1, build_tools/update-dependencies.sh
         with:
           merge-multiple: true
           path: /tmp/artifacts
       - name: List artifacts
         run: find /tmp/artifacts -type f
       - name: Create draft release
-        uses: softprops/action-gh-release@a06a81a03ee405af7f2048a818ed3f03bbf83c7b # v2.5.0, build_tools/update-dependencies.sh
+        uses: softprops/action-gh-release@b4309332981a82ec1c5618f44dd2e27cc8bfbfda # v3.0.0, build_tools/update-dependencies.sh
         with:
           tag_name: ${{ inputs.version }}
           name: fish ${{ inputs.version }}

.github/workflows/test.yml

@@ -32,14 +32,6 @@ jobs:
     - name: make fish_run_tests
       run: |
         make -C build VERBOSE=1 fish_run_tests
-    - name: translation updates
-      run: |
-        # Generate PO files. This should not result it a change in the repo if all translations are
-        # up to date.
-        # Ensure that fish is available as an executable.
-        PATH="$PWD/build:$PATH" build_tools/update_translations.fish
-        # Show diff output. Fail if there is any.
-        git --no-pager diff --exit-code || { echo 'There are uncommitted changes after regenerating the gettext PO files. Make sure to update them via `build_tools/update_translations.fish` after changing source files.'; exit 1; }
 
   ubuntu-32bit-static-pcre2:
     runs-on: ubuntu-latest
@@ -159,21 +151,26 @@ jobs:
         shell: msys2 {0}
     steps:
     - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2, build_tools/update-dependencies.sh
-    - uses: msys2/setup-msys2@4f806de0a5a7294ffabaff804b38a9b435a73bda # v2.30.0, build_tools/update-dependencies.sh
+    - uses: msys2/setup-msys2@e9898307ac31d1a803454791be09ab9973336e1c # v2.31.1, build_tools/update-dependencies.sh
       with:
         update: true
         msystem: MSYS
+      id: msys2
     - name: Install deps
       # Not using setup-msys2 `install` option to make it easier to copy/paste
       run: |
-        pacman --noconfirm -S --needed git rust
+        pacman --noconfirm -S --needed git rust python3 diffutils tmux
+    - name: rebase
+      env:
+        MSYS2_LOCATION: ${{ steps.msys2.outputs.msys2-location }}
+      shell: cmd
+      run: |
+        "%MSYS2_LOCATION%\usr\bin\dash" /usr/bin/rebaseall -p -v
     - name: cargo build
       run: |
         cargo build
-    - name: smoketest
-      # We can't run `build_tools/check.sh` yet, there are just too many failures
-      # so this is just a quick check to make sure that fish can swim
+    - name: tests
+      env:
+        FISH_CHECK_LINT: false
       run: |
-        set -x
-        [ "$(target/debug/fish.exe -c 'echo (math 1 + 1)')" = 2 ]
-        cargo test
+        cargo xtask check

.gitignore

@@ -20,7 +20,6 @@
 *.o
 *.obj
 *.orig
-!tests/*.out
 *.out
 *.pch
 *.slo
@@ -36,46 +35,31 @@
 Desktop.ini
 Thumbs.db
 ehthumbs.db
+__pycache__/
 
 .directory
 .fuse_hidden*
 
 
-# Directories that only contain transitory files from building and testing.
-/doc/
-/share/man/
-/share/doc/
-/test/
-/user_doc/
-
-# File names that can appear in the project root that represent artifacts from
-# building and testing.
-/FISH-BUILD-VERSION-FILE
-/command_list.txt
-/command_list_toc.txt
-/compile_commands.json
-/doc.h
+# Artifacts from in-tree builds ("cmake .").
+/build.ninja
+/cargo/
+/CMakeCache.txt
+/CMakeFiles/
+/cmake_install.cmake
 /fish
-/fish.pc
 /fish_indent
 /fish_key_reader
-/fish_tests
-/lexicon.txt
-/lexicon_filter
-/toc.txt
-/version
-fish-build-version-witness.txt
-__pycache__
+/fish-localization-map-cache/
+/fish.pc
+/fish.pc.noversion
+/.ninja_log
 
 # File names that can appear below the project root that represent artifacts
 # from building and testing.
-/doc_src/commands.hdr
-/doc_src/index.hdr
-/po/*.gmo
 /share/__fish_build_paths.fish
-/share/pkgconfig
-/tests/*.tmp.*
 /tests/.last-check-all-files
+/.venv/
 
 # xcode
 ## Build generated
@@ -83,24 +67,19 @@ __pycache__
 *.xccheckout
 *.xcscmblueprin
 .vscode
-/DerivedData/
 /build/
+/DerivedData/
 /tags
-xcuserdata/
+/xcuserdata/
 
 # Generated by Cargo
 # will have compiled files and executables
-debug/
-target/
-
+/target/
 # These are backup files generated by rustfmt
 **/*.rs.bk
 
-# MSVC Windows builds of rustc generate these, which store debugging information
-*.pdb
-
 # Generated by clangd
-/.cache
+/.cache/
 
 # JetBrains editors.
 .idea/

AGENTS.md

@@ -1,3 +0,0 @@
-# Coding style
-
-- Use comments sparingly. Don't explain what the code is doing, rather explain why.

CHANGELOG.rst

@@ -1,3 +1,129 @@
+fish ?.?.? (released ???)
+=========================
+
+Deprecations and removed features
+---------------------------------
+- `--command` and `--path` options in `complete` no longer unescape their value.
+
+Interactive improvements
+------------------------
+- On the first run after upgrading from an older version, fish will try harder to check if the current theme matches a historical default, in which case fish won't create ``~/.config/fish/conf.d/fish_frozen_theme.fish``.
+  This means that on systems where fish version 3.x was installed originally, the update will avoid creating that file (:issue:`12725`).
+- ``fish_hg_prompt``, ``fish_git_prompt`` and ``fish_fossil_prompt`` now strip control characters from VCS state read off disk, matching ``prompt_pwd``.
+- The sample informative and minimalist prompts now use ``prompt_pwd`` instead of printing ``$PWD`` directly.
+- ``bind`` shows the file where bindings were defined (:issue:`12504`).
+- Abbreviations with ``--position=anywhere`` can now be completed in argument position, not just in command position (:issue:`12630`).
+
+Other improvements
+------------------
+- ``cd`` supports the ``-L`` and ``-P`` options, like other shells, to allow specifying whether symbolic links (symlinks) are resolved when changing directories (:issue:`7206`).
+
+For distributors and developers
+-------------------------------
+- With the exception of the ``$CMAKE_INSTALL_PREFIX/share/fish/man`` directory, fish no longer installs files to ``$CMAKE_INSTALL_PREFIX/share/fish``.
+  In particular, this means that both
+  ``$CMAKE_INSTALL_PREFIX/share/fish/completions`` and
+  ``$CMAKE_INSTALL_PREFIX/share/fish/functions``
+  should now be empty.
+  If another package installs completions or functions to those directories,
+  they should be changed to install to
+  ``extra_completionsdir`` (typically ``$CMAKE_INSTALL_PREFIX/share/fish/vendor_completions.d``) or
+  ``extra_functionsdir`` (typically ``$CMAKE_INSTALL_PREFIX/share/fish/vendor_functions.d``)
+  instead.
+  The old location has been ignored since fish 4.2.
+
+Regression fixes:
+-----------------
+- (from 4.4.0) Vi mode ``c,W`` key binding wrongly deleted trailing spaces (:issue:`12790`).
+- (from 4.4.0) Vi mode ``x`` in :doc:`builtin read <cmds/read>` (:issue:`12724`).
+- (from 4.3.3) Repeated tab would sometimes insert smartcase completions redundantly.
+
+fish 4.7.1 (released May 08, 2026)
+==================================
+
+This release fixes a regression in 4.7.0 that caused the web config (``fish_config``) to fail to start (:issue:`12717`).
+
+fish 4.7.0 (released May 05, 2026)
+==================================
+
+Deprecations and removed features
+---------------------------------
+- The default theme (i.e. the ``fish_color_*`` variables) is no longer set in non-interactive shells.
+
+Interactive improvements
+------------------------
+- :doc:`prompt_pwd <cmds/prompt_pwd>` now strips control characters.
+- Repaint events (as triggered by changes to color variables or by event handlers running ``commandline -f repaint``) no longer reset the completion pager and other transient UI states (:issue:`12683`).
+- :envvar:`fish_color_valid_path` now respects background and underline colors (:issue:`12622`).
+- :doc:`funced <cmds/funced>` will no longer lose work if there are parse errors multiple times without new changes to the file.
+- Fixed a case where directory completions were sorted in a surprising order (:issue:`12695`).
+- When at the command token, the :kbd:`alt-o` binding will now open read-only files too (:issue:`12671`).
+- Private mode in-memory history (``set fish_history``) is no longer shared with :doc:`builtin read <cmds/read>` (:issue:`12662`).
+
+Other improvements
+------------------
+- History is no longer corrupted with NUL bytes when fish receives SIGTERM or SIGHUP (:issue:`10300`).
+- :doc:`fish_update_completions <cmds/fish_update_completions>` now handles groff ``\X'...'`` device control escapes, fixing completion generation for man pages produced by help2man 1.50 and later (such as coreutils 9.10).
+- Removing history entries via the :doc:`web-based config <cmds/fish_config>` is more intuitive.
+- If :envvar:`XDG_DATA_DIRS` is empty, the default value is assumed, which means that fish will now also use configuration from paths like ``$PREFIX/share/fish/vendor_completions.d`` (:issue:`11349`).
+- Some internal file descriptors were moved to number 10 or higher, to reduce risk of clashes with those used by the user in scripts.
+- The wording of error messages has been made consistent, especially for builtin subcommands (:issue:`12556`).
+
+For distributors and developers
+-------------------------------
+- When the default global config directory (``$PREFIX/etc/fish``) exists but has been overridden via ``-DCMAKE_INSTALL_SYSCONFDIR``, fish will now respect that override (:issue:`10748`).
+- ``build_tools/update_translations.fish`` has been replaced by ``cargo xtask gettext {check,new,update}`` (:issue:`12676`).
+- ``cargo xtask shellcheck`` to lint shell-scripts.
+
+Regression fixes:
+-----------------
+- (from 4.6) Vi mode ``dl`` (:issue:`12461`).
+- (from 4.6) Backspace after newline (:issue:`12583`).
+- (from 4.3.3) Long options were spuriously completed after typing short options (85e76ba3561).
+- (from 3.2) ``nosuchcommand || echo hello`` executes the right hand side again (:issue:`12654`).
+
+fish 4.6.0 (released March 28, 2026)
+====================================
+
+Notable improvements and fixes
+------------------------------
+- New Spanish translations (:issue:`12489`).
+- New Japanese translations (:issue:`12499`).
+
+Deprecations and removed features
+---------------------------------
+- The default width for emoji is switched from 1 to 2, improving the experience for users connecting to old systems from modern desktops. Users of old desktops who notice that lines containing emoji are misaligned can set ``$fish_emoji_width`` back to 1 (:issue:`12562`).
+
+Interactive improvements
+------------------------
+- The tab completion pager now left-justifies the description of each column (:issue:`12546`).
+- fish now supports the ``SHELL_PROMPT_PREFIX``, ``SHELL_PROMPT_SUFFIX``, and ``SHELL_WELCOME`` environment variables. The prefix and suffix are automatically prepended and appended to the left prompt, and the welcome message is displayed on startup after the greeting.
+  These variables are set by systemd's ``run0`` for example (:issue:`10924`).
+
+Improved terminal support
+-------------------------
+- ``set_color`` is able to turn off italics, reverse mode, strikethrough and underline individually (e.g. ``--italics=off``).
+- ``set_color`` learned the foreground (``--foreground`` or ``-f``) and reset (``--reset``) options.
+- An error caused by slow terminal responses at macOS startup has been addressed (:issue:`12571`).
+
+Other improvements
+------------------
+- Signals like ``SIGWINCH`` (as sent on terminal resize) no longer interrupt builtin output (:issue:`12496`).
+- For compatibility with Bash, fish now accepts ``|&`` as alternate spelling of ``&|``, for piping both standard output and standard error (:issue:`11516`).
+- ``fish_indent`` now preserves comments and newlines immediately preceding a brace block (``{ }``) (:issue:`12505`).
+- A crash when suspending certain pipelines with :kbd:`ctrl-z` has been fixed (:issue:`12301`).
+
+For distributors and developers
+-------------------------------
+- ``cargo xtask`` subcommands no longer panic on test failures.
+
+Regression fixes:
+-----------------
+- (from 4.5.0) Intermediate ``⏎`` artifact when redrawing prompt (:issue:`12476`).
+- (from 4.4.0) ``history`` honors explicitly specified ``--color=`` again (:issue:`12512`).
+- (from 4.4.0) Vi mode ``dl`` and ``dh`` (:issue:`12461`).
+- (from 4.3.0) Error completing of commands starting with ``-`` (:issue:`12522`).
+
 fish 4.5.0 (released February 17, 2026)
 =======================================
 
@@ -39,7 +165,7 @@ New or improved bindings
 - Vi mode word movements (``w``, ``W``, ``e``, and ``E``) are now largely in line with Vim. The only exception is that underscores are treated as word separators (:issue:`12269`).
 - New special input functions to support these movements: ``forward-word-vi``, ``kill-word-vi``, ``forward-bigword-vi``, ``kill-bigword-vi``, ``forward-word-end``, ``backward-word-end``, ``forward-bigword-end``, ``backward-bigword-end``, ``kill-a-word``, ``kill-inner-word``, ``kill-a-bigword``, and ``kill-inner-bigword``.
 - Vi mode key bindings now support counts for movement and deletion commands (e.g. `d3w` or `3l`), via a new operator mode (:issue:`2192`).
-- New ``catpuccin-*`` color themes.
+- New ``catppuccin-*`` color themes.
 
 Improved terminal support
 -------------------------
@@ -1334,7 +1460,7 @@ Deprecations and removed features
 
   Like ``stderr-nocaret``, they will eventually be made read-only.
 - Most ``string`` subcommands no longer append a newline to their input if the input didn't have one (:issue:`8473`, :issue:`3847`)
-- Fish's escape sequence removal (like for ``string length --visible`` or to figure out how wide the prompt is) no longer has special support for non-standard color sequences like from Data General terminals, e.g. the Data General Dasher D220 from 1984. This removes a bunch of work in the common case, allowing ``string length --visible`` to be much faster with unknown escape sequences. We don't expect anyone to have ever used fish with such a terminal (:issue:`8769`).
+- fish's escape sequence removal (like for ``string length --visible`` or to figure out how wide the prompt is) no longer has special support for non-standard color sequences like from Data General terminals, e.g. the Data General Dasher D220 from 1984. This removes a bunch of work in the common case, allowing ``string length --visible`` to be much faster with unknown escape sequences. We don't expect anyone to have ever used fish with such a terminal (:issue:`8769`).
 - Code to upgrade universal variables from fish before 3.0 has been removed. Users who upgrade directly from fish versions 2.7.1 or before will have to set their universal variables & abbreviations again. (:issue:`8781`)
 - The meaning of an empty color variable has changed (:issue:`8793`). Previously, when a variable was set but empty, it would be interpreted as the "normal" color. Now, empty color variables cause the same effect as unset variables - the general highlighting variable for that type is used instead. For example::
 
@@ -1345,14 +1471,14 @@ Deprecations and removed features
 
   This makes it easier to make self-contained color schemes that don't accidentally use color that was set before.
   ``fish_config`` has been adjusted to set known color variables that a theme doesn't explicitly set to empty.
-- ``eval`` is now a reserved keyword, so it can't be used as a function name. This follows ``set`` and ``read``, and is necessary because it can't be cleanly shadowed by a function - at the very least ``eval set -l argv foo`` breaks. Fish will ignore autoload files for it, so left over ``eval.fish`` from previous fish versions won't be loaded.
+- ``eval`` is now a reserved keyword, so it can't be used as a function name. This follows ``set`` and ``read``, and is necessary because it can't be cleanly shadowed by a function - at the very least ``eval set -l argv foo`` breaks. fish will ignore autoload files for it, so left over ``eval.fish`` from previous fish versions won't be loaded.
 - The git prompt in informative mode now defaults to skipping counting untracked files, as this was extremely slow. To turn it on, set :envvar:`__fish_git_prompt_showuntrackedfiles` or set the git config value "bash.showuntrackedfiles" to ``true`` explicitly (which can be done for individual repositories). The "informative+vcs" sample prompt already skipped display of untracked files, but didn't do so in a way that skipped the computation, so it should be quite a bit faster in many cases (:issue:`8980`).
 - The ``__terlar_git_prompt`` function, used by the "Terlar" sample prompt, has been rebuilt as a configuration of the normal ``fish_git_prompt`` to ease maintenance, improve performance and add features (like reading per-repo git configuration). Some slight changes remain; users who absolutely must have the same behavior are encouraged to copy the old function (:issue:`9011`, :issue:`7918`, :issue:`8979`).
 
 Scripting improvements
 ----------------------
 - Quoted command substitution that directly follow a variable expansion (like ``echo "$var$(echo x)"``) no longer affect the variable expansion (:issue:`8849`).
-- Fish now correctly expands command substitutions that are preceded by an escaped dollar (like ``echo \$(echo)``). This regressed in version 3.4.0.
+- fish now correctly expands command substitutions that are preceded by an escaped dollar (like ``echo \$(echo)``). This regressed in version 3.4.0.
 - ``math`` can now handle underscores (``_``) as visual separators in numbers (:issue:`8611`, :issue:`8496`)::
 
     math 5 + 2_123_252
@@ -1372,7 +1498,7 @@ Scripting improvements
 
 Interactive improvements
 ------------------------
-- Fish now reports a special error if a command wasn't found and there is a non-executable file by that name in :envvar:`PATH` (:issue:`8804`).
+- fish now reports a special error if a command wasn't found and there is a non-executable file by that name in :envvar:`PATH` (:issue:`8804`).
 - ``less`` and other interactive commands would occasionally be stopped when run in a pipeline with fish functions; this has been fixed (:issue:`8699`).
 - Case-changing autosuggestions generated mid-token now correctly append only the suffix, instead of duplicating the token (:issue:`8820`).
 - ``ulimit`` learned a number of new options for the resource limits available on Linux, FreeBSD ande NetBSD, and returns a specific warning if the limit specified is not available on the active operating system (:issue:`8823`, :issue:`8786`).
@@ -1382,9 +1508,9 @@ Interactive improvements
 - Since fish 3.2.0, pressing :kbd:`ctrl-d` while a command is running would end up inserting a space into the next commandline, which has been fixed (:issue:`8871`).
 - A bug that caused multi-line prompts to be moved down a line when pasting or switching modes has been fixed (:issue:`3481`).
 - The Web-based configuration system no longer strips too many quotes in the abbreviation display (:issue:`8917`, :issue:`8918`).
-- Fish started with ``--no-config`` will now use the default keybindings (:issue:`8493`)
+- fish started with ``--no-config`` will now use the default keybindings (:issue:`8493`)
 - When fish inherits a :envvar:`USER` environment variable value that doesn't correspond to the current effective user ID, it will now correct it in all cases (:issue:`8879`, :issue:`8583`).
-- Fish sets a new :envvar:`EUID` variable containing the current effective user id (:issue:`8866`).
+- fish sets a new :envvar:`EUID` variable containing the current effective user id (:issue:`8866`).
 - ``history search`` no longer interprets the search term as an option (:issue:`8853`)
 - The status message when a job terminates should no longer be erased by a multiline prompt (:issue:`8817`)
 
@@ -1645,7 +1771,7 @@ Improved terminal support
 
 Other improvements
 ------------------
-- Fish's test suite now uses ``ctest``, and has become much faster to run. It is now also possible to run only specific tests with targets named ``test_$filename`` - ``make test_set.fish`` only runs the set.fish test. (:issue:`7851`)
+- fish's test suite now uses ``ctest``, and has become much faster to run. It is now also possible to run only specific tests with targets named ``test_$filename`` - ``make test_set.fish`` only runs the set.fish test. (:issue:`7851`)
 - The HTML version of the documentation now includes copy buttons for code examples (:issue:`8218`).
 - The HTML version of the documentation and the web-based configuration tool now pick more modern system fonts instead of falling back to Arial and something like Courier New most of the time (:issue:`8632`).
 - The Debian & Ubuntu package linked from fishshell.com is now a single package, rather than split into ``fish`` and ``fish-common`` (:issue:`7845`).
@@ -3244,7 +3370,7 @@ Other fixes and improvements
    variables (:issue:`4200`, :issue:`4341`), executing functions, globs (:issue:`4579`),
    ``string`` reading from standard input (:issue:`4610`), and slicing history
    (in particular, ``$history[1]`` for the last executed command).
--  Fish’s internal wcwidth function has been updated to deal with newer
+-  fish’s internal wcwidth function has been updated to deal with newer
    Unicode, and the width of some characters can be configured via the
    ``fish_ambiguous_width`` (:issue:`5149`) and ``fish_emoji_width`` (:issue:`2652`)
    variables. Alternatively, a new build-time option INTERNAL_WCWIDTH
@@ -3281,7 +3407,7 @@ For distributors and developers
    standard sh instead.
 -  The ``hostname`` command is no longer required for fish to operate.
 
-–
+-
 
 fish 2.7.1 (released December 23, 2017)
 =======================================
@@ -3293,7 +3419,7 @@ session (:issue:`4521`).
 If you are upgrading from version 2.6.0 or before, please also review
 the release notes for 2.7.0 and 2.7b1 (included below).
 
-–
+-
 
 fish 2.7.0 (released November 23, 2017)
 =======================================
@@ -3305,7 +3431,7 @@ from version 2.6.0 or before, please also review the release notes for
 Xcode builds and macOS packages could not be produced with 2.7b1, but
 this is fixed in 2.7.0.
 
-–
+-
 
 fish 2.7b1 (released October 31, 2017)
 ======================================
@@ -4000,13 +4126,13 @@ Backward-incompatible changes
 Other notable fixes and improvements
 ------------------------------------
 
--  Fish no longer silences errors in config.fish (:issue:`2702`)
+-  fish no longer silences errors in config.fish (:issue:`2702`)
 -  Directory autosuggestions will now descend as far as possible if
    there is only one child directory (:issue:`2531`)
 -  Add support for bright colors (:issue:`1464`)
 -  Allow Ctrl-J (``\cj``) to be bound separately from Ctrl-M
    (``\cm``) (:issue:`217`)
--  psub now has a “-s”/“–suffix” option to name the temporary file with
+-  psub now has a “-s”/“-suffix” option to name the temporary file with
    that suffix
 -  Enable 24-bit colors on select terminals (:issue:`2495`)
 -  Support for SVN status in the prompt (:issue:`2582`)
@@ -4030,13 +4156,13 @@ Other notable fixes and improvements
       systemd-analyze, localectl, timedatectl
    -  and more
 
--  Fish no longer has a function called sgrep, freeing it for user
+-  fish no longer has a function called sgrep, freeing it for user
    customization (:issue:`2245`)
 -  A rewrite of the completions for cd, fixing a few bugs (:issue:`2299`, :issue:`2300`,
    :issue:`562`)
 -  Linux VTs now run in a simplified mode to avoid issues (:issue:`2311`)
 -  The vi-bindings now inherit from the emacs bindings
--  Fish will also execute ``fish_user_key_bindings`` when in vi-mode
+-  fish will also execute ``fish_user_key_bindings`` when in vi-mode
 -  ``funced`` will now also check $VISUAL (:issue:`2268`)
 -  A new ``suspend`` function (:issue:`2269`)
 -  Subcommand completion now works better with split /usr (:issue:`2141`)
@@ -4105,7 +4231,7 @@ Other notable fixes and improvements
 
 -  New documentation design (:issue:`1662`), which requires a Doxygen version
    1.8.7 or newer to build.
--  Fish now defines a default directory for other packages to provide
+-  fish now defines a default directory for other packages to provide
    completions. By default this is
    ``/usr/share/fish/vendor-completions.d``; on systems with
    ``pkgconfig`` installed this path is discoverable with
@@ -4396,7 +4522,7 @@ Other Notable Fixes
 -  xsel is no longer built as part of fish. It will still be invoked if
    installed separately :issue:`633`
 -  \__fish_filter_mime no longer spews :issue:`628`
--  The –no-execute option to fish no longer falls over when reaching the
+-  The -no-execute option to fish no longer falls over when reaching the
    end of a block :issue:`624`
 -  fish_config knows how to find fish even if it’s not in the $PATH :issue:`621`
 -  A leading space now prevents writing to history, as is done in bash

CMakeLists.txt

@@ -53,36 +53,36 @@ add_definitions(-DCMAKE_SOURCE_DIR="${REAL_CMAKE_SOURCE_DIR}")
 
 set(build_types Release RelWithDebInfo Debug "")
 if(NOT "${CMAKE_BUILD_TYPE}" IN_LIST build_types)
-  message(WARNING "Unsupported build type ${CMAKE_BUILD_TYPE}. If this doesn't build, try one of Release, RelWithDebInfo or Debug")
+    message(WARNING "Unsupported build type ${CMAKE_BUILD_TYPE}. If this doesn't build, try one of Release, RelWithDebInfo or Debug")
 endif()
 
 add_custom_target(
-  fish ALL
-  COMMAND
-    "${CMAKE_COMMAND}" -E
-      env ${VARS_FOR_CARGO}
-        ${Rust_CARGO}
-          build --bin fish
-          $<$<CONFIG:Release>:--release>
-          $<$<CONFIG:RelWithDebInfo>:--profile=release-with-debug>
-          --target ${Rust_CARGO_TARGET}
-          --no-default-features
-          --features=${FISH_CARGO_FEATURES}
-          ${CARGO_FLAGS}
-    &&
-    "${CMAKE_COMMAND}" -E
-      copy "${rust_target_dir}/${rust_profile}/fish" "${CMAKE_CURRENT_BINARY_DIR}"
-  WORKING_DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}"
-  USES_TERMINAL
+    fish ALL
+    COMMAND
+        "${CMAKE_COMMAND}" -E
+            env ${VARS_FOR_CARGO}
+                ${Rust_CARGO}
+                    build --bin fish
+                    $<$<CONFIG:Release>:--release>
+                    $<$<CONFIG:RelWithDebInfo>:--profile=release-with-debug>
+                    --target ${Rust_CARGO_TARGET}
+                    --no-default-features
+                    --features=${FISH_CARGO_FEATURES}
+                    ${CARGO_FLAGS}
+        &&
+        "${CMAKE_COMMAND}" -E
+            copy "${rust_target_dir}/${rust_profile}/fish" "${CMAKE_CURRENT_BINARY_DIR}"
+    WORKING_DIRECTORY "${CMAKE_CURRENT_SOURCE_DIR}"
+    USES_TERMINAL
 )
 
 function(CREATE_LINK target)
-  add_custom_target(
-    ${target} ALL
-    DEPENDS fish
-    COMMAND ln -f fish ${target}
-    WORKING_DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}"
-  )
+    add_custom_target(
+        ${target} ALL
+        DEPENDS fish
+        COMMAND ln -f fish ${target}
+        WORKING_DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}"
+    )
 endfunction(CREATE_LINK)
 
 # Define fish_indent.

CONTRIBUTING.rst

@@ -1,10 +1,10 @@
 ####################
-Contributing To Fish
+Contributing To fish
 ####################
 
 This document tells you how you can contribute to fish.
 
-Fish is free and open source software, distributed under the terms of the GPLv2.
+fish is free and open source software, distributed under the terms of the GPLv2.
 
 Contributions are welcome, and there are many ways to contribute!
 
@@ -21,7 +21,7 @@ Archives are available at https://lists.sr.ht/~krobelus/fish-shell/.
 GitHub
 ======
 
-Fish is available on GitHub, at https://github.com/fish-shell/fish-shell.
+fish is available on GitHub, at https://github.com/fish-shell/fish-shell.
 
 First, you'll need an account there, and you'll need a git clone of fish.
 Fork it on GitHub and then run::
@@ -133,14 +133,14 @@ For formatting, we use:
 - ``fish_indent`` (shipped with fish) for fish script
 - ``ruff format`` for Python
 
-To reformat files, there is a script
+To reformat files, there is an xtask
 
 ::
 
-   build_tools/style.fish --all
-   build_tools/style.fish somefile.rs some.fish
+   cargo xtask format --all
+   cargo xtask format somefile.rs some.fish
 
-Fish Script Style Guide
+fish Script Style Guide
 -----------------------
 
 1. All fish scripts, such as those in the *share/functions* and *tests*
@@ -154,7 +154,7 @@ Fish Script Style Guide
    for public vars or ``_fish`` for private vars to minimize the
    possibility of name clashes with user defined vars.
 
-Configuring Your Editor for Fish Scripts
+Configuring Your Editor for fish Scripts
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
 If you use Vim: Install `vim-fish <https://github.com/dag/vim-fish>`__,
@@ -253,7 +253,7 @@ To run all tests and linters, use::
 Contributing Translations
 =========================
 
-Fish uses GNU gettext to translate messages from English to other languages.
+fish uses GNU gettext to translate messages from English to other languages.
 We use custom tools for extracting messages from source files and to localize at runtime.
 This means that we do not have a runtime dependency on the gettext library.
 It also means that some features are not supported, such as message context and plurals.
@@ -271,13 +271,14 @@ Adding translations for a new language
 --------------------------------------
 
 Creating new translations requires the Gettext tools.
-More specifically, you will need ``msguniq`` and ``msgmerge`` for creating translations for a new
-language.
-To create a new translation, run::
+More specifically, you will need ``msguniq``, ``msgmerge``, and ``msgattrib``
+for creating translations for a new language.
+To create a PO file for a new language ``ll_CC``, run::
 
-    build_tools/update_translations.fish localization/po/ll_CC.po
+    cargo xtask gettext new ll_CC
 
-This will create a new PO file containing all messages available for translation.
+This will create a new PO file in ``localization/po/``
+containing all messages available for translation.
 If the file already exists, it will be updated.
 
 After modifying a PO file, you can recompile fish, and it will integrate the modifications you made.
@@ -347,10 +348,12 @@ Modifications to strings in source files
 ----------------------------------------
 
 If a string changes in the sources, the old translations will no longer work.
-They will be preserved in the PO files, but commented-out (starting with ``#~``).
 If you add/remove/change a translatable strings in a source file,
-run ``build_tools/update_translations.fish`` to propagate this to all translation files (``localization/po/*.po``).
+run ``cargo xtask gettext update`` to propagate this to all translation files (``localization/po/*.po``).
 This is only relevant for developers modifying the source files of fish or fish scripts.
+Note translations for messages which are no longer present in the sources will be deleted from the PO files.
+If the source string changed in a way which should not affect translations,
+consider updating the ``msgid`` in the PO files such that translations are preserved.
 
 Setting Code Up For Translations
 --------------------------------

Cargo.toml

@@ -2,26 +2,30 @@
 members = ["crates/*"]
 
 [workspace.package]
+edition = "2024"
 # To build revisions that use Corrosion (those before 2024-01), use CMake 3.19, Rustc 1.78 and Rustup 1.27.
 rust-version = "1.85"
-edition = "2024"
 repository = "https://github.com/fish-shell/fish-shell"
 # see doc_src/license.rst for details
 # don't forget to update COPYING and debian/copyright too
 license = "GPL-2.0-only AND LGPL-2.0-or-later AND MIT AND PSF-2.0"
 
 [workspace.dependencies]
+anstyle = "1.0.13"
+anyhow = "1.0.102"
 assert_matches = "1.5.0"
 bitflags = "2.5.0"
 cc = "1.0.94"
 cfg-if = "1.0.3"
 clap = { version = "4.5.54", features = ["derive"] }
+clap_complete = { version = "4.6.4", features = ["unstable-dynamic"] }
 errno = "0.3.0"
 fish-build-helper = { path = "crates/build-helper" }
 fish-build-man-pages = { path = "crates/build-man-pages" }
 fish-color = { path = "crates/color" }
 fish-common = { path = "crates/common" }
 fish-fallback = { path = "crates/fallback" }
+fish-feature-flags = { path = "crates/feature-flags" }
 fish-gettext = { path = "crates/gettext" }
 fish-gettext-extraction = { path = "crates/gettext-extraction" }
 fish-gettext-maps = { path = "crates/gettext-maps" }
@@ -30,52 +34,54 @@ fish-printf = { path = "crates/printf", features = ["widestring"] }
 fish-tempfile = { path = "crates/tempfile" }
 fish-util = { path = "crates/util" }
 fish-wcstringutil = { path = "crates/wcstringutil" }
+fish-wgetopt = { path = "crates/wgetopt" }
 fish-widecharwidth = { path = "crates/widecharwidth" }
 fish-widestring = { path = "crates/widestring" }
-fish-wgetopt = { path = "crates/wgetopt" }
+ignore = "0.4.25"
 itertools = "0.14.0"
 libc = "0.2.177"
 # lru pulls in hashbrown by default, which uses a faster (though less DoS resistant) hashing algo.
 # disabling default features uses the stdlib instead, but it doubles the time to rewrite the history
 # files as of 22 April 2024.
-lru = "0.16.2"
+lru = "0.18.0"
 nix = { version = "0.31.1", default-features = false, features = [
-    "event",
-    "fs",
-    "inotify",
-    "hostname",
-    "resource",
-    "process",
-    "signal",
-    "term",
-    "user",
+  "event",
+  "fs",
+  "hostname",
+  "inotify",
+  "process",
+  "resource",
+  "signal",
+  "term",
+  "user",
 ] }
 num-traits = "0.2.19"
 once_cell = "1.19.0"
 pcre2 = { git = "https://github.com/fish-shell/rust-pcre2", tag = "0.2.9-utf32", default-features = false, features = [
-    "utf32",
+  "utf32",
 ] }
 phf = { version = "0.13", default-features = false }
 phf_codegen = "0.13"
 portable-atomic = { version = "1", default-features = false, features = [
-    "fallback",
+  "fallback",
 ] }
 proc-macro2 = "1.0"
-rand = { version = "0.9.2", default-features = false, features = [
-    "small_rng",
-    "thread_rng",
-] }
+rand = { version = "0.10.1", default-features = false, features = ["thread_rng"] }
+regex = "1.12.3"
 rsconf = "0.3.0"
 rust-embed = { version = "8.11.0", features = [
-    "deterministic-timestamps",
-    "include-exclude",
-    "interpolate-folder-path",
+  "deterministic-timestamps",
+  "include-exclude",
+  "interpolate-folder-path",
 ] }
+rustc_version = "0.4.1"
 serial_test = { version = "3", default-features = false }
-widestring = "1.2.0"
+strum_macros = "0.28.0"
 unicode-segmentation = "1.12.0"
 unicode-width = "0.2.0"
 unix_path = "1.0.1"
+walkdir = "2.5.0"
+widestring = "1.2.0"
 xterm-color = "1.0.1"
 
 [profile.release]
@@ -88,7 +94,7 @@ debug = true
 
 [package]
 name = "fish"
-version = "4.5.0"
+version = "4.7.1"
 edition.workspace = true
 rust-version.workspace = true
 default-run = "fish"
@@ -106,6 +112,7 @@ fish-build-man-pages = { workspace = true, optional = true }
 fish-color.workspace = true
 fish-common.workspace = true
 fish-fallback.workspace = true
+fish-feature-flags.workspace = true
 fish-gettext = { workspace = true, optional = true }
 fish-gettext-extraction = { workspace = true, optional = true }
 fish-printf.workspace = true
@@ -124,7 +131,7 @@ num-traits.workspace = true
 once_cell.workspace = true
 pcre2.workspace = true
 rand.workspace = true
-unicode-width.workspace = true
+strum_macros.workspace = true
 xterm-color.workspace = true
 
 [target.'cfg(not(target_has_atomic = "64"))'.dependencies]
@@ -132,16 +139,16 @@ portable-atomic.workspace = true
 
 [target.'cfg(windows)'.dependencies]
 rust-embed = { workspace = true, features = [
-    "deterministic-timestamps",
-    "debug-embed",
-    "include-exclude",
-    "interpolate-folder-path",
+  "deterministic-timestamps",
+  "debug-embed",
+  "include-exclude",
+  "interpolate-folder-path",
 ] }
 [target.'cfg(not(windows))'.dependencies]
 rust-embed = { workspace = true, features = [
-    "deterministic-timestamps",
-    "include-exclude",
-    "interpolate-folder-path",
+  "deterministic-timestamps",
+  "include-exclude",
+  "interpolate-folder-path",
 ] }
 
 [dev-dependencies]
@@ -153,6 +160,7 @@ fish-build-helper.workspace = true
 fish-gettext-mo-file-parser.workspace = true
 phf_codegen = { workspace = true, optional = true }
 rsconf.workspace = true
+rustc_version.workspace = true
 
 [target.'cfg(windows)'.build-dependencies]
 unix_path.workspace = true
@@ -177,22 +185,20 @@ path = "src/bin/fish_key_reader.rs"
 default = ["embed-manpages", "localize-messages"]
 benchmark = []
 embed-manpages = ["dep:fish-build-man-pages"]
+# This feature is used to enable extracting messages from the source code for localization.
+# It only needs to be enabled if updating these messages (and the corresponding PO files) is
+# desired. This happens for the `gettext` xtask, which is also invoked via `cargo xtask check`.
+# There should not be a need to enable this feature manually.
+gettext-extract = ["dep:fish-gettext-extraction"]
 # Enable gettext localization at runtime. Requires the `msgfmt` tool to generate catalog data at
 # build time.
 localize-messages = ["dep:fish-gettext"]
-# This feature is used to enable extracting messages from the source code for localization.
-# It only needs to be enabled if updating these messages (and the corresponding PO files) is
-# desired. This happens when running tests via `cargo xtask check` and when calling
-# `build_tools/update_translations.fish`, so there should not be a need to enable it manually.
-gettext-extract = ["dep:fish-gettext-extraction"]
 
 # The following features are auto-detected by the build-script and should not be enabled manually.
 tsan = []
 
 [workspace.lints]
-rust.non_camel_case_types = "allow"
-rust.non_upper_case_globals = "allow"
-rust.unknown_lints = "allow"
+rust.unknown_lints = { level = "allow", priority = -1 }
 rust.unstable_name_collisions = "allow"
 rustdoc.private_intra_doc_links = "allow"
 
@@ -224,6 +230,8 @@ unused_trait_names = "warn"
 # In the future, they might change to flag other methods of printing.
 print_stdout = "deny"
 print_stderr = "deny"
+# usage in tests is fine since it avoids interacting with TopicMonitor
+# and is configured in clippy.toml with `allow-print-in-tests`
 
 [lints]
 workspace = true

README.rst

@@ -1,9 +1,5 @@
-.. |Cirrus CI| image:: https://api.cirrus-ci.com/github/fish-shell/fish-shell.svg?branch=master
-      :target: https://cirrus-ci.com/github/fish-shell/fish-shell
-      :alt: Cirrus CI Build Status
-
-`fish <https://fishshell.com/>`__ - the friendly interactive shell |Build Status| |Cirrus CI|
-=============================================================================================
+`fish <https://fishshell.com/>`__ - the friendly interactive shell |Build Status|
+=================================================================================
 
 fish is a smart and user-friendly command line shell for macOS, Linux,
 and the rest of the family. fish includes features like syntax
@@ -31,7 +27,7 @@ macOS
 
 fish can be installed:
 
--  using `Homebrew <http://brew.sh/>`__: ``brew install fish``
+-  using `Homebrew <https://brew.sh/>`__: ``brew install fish``
 -  using `MacPorts <https://www.macports.org/>`__:
    ``sudo port install fish``
 -  using the `installer from fishshell.com <https://fishshell.com/>`__
@@ -66,7 +62,7 @@ Windows
    for Linux with the instructions for the appropriate distribution
    listed above under “Packages for Linux”, or from source with the
    instructions below.
--  Fish can also be installed on all versions of Windows using
+-  fish can also be installed on all versions of Windows using
    `Cygwin <https://cygwin.com/>`__ or `MSYS2 <https://github.com/Berrysoft/fish-msys2>`__.
 
 Building from source
@@ -221,5 +217,5 @@ There is also a fish tag on Stackoverflow, but it is typically a poor fit.
 Found a bug? Have an awesome idea? Please `open an
 issue <https://github.com/fish-shell/fish-shell/issues/new>`__.
 
-.. |Build Status| image:: https://github.com/fish-shell/fish-shell/workflows/make%20test/badge.svg
-   :target: https://github.com/fish-shell/fish-shell/actions
+.. |Build Status| image:: https://github.com/fish-shell/fish-shell/actions/workflows/test.yml/badge.svg
+   :target: https://github.com/fish-shell/fish-shell/actions/workflows/test.yml

build.rs

@@ -6,6 +6,10 @@
 use std::path::{Path, PathBuf};
 
 fn main() {
+    let is_nightly =
+        rustc_version::version_meta().unwrap().channel == rustc_version::Channel::Nightly;
+    rsconf::declare_cfg("nightly", is_nightly);
+
     setup_paths();
 
     // Add our default to enable tools that don't go through CMake, like "cargo test" and the

build_tools/check.sh

@@ -8,11 +8,29 @@ if [ "$FISH_CHECK_LINT" = false ]; then
     lint=false
 fi
 
+case "$(uname)" in
+    MSYS*)
+        is_cygwin=true
+        cygwin_var=MSYS
+        ;;
+    CYGWIN*)
+        is_cygwin=true
+        cygwin_var=CYGWIN
+        ;;
+    *)
+        is_cygwin=false
+        ;;
+esac
+
 check_dependency_versions=false
 if [ "${FISH_CHECK_DEPENDENCY_VERSIONS:-false}" != false ]; then
     check_dependency_versions=true
 fi
 
+green='\e[0;32m'
+yellow='\e[1;33m'
+reset='\e[m'
+
 if $check_dependency_versions; then
     command -v curl
     command -v jq
@@ -42,6 +60,7 @@ cargo() {
     fi
 }
 
+# shellcheck disable=2317,2329
 cleanup () {
     if [ -n "$gettext_template_dir" ] && [ -e "$gettext_template_dir" ]; then
         rm -r "$gettext_template_dir"
@@ -71,6 +90,7 @@ fi
 
 gettext_template_dir=$(mktemp -d)
 (
+    # shellcheck disable=2030
     export FISH_GETTEXT_EXTRACTION_DIR="$gettext_template_dir"
     cargo build --workspace --all-targets --features=gettext-extract
 )
@@ -78,17 +98,65 @@ if $lint; then
     if command -v cargo-deny >/dev/null; then
         cargo deny --all-features --locked --exclude-dev check licenses
     fi
-    PATH="$build_dir:$PATH" "$workspace_root/build_tools/style.fish" --all --check
-    for features in "" --no-default-features; do
+
+    if command -v shellcheck >/dev/null || { test -n "$CI" && ! $is_cygwin; }; then
+        cargo xtask shellcheck
+    fi
+
+    PATH="$build_dir:$PATH" cargo xtask format --all --check
+    for features in "" --no-default-features --all-features; do
         cargo clippy --workspace --all-targets $features
     done
+
+    cargo xtask gettext --rust-extraction-dir="$gettext_template_dir" check
 fi
-cargo test --no-default-features --workspace --all-targets
+
+# When running `cargo test`, some binaries (e.g. `fish_gettext_extraction`)
+# are dynamically linked against Rust's `std-xxx.dll` instead of being
+# statically link as they usually are.
+# On Cygwin, `PATH`is not properly updated to point to the `std-xxx.dll`
+# location, so we have to do it manually.
+# See:
+# - https://github.com/rust-lang/rust/issues/149050
+# - https://github.com/msys2/MSYS2-packages/issues/5784
+(
+    if $is_cygwin; then
+        PATH="$PATH:$(rustc --print target-libdir)"
+        export PATH
+    fi
+    cargo test --no-default-features --workspace --all-targets
+)
 cargo test --doc --workspace
+
 if $lint; then
     cargo doc --workspace --no-deps
 fi
-FISH_GETTEXT_EXTRACTION_DIR=$gettext_template_dir "$workspace_root/tests/test_driver.py" "$build_dir"
+
+system_tests() {
+    "$workspace_root/tests/test_driver.py" "$build_dir" "$@"
+}
+
+if $is_cygwin; then
+    # shellcheck disable=2059
+    printf "=== Running ${green}integration tests ${yellow}with${green} symlinks${reset}\n"
+    (
+        export "$cygwin_var"=winsymlinks
+        system_tests
+    )
+
+    # shellcheck disable=2059
+    printf "=== Running ${green}integration tests ${yellow}without${green} symlinks${reset}\n"
+    (
+        # Only redo the tests that use `ln` to saves some time
+        export "$cygwin_var"=
+        # shellcheck disable=2046
+        system_tests $(grep -l -E '\bln\b' -r tests/checks/)
+    )
+else
+    # shellcheck disable=2059
+    printf "=== Running ${green}integration tests${reset}\n"
+    system_tests
+fi
 
 exit
 }

build_tools/fish_xgettext.fish

@@ -1,147 +0,0 @@
-#!/usr/bin/env fish
-#
-# Tool to generate gettext messages template file.
-# Writes to stdout.
-# Intended to be called from `update_translations.fish`.
-
-argparse use-existing-template= -- $argv
-or exit $status
-
-begin
-    # Write header. This is required by msguniq.
-    # Note that this results in the file being overwritten.
-    # This is desired behavior, to get rid of the results of prior invocations
-    # of this script.
-    begin
-        echo 'msgid ""'
-        echo 'msgstr ""'
-        echo '"Content-Type: text/plain; charset=UTF-8\n"'
-        echo ""
-    end
-
-    set -g workspace_root (path resolve (status dirname)/..)
-
-    set -l rust_extraction_dir
-    if set -l --query _flag_use_existing_template
-        set rust_extraction_dir $_flag_use_existing_template
-    else
-        set rust_extraction_dir (mktemp -d)
-        # We need to build to ensure that the proc macro for extracting strings runs.
-        FISH_GETTEXT_EXTRACTION_DIR=$rust_extraction_dir cargo check --features=gettext-extract
-        or exit 1
-    end
-
-    function mark_section
-        set -l section_name $argv[1]
-        echo 'msgid "fish-section-'$section_name'"'
-        echo 'msgstr ""'
-        echo ''
-    end
-
-    mark_section tier1-from-rust
-
-    # Get rid of duplicates and sort.
-    find $rust_extraction_dir -type f -exec cat {} + | msguniq --no-wrap --sort-output
-    or exit 1
-
-    if not set -l --query _flag_use_existing_template
-        rm -r $rust_extraction_dir
-    end
-
-    function extract_fish_script_messages_impl
-        set -l regex $argv[1]
-        set -e argv[1]
-        # Using xgettext causes more trouble than it helps.
-        # This is due to handling of escaping in fish differing from formats xgettext understands
-        # (e.g. POSIX shell strings).
-        # We work around this issue by manually writing the file content.
-
-        # Steps:
-        # 1. We extract strings to be translated from the relevant files and drop the rest. This step
-        #    depends on the regex matching the entire line, and the first capture group matching the
-        #    string.
-        # 2. We unescape. This gets rid of some escaping necessary in fish strings.
-        # 3. The resulting strings are sorted alphabetically. This step is optional. Not sorting would
-        #    result in strings from the same file appearing together. Removing duplicates is also
-        #    optional, since msguniq takes care of that later on as well.
-        # 4. Single backslashes are replaced by double backslashes. This results in the backslashes
-        #    being interpreted as literal backslashes by gettext tooling.
-        # 5. Double quotes are escaped, such that they are not interpreted as the start or end of
-        #    a msgid.
-        # 6. We transform the string into the format expected in a PO file.
-        cat $argv |
-            string replace --filter --regex $regex '$1' |
-            string unescape |
-            sort -u |
-            sed -E -e 's_\\\\_\\\\\\\\_g' -e 's_"_\\\\"_g' -e 's_^(.*)$_msgid "\1"\nmsgstr ""\n_'
-    end
-
-    function extract_fish_script_messages
-        set -l tier $argv[1]
-        set -e argv[1]
-        if not set -q argv[1]
-            return
-        end
-        # This regex handles explicit requests to translate a message. These are more important to translate
-        # than messages which should be implicitly translated.
-        set -l explicit_regex '.*\( *_ (([\'"]).+?(?<!\\\\)\\2) *\).*'
-        mark_section "$tier-from-script-explicitly-added"
-        extract_fish_script_messages_impl $explicit_regex $argv
-
-        # This regex handles descriptions for `complete` and `function` statements. These messages are not
-        # particularly important to translate. Hence the "implicit" label.
-        set -l implicit_regex '^(?:\s|and |or )*(?:complete|function).*? (?:-d|--description) (([\'"]).+?(?<!\\\\)\\2).*'
-        mark_section "$tier-from-script-implicitly-added"
-        extract_fish_script_messages_impl $implicit_regex $argv
-    end
-
-    set -g share_dir $workspace_root/share
-
-    set -l tier1 $share_dir/config.fish
-    set -l tier2
-    set -l tier3
-
-    for file in $share_dir/completions/*.fish $share_dir/functions/*.fish
-        # set -l tier (string match -r '^# localization: .*' <$file)
-        set -l tier (string replace -rf -m1 \
-            '^# localization: (.*)$' '$1' <$file)
-        if set -q tier[1]
-            switch "$tier"
-                case tier1 tier2 tier3
-                    set -a $tier $file
-                case 'skip*'
-                case '*'
-                    echo >&2 "$file:1 unexpected localization tier: $tier"
-                    exit 1
-            end
-            continue
-        end
-        set -l dirname (path basename (path dirname $file))
-        set -l command_name (path basename --no-extension $file)
-        if test $dirname = functions &&
-                string match -q -- 'fish_*' $command_name
-            set -a tier1 $file
-            continue
-        end
-        if test $dirname != completions
-            echo >&2 "$file:1 missing localization tier for function file"
-            exit 1
-        end
-        if test -e $workspace_root/doc_src/cmds/$command_name.rst
-            set -a tier1 $file
-        else
-            set -a tier3 $file
-        end
-    end
-
-    extract_fish_script_messages tier1 $tier1
-    extract_fish_script_messages tier2 $tier2
-    extract_fish_script_messages tier3 $tier3
-end |
-    # At this point, all extracted strings have been written to stdout,
-    # starting with the ones taken from the Rust sources,
-    # followed by strings explicitly marked for translation in fish scripts,
-    # and finally the strings from fish scripts which get translated implicitly.
-    # Because we do not eliminate duplicates across these categories,
-    # we do it here, since other gettext tools expect no duplicates.
-    msguniq --no-wrap

build_tools/make_linux_packages.sh

@@ -0,0 +1,83 @@
+#!/bin/sh
+
+# This script takes a source tarball (from build_tools/make_tarball.sh) and a vendor tarball (from
+# build_tools/make_vendor_tarball.sh, generated if not present), and produces:
+# * Appropriately-named symlinks to look like a Debian package
+# * Debian .changes and .dsc files with plain names ($version-1) and supported Ubuntu prefixes
+#   ($version-1~somedistro)
+# * An RPM spec file
+# By default, input and output files go in ~/fish_built, but this can be controlled with the
+# FISH_ARTEFACT_PATH environment variable.
+
+{
+
+set -e
+
+version=$1
+[ -n "$version" ] || { echo "Version number required as argument" >&2; exit 1; }
+
+
+[ -n "$DEB_SIGN_KEYID$DEB_SIGN_KEYFILE" ] ||
+    echo "Warning: neither DEB_SIGN_KEYID or DEB_SIGN_KEYFILE environment variables are set; you
+will need a signing key for the author of the most recent debian/changelog entry." >&2
+
+workpath=${FISH_ARTEFACT_PATH:-~/fish_built}
+source_tarball="$workpath"/fish-"$version".tar.xz
+vendor_tarball="$workpath"/fish-"$version"-vendor.tar.xz
+
+[ -e "$source_tarball" ] || { echo "Missing source tarball, expected at $source_tarball" >&2; exit 1; }
+cd "$workpath"
+
+# Unpack the sources
+tar xf "$source_tarball"
+sourcepath="$workpath"/fish-"$version"
+
+# Generate the vendor tarball if it is not already present
+[ -e "$vendor_tarball" ] || (cd "$sourcepath"; build_tools/make_vendor_tarball.sh;)
+
+# This step requires network access, so do it early in case it fails
+# sh has no real array support
+ubuntu_versions=$(uv run --script "$sourcepath"/build_tools/supported_ubuntu_versions.py)
+
+# Write the specfile
+[ -e "$workpath"/fish.spec ] && { echo "Cowardly refusing to overwite an existing fish.spec" >&2;
+exit 1; }
+rpmversion=$(echo "$version" |sed -e 's/-/+/' -e 's/-/./g')
+sed -e "s/@version@/$version/g" -e "s/@rpmversion@/$rpmversion/g" \
+    < "$sourcepath"/fish.spec.in > "$workpath"/fish.spec
+
+# Make the symlinks for Debian
+ln -s "$source_tarball" "$workpath"/fish_"$version".orig.tar.xz
+ln -s "$vendor_tarball" "$workpath"/fish_"$version".orig-cargo-vendor.tar.xz
+
+# Set up the Debian source tree
+cd "$sourcepath"
+mkdir cargo-vendor
+tar -C cargo-vendor -x -f "$vendor_tarball"
+cp -r contrib/debian debian
+
+# The vendor tarball contains a new .cargo/config.toml, which has the
+# vendoring overrides appended to it. dpkg-source will add this as a
+# patch using the flags in debian/
+cp cargo-vendor/.cargo/config.toml .cargo/config.toml
+
+# Update the Debian changelog
+# The release scripts do this for release builds - skip if it has already been done
+if head -n1 debian/changelog | grep --invert-match --quiet --fixed-strings "$version"; then
+    debchange --newversion "$version-1" --distribution unstable "Snapshot build"
+fi
+
+# Builds the "plain" Debian package
+# debuild runs lintian, which takes ten minutes to run over the vendor directories
+# just use dpkg-buildpackage directly
+dpkg-buildpackage --build=source -d
+
+# Build the Ubuntu packages
+# deb-reversion does not work on source packages, so do the whole thing ourselves
+for series in $ubuntu_versions; do
+    sed -i -e "1 s/$version-1)/$version-1~$series)/" -e "1 s/unstable/$series/" debian/changelog
+    dpkg-buildpackage --build=source -d
+    sed -i -e "1 s/$version-1~$series)/$version-1)/" -e "1 s/$series/unstable/" debian/changelog
+done
+
+}

build_tools/make_tarball.sh

@@ -15,9 +15,8 @@ tmpdir=$(mktemp -d)
 manifest=$tmpdir/Cargo.toml
 lockfile=$tmpdir/Cargo.lock
 
-sed "s/^version = \".*\"\$/version = \"$VERSION\"/g" Cargo.toml \
- >"$manifest"
-awk -v version=$VERSION '
+sed "s/^version = \".*\"\$/version = \"$VERSION\"/g" Cargo.toml >"$manifest"
+awk -v version="$VERSION" '
     /^name = "fish"$/ { ok=1 }
     ok == 1 && /^version = ".*"$/ {
         ok = 2;

build_tools/make_vendor_tarball.sh

@@ -8,20 +8,6 @@
 # Exit on error
 set -e
 
-# We need GNU tar as that supports the --mtime and --transform options
-TAR=notfound
-for try in tar gtar gnutar; do
-    if $try -Pcf /dev/null --mtime now /dev/null >/dev/null 2>&1; then
-        TAR=$try
-        break
-    fi
-done
-
-if [ "$TAR" = "notfound" ]; then
-    echo 'No suitable tar (supporting --mtime) found as tar/gtar/gnutar in PATH'
-    exit 1
-fi
-
 # Get the current directory, which we'll use for telling Cargo where to find the sources
 wd="$PWD"
 

build_tools/osx_package_scripts/add-shell

@@ -10,8 +10,8 @@ fi
 
 scriptname=$(basename "$0")
 if [ "$(id -u)" -ne 0 ]; then
-     echo "${scriptname} must be run as root"
-     exit 1
+    echo "${scriptname} must be run as root"
+    exit 1
 fi
 
 file=/etc/shells

build_tools/osx_package_scripts/preinstall

@@ -2,6 +2,6 @@
 
 echo "Removing any previous installation"
 pkgutil --pkg-info "${INSTALL_PKG_SESSION_ID}" && pkgutil --only-files --files "${INSTALL_PKG_SESSION_ID}" | while read -r installed
-     do rm -v "${DSTVOLUME}${installed}"
+    do rm -v "${DSTVOLUME}${installed}"
 done
 echo "... removed"

build_tools/release-notes.sh

@@ -47,8 +47,8 @@ if test -z "$CI" || [ "$(git -C "$workspace_root" tag | wc -l)" -gt 1 ]; then {
             num_authors=$(wc -l <"$relnotes_tmp/committers-now")
             num_new_authors=$(wc -l <"$relnotes_tmp/committers-new")
             printf %s \
-                "This release comprises $num_commits commits since $previous_version," \
-                " contributed by $num_authors authors, $num_new_authors of which are new committers."
+                "This release brings $num_commits new commits since $previous_version," \
+                " contributed by $num_authors authors, $num_new_authors of which are new faces."
             echo
             echo
         )
@@ -86,10 +86,13 @@ if test -z "$CI" || [ "$(git -C "$workspace_root" tag | wc -l)" -gt 1 ]; then {
     echo
     echo 'Download links:'
     echo 'To download the source code for fish, we suggest the file named ``fish-'"$version"'.tar.xz``.'
+    # shellcheck disable=2016
     echo 'The file downloaded from ``Source code (tar.gz)`` will not build correctly.'
+    # shellcheck disable=2016
     echo 'A GPG signature using `this key <'"${FISH_GPG_PUBLIC_KEY_URL:-???}"'>`__ is available as ``fish-'"$version"'.tar.xz.asc``.'
     echo
     echo 'The files called ``fish-'"$version"'-linux-*.tar.xz`` contain'
+    # shellcheck disable=2016
     echo '`standalone fish binaries <https://github.com/fish-shell/fish-shell/?tab=readme-ov-file#building-fish-with-cargo>`__'
     echo 'for any Linux with the given CPU architecture.'
 } >"$relnotes_tmp/fake-workspace"/CHANGELOG.rst

build_tools/release.sh

@@ -72,7 +72,7 @@ integration_branch=$(
         --format='%(refname:strip=2)'
 )
 [ -n "$integration_branch" ] ||
-    git merge-base --is-ancestor $remote/master HEAD
+    git merge-base --is-ancestor "$remote"/master HEAD
 
 sed -n 1p CHANGELOG.rst | grep -q '^fish .*(released .*)$'
 sed -n 2p CHANGELOG.rst | grep -q '^===*$'
@@ -113,9 +113,9 @@ CreateCommit "Release $version"
 # Tags must be full objects, not lightweight tags, for
 # git_version-gen.sh to work.
 git -c "user.signingKey=$committer" \
-    tag --sign --message="Release $version" $version
+    tag --sign --message="Release $version" "$version"
 
-git push $remote $version
+git push "$remote" "$version"
 
 TIMEOUT=
 gh() {
@@ -173,6 +173,7 @@ actual_tag_oid=$(git ls-remote "$remote" |
 (
     cd "$tmpdir/local-tarball/fish-$version"
     uv --no-managed-python venv
+    # shellcheck disable=1091
     . .venv/bin/activate
     cmake -GNinja -DCMAKE_BUILD_TYPE=Debug .
     ninja doc
@@ -180,14 +181,17 @@ actual_tag_oid=$(git ls-remote "$remote" |
 CopyDocs() {
     rm -rf "$fish_site/site/docs/$1"
     cp -r "$tmpdir/local-tarball/fish-$version/cargo/fish-docs/html" "$fish_site/site/docs/$1"
-    git -C $fish_site add "site/docs/$1"
+    git -C "$fish_site" add "site/docs/$1"
 }
 minor_version=${version%.*}
 CopyDocs "$minor_version"
 latest_release=$(
     releases=$(git tag | grep '^[0-9]*\.[0-9]*\.[0-9]*.*' |
-        sed $(: "De-prioritize release candidates (1.2.3-rc0)") \
-        's/-/~/g' | LC_ALL=C sort --version-sort)
+        sed '
+            # De-prioritize release candidates (1.2.3-rc0)
+            s/-/~/g
+        ' | LC_ALL=C sort --version-sort
+    )
     printf %s\\n "$releases" | tail -1
 )
 if [ "$version" = "$latest_release" ]; then
@@ -251,6 +255,28 @@ do
     sleep 20
 done
 
+milestone_version="$(
+    if echo "$version" | grep -q '\.0$'; then
+        echo "$minor_version"
+    else
+        echo "$version"
+    fi
+)"
+milestone_number() {
+    gh_api_repo milestones?state=open |
+        jq --arg name "fish $1" '
+            .[] | select(.title == $name) | .number
+        '
+}
+gh_api_repo milestones/"$(milestone_number "$milestone_version")" \
+    --method PATCH --raw-field state=closed
+next_minor_version=$(echo "$minor_version" |
+    awk -F. '{ printf "%s.%s", $1, $2+1 }')
+if [ -z "$(milestone_number "$next_minor_version")" ]; then
+    gh_api_repo milestones --method POST \
+        --raw-field title="fish $next_minor_version"
+fi
+
 (
     cd "$fish_site"
     make new-release
@@ -272,7 +298,7 @@ done
 )
 
 if [ -n "$integration_branch" ]; then {
-    git push $remote "$version^{commit}":refs/heads/$integration_branch
+    git push "$remote" "$version^{commit}:refs/heads/$integration_branch"
 } else {
     changelog=$(cat - CHANGELOG.rst <<EOF
 fish ?.?.? (released ???)
@@ -283,32 +309,9 @@ EOF
     printf %s\\n "$changelog" >CHANGELOG.rst
     git add CHANGELOG.rst
     CreateCommit "start new cycle"
-    git push $remote HEAD:master
+    git push "$remote" HEAD:master
 } fi
 
-milestone_version="$(
-    if echo "$version" | grep -q '\.0$'; then
-        echo "$minor_version"
-    else
-        echo "$version"
-    fi
-)"
-milestone_number() {
-    gh_api_repo milestones?state=open |
-        jq --arg name "fish $1" '
-            .[] | select(.title == $name) | .number
-        '
-}
-gh_api_repo milestones/"$(milestone_number "$milestone_version")" \
-    --method PATCH --raw-field state=closed
-
-next_minor_version=$(echo "$minor_version" |
-    awk -F. '{ printf "%s.%s", $1, $2+1 }')
-if [ -z "$(milestone_number "$next_minor_version")" ]; then
-    gh_api_repo milestones --method POST \
-        --raw-field title="fish $next_minor_version"
-fi
-
 exit
 
 }

build_tools/style.fish

@@ -1,123 +0,0 @@
-#!/usr/bin/env fish
-#
-# This runs Python files, fish scripts (*.fish), and Rust files
-# through their respective code formatting programs.
-#
-# `--all`: Format all eligible files instead of the ones specified as arguments.
-# `--check`: Instead of reformatting, fail if a file is not formatted correctly.
-# `--force`: Proceed without asking if uncommitted changes are detected.
-#            Only relevant if `--all` is specified but `--check` is not specified.
-
-set -l fish_files
-set -l python_files
-set -l rust_files
-set -l all no
-
-argparse all check force -- $argv
-or exit $status
-
-if set -l -q _flag_all
-    set all yes
-    if set -q argv[1]
-        echo "Unexpected arguments: '$argv'"
-        exit 1
-    end
-end
-
-set -l workspace_root (realpath (status dirname)/..)
-
-if test $all = yes
-    if not set -l -q _flag_force; and not set -l -q _flag_check
-        # Potential for false positives: Not all fish files are formatted, see the `fish_files`
-        # definition below.
-        set -l relevant_uncommitted_changes (git status --porcelain --short --untracked-files=all | sed -e 's/^ *[^ ]* *//' | grep -E '.*\.(fish|py|rs)$')
-        if set -q relevant_uncommitted_changes[1]
-            for changed_file in $relevant_uncommitted_changes
-                echo $changed_file
-            end
-            echo
-            echo 'You have uncommitted changes (listed above). Are you sure you want to restyle?'
-            read -P 'y/N? ' -n1 -l ans
-            if not string match -qi y -- $ans
-                exit 1
-            end
-        end
-    end
-    set fish_files $workspace_root/{benchmarks,build_tools,etc,share}/**.fish
-    set python_files $workspace_root
-else
-    # Format the files specified as arguments.
-    set -l files $argv
-    set fish_files (string match -r '^.*\.fish$' -- $files)
-    set python_files (string match -r '^.*\.py$' -- $files)
-    set rust_files (string match -r '^.*\.rs$' -- $files)
-end
-
-set -l red (set_color red)
-set -l green (set_color green)
-set -l yellow (set_color yellow)
-set -l normal (set_color normal)
-
-function die -V red -V normal
-    echo $red$argv[1]$normal
-    exit 1
-end
-
-if set -q fish_files[1]
-    if not type -q fish_indent
-        echo
-        echo $yellow'Could not find `fish_indent` in `$PATH`.'$normal
-        exit 127
-    end
-    echo === Running "$green"fish_indent"$normal"
-    if set -l -q _flag_check
-        fish_indent --check -- $fish_files
-        or die "Fish files are not formatted correctly."
-    else
-        fish_indent -w -- $fish_files
-    end
-end
-
-if set -q python_files[1]
-    if not type -q ruff
-        echo
-        echo $yellow'Please install `ruff` to style python'$normal
-        exit 127
-    end
-    echo === Running "$green"ruff format"$normal"
-    if set -l -q _flag_check
-        ruff format --check $python_files
-        or die "Python files are not formatted correctly."
-    else
-        ruff format $python_files
-    end
-end
-
-if test $all = yes; or set -q rust_files[1]
-    if not cargo fmt --version >/dev/null
-        echo
-        echo $yellow'Please install "rustfmt" to style Rust, e.g. via:'
-        echo "rustup component add rustfmt"$normal
-        exit 127
-    end
-
-    set -l edition_spec string match -r '^edition\s*=.*'
-    test "$($edition_spec <Cargo.toml)" = "$($edition_spec <.rustfmt.toml)"
-    or die "Cargo.toml and .rustfmt.toml use different editions"
-
-    echo === Running "$green"rustfmt"$normal"
-    if set -l -q _flag_check
-        if test $all = yes
-            cargo fmt --all --check
-        else
-            rustfmt --check --files-with-diff $rust_files
-        end
-        or die "Rust files are not formatted correctly."
-    else
-        if test $all = yes
-            cargo fmt --all
-        else
-            rustfmt $rust_files
-        end
-    end
-end

build_tools/update-dependencies.sh

@@ -3,7 +3,6 @@
 set -ex
 
 command -v curl
-command -v gcloud
 command -v jq
 command -v rustup
 command -v updatecli
@@ -42,17 +41,18 @@ uv lock --upgrade --exclude-newer="$(date --date='7 days ago' --iso-8601)"
 
 from_gh() {
     repo=$1
-    path=$2
-    destination=$3
-    contents=$(curl -fsS https://raw.githubusercontent.com/"${repo}"/refs/heads/master/"${path}")
-    printf '%s\n' >"$destination" "$contents"
+    branch=$2
+    path=$3
+    destination=$4
+    contents=$(curl -fsS https://raw.githubusercontent.com/"${repo}"/refs/heads/"${branch}"/"${path}")
+    printf '%s\n' "$contents" >"$destination"
 }
 
-from_gh ridiculousfish/widecharwidth widechar_width.rs crates/widecharwidth/src/widechar_width.rs
-from_gh ridiculousfish/littlecheck littlecheck/littlecheck.py tests/littlecheck.py
-from_gh catppuccin/fish 'themes/Catppuccin Frappe.theme' share/themes/catppuccin-frappe.theme
-from_gh catppuccin/fish 'themes/Catppuccin Macchiato.theme' share/themes/catppuccin-macchiato.theme
-from_gh catppuccin/fish 'themes/Catppuccin Mocha.theme' share/themes/catppuccin-mocha.theme
+from_gh ridiculousfish/widecharwidth master widechar_width.rs crates/widecharwidth/src/widechar_width.rs
+from_gh ridiculousfish/littlecheck master littlecheck/littlecheck.py tests/littlecheck.py
+from_gh catppuccin/fish main themes/catppuccin-frappe.theme share/themes/catppuccin-frappe.theme
+from_gh catppuccin/fish main themes/catppuccin-macchiato.theme share/themes/catppuccin-macchiato.theme
+from_gh catppuccin/fish main themes/catppuccin-mocha.theme share/themes/catppuccin-mocha.theme
 
 # Update Cargo.lock
 cargo update

build_tools/update_translations.fish

@@ -1,156 +0,0 @@
-#!/usr/bin/env fish
-
-# Updates the files used for gettext translations.
-# By default, the whole xgettext + msgmerge pipeline runs,
-# which extracts the messages from the source files into $template_file,
-# and updates the PO files for each language from that.
-#
-# Use cases:
-# For developers:
-#   - Run with no args to update all PO files after making changes to Rust/fish sources.
-# For translators:
-#   - Specify the language you want to work on as an argument, which must be a file in the
-#     localization/po/ directory. You can specify a language which does not have translations
-#     yet by specifying the name of a file which does not yet exist.
-#     Make sure to follow the naming convention.
-# For testing:
-#   - Specify `--dry-run` to see if any updates to the PO files would by applied by this script.
-#     If this flag is specified, the script will exit with an error if there are outstanding
-#     changes, and will display the diff. Do not specify other flags if `--dry-run` is specified.
-#
-# Specify `--use-existing-template=DIR` to prevent running cargo for extracting an up-to-date
-# version of the localized strings. This flag is intended for testing setups which make it
-# inconvenient to run cargo here, but run it in an earlier step to ensure up-to-date values.
-# This argument is passed on to the `fish_xgettext.fish` script and has no other uses.
-# `DIR` must be the path to a gettext template file generated from our compilation process.
-# It can be obtained by running:
-#   set -l DIR (mktemp -d)
-#   FISH_GETTEXT_EXTRACTION_DIR=$DIR cargo check --features=gettext-extract
-
-# The sort utility is locale-sensitive.
-# Ensure that sorting output is consistent by setting LC_ALL here.
-set -gx LC_ALL C.UTF-8
-
-set -l build_tools (status dirname)
-set -l po_dir $build_tools/../localization/po
-
-set -l extract
-
-argparse dry-run use-existing-template= -- $argv
-or exit $status
-
-if test -z $argv[1]
-    # Update everything if not specified otherwise.
-    set -g po_files $po_dir/*.po
-else
-    set -l po_dir_id (stat --format='%d:%i' -- $po_dir)
-    for arg in $argv
-        set -l arg_dir_id (stat --format='%d:%i' -- (dirname $arg) 2>/dev/null)
-        if test $po_dir_id != "$arg_dir_id"
-            echo "Argument $arg is not a file in the directory $(realpath $po_dir)."
-            echo "Non-option arguments must specify paths to files in this directory."
-            echo ""
-            echo "If you want to add a new language to the translations not the following:"
-            echo "The filename must identify a language, with a two letter ISO 639-1 language code of the target language (e.g. 'pt' for Portuguese), and use the file extension '.po'."
-            echo "Optionally, you can specify a regional variant (e.g. 'pt_BR')."
-            echo "So valid filenames are of the shape 'll.po' or 'll_CC.po'."
-            exit 1
-        end
-        if not basename $arg | grep -qE '^[a-z]{2,3}(_[A-Z]{2})?\.po$'
-            echo "Filename does not match the expected format ('ll.po' or 'll_CC.po')."
-            exit 1
-        end
-    end
-    set -g po_files $argv
-end
-
-set -g template_file (mktemp)
-# Protect from externally set $tmpdir leaking into this script.
-set -g tmpdir
-
-function cleanup_exit
-    set -l exit_status $status
-
-    rm $template_file
-
-    if set -g --query tmpdir[1]
-        rm -r $tmpdir
-    end
-
-    exit $exit_status
-end
-
-if set -l --query extract
-    set -l xgettext_args
-    if set -l --query _flag_use_existing_template
-        set xgettext_args --use-existing-template=$_flag_use_existing_template
-    end
-    $build_tools/fish_xgettext.fish $xgettext_args >$template_file
-    or cleanup_exit
-end
-
-if set -l --query _flag_dry_run
-    # On a dry run, we do not modify localization/po/ but write to a temporary directory instead
-    # and check if there is a difference between localization/po/ and the tmpdir after re-generating
-    # the PO files.
-    set -g tmpdir (mktemp -d)
-
-    # Ensure tmpdir has the same initial state as the po dir.
-    cp -r $po_dir/* $tmpdir
-end
-
-# This is used to identify lines which should be set here via $header_lines.
-# Make sure that this prefix does not appear elsewhere in the file and only contains characters
-# without special meaning in a sed pattern.
-set -g header_prefix "# fish-note-sections: "
-
-function print_header
-    set -l header_lines \
-        "Translations are divided into sections, each starting with a fish-section-* pseudo-message." \
-        "The first few sections are more important." \
-        "Ignore the tier3 sections unless you have a lot of time."
-    for line in $header_lines
-        printf '%s%s\n' $header_prefix $line
-    end
-end
-
-function merge_po_files --argument-names template_file po_file
-    msgmerge --no-wrap --update --no-fuzzy-matching --backup=none --quiet \
-        $po_file $template_file
-    or cleanup_exit
-    set -l new_po_file (mktemp) # TODO Remove on failure.
-    # Remove obsolete messages instead of keeping them as #~ entries.
-    and msgattrib --no-wrap --no-obsolete -o $new_po_file $po_file
-    or cleanup_exit
-
-    begin
-        print_header
-        # Paste PO file without old header lines.
-        sed '/^'$header_prefix'/d' $new_po_file
-    end >$po_file
-    rm $new_po_file
-end
-
-for po_file in $po_files
-    if set --query tmpdir[1]
-        set po_file $tmpdir/(basename $po_file)
-    end
-    if test -e $po_file
-        merge_po_files $template_file $po_file
-    else
-        begin
-            print_header
-            cat $template_file
-        end >$po_file
-    end
-end
-
-if set -g --query tmpdir[1]
-    diff -ur $po_dir $tmpdir
-    or begin
-        echo ERROR: translations in localization/po/ are stale. Try running build_tools/update_translations.fish
-        cleanup_exit
-    end
-end
-
-cleanup_exit

build_tools/version-available-in-debian.sh

@@ -1,4 +1,4 @@
-#!/bin/bash
+#!/usr/bin/env bash
 
 set -euo pipefail
 
@@ -11,6 +11,6 @@ codename=$(
 curl -fsS https://sources.debian.org/api/src/"${package}"/ |
     jq -r --arg codename "${codename}" '
         .versions[] | select(.suites[] == $codename) | .version' |
-    sed 's/^\([0-9]\+\.[0-9]\+\).*/\1/' |
+    sed -E 's/^([0-9]+\.[0-9]+).*/\1/' |
     sort --version-sort |
     tail -1

clippy.toml

@@ -0,0 +1 @@
+allow-print-in-tests = true

cmake/Docs.cmake

@@ -51,9 +51,10 @@ endif()
 add_feature_info(Documentation WITH_DOCS "user manual and documentation")
 
 if(WITH_DOCS)
-    add_custom_target(doc ALL
-                      DEPENDS sphinx-docs sphinx-manpages)
+    add_custom_target(doc ALL DEPENDS sphinx-docs sphinx-manpages)
     # Group docs targets into a DocsTargets folder
-    set_property(TARGET doc sphinx-docs sphinx-manpages
-                 PROPERTY FOLDER cmake/DocTargets)
+    set_property(
+        TARGET doc sphinx-docs sphinx-manpages
+        PROPERTY FOLDER cmake/DocTargets
+    )
 endif()

cmake/FindRust.cmake

@@ -22,17 +22,17 @@ foreach(_VAR ${_Rust_USER_VARS})
 endforeach()
 
 if (NOT DEFINED Rust_CARGO_CACHED)
-  find_program(Rust_CARGO_CACHED cargo PATHS "$ENV{HOME}/.cargo/bin")
+    find_program(Rust_CARGO_CACHED cargo PATHS "$ENV{HOME}/.cargo/bin")
 endif()
 
 if (NOT EXISTS "${Rust_CARGO_CACHED}")
-  message(FATAL_ERROR "The cargo executable ${Rust_CARGO_CACHED} was not found. "
-    "Consider setting `Rust_CARGO_CACHED` to the absolute path of `cargo`."
-  )
+    message(FATAL_ERROR "The cargo executable ${Rust_CARGO_CACHED} was not found. "
+        "Consider setting `Rust_CARGO_CACHED` to the absolute path of `cargo`."
+    )
 endif()
 
 if (NOT DEFINED Rust_COMPILER_CACHED)
-  find_program(Rust_COMPILER_CACHED rustc PATHS "$ENV{HOME}/.cargo/bin")
+    find_program(Rust_COMPILER_CACHED rustc PATHS "$ENV{HOME}/.cargo/bin")
 endif()
 
 
@@ -45,31 +45,31 @@ endif()
 # Figure out the target by just using the host target.
 # If you want to cross-compile, you'll have to set Rust_CARGO_TARGET
 if(NOT Rust_CARGO_TARGET_CACHED)
-  execute_process(
-    COMMAND "${Rust_COMPILER_CACHED}" --version --verbose
-    OUTPUT_VARIABLE _RUSTC_VERSION_RAW
-    RESULT_VARIABLE _RUSTC_VERSION_RESULT
-  )
-
-  if(NOT ( "${_RUSTC_VERSION_RESULT}" EQUAL "0" ))
-    message(FATAL_ERROR "Failed to get rustc version.\n"
-      "${Rust_COMPILER} --version failed with error: `${_RUSTC_VERSION_RESULT}`")
-  endif()
-
-  if (_RUSTC_VERSION_RAW MATCHES "host: ([a-zA-Z0-9_\\-]*)\n")
-    set(Rust_DEFAULT_HOST_TARGET "${CMAKE_MATCH_1}")
-  else()
-    message(FATAL_ERROR
-      "Failed to parse rustc host target. `rustc --version --verbose` evaluated to:\n${_RUSTC_VERSION_RAW}"
+    execute_process(
+        COMMAND "${Rust_COMPILER_CACHED}" --version --verbose
+        OUTPUT_VARIABLE _RUSTC_VERSION_RAW
+        RESULT_VARIABLE _RUSTC_VERSION_RESULT
     )
-  endif()
 
-  if(CMAKE_CROSSCOMPILING)
-    message(FATAL_ERROR "CMake is in cross-compiling mode."
-      "Manually set `Rust_CARGO_TARGET`."
-    )
-  endif()
-  set(Rust_CARGO_TARGET_CACHED "${Rust_DEFAULT_HOST_TARGET}" CACHE STRING "Target triple")
+    if(NOT ( "${_RUSTC_VERSION_RESULT}" EQUAL "0" ))
+        message(FATAL_ERROR "Failed to get rustc version.\n"
+            "${Rust_COMPILER} --version failed with error: `${_RUSTC_VERSION_RESULT}`")
+    endif()
+
+    if (_RUSTC_VERSION_RAW MATCHES "host: ([a-zA-Z0-9_\\-]*)\n")
+        set(Rust_DEFAULT_HOST_TARGET "${CMAKE_MATCH_1}")
+    else()
+        message(FATAL_ERROR
+            "Failed to parse rustc host target. `rustc --version --verbose` evaluated to:\n${_RUSTC_VERSION_RAW}"
+        )
+    endif()
+
+    if(CMAKE_CROSSCOMPILING)
+        message(FATAL_ERROR "CMake is in cross-compiling mode."
+            "Manually set `Rust_CARGO_TARGET`."
+        )
+    endif()
+    set(Rust_CARGO_TARGET_CACHED "${Rust_DEFAULT_HOST_TARGET}" CACHE STRING "Target triple")
 endif()
 
 # Set the input variables as non-cache variables so that the variables are available after

cmake/Install.cmake

@@ -14,32 +14,36 @@ set(rel_completionsdir "fish/vendor_completions.d")
 set(rel_functionsdir "fish/vendor_functions.d")
 set(rel_confdir "fish/vendor_conf.d")
 
-set(extra_completionsdir
-    "${datadir}/${rel_completionsdir}"
-    CACHE STRING "Path for extra completions")
+set(
+    extra_completionsdir "${datadir}/${rel_completionsdir}"
+    CACHE STRING "Path for extra completions"
+)
 
-set(extra_functionsdir
-    "${datadir}/${rel_functionsdir}"
-    CACHE STRING "Path for extra functions")
+set(
+    extra_functionsdir "${datadir}/${rel_functionsdir}"
+    CACHE STRING "Path for extra functions"
+)
 
-set(extra_confdir
-    "${datadir}/${rel_confdir}"
-    CACHE STRING "Path for extra configuration")
+set(
+    extra_confdir "${datadir}/${rel_confdir}"
+    CACHE STRING "Path for extra configuration"
+)
 
 
 # These are the man pages that go in system manpath; all manpages go in the fish-specific manpath.
-set(MANUALS ${SPHINX_OUTPUT_DIR}/man/man1/fish.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish_indent.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish_key_reader.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish-doc.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish-tutorial.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish-language.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish-interactive.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish-terminal-compatibility.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish-completions.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish-prompt-tutorial.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish-for-bash-users.1
-            ${SPHINX_OUTPUT_DIR}/man/man1/fish-faq.1
+set(MANUALS
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish_indent.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish_key_reader.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish-doc.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish-tutorial.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish-language.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish-interactive.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish-terminal-compatibility.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish-completions.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish-prompt-tutorial.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish-for-bash-users.1
+    ${SPHINX_OUTPUT_DIR}/man/man1/fish-faq.1
 )
 
 # Determine which man page we don't want to install.
@@ -48,40 +52,49 @@ set(MANUALS ${SPHINX_OUTPUT_DIR}/man/man1/fish.1
 # On other operating systems, don't install a realpath man page, as they almost all have a realpath
 # command, while macOS does not.
 if(${CMAKE_SYSTEM_NAME} MATCHES "Darwin")
-  set(CONDEMNED_PAGE "open.1")
+    set(CONDEMNED_PAGE "open.1")
 else()
-  set(CONDEMNED_PAGE "realpath.1")
+    set(CONDEMNED_PAGE "realpath.1")
 endif()
 
 # Define a function to help us create directories.
 function(FISH_CREATE_DIRS)
-  foreach(dir ${ARGV})
-    install(DIRECTORY DESTINATION ${dir})
-  endforeach(dir)
+    foreach(dir ${ARGV})
+        install(DIRECTORY DESTINATION ${dir})
+    endforeach(dir)
 endfunction(FISH_CREATE_DIRS)
 
 function(FISH_TRY_CREATE_DIRS)
-  foreach(dir ${ARGV})
-    if(NOT IS_ABSOLUTE ${dir})
-      set(abs_dir "\$ENV{DESTDIR}\${CMAKE_INSTALL_PREFIX}/${dir}")
-    else()
-      set(abs_dir "\$ENV{DESTDIR}${dir}")
-    endif()
-    install(SCRIPT CODE "EXECUTE_PROCESS(COMMAND mkdir -p ${abs_dir} OUTPUT_QUIET ERROR_QUIET)
-                         execute_process(COMMAND chmod 755 ${abs_dir} OUTPUT_QUIET ERROR_QUIET)
-                        ")
-  endforeach()
+    foreach(dir ${ARGV})
+        if(NOT IS_ABSOLUTE ${dir})
+            set(abs_dir "\$ENV{DESTDIR}\${CMAKE_INSTALL_PREFIX}/${dir}")
+        else()
+            set(abs_dir "\$ENV{DESTDIR}${dir}")
+        endif()
+        install(SCRIPT CODE "
+            EXECUTE_PROCESS(COMMAND mkdir -p ${abs_dir} OUTPUT_QUIET ERROR_QUIET)
+            execute_process(COMMAND chmod 755 ${abs_dir} OUTPUT_QUIET ERROR_QUIET)
+        ")
+    endforeach()
 endfunction(FISH_TRY_CREATE_DIRS)
 
-install(PROGRAMS ${CMAKE_CURRENT_BINARY_DIR}/fish
-        PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ
-                    GROUP_EXECUTE WORLD_READ WORLD_EXECUTE
-        DESTINATION ${bindir})
+install(
+    PROGRAMS ${CMAKE_CURRENT_BINARY_DIR}/fish
+    PERMISSIONS
+        OWNER_READ
+        OWNER_WRITE
+        OWNER_EXECUTE
+        GROUP_READ
+        GROUP_EXECUTE
+        WORLD_READ
+        WORLD_EXECUTE
+    DESTINATION ${bindir}
+)
 
 if(NOT IS_ABSOLUTE ${bindir})
-  set(abs_bindir "\$ENV{DESTDIR}\${CMAKE_INSTALL_PREFIX}/${bindir}")
+    set(abs_bindir "\$ENV{DESTDIR}\${CMAKE_INSTALL_PREFIX}/${bindir}")
 else()
-  set(abs_bindir "\$ENV{DESTDIR}${bindir}")
+    set(abs_bindir "\$ENV{DESTDIR}${bindir}")
 endif()
 install(CODE "file(CREATE_LINK ${abs_bindir}/fish ${abs_bindir}/fish_indent)")
 install(CODE "file(CREATE_LINK ${abs_bindir}/fish ${abs_bindir}/fish_key_reader)")
@@ -90,87 +103,68 @@ fish_create_dirs(${sysconfdir}/fish/conf.d ${sysconfdir}/fish/completions
     ${sysconfdir}/fish/functions)
 install(FILES etc/config.fish DESTINATION ${sysconfdir}/fish/)
 
-fish_create_dirs(${rel_datadir}/fish ${rel_datadir}/fish/completions
-                 ${rel_datadir}/fish/functions
-                 ${rel_datadir}/fish/man/man1 ${rel_datadir}/fish/tools
-                 ${rel_datadir}/fish/tools/web_config
-                 ${rel_datadir}/fish/tools/web_config/js
-                 ${rel_datadir}/fish/prompts
-                 ${rel_datadir}/fish/themes
-                 )
+fish_create_dirs(
+    ${rel_datadir}/fish
+    ${rel_datadir}/fish/man/man1
+)
 
+# This file is embedded in the executable by rust-embed and never read from the filesystem
 configure_file(share/__fish_build_paths.fish.in share/__fish_build_paths.fish)
-install(FILES share/config.fish
-              ${CMAKE_CURRENT_BINARY_DIR}/share/__fish_build_paths.fish
-        DESTINATION ${rel_datadir}/fish)
 
 # Create only the vendor directories inside the prefix (#5029 / #6508)
-fish_create_dirs(${rel_datadir}/fish/vendor_completions.d ${rel_datadir}/fish/vendor_functions.d
-    ${rel_datadir}/fish/vendor_conf.d)
+fish_create_dirs(
+    ${rel_datadir}/fish/vendor_completions.d
+    ${rel_datadir}/fish/vendor_functions.d
+    ${rel_datadir}/fish/vendor_conf.d
+)
 
 fish_try_create_dirs(${rel_datadir}/pkgconfig)
 configure_file(fish.pc.in fish.pc.noversion @ONLY)
 
-add_custom_command(OUTPUT fish.pc
+add_custom_command(
+    OUTPUT fish.pc
     COMMAND sed '/Version/d' fish.pc.noversion > fish.pc
     COMMAND printf "Version: " >> fish.pc
     COMMAND ${CMAKE_CURRENT_SOURCE_DIR}/build_tools/git_version_gen.sh >> fish.pc
     WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
-    DEPENDS ${CMAKE_CURRENT_BINARY_DIR}/fish.pc.noversion)
+    DEPENDS ${CMAKE_CURRENT_BINARY_DIR}/fish.pc.noversion
+)
 
 add_custom_target(build_fish_pc ALL DEPENDS fish.pc)
 
-install(FILES ${CMAKE_CURRENT_BINARY_DIR}/fish.pc
-        DESTINATION ${rel_datadir}/pkgconfig)
-
-install(DIRECTORY share/completions/
-        DESTINATION ${rel_datadir}/fish/completions
-        FILES_MATCHING PATTERN "*.fish")
-
-install(DIRECTORY share/functions/
-        DESTINATION ${rel_datadir}/fish/functions
-        FILES_MATCHING PATTERN "*.fish")
-
-install(DIRECTORY share/prompts/
-        DESTINATION ${rel_datadir}/fish/prompts
-        FILES_MATCHING PATTERN "*.fish")
-
-install(DIRECTORY share/themes/
-        DESTINATION ${rel_datadir}/fish/themes
-        FILES_MATCHING PATTERN "*.theme")
+install(
+    FILES ${CMAKE_CURRENT_BINARY_DIR}/fish.pc
+    DESTINATION ${rel_datadir}/pkgconfig
+)
 
 # CONDEMNED_PAGE is managed by the conditional above
 # Building the man pages is optional: if sphinx isn't installed, they're not built
-install(DIRECTORY ${SPHINX_OUTPUT_DIR}/man/man1/
-        DESTINATION ${rel_datadir}/fish/man/man1
-        FILES_MATCHING
-        PATTERN "*.1"
-        PATTERN ${CONDEMNED_PAGE} EXCLUDE)
-
-install(PROGRAMS share/tools/create_manpage_completions.py
-        DESTINATION ${rel_datadir}/fish/tools/)
-
-install(DIRECTORY share/tools/web_config
-        DESTINATION ${rel_datadir}/fish/tools/
-        FILES_MATCHING
-        PATTERN "*.png"
-        PATTERN "*.css"
-        PATTERN "*.html"
-        PATTERN "*.py"
-        PATTERN "*.js")
+install(
+    DIRECTORY ${SPHINX_OUTPUT_DIR}/man/man1/
+    DESTINATION ${rel_datadir}/fish/man/man1
+    FILES_MATCHING
+    PATTERN "*.1"
+    PATTERN ${CONDEMNED_PAGE} EXCLUDE
+)
 
 # Building the man pages is optional: if Sphinx isn't installed, they're not built
 install(FILES ${MANUALS} DESTINATION ${mandir}/man1/ OPTIONAL)
-install(DIRECTORY ${SPHINX_OUTPUT_DIR}/html/ # Trailing slash is important!
-        DESTINATION ${docdir} OPTIONAL)
+install(
+    DIRECTORY ${SPHINX_OUTPUT_DIR}/html/ # Trailing slash is important!
+    DESTINATION ${docdir} OPTIONAL
+)
 install(FILES CHANGELOG.rst DESTINATION ${docdir})
 
 # Group install targets into a InstallTargets folder
-set_property(TARGET build_fish_pc
-             PROPERTY FOLDER cmake/InstallTargets)
+set_property(
+    TARGET build_fish_pc
+    PROPERTY FOLDER cmake/InstallTargets
+)
 
 # Make a target build_root that installs into the buildroot directory, for testing.
 set(BUILDROOT_DIR ${CMAKE_CURRENT_BINARY_DIR}/buildroot)
-add_custom_target(build_root
-                  COMMAND DESTDIR=${BUILDROOT_DIR} ${CMAKE_COMMAND}
-                          --build ${CMAKE_CURRENT_BINARY_DIR} --target install)
+add_custom_target(
+    build_root
+    COMMAND DESTDIR=${BUILDROOT_DIR} ${CMAKE_COMMAND}
+        --build ${CMAKE_CURRENT_BINARY_DIR} --target install
+)

cmake/PCRE2.cmake

@@ -1,9 +1,10 @@
 set(FISH_USE_SYSTEM_PCRE2 ON CACHE BOOL
-  "Try to use PCRE2 from the system, instead of the pcre2-sys version")
+    "Try to use PCRE2 from the system, instead of the pcre2-sys version"
+)
 
 if(FISH_USE_SYSTEM_PCRE2)
-  message(STATUS "Trying to use PCRE2 from the system")
+    message(STATUS "Trying to use PCRE2 from the system")
 else()
-  message(STATUS "Forcing static build of PCRE2")
-  set(FISH_PCRE2_BUILDFLAG "PCRE2_SYS_STATIC=1")
+    message(STATUS "Forcing static build of PCRE2")
+    set(FISH_PCRE2_BUILDFLAG "PCRE2_SYS_STATIC=1")
 endif(FISH_USE_SYSTEM_PCRE2)

cmake/Tests.cmake

@@ -1,27 +1,27 @@
 FILE(GLOB FISH_CHECKS CONFIGURE_DEPENDS ${CMAKE_SOURCE_DIR}/tests/checks/*.fish)
 foreach(CHECK ${FISH_CHECKS})
-  get_filename_component(CHECK_NAME ${CHECK} NAME)
-  add_custom_target(
-    test_${CHECK_NAME}
-    COMMAND ${CMAKE_SOURCE_DIR}/tests/test_driver.py ${CMAKE_CURRENT_BINARY_DIR}
-                checks/${CHECK_NAME}
-    WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}/tests
-    DEPENDS fish fish_indent fish_key_reader
-    USES_TERMINAL
-  )
+    get_filename_component(CHECK_NAME ${CHECK} NAME)
+    add_custom_target(
+        test_${CHECK_NAME}
+        COMMAND ${CMAKE_SOURCE_DIR}/tests/test_driver.py ${CMAKE_CURRENT_BINARY_DIR}
+            checks/${CHECK_NAME}
+        WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}/tests
+        DEPENDS fish fish_indent fish_key_reader
+        USES_TERMINAL
+    )
 endforeach(CHECK)
 
 FILE(GLOB PEXPECTS CONFIGURE_DEPENDS ${CMAKE_SOURCE_DIR}/tests/pexpects/*.py)
 foreach(PEXPECT ${PEXPECTS})
-  get_filename_component(PEXPECT ${PEXPECT} NAME)
-  add_custom_target(
-    test_${PEXPECT}
-    COMMAND ${CMAKE_SOURCE_DIR}/tests/test_driver.py ${CMAKE_CURRENT_BINARY_DIR}
-                pexpects/${PEXPECT}
-    WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}/tests
-    DEPENDS fish fish_indent fish_key_reader
-    USES_TERMINAL
-  )
+    get_filename_component(PEXPECT ${PEXPECT} NAME)
+    add_custom_target(
+        test_${PEXPECT}
+        COMMAND ${CMAKE_SOURCE_DIR}/tests/test_driver.py ${CMAKE_CURRENT_BINARY_DIR}
+            pexpects/${PEXPECT}
+        WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}/tests
+        DEPENDS fish fish_indent fish_key_reader
+        USES_TERMINAL
+    )
 endforeach(PEXPECT)
 
 # Rust stuff.
@@ -30,14 +30,20 @@ if(DEFINED ASAN)
     # Rust w/ -Zsanitizer=address requires explicitly specifying the --target triple or else linker
     # errors pertaining to asan symbols will ensue.
     if(NOT DEFINED Rust_CARGO_TARGET)
-        message(FATAL_ERROR "ASAN requires defining the CMake variable Rust_CARGO_TARGET to the
-            intended target triple")
+        message(
+            FATAL_ERROR
+            "ASAN requires defining the CMake variable Rust_CARGO_TARGET to the
+            intended target triple"
+        )
     endif()
 endif()
 if(DEFINED TSAN)
     if(NOT DEFINED Rust_CARGO_TARGET)
-        message(FATAL_ERROR "TSAN requires defining the CMake variable Rust_CARGO_TARGET to the
-            intended target triple")
+        message(
+            FATAL_ERROR
+            "TSAN requires defining the CMake variable Rust_CARGO_TARGET to the
+            intended target triple"
+        )
     endif()
 endif()
 
@@ -53,10 +59,10 @@ endif()
 
 # The top-level test target is "fish_run_tests".
 add_custom_target(fish_run_tests
-  # TODO: This should be replaced with a unified solution, possibly build_tools/check.sh.
-  COMMAND ${CMAKE_SOURCE_DIR}/tests/test_driver.py ${max_concurrency_flag} ${CMAKE_CURRENT_BINARY_DIR}
-  COMMAND env ${VARS_FOR_CARGO}
-            ${Rust_CARGO}
+    # TODO: This should be replaced with a unified solution, possibly build_tools/check.sh.
+    COMMAND ${CMAKE_SOURCE_DIR}/tests/test_driver.py ${max_concurrency_flag} ${CMAKE_CURRENT_BINARY_DIR}
+    COMMAND
+        env ${VARS_FOR_CARGO} ${Rust_CARGO}
             test
             --no-default-features
             --features=${FISH_CARGO_FEATURES}
@@ -64,7 +70,7 @@ add_custom_target(fish_run_tests
             --workspace
             --target-dir ${rust_target_dir}
             ${cargo_test_flags}
-  WORKING_DIRECTORY "${CMAKE_SOURCE_DIR}"
-  DEPENDS fish fish_indent fish_key_reader
-  USES_TERMINAL
+    WORKING_DIRECTORY "${CMAKE_SOURCE_DIR}"
+    DEPENDS fish fish_indent fish_key_reader
+    USES_TERMINAL
 )

contrib/debian/changelog

@@ -1,3 +1,27 @@
+fish (4.7.1-1) stable; urgency=medium
+
+  * Release of new version 4.7.1.
+
+  See https://github.com/fish-shell/fish-shell/releases/tag/4.7.1 for details.
+
+ -- Johannes Altmanninger <aclopte@gmail.com>  Fri, 08 May 2026 00:02:14 +0800
+
+fish (4.7.0-1) stable; urgency=medium
+
+  * Release of new version 4.7.0.
+
+  See https://github.com/fish-shell/fish-shell/releases/tag/4.7.0 for details.
+
+ -- Johannes Altmanninger <aclopte@gmail.com>  Tue, 05 May 2026 15:24:27 +0800
+
+fish (4.6.0-1) stable; urgency=medium
+
+  * Release of new version 4.6.0.
+
+  See https://github.com/fish-shell/fish-shell/releases/tag/4.6.0 for details.
+
+ -- Johannes Altmanninger <aclopte@gmail.com>  Sat, 28 Mar 2026 12:56:37 +0800
+
 fish (4.5.0-1) stable; urgency=medium
 
   * Release of new version 4.5.0.

contrib/debian/control

@@ -10,10 +10,12 @@ Build-Depends: debhelper-compat (= 13),
  gettext,
  libpcre2-dev,
  rustc (>= 1.85) | rustc-web (>= 1.85) | rustc-1.85,
+# pkg-config is needed for the pcre2 crate to find the pcre2 system library
+ pkgconf | pkg-config,
  python3-sphinx,
 # Test dependencies
  locales-all,
- man-db,
+ man-db | man,
  python3
 # 4.6.2 is Debian 12/Ubuntu Noble 24.04; Ubuntu Jammy is 4.6.0.1
 Standards-Version: 4.6.2
@@ -26,8 +28,8 @@ Architecture: any
 # for col and lock
 Depends: bsdextrautils,
  file,
-# for man
- man-db,
+# for showing built-in help pages
+ man-db | man,
 # for kill
  procps,
  python3 (>=3.5),

contrib/shell.nix

@@ -0,0 +1,33 @@
+# Environment containing all dependencies needed for
+# - building fish,
+# - building documentation,
+# - running all tests,
+# - formatting and checking lints.
+#
+# enter interactive bash shell:
+#   nix-shell contrib/shell.nix
+#
+# using system nixpkgs (otherwise fetches pinned version):
+#   nix-shell contrib/shell.nix --arg pkgs 'import <nixpkgs> {}'
+#
+# run single command:
+#   nix-shell contrib/shell --run "cargo xtask check"
+{ pkgs ? (import (builtins.fetchTarball {
+  url = "https://github.com/NixOS/nixpkgs/archive/nixos-25.11.tar.gz";
+  sha256 = "1ia5kjykm9xmrpwbzhbaf4cpwi3yaxr7shl6amj8dajvgbyh2yh4";
+}) { }), ... }:
+pkgs.mkShell {
+  buildInputs = [
+    (pkgs.python3.withPackages (pyPkgs: [ pyPkgs.pexpect ]))
+    pkgs.cargo
+    pkgs.clippy
+    pkgs.cmake
+    pkgs.gettext
+    pkgs.pcre2
+    pkgs.procps # tests use pgrep/pkill
+    pkgs.ruff
+    pkgs.rustc
+    pkgs.rustfmt
+    pkgs.sphinx
+  ];
+}

crates/build-helper/Cargo.toml

@@ -1,8 +1,8 @@
 [package]
 name = "fish-build-helper"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 

crates/build-helper/src/lib.rs

@@ -38,6 +38,14 @@ pub fn fish_doc_dir() -> Cow<'static, Path> {
     cargo_target_dir().join("fish-docs").into()
 }
 
+fn l10n_dir() -> Cow<'static, Path> {
+    workspace_root().join("localization").into()
+}
+
+pub fn po_dir() -> Cow<'static, Path> {
+    l10n_dir().join("po").into()
+}
+
 // TODO Move this to rsconf
 pub fn rebuild_if_path_changed<P: AsRef<Path>>(path: P) {
     rsconf::rebuild_if_path_changed(path.as_ref().to_str().unwrap());
@@ -90,14 +98,14 @@ pub fn target_os_is_cygwin() -> bool {
 
 #[macro_export]
 macro_rules! as_os_strs {
-    [ $( $x:expr, )* ] => {
+    [ $( $x:expr ),* $(,)? ] => {
         {
             use std::ffi::OsStr;
             fn as_os_str<S: AsRef<OsStr> + ?Sized>(s: &S) -> &OsStr {
                 s.as_ref()
             }
-            &[
-                $( as_os_str($x), )*
+            [
+                $( as_os_str($x) ),*
             ]
         }
     }

crates/build-man-pages/Cargo.toml

@@ -1,8 +1,8 @@
 [package]
 name = "fish-build-man-pages"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 

crates/color/Cargo.toml

@@ -1,8 +1,8 @@
 [package]
 name = "fish-color"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 

crates/common/Cargo.toml

@@ -1,16 +1,21 @@
 [package]
 name = "fish-common"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 
 [dependencies]
 bitflags.workspace = true
+fish-feature-flags.workspace = true
 fish-widestring.workspace = true
 libc.workspace = true
 nix.workspace = true
 
+[build-dependencies]
+fish-build-helper.workspace = true
+rsconf.workspace = true
+
 [lints]
 workspace = true

crates/common/build.rs

@@ -0,0 +1,5 @@
+use fish_build_helper::target_os_is_apple;
+
+fn main() {
+    rsconf::declare_cfg("apple", target_os_is_apple());
+}

crates/fallback/Cargo.toml

@@ -1,8 +1,8 @@
 [package]
 name = "fish-fallback"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 

crates/fallback/src/lib.rs

@@ -8,12 +8,12 @@
 use std::cmp;
 use std::sync::{
     LazyLock,
-    atomic::{AtomicIsize, Ordering},
+    atomic::{AtomicUsize, Ordering},
 };
 
 /// Width of ambiguous East Asian characters and, as of TR11, all private-use characters.
 /// 1 is the typical default, but we accept any non-negative override via `$fish_ambiguous_width`.
-pub static FISH_AMBIGUOUS_WIDTH: AtomicIsize = AtomicIsize::new(1);
+pub static FISH_AMBIGUOUS_WIDTH: AtomicUsize = AtomicUsize::new(1);
 
 /// Width of emoji characters.
 ///
@@ -25,68 +25,33 @@
 /// Valid values are 1, and 2. 1 is the typical emoji width used in Unicode 8 while some newer
 /// terminals use a width of 2 since Unicode 9.
 // For some reason, this is declared here and exposed here, but is set in `env_dispatch`.
-pub static FISH_EMOJI_WIDTH: AtomicIsize = AtomicIsize::new(1);
+pub static FISH_EMOJI_WIDTH: AtomicUsize = AtomicUsize::new(2);
 
 static WC_LOOKUP_TABLE: LazyLock<WcLookupTable> = LazyLock::new(WcLookupTable::new);
 
-/// A safe wrapper around the system `wcwidth()` function
-#[cfg(not(cygwin))]
-pub fn wcwidth(c: char) -> isize {
-    unsafe extern "C" {
-        pub unsafe fn wcwidth(c: libc::wchar_t) -> libc::c_int;
-    }
-
-    const {
-        assert!(size_of::<libc::wchar_t>() >= size_of::<char>());
-    }
-
-    let width = unsafe { wcwidth(c as libc::wchar_t) };
-    isize::try_from(width).unwrap()
-}
-
-// Big hack to use our versions of wcswidth where we know them to be broken, which is
-// EVERYWHERE (https://github.com/fish-shell/fish-shell/issues/2199)
-pub fn fish_wcwidth(c: char) -> isize {
-    // The system version of wcwidth should accurately reflect the ability to represent characters
-    // in the console session, but knows nothing about the capabilities of other terminal emulators
-    // or ttys. Use it from the start only if we are logged in to the physical console.
-    #[cfg(not(cygwin))]
-    if fish_common::is_console_session() {
-        return wcwidth(c);
-    }
-
+pub fn fish_wcwidth(c: char) -> Option<usize> {
     // Check for VS16 which selects emoji presentation. This "promotes" a character like U+2764
     // (width 1) to an emoji (probably width 2). So treat it as width 1 so the sums work. See #2652.
     // VS15 selects text presentation.
     let variation_selector_16 = '\u{FE0F}';
     let variation_selector_15 = '\u{FE0E}';
     if c == variation_selector_16 {
-        return 1;
+        return Some(1);
     } else if c == variation_selector_15 {
-        return 0;
+        return Some(0);
     }
 
     // Check for Emoji_Modifier property. Only the Fitzpatrick modifiers have this, in range
     // 1F3FB..1F3FF. This is a hack because such an emoji appearing on its own would be drawn as
     // width 2, but that's unlikely to be useful. See #8275.
     if ('\u{1F3FB}'..='\u{1F3FF}').contains(&c) {
-        return 0;
+        return Some(0);
     }
 
     let width = WC_LOOKUP_TABLE.classify(c);
-    match width {
-        WcWidth::NonCharacter | WcWidth::NonPrint | WcWidth::Combining | WcWidth::Unassigned => {
-            #[cfg(not(cygwin))]
-            {
-                // Fall back to system wcwidth in this case.
-                wcwidth(c)
-            }
-            #[cfg(cygwin)]
-            {
-                // No system wcwidth for UTF-32 on cygwin.
-                0
-            }
-        }
+    Some(match width {
+        WcWidth::NonPrint => return None,
+        WcWidth::NonCharacter | WcWidth::Combining | WcWidth::Unassigned => 0,
         WcWidth::Ambiguous | WcWidth::PrivateUse => {
             // TR11: "All private-use characters are by default classified as Ambiguous".
             FISH_AMBIGUOUS_WIDTH.load(Ordering::Relaxed)
@@ -94,26 +59,25 @@ pub fn fish_wcwidth(c: char) -> isize {
         WcWidth::One => 1,
         WcWidth::Two => 2,
         WcWidth::WidenedIn9 => FISH_EMOJI_WIDTH.load(Ordering::Relaxed),
-    }
+    })
 }
 
-/// fish's internal versions of wcwidth and wcswidth, which can use an internal implementation if
-/// the system one is busted.
-pub fn fish_wcswidth(s: &wstr) -> isize {
-    // ascii fast path; empty iterator returns true for .all()
-    if s.chars().all(|c| c.is_ascii() && !c.is_ascii_control()) {
-        return s.len() as isize;
+pub fn fish_wcswidth(s: &wstr) -> Option<usize> {
+    fish_wcswidth_canonicalizing(s, std::convert::identity)
+}
+
+pub fn fish_wcswidth_canonicalizing(s: &wstr, canonicalize: fn(char) -> char) -> Option<usize> {
+    let chars = s.chars().map(canonicalize);
+    // ascii fast path
+    if chars.clone().all(|c| c.is_ascii() && !c.is_ascii_control()) {
+        return Some(s.len());
     }
 
     let mut result = 0;
-    for c in s.chars() {
-        let w = fish_wcwidth(c);
-        if w < 0 {
-            return -1;
-        }
-        result += w;
+    for c in chars {
+        result += fish_wcwidth(c)?;
     }
-    result
+    Some(result)
 }
 
 pub fn wcscasecmp(lhs: &wstr, rhs: &wstr) -> cmp::Ordering {

crates/feature-flags/Cargo.toml

@@ -0,0 +1,13 @@
+[package]
+name = "fish-feature-flags"
+version = "0.0.0"
+edition.workspace = true
+rust-version.workspace = true
+repository.workspace = true
+license.workspace = true
+
+[dependencies]
+fish-widestring.workspace = true
+
+[lints]
+workspace = true

crates/feature-flags/src/lib.rs

@@ -1,15 +1,14 @@
 //! Flags to enable upcoming features
 
-use crate::prelude::*;
-use std::sync::atomic::AtomicBool;
-use std::sync::atomic::Ordering;
-
-#[cfg(test)]
-use std::cell::RefCell;
+use fish_widestring::{L, WExt as _, wstr};
+use std::{
+    cell::RefCell,
+    sync::atomic::{AtomicBool, Ordering},
+};
 
 /// The list of flags.
 #[repr(u8)]
-#[derive(Clone, Copy)]
+#[derive(Clone, Copy, PartialEq, Eq)]
 pub enum FeatureFlag {
     /// Whether ^ is supported for stderr redirection.
     StderrNoCaret,
@@ -63,10 +62,10 @@ pub struct FeatureMetadata {
     pub description: &'static wstr,
 
     /// Default flag value.
-    pub default_value: bool,
+    default_value: bool,
 
     /// Whether the value can still be changed or not.
-    pub read_only: bool,
+    read_only: bool,
 }
 
 /// The metadata, indexed by flag.
@@ -156,31 +155,26 @@ pub struct FeatureMetadata {
 ];
 
 thread_local!(
-    #[cfg(test)]
-    static LOCAL_FEATURES: RefCell<Option<Features>> = const { RefCell::new(None) };
+    static LOCAL_OVERRIDE_STACK: RefCell<Vec<(FeatureFlag, bool)>> =
+        const { RefCell::new(Vec::new()) };
 );
 
 /// The singleton shared feature set.
 static FEATURES: Features = Features::new();
 
 /// Perform a feature test on the global set of features.
-pub fn test(flag: FeatureFlag) -> bool {
-    #[cfg(test)]
-    {
-        LOCAL_FEATURES.with(|fc| fc.borrow().as_ref().unwrap_or(&FEATURES).test(flag))
+pub fn feature_test(flag: FeatureFlag) -> bool {
+    if let Some(value) = LOCAL_OVERRIDE_STACK.with(|stack| {
+        for &(overridden_feature, value) in stack.borrow().iter().rev() {
+            if flag == overridden_feature {
+                return Some(value);
+            }
+        }
+        None
+    }) {
+        return value;
     }
-    #[cfg(not(test))]
-    {
-        FEATURES.test(flag)
-    }
-}
-
-pub use test as feature_test;
-
-/// Set a flag.
-#[cfg(test)]
-pub fn set(flag: FeatureFlag, value: bool) {
-    LOCAL_FEATURES.with(|fc| fc.borrow().as_ref().unwrap_or(&FEATURES).set(flag, value));
+    FEATURES.test(flag)
 }
 
 /// Parses a comma-separated feature-flag string, updating ourselves with the values.
@@ -188,20 +182,7 @@ pub fn set(flag: FeatureFlag, value: bool) {
 /// The special group name "all" may be used for those who like to live on the edge.
 /// Unknown features are silently ignored.
 pub fn set_from_string<'a>(str: impl Into<&'a wstr>) {
-    let wstr: &wstr = str.into();
-    #[cfg(test)]
-    {
-        LOCAL_FEATURES.with(|fc| {
-            fc.borrow()
-                .as_ref()
-                .unwrap_or(&FEATURES)
-                .set_from_string(wstr);
-        });
-    }
-    #[cfg(not(test))]
-    {
-        FEATURES.set_from_string(wstr);
-    }
+    FEATURES.set_from_string(str.into());
 }
 
 impl Features {
@@ -237,19 +218,14 @@ fn set(&self, flag: FeatureFlag, value: bool) {
     }
 
     fn set_from_string(&self, str: &wstr) {
-        let whitespace = L!("\t\n\0x0B\0x0C\r ").as_char_slice();
-        for entry in str.as_char_slice().split(|c| *c == ',') {
+        for entry in str.split(',') {
+            let entry = entry.trim();
             if entry.is_empty() {
                 continue;
             }
 
-            // Trim leading and trailing whitespace
-            let entry = &entry[entry.iter().take_while(|c| whitespace.contains(c)).count()..];
-            let entry =
-                &entry[..entry.len() - entry.iter().take_while(|c| whitespace.contains(c)).count()];
-
             // A "no-" prefix inverts the sense.
-            let (name, value) = match entry.strip_prefix(L!("no-").as_char_slice()) {
+            let (name, value) = match entry.strip_prefix("no-") {
                 Some(suffix) => (suffix, false),
                 None => (entry, true),
             };
@@ -275,28 +251,20 @@ fn set_from_string(&self, str: &wstr) {
     }
 }
 
-#[cfg(test)]
-pub fn scoped_test(flag: FeatureFlag, value: bool, test_fn: impl FnOnce()) {
-    LOCAL_FEATURES.with(|fc| {
-        assert!(
-            fc.borrow().is_none(),
-            "scoped_test() does not support nesting"
-        );
-
-        let f = Features::new();
-        f.set(flag, value);
-        *fc.borrow_mut() = Some(f);
-
+/// Run code with a feature overridden.
+/// This should only be used in tests.
+pub fn with_overridden_feature(flag: FeatureFlag, value: bool, test_fn: impl FnOnce()) {
+    LOCAL_OVERRIDE_STACK.with(|stack| {
+        stack.borrow_mut().push((flag, value));
         test_fn();
-
-        *fc.borrow_mut() = None;
+        stack.borrow_mut().pop();
     });
 }
 
 #[cfg(test)]
 mod tests {
-    use super::{FeatureFlag, Features, METADATA, scoped_test, set, test};
-    use crate::prelude::*;
+    use super::{FeatureFlag, Features, METADATA, feature_test, with_overridden_feature};
+    use fish_widestring::L;
 
     #[test]
     fn test_feature_flags() {
@@ -322,25 +290,19 @@ fn test_feature_flags() {
     }
 
     #[test]
-    fn test_scoped() {
-        scoped_test(FeatureFlag::QuestionMarkNoGlob, true, || {
-            assert!(test(FeatureFlag::QuestionMarkNoGlob));
+    fn test_overridden_feature() {
+        with_overridden_feature(FeatureFlag::QuestionMarkNoGlob, true, || {
+            assert!(feature_test(FeatureFlag::QuestionMarkNoGlob));
         });
 
-        set(FeatureFlag::QuestionMarkNoGlob, true);
-
-        scoped_test(FeatureFlag::QuestionMarkNoGlob, false, || {
-            assert!(!test(FeatureFlag::QuestionMarkNoGlob));
+        with_overridden_feature(FeatureFlag::QuestionMarkNoGlob, false, || {
+            assert!(!feature_test(FeatureFlag::QuestionMarkNoGlob));
         });
 
-        set(FeatureFlag::QuestionMarkNoGlob, false);
-    }
-
-    #[test]
-    #[should_panic]
-    fn test_nested_scopes_not_supported() {
-        scoped_test(FeatureFlag::QuestionMarkNoGlob, true, || {
-            scoped_test(FeatureFlag::QuestionMarkNoGlob, false, || {});
+        with_overridden_feature(FeatureFlag::QuestionMarkNoGlob, false, || {
+            with_overridden_feature(FeatureFlag::QuestionMarkNoGlob, true, || {
+                assert!(feature_test(FeatureFlag::QuestionMarkNoGlob));
+            });
         });
     }
 }

crates/gettext-extraction/Cargo.toml

@@ -1,11 +1,11 @@
 [package]
 name = "fish-gettext-extraction"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
+description = "proc-macro for extracting strings for gettext translation"
 repository.workspace = true
 license.workspace = true
-description = "proc-macro for extracting strings for gettext translation"
 
 [lib]
 proc-macro = true

crates/gettext-maps/Cargo.toml

@@ -1,8 +1,8 @@
 [package]
 name = "fish-gettext-maps"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 

crates/gettext-maps/build.rs

@@ -11,11 +11,7 @@ fn main() {
         PathBuf::from(fish_build_helper::fish_build_dir()).join("fish-localization-map-cache");
     embed_localizations(&cache_dir);
 
-    fish_build_helper::rebuild_if_path_changed(
-        fish_build_helper::workspace_root()
-            .join("localization")
-            .join("po"),
-    );
+    fish_build_helper::rebuild_if_path_changed(fish_build_helper::po_dir());
 }
 
 fn embed_localizations(cache_dir: &Path) {
@@ -25,10 +21,6 @@ fn embed_localizations(cache_dir: &Path) {
         io::{BufWriter, Write as _},
     };
 
-    let po_dir = fish_build_helper::workspace_root()
-        .join("localization")
-        .join("po");
-
     // Ensure that the directory is created, because clippy cannot compile the code if the
     // directory does not exist.
     std::fs::create_dir_all(cache_dir).unwrap();
@@ -56,7 +48,7 @@ fn embed_localizations(cache_dir: &Path) {
         Ok(output) => {
             let has_check_format =
                 String::from_utf8_lossy(&output.stdout).contains("--check-format");
-            for dir_entry_result in po_dir.read_dir().unwrap() {
+            for dir_entry_result in fish_build_helper::po_dir().read_dir().unwrap() {
                 let dir_entry = dir_entry_result.unwrap();
                 let po_file_path = dir_entry.path();
                 if po_file_path.extension() != Some(OsStr::new("po")) {
@@ -147,7 +139,7 @@ fn to_raw_str(s: &str) -> String {
                 write!(
                     &mut cached_map_file,
                     "static {}: phf::Map<&'static str, &'static str> = {}",
-                    &map_name,
+                    map_name,
                     single_language_localization_map.build()
                 )
                 .unwrap();

crates/gettext-mo-file-parser/Cargo.toml

@@ -1,8 +1,8 @@
 [package]
 name = "fish-gettext-mo-file-parser"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 

crates/gettext/Cargo.toml

@@ -1,8 +1,8 @@
 [package]
 name = "fish-gettext"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 

crates/printf/Cargo.toml

@@ -1,10 +1,10 @@
 [package]
 name = "fish-printf"
+version = "0.2.1"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.2.1"
-repository.workspace = true
 description = "printf implementation, based on musl"
+repository.workspace = true
 license = "MIT"
 
 [dependencies]

crates/printf/src/tests.rs

@@ -400,8 +400,10 @@ fn test_char() {
 
 #[test]
 fn test_ptr() {
-    assert_fmt!("%p", core::ptr::null::<u8>() => "0");
-    assert_fmt!("%p", 0xDEADBEEF_usize as *const u8 => "0xdeadbeef");
+    assert_fmt!("%p", core::ptr::null::<()>() => "0");
+
+    let tmp = core::ptr::without_provenance::<()>(0xDEADBEEF);
+    assert_fmt!("%p", tmp => "0xdeadbeef");
 }
 
 #[test]

crates/tempfile/Cargo.toml

@@ -1,13 +1,13 @@
 [package]
 name = "fish-tempfile"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 
 [dependencies]
-nix = { workspace = true, features = ["fs", "feature"] }
+nix = { workspace = true, features = ["feature", "fs"] }
 rand.workspace = true
 
 [lints]

crates/util/Cargo.toml

@@ -1,13 +1,16 @@
 [package]
 name = "fish-util"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 
 [dependencies]
+errno.workspace = true
 fish-widestring.workspace = true
+libc.workspace = true
+nix.workspace = true
 rand.workspace = true
 
 [lints]

crates/util/src/lib.rs

@@ -1,9 +1,15 @@
 //! Generic utilities library.
 
+use errno::errno;
 use fish_widestring::prelude::*;
 use rand::{SeedableRng as _, rngs::SmallRng};
-use std::cmp::Ordering;
-use std::time;
+use std::{
+    cmp::Ordering,
+    ffi::CStr,
+    io::Write as _,
+    os::fd::{BorrowedFd, RawFd},
+    time,
+};
 
 /// Compares two wide character strings with an (arguably) intuitive ordering. This function tries
 /// to order strings in a way which is intuitive to humans with regards to sorting strings
@@ -57,14 +63,23 @@ pub fn wcsfilecmp(a: &wstr, b: &wstr) -> Ordering {
             continue;
         }
 
-        // Sort dashes after Z - see #5634
-        let mut acl = if ac == '-' { '[' } else { ac };
-        let mut bcl = if bc == '-' { '[' } else { bc };
+        let transform = |c| {
+            // Sort dashes after Z - see #5634
+            if c == '-' {
+                return '[';
+            }
+            if c == '/' {
+                return '\0';
+            }
+            c
+        };
+        let ac = transform(ac);
+        let bc = transform(bc);
         // TODO Compare the tail (enabled by Rust's Unicode support).
-        acl = acl.to_uppercase().next().unwrap();
-        bcl = bcl.to_uppercase().next().unwrap();
+        let ac = ac.to_uppercase().next().unwrap();
+        let bc = bc.to_uppercase().next().unwrap();
 
-        match acl.cmp(&bcl) {
+        match ac.cmp(&bc) {
             Ordering::Equal => {
                 ai += 1;
                 bi += 1;
@@ -127,9 +142,9 @@ pub fn wcsfilecmp_glob(a: &wstr, b: &wstr) -> Ordering {
         }
 
         // TODO Compare the tail (enabled by Rust's Unicode support).
-        let acl = ac.to_lowercase().next().unwrap();
-        let bcl = bc.to_lowercase().next().unwrap();
-        match acl.cmp(&bcl) {
+        let ac = ac.to_lowercase().next().unwrap();
+        let bc = bc.to_lowercase().next().unwrap();
+        match ac.cmp(&bc) {
             Ordering::Equal => {
                 ai += 1;
                 bi += 1;
@@ -234,6 +249,26 @@ fn wcsfilecmp_leading_digits(a: &wstr, b: &wstr) -> (Ordering, usize, usize) {
     (ret, ai, bi)
 }
 
+pub fn write_to_fd(input: &[u8], fd: RawFd) -> nix::Result<usize> {
+    nix::unistd::write(unsafe { BorrowedFd::borrow_raw(fd) }, input)
+}
+
+/// Prints the provided string, followed by a colon, space, and the string representation of the
+/// current errno via [`libc::strerror`].
+pub fn perror(s: &str) {
+    let e = errno().0;
+    let mut stderr = std::io::stderr().lock();
+    if !s.is_empty() {
+        let _ = write!(stderr, "{s}: ");
+    }
+    let slice = unsafe {
+        let msg = libc::strerror(e);
+        CStr::from_ptr(msg).to_bytes()
+    };
+    let _ = stderr.write_all(slice);
+    let _ = stderr.write_all(b"\n");
+}
+
 #[cfg(test)]
 mod tests {
     use super::wcsfilecmp;
@@ -288,5 +323,8 @@ macro_rules! validate {
         validate!("a00b", "a0b", Ordering::Less);
         validate!("a0b", "a00b", Ordering::Greater);
         validate!("a-b", "azb", Ordering::Greater);
+        validate!("a", "a b", Ordering::Less);
+        validate!("a/", "a b/", Ordering::Less);
+        validate!("a/b", "a b", Ordering::Less); // Note this is arbitrary.
     }
 }

crates/wcstringutil/Cargo.toml

@@ -1,13 +1,12 @@
 [package]
 name = "fish-wcstringutil"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 
 [dependencies]
-fish-common.workspace = true
 fish-fallback.workspace = true
 fish-widestring.workspace = true
 

crates/wcstringutil/src/lib.rs

@@ -1,8 +1,7 @@
 //! Helper functions for working with wcstring.
 
-use fish_common::{get_ellipsis_char, get_ellipsis_str};
 use fish_fallback::{fish_wcwidth, lowercase, lowercase_rev, wcscasecmp, wcscasecmp_fuzzy};
-use fish_widestring::{decode_byte_from_char, prelude::*};
+use fish_widestring::{ELLIPSIS_CHAR, prelude::*};
 
 /// Return the number of newlines in a string.
 pub fn count_newlines(s: &wstr) -> usize {
@@ -336,30 +335,6 @@ pub fn string_fuzzy_match_string(
     StringFuzzyMatch::try_create(string, match_against, anchor_start)
 }
 
-/// Implementation of wcs2bytes that accepts a callback.
-/// The first argument can be either a `&str` or `&wstr`.
-/// This invokes `func` with byte slices containing the UTF-8 encoding of the characters in the
-/// input, doing one invocation per character.
-/// If `func` returns false, it stops; otherwise it continues.
-/// Return false if the callback returned false, otherwise true.
-pub fn str2bytes_callback(input: impl IntoCharIter, mut func: impl FnMut(&[u8]) -> bool) -> bool {
-    // A `char` represents an Unicode scalar value, which takes up at most 4 bytes when encoded in UTF-8.
-    let mut converted = [0_u8; 4];
-
-    for c in input.chars() {
-        let bytes = if let Some(byte) = decode_byte_from_char(c) {
-            converted[0] = byte;
-            &converted[..=0]
-        } else {
-            c.encode_utf8(&mut converted).as_bytes()
-        };
-        if !func(bytes) {
-            return false;
-        }
-    }
-    true
-}
-
 /// Split a string by runs of any of the separator characters provided in `seps`.
 /// Note the delimiters are the characters in `seps`, not `seps` itself.
 /// `seps` may contain the NUL character.
@@ -473,32 +448,13 @@ pub fn split_about<'haystack>(
     output
 }
 
-#[derive(Eq, PartialEq)]
-pub enum EllipsisType {
-    None,
-    // Prefer niceness over minimalness
-    Prettiest,
-    // Make every character count ($ instead of ...)
-    Shortest,
-}
-
-pub fn truncate(input: &wstr, max_len: usize, etype: Option<EllipsisType>) -> WString {
-    let etype = etype.unwrap_or(EllipsisType::Prettiest);
+// TODO: This should work on render width rather than the number of codepoints.
+pub fn truncate(input: &wstr, max_len: usize) -> WString {
     if input.len() <= max_len {
         return input.to_owned();
     }
-
-    if etype == EllipsisType::None {
-        return input[..max_len].to_owned();
-    }
-    if etype == EllipsisType::Prettiest {
-        let ellipsis_str = get_ellipsis_str();
-        let mut output = input[..max_len - ellipsis_str.len()].to_owned();
-        output += ellipsis_str;
-        return output;
-    }
     let mut output = input[..max_len - 1].to_owned();
-    output.push(get_ellipsis_char());
+    output.push(ELLIPSIS_CHAR);
     output
 }
 
@@ -565,12 +521,12 @@ fn next(&mut self) -> Option<Self::Item> {
     }
 }
 
-/// Like fish_wcwidth, but returns 0 for characters with no real width instead of -1.
+/// Like fish_wcwidth, but returns 0 for characters with no real width instead of none.
 pub fn fish_wcwidth_visible(c: char) -> isize {
     if c == '\x08' {
         return -1;
     }
-    fish_wcwidth(c).max(0)
+    fish_wcwidth(c).unwrap_or_default().try_into().unwrap()
 }
 
 #[cfg(test)]

crates/wgetopt/Cargo.toml

@@ -1,8 +1,8 @@
 [package]
 name = "fish-wgetopt"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 

crates/widecharwidth/Cargo.toml

@@ -1,8 +1,8 @@
 [package]
 name = "fish-widecharwidth"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license = "CC0-1.0"
 

crates/widestring/Cargo.toml

@@ -1,12 +1,14 @@
 [package]
 name = "fish-widestring"
+version = "0.0.0"
 edition.workspace = true
 rust-version.workspace = true
-version = "0.0.0"
 repository.workspace = true
 license.workspace = true
 
 [dependencies]
+libc.workspace = true
+unicode-width.workspace = true
 widestring.workspace = true
 
 [lints]

crates/widestring/src/lib.rs

@@ -6,13 +6,34 @@
 
 pub mod word_char;
 
-use std::{iter, slice};
+use std::{
+    ffi::{CStr, CString, OsStr, OsString},
+    iter,
+    os::unix::ffi::{OsStrExt as _, OsStringExt as _},
+    slice,
+};
 pub use widestring::{Utf32Str as wstr, Utf32String as WString, utf32str as L, utfstr::CharsUtf32};
 
 pub mod prelude {
     pub use crate::{IntoCharIter, L, ToWString, WExt, WString, wstr};
 }
 
+// Highest legal ASCII value.
+pub const ASCII_MAX: char = 127 as char;
+
+// Highest legal 16-bit Unicode value.
+pub const UCS2_MAX: char = '\u{FFFF}';
+
+// Highest legal byte value.
+pub const BYTE_MAX: char = 0xFF as char;
+
+// Unicode BOM value.
+pub const UTF8_BOM_WCHAR: char = '\u{FEFF}';
+
+/// The character to use where the text has been truncated.
+pub const ELLIPSIS_CHAR: char = '\u{2026}'; // ('…')
+
+pub const SPECIAL_KEY_ENCODE_BASE: char = '\u{F500}';
 // These are in the Unicode private-use range. We really shouldn't use this
 // range but have little choice in the matter given how our lexer/parser works.
 // We can't use non-characters for these two ranges because there are only 66 of
@@ -25,9 +46,79 @@ pub mod prelude {
 // Note: We don't use the highest 8 bit range (0xF800 - 0xF8FF) because we know
 // of at least one use of a codepoint in that range: the Apple symbol (0xF8FF)
 // on Mac OS X. See http://www.unicode.org/faq/private_use.html.
-pub const ENCODE_DIRECT_BASE: char = '\u{F600}';
+pub const ENCODE_DIRECT_BASE: char = char_offset(SPECIAL_KEY_ENCODE_BASE, 256);
 pub const ENCODE_DIRECT_END: char = char_offset(ENCODE_DIRECT_BASE, 256);
 
+// Use Unicode "non-characters" for internal characters as much as we can. This
+// gives us 32 "characters" for internal use that we can guarantee should not
+// appear in our input stream. See http://www.unicode.org/faq/private_use.html.
+pub const RESERVED_CHAR_BASE: char = '\u{FDD0}';
+pub const RESERVED_CHAR_END: char = '\u{FDF0}';
+// Split the available non-character values into two ranges to ensure there are
+// no conflicts among the places we use these special characters.
+pub const EXPAND_RESERVED_BASE: char = RESERVED_CHAR_BASE;
+pub const EXPAND_RESERVED_END: char = char_offset(EXPAND_RESERVED_BASE, 16);
+pub const WILDCARD_RESERVED_BASE: char = EXPAND_RESERVED_END;
+pub const WILDCARD_RESERVED_END: char = char_offset(WILDCARD_RESERVED_BASE, 16);
+// Make sure the ranges defined above don't exceed the range for non-characters.
+// This is to make sure we didn't do something stupid in subdividing the
+// Unicode range for our needs.
+const _: () = assert!(WILDCARD_RESERVED_END <= RESERVED_CHAR_END);
+
+/// Character representing any character except '/' (slash).
+pub const ANY_CHAR: char = char_offset(WILDCARD_RESERVED_BASE, 0);
+/// Character representing any character string not containing '/' (slash).
+pub const ANY_STRING: char = char_offset(WILDCARD_RESERVED_BASE, 1);
+/// Character representing any character string.
+pub const ANY_STRING_RECURSIVE: char = char_offset(WILDCARD_RESERVED_BASE, 2);
+/// This is a special pseudo-char that is not used other than to mark the
+/// end of the special characters so we can sanity check the enum range.
+#[allow(dead_code)]
+pub const ANY_SENTINEL: char = char_offset(WILDCARD_RESERVED_BASE, 3);
+
+/// Character representing a home directory.
+pub const HOME_DIRECTORY: char = char_offset(EXPAND_RESERVED_BASE, 0);
+/// Character representing process expansion for %self.
+pub const PROCESS_EXPAND_SELF: char = char_offset(EXPAND_RESERVED_BASE, 1);
+/// Character representing variable expansion.
+pub const VARIABLE_EXPAND: char = char_offset(EXPAND_RESERVED_BASE, 2);
+/// Character representing variable expansion into a single element.
+pub const VARIABLE_EXPAND_SINGLE: char = char_offset(EXPAND_RESERVED_BASE, 3);
+/// Character representing the start of a bracket expansion.
+pub const BRACE_BEGIN: char = char_offset(EXPAND_RESERVED_BASE, 4);
+/// Character representing the end of a bracket expansion.
+pub const BRACE_END: char = char_offset(EXPAND_RESERVED_BASE, 5);
+/// Character representing separation between two bracket elements.
+pub const BRACE_SEP: char = char_offset(EXPAND_RESERVED_BASE, 6);
+/// Character that takes the place of any whitespace within non-quoted text in braces
+pub const BRACE_SPACE: char = char_offset(EXPAND_RESERVED_BASE, 7);
+/// Separate subtokens in a token with this character.
+pub const INTERNAL_SEPARATOR: char = char_offset(EXPAND_RESERVED_BASE, 8);
+/// Character representing an empty variable expansion. Only used transitively while expanding
+/// variables.
+pub const VARIABLE_EXPAND_EMPTY: char = char_offset(EXPAND_RESERVED_BASE, 9);
+
+const _: () = assert!(
+    EXPAND_RESERVED_END as u32 > VARIABLE_EXPAND_EMPTY as u32,
+    "Characters used in expansions must stay within private use area"
+);
+
+/// The string represented by PROCESS_EXPAND_SELF
+pub const PROCESS_EXPAND_SELF_STR: &wstr = L!("%self");
+
+/// Return true if the character is in a range reserved for fish's private use.
+///
+/// NOTE: This is used when tokenizing the input. It is also used when reading input, before
+/// tokenization, to replace such chars with REPLACEMENT_WCHAR if they're not part of a quoted
+/// string. We don't want external input to be able to feed reserved characters into our
+/// lexer/parser or code evaluator.
+//
+// TODO: Actually implement the replacement as documented above.
+pub fn fish_reserved_codepoint(c: char) -> bool {
+    (c >= RESERVED_CHAR_BASE && c < RESERVED_CHAR_END)
+        || (c >= SPECIAL_KEY_ENCODE_BASE && c < ENCODE_DIRECT_END)
+}
+
 /// Encode a literal byte in a UTF-32 character. This is required for e.g. the echo builtin, whose
 /// escape sequences can be used to construct raw byte sequences which are then interpreted as e.g.
 /// UTF-8 by the terminal. If we were to interpret each of those bytes as a codepoint and encode it
@@ -40,6 +131,86 @@ pub fn encode_byte_to_char(byte: u8) -> char {
         .expect("private-use codepoint should be valid char")
 }
 
+/// Returns a newly allocated multibyte character string equivalent of the specified wide character
+/// string.
+///
+/// This function decodes illegal character sequences in a reversible way using the private use
+/// area.
+pub fn wcs2bytes(input: impl IntoCharIter) -> Vec<u8> {
+    let mut result = vec![];
+    wcs2bytes_appending(&mut result, input);
+    result
+}
+
+pub fn wcs2osstring(input: &wstr) -> OsString {
+    if input.is_empty() {
+        return OsString::new();
+    }
+
+    let mut result = vec![];
+    wcs2bytes_appending(&mut result, input);
+    OsString::from_vec(result)
+}
+
+/// Same as [`wcs2bytes`]. Meant to be used when we need a zero-terminated string to feed legacy APIs.
+/// Note: if `input` contains any interior NUL bytes, the result will be truncated at the first!
+pub fn wcs2zstring(input: &wstr) -> CString {
+    if input.is_empty() {
+        return CString::default();
+    }
+
+    let mut vec = Vec::with_capacity(input.len() + 1);
+    str2bytes_callback(input, |buff| {
+        vec.extend_from_slice(buff);
+        true
+    });
+    vec.push(b'\0');
+
+    match CString::from_vec_with_nul(vec) {
+        Ok(cstr) => cstr,
+        Err(err) => {
+            // `input` contained a NUL in the middle; we can retrieve `vec`, though
+            let mut vec = err.into_bytes();
+            let pos = vec.iter().position(|c| *c == b'\0').unwrap();
+            vec.truncate(pos + 1);
+            // Safety: We truncated after the first NUL
+            unsafe { CString::from_vec_with_nul_unchecked(vec) }
+        }
+    }
+}
+
+/// Like [`wcs2bytes`], but appends to `output` instead of returning a new string.
+pub fn wcs2bytes_appending(output: &mut Vec<u8>, input: impl IntoCharIter) {
+    str2bytes_callback(input, |buff| {
+        output.extend_from_slice(buff);
+        true
+    });
+}
+
+/// Implementation of wcs2bytes that accepts a callback.
+/// The first argument can be either a `&str` or `&wstr`.
+/// This invokes `func` with byte slices containing the UTF-8 encoding of the characters in the
+/// input, doing one invocation per character.
+/// If `func` returns false, it stops; otherwise it continues.
+/// Return false if the callback returned false, otherwise true.
+pub fn str2bytes_callback(input: impl IntoCharIter, mut func: impl FnMut(&[u8]) -> bool) -> bool {
+    // A `char` represents an Unicode scalar value, which takes up at most 4 bytes when encoded in UTF-8.
+    let mut converted = [0_u8; 4];
+
+    for c in input.chars() {
+        let bytes = if let Some(byte) = decode_byte_from_char(c) {
+            converted[0] = byte;
+            &converted[..=0]
+        } else {
+            c.encode_utf8(&mut converted).as_bytes()
+        };
+        if !func(bytes) {
+            return false;
+        }
+    }
+    true
+}
+
 /// Decode a literal byte from a UTF-32 character.
 pub fn decode_byte_from_char(c: char) -> Option<u8> {
     if c >= ENCODE_DIRECT_BASE && c < ENCODE_DIRECT_END {
@@ -53,6 +224,278 @@ pub fn decode_byte_from_char(c: char) -> Option<u8> {
     }
 }
 
+/// A trait to make it more convenient to pass ascii/Unicode strings to functions that can take
+/// non-Unicode values. The result is nul-terminated and can be passed to OS functions.
+///
+/// This is only implemented for owned types where an owned instance will skip allocations (e.g.
+/// `CString` can return `self`) but not implemented for owned instances where a new allocation is
+/// always required (e.g. implemented for `&wstr` but not `WideString`) because you might as well be
+/// left with the original item if we're going to allocate from scratch in all cases.
+pub trait ToCString {
+    /// Correctly convert to a nul-terminated [`CString`] that can be passed to OS functions.
+    fn to_cstring(self) -> CString;
+}
+
+impl ToCString for CString {
+    fn to_cstring(self) -> CString {
+        self
+    }
+}
+
+impl ToCString for &CStr {
+    fn to_cstring(self) -> CString {
+        self.to_owned()
+    }
+}
+
+/// Safely converts from `&wstr` to a `CString` to a nul-terminated `CString` that can be passed to
+/// OS functions, taking into account non-Unicode values that have been shifted into the private-use
+/// range by using [`wcs2zstring()`].
+impl ToCString for &wstr {
+    /// The wide string may contain non-Unicode bytes mapped to the private-use Unicode range, so we
+    /// have to use [`wcs2zstring()`](self::wcs2zstring) to convert it correctly.
+    fn to_cstring(self) -> CString {
+        self::wcs2zstring(self)
+    }
+}
+
+/// Safely converts from `&WString` to a nul-terminated `CString` that can be passed to OS
+/// functions, taking into account non-Unicode values that have been shifted into the private-use
+/// range by using [`wcs2zstring()`].
+impl ToCString for &WString {
+    fn to_cstring(self) -> CString {
+        self.as_utfstr().to_cstring()
+    }
+}
+
+/// Convert a (probably ascii) string to CString that can be passed to OS functions.
+impl ToCString for Vec<u8> {
+    fn to_cstring(mut self) -> CString {
+        self.push(b'\0');
+        CString::from_vec_with_nul(self).unwrap()
+    }
+}
+
+/// Convert a (probably ascii) string to nul-terminated CString that can be passed to OS functions.
+impl ToCString for &[u8] {
+    fn to_cstring(self) -> CString {
+        CString::new(self).unwrap()
+    }
+}
+
+mod decoder {
+    use crate::{ENCODE_DIRECT_BASE, ENCODE_DIRECT_END, char_offset, wstr};
+    use buffer::Buffer;
+    use std::{char::REPLACEMENT_CHARACTER, ops::Range};
+    use widestring::utfstr::CharsUtf32;
+
+    mod buffer {
+        // The size required for a PUA-encoded character from our special PUA range - 1,
+        // since that is the maximum number characters our look-ahead needs to check.
+        const MAX_SIZE: usize = 2;
+        pub(super) struct Buffer {
+            buffer: [char; MAX_SIZE],
+            length: usize,
+        }
+
+        impl Buffer {
+            pub(super) fn empty() -> Self {
+                Self {
+                    buffer: ['\0'; MAX_SIZE],
+                    length: 0,
+                }
+            }
+
+            pub(super) fn push(&mut self, c: char) {
+                self.buffer[self.length] = c;
+                self.length += 1;
+            }
+
+            pub(super) fn pop(&mut self) -> Option<char> {
+                if self.length == 0 {
+                    return None;
+                }
+                self.length -= 1;
+                Some(self.buffer[self.length])
+            }
+
+            pub(super) fn pop_front(&mut self) -> Option<char> {
+                if self.length == 0 {
+                    return None;
+                }
+                self.buffer.rotate_left(1);
+                self.length -= 1;
+                Some(self.buffer[MAX_SIZE - 1])
+            }
+        }
+    }
+
+    const PUA_ENCODE_RANGE: Range<char> = ENCODE_DIRECT_BASE..ENCODE_DIRECT_END;
+    const ENCODED_PUA_CHAR_FIRST_CHAR: char = char_offset(ENCODE_DIRECT_BASE, 0xef);
+    // The second UTF-8 byte of a character in our special PUA range is in the range
+    // 0x98..0x9c.
+    const ENCODED_PUA_CHAR_SECOND_CHAR_RANGE: Range<char> =
+        char_offset(ENCODE_DIRECT_BASE, 0x98)..char_offset(ENCODE_DIRECT_BASE, 0x9c);
+
+    /// This serves as the data container for building a double-ended iterator which decodes our
+    /// PUA-encoded chars into a char iterator where each encoded non-UTF-8 byte is replaced by the
+    /// replacement character, and each encoded PUA codepoint is turned back into a single char
+    /// whose value is the original PUA codepoint.
+    ///
+    /// The latter part makes the decoding logic somewhat complicated, because encoded PUA chars
+    /// take up 3 chars in our encoding. Therefore, in some cases, we need to take more than 1 char
+    /// from the `encoded_chars` iterator before we know whether to decode 3 chars together into a
+    /// single char, or whether the chars should be replaced by the replacement char individually.
+    /// In cases where we took more than 1 char and then notice that individual replacement is
+    /// warranted, we return a replacement char for the first char we took from the iterator, and
+    /// cache the 1 or 2 other chars we read in `buffer_front` or `buffer_back`, depending on the
+    /// reading direction. Buffers store elements in such an order that getting the next character
+    /// requires `pop` when using the buffer associated with the current reading direction, and
+    /// `pop_front` when using the other buffer. This is done to optimize the common case of
+    /// iterating in a single direction.
+    ///
+    /// The buffers have to be considered before taking more chars from `encoded_chars`.
+    /// If the iterator is only read in one direction, the buffer for the other direction will not
+    /// be used. But because it's possible that the iterator is read from both ends, it can happen
+    /// that when `encoded_chars` runs out, the buffer for the opposite reading direction is
+    /// non-empty. In the [`Self::next`] and [`Self::next_back`] implementations, this logic is
+    /// encapsulated into closures for getting the next char from the appropriate source.
+    /// At most 2 chars will ever be stored in a buffer, so they are implemented using a fixed-size
+    /// array, requiring no heap allocations.
+    ///
+    /// Note that in most cases, we can avoid using the buffers, and simply forward the char
+    /// obtained from `encoded_chars`. Only chars in [`PUA_ENCODE_RANGE`] can possibly encode PUA
+    /// chars, so if we read any other char, we know that it's not part of such an encoding and can
+    /// return it directly. If we read in the forward direction, we can also exploit knowledge about
+    /// the possible values of our PUA encoding. Specifically, the first char in such an encoding
+    /// will always be [`ENCODED_PUA_CHAR_FIRST_CHAR`], and the second char will be in the range
+    /// [`ENCODED_PUA_CHAR_SECOND_CHAR_RANGE`].
+    pub(super) struct Decoder<'a> {
+        encoded_chars: CharsUtf32<'a>,
+        buffer_front: Buffer,
+        buffer_back: Buffer,
+    }
+
+    impl<'a> Decoder<'a> {
+        pub(super) fn new(encoded_str: &'a wstr) -> Self {
+            Self {
+                encoded_chars: encoded_str.chars(),
+                buffer_front: Buffer::empty(),
+                buffer_back: Buffer::empty(),
+            }
+        }
+    }
+
+    fn try_pua_decoding(encoding: &[char; 3]) -> Option<char> {
+        let mut bytes = [0u8; 3];
+        for (index, &c) in encoding.iter().enumerate() {
+            bytes[index] = super::decode_byte_from_char(c)?;
+        }
+        let first_decoded_char =
+            std::str::from_utf8(&bytes).ok()?.chars().next().expect(
+                "Non-empty byte slice which is valid UTF-8 must result in at least one char.",
+            );
+        // For strings whose width we compute, we only expect invalid UTF-8 and codepoints from the
+        // PUA encoding range to be PUA encoded.
+        // If we reach this point, the encoded bytes are valid UTF-8, so the only remaining
+        // expected case are codepoints from the PUA encoding range.
+        // These all take 3 bytes to represent in UTF-8, so if we check that the first parsed
+        // codepoint is in the expected range, we know that exactly 3 bytes were consumed for
+        // parsing this codepoint.
+        assert!(PUA_ENCODE_RANGE.contains(&first_decoded_char));
+        Some(first_decoded_char)
+    }
+
+    fn replace_if_pua_encoded(c: char) -> char {
+        if PUA_ENCODE_RANGE.contains(&c) {
+            REPLACEMENT_CHARACTER
+        } else {
+            c
+        }
+    }
+
+    impl<'a> Iterator for Decoder<'a> {
+        type Item = char;
+
+        fn next(&mut self) -> Option<Self::Item> {
+            let mut get_next_char = || {
+                self.buffer_front
+                    .pop()
+                    .or_else(|| self.encoded_chars.next())
+                    .or_else(|| self.buffer_back.pop_front())
+            };
+            let c_0 = get_next_char()?;
+            if c_0 != ENCODED_PUA_CHAR_FIRST_CHAR {
+                return Some(replace_if_pua_encoded(c_0));
+            }
+            if let Some(c_1) = get_next_char() {
+                if ENCODED_PUA_CHAR_SECOND_CHAR_RANGE.contains(&c_1) {
+                    if let Some(c_2) = get_next_char() {
+                        if let Some(decoded_pua_char) = try_pua_decoding(&[c_0, c_1, c_2]) {
+                            return Some(decoded_pua_char);
+                        }
+                        self.buffer_front.push(c_2);
+                    }
+                }
+                self.buffer_front.push(c_1);
+            }
+            // If decoding 3 consecutive PUA chars into the encoded PUA char fails, `c_0` should be
+            // returned. `c_0` is `ENCODED_PUA_CHAR_FIRST_CHAR` if we reach this point, so return
+            // the `REPLACEMENT_CHARACTER` in these cases.
+            Some(REPLACEMENT_CHARACTER)
+        }
+    }
+
+    impl<'a> DoubleEndedIterator for Decoder<'a> {
+        fn next_back(&mut self) -> Option<Self::Item> {
+            let mut get_next_char = || {
+                self.buffer_back
+                    .pop()
+                    .or_else(|| self.encoded_chars.next_back())
+                    .or_else(|| self.buffer_front.pop_front())
+            };
+            let c_2 = get_next_char()?;
+            if !PUA_ENCODE_RANGE.contains(&c_2) {
+                return Some(c_2);
+            }
+            if let Some(c_1) = get_next_char() {
+                if PUA_ENCODE_RANGE.contains(&c_1) {
+                    if let Some(c_0) = get_next_char() {
+                        if let Some(decoded_pua_char) = try_pua_decoding(&[c_0, c_1, c_2]) {
+                            return Some(decoded_pua_char);
+                        }
+                        self.buffer_back.push(c_0);
+                    }
+                    self.buffer_back.push(c_1);
+                }
+            }
+            // If decoding 3 consecutive PUA chars into the encoded PUA char fails, `c_2` should be
+            // returned. `c_2` is in `PUA_ENCODE_RANGE` if we reach this point, so return the
+            // `REPLACEMENT_CHARACTER` in these cases.
+            Some(REPLACEMENT_CHARACTER)
+        }
+    }
+}
+
+/// Only exists for tests. Exported because the encoding functionality is not available in this
+/// crate. Do not use for non-testing purposes.
+pub fn decode_with_replacement(encoded_str: &wstr) -> impl DoubleEndedIterator<Item = char> {
+    decoder::Decoder::new(encoded_str)
+}
+
+/// Takes a PUA-encoded string, decodes it by restoring encoded PUA codepoints and replacing encoded
+/// non-UTF-8 bytes by the replacement character U+FFFD.
+/// The result is passed to the [`unicode_width`] crate, which will compute its width, which will be
+/// the return value of this function.
+pub fn decoded_width(encoded_str: &wstr) -> usize {
+    // TODO: Avoid constructing String by using `unicode_width::char_iter_width` once that is
+    // available in a released version of the crate (it's already on the crate's master branch).
+    use unicode_width::UnicodeWidthStr as _;
+    decoder::Decoder::new(encoded_str)
+        .collect::<String>()
+        .width()
+}
+
 pub const fn char_offset(base: char, offset: u32) -> char {
     match char::from_u32(base as u32 + offset) {
         Some(c) => c,
@@ -60,6 +503,82 @@ pub const fn char_offset(base: char, offset: u32) -> char {
     }
 }
 
+/// Encodes the bytes in `input` into a [`WString`], encoding non-UTF-8 bytes into private-use-area
+/// code-points. Bytes which would be parsed into our reserved PUA range are encoded individually,
+/// to allow for correct round-tripping.
+pub fn bytes2wcstring(mut input: &[u8]) -> WString {
+    if input.is_empty() {
+        return WString::new();
+    }
+
+    let mut result = WString::with_capacity(input.len());
+
+    fn append_escaped_str(output: &mut WString, input: &str) {
+        for (i, c) in input.char_indices() {
+            if fish_reserved_codepoint(c) {
+                for byte in &input.as_bytes()[i..i + c.len_utf8()] {
+                    output.push(encode_byte_to_char(*byte));
+                }
+            } else {
+                output.push(c);
+            }
+        }
+    }
+
+    while !input.is_empty() {
+        match std::str::from_utf8(input) {
+            Ok(parsed_str) => {
+                append_escaped_str(&mut result, parsed_str);
+                // The entire remaining input could be parsed, so we are done.
+                break;
+            }
+            Err(e) => {
+                let (valid, after_valid) = input.split_at(e.valid_up_to());
+                // SAFETY: The previous `str::from_utf8` call established that the prefix `valid`
+                // is valid UTF-8. This prefix may be empty.
+                let parsed_str = unsafe { std::str::from_utf8_unchecked(valid) };
+                append_escaped_str(&mut result, parsed_str);
+                // The length of the prefix of `after_valid` which is invalid UTF-8.
+                // The remaining bytes of `input` (if any) will be parsed in subsequent iterations
+                // of the loop, starting from the first byte that starts a valid UTF-8-encoded codepoint.
+                // `error_len` can return `None`, if it sees a byte sequence that could be the
+                // prefix of a valid code-point encoding at the end of the byte slice.
+                // This is useful when the input is chunked, but we don't do that, so in this case
+                // we use our custom encoding for all remaining bytes (at most 3).
+                let error_len = e.error_len().unwrap_or(after_valid.len());
+                for byte in &after_valid[..error_len] {
+                    result.push(encode_byte_to_char(*byte));
+                }
+                input = &after_valid[error_len..];
+            }
+        }
+    }
+    result
+}
+
+/// Use this rather than [`WString::from_str`] when the input could contain PUA bytes we use to
+/// encode non-UTF-8 bytes. Otherwise, when decoding the resulting [`WString`], the PUA bytes in
+/// the input would be converted to non-UTF-8 bytes.
+pub fn str2wcstring<S: AsRef<str>>(input: S) -> WString {
+    bytes2wcstring(input.as_ref().as_bytes())
+}
+
+pub fn cstr2wcstring<C: AsRef<CStr>>(input: C) -> WString {
+    bytes2wcstring(input.as_ref().to_bytes())
+}
+
+pub fn osstr2wcstring<O: AsRef<OsStr>>(input: O) -> WString {
+    bytes2wcstring(input.as_ref().as_bytes())
+}
+
+/// # SAFETY
+///
+/// `input` must point to a valid NUL-terminated string.
+pub unsafe fn charptr2wcstring(input: *const libc::c_char) -> WString {
+    let input: &[u8] = unsafe { CStr::from_ptr(input).to_bytes() };
+    bytes2wcstring(input)
+}
+
 /// Finds `needle` in a `haystack` and returns the index of the first matching element, if any.
 ///
 /// # Examples

crates/xtask/Cargo.toml

@@ -1,11 +1,19 @@
 [package]
 name = "xtask"
 version = "0.0.0"
-rust-version.workspace = true
 edition.workspace = true
+rust-version.workspace = true
 repository.workspace = true
 
 [dependencies]
+anstyle.workspace = true
+anyhow.workspace = true
 clap.workspace = true
+clap_complete.workspace = true
 fish-build-helper.workspace = true
+fish-common.workspace = true
 fish-tempfile.workspace = true
+fish-widestring.workspace = true
+ignore.workspace = true
+pcre2.workspace = true
+walkdir.workspace = true

crates/xtask/src/format.rs

@@ -0,0 +1,182 @@
+use anstyle::{AnsiColor, Style};
+use anyhow::{Context, Result, bail};
+use clap::Args;
+use std::{
+    io::{ErrorKind, Write},
+    path::PathBuf,
+    process::{Command, Stdio},
+};
+
+use crate::files_with_extension;
+
+const GREEN: Style = AnsiColor::Green.on_default();
+const YELLOW: Style = AnsiColor::Yellow.on_default();
+
+#[derive(Args)]
+pub struct FormatArgs {
+    /// Consider all eligible files.
+    #[arg(long)]
+    all: bool,
+    /// Report files which are not formatted as expected, without modifying any files.
+    #[arg(long)]
+    check: bool,
+    /// Format files even if uncommitted changes are detected.
+    #[arg(long)]
+    force: bool,
+    paths: Vec<PathBuf>,
+}
+
+pub fn format(args: FormatArgs) -> Result<()> {
+    if !args.all && args.paths.is_empty() {
+        println!(
+            "{YELLOW}warning: No paths specified. Nothing to do. Use the \"--all\" flag to consider all eligible files.{YELLOW:#}"
+        );
+        return Ok(());
+    }
+    if !args.force && !args.check {
+        match Command::new("git")
+            .args(["status", "--porcelain", "--short", "--untracked-files=all"])
+            .output()
+        {
+            Ok(output) => {
+                if !output.stdout.is_empty() {
+                    std::io::stdout()
+                        .write_all(&output.stdout)
+                        .context("Could not write to stdout.")?;
+                    print!(
+                        "You have uncommitted changes (listed above). Are you sure you want to format? (y/N): "
+                    );
+                    std::io::stdout()
+                        .flush()
+                        .context("Could not flush stdout.")?;
+                    let mut response = String::new();
+                    std::io::stdin()
+                        .read_line(&mut response)
+                        .context("Could not read from stdin.")?;
+                    if response.trim_end() != "y" {
+                        println!("Exiting without formatting.");
+                        return Ok(());
+                    }
+                }
+            }
+            Err(e) => {
+                if e.kind() == ErrorKind::NotFound {
+                    println!(
+                        "{YELLOW}warning: Did not find git, will proceed without checking for unstaged changes.{YELLOW:#}"
+                    )
+                } else {
+                    bail!("Failed to run git status:\n{e}");
+                }
+            }
+        }
+    }
+    format_fish(&args)?;
+    format_python(&args)?;
+    format_rust(&args)?;
+    Ok(())
+}
+
+fn run_formatter(formatter: &mut Command, name: &str) -> Result<()> {
+    println!("=== Running {GREEN}{name}{GREEN:#}");
+    match formatter.status() {
+        Ok(exit_status) => {
+            if exit_status.success() {
+                Ok(())
+            } else {
+                bail!("{name:?}: Files are not formatted correctly.");
+            }
+        }
+        Err(e) => {
+            if e.kind() == std::io::ErrorKind::NotFound {
+                eprintln!(
+                    "{YELLOW}Formatter not found: {name:?}. Skipping associated files.{YELLOW:#}"
+                );
+                Ok(())
+            } else {
+                Err(e).with_context(|| format!("Error occurred while running {name:?}"))
+            }
+        }
+    }
+}
+
+fn format_fish(args: &FormatArgs) -> Result<()> {
+    let mut fish_paths = files_with_extension(&args.paths, "fish")?;
+    if args.all {
+        let workspace_root = fish_build_helper::workspace_root();
+        let fish_formatting_dirs = ["benchmarks", "build_tools", "etc", "share"];
+        fish_paths.extend(files_with_extension(
+            fish_formatting_dirs
+                .iter()
+                .map(|dir_name| workspace_root.join(dir_name)),
+            "fish",
+        )?);
+    };
+    if fish_paths.is_empty() {
+        return Ok(());
+    }
+    // TODO: make `fish_indent` available as a Rust library function, to avoid needing a
+    // `fish_indent` binary in `$PATH`.
+    let mut formatter = Command::new("fish_indent");
+    if args.check {
+        formatter.arg("--check");
+    } else {
+        formatter.arg("-w");
+    }
+    formatter.arg("--");
+    formatter.args(fish_paths);
+    run_formatter(&mut formatter, "fish_indent")
+}
+
+fn format_python(args: &FormatArgs) -> Result<()> {
+    let mut formatter = Command::new("ruff");
+    formatter.arg("format");
+    if args.check {
+        formatter.arg("--check");
+    }
+    let mut python_files = files_with_extension(&args.paths, "py")?;
+
+    if args.all {
+        python_files.push(fish_build_helper::workspace_root().to_owned());
+    };
+    if python_files.is_empty() {
+        return Ok(());
+    }
+    formatter.args(python_files);
+    run_formatter(&mut formatter, "ruff format")
+}
+
+fn format_rust(args: &FormatArgs) -> Result<()> {
+    let rustfmt_status = Command::new("cargo")
+        .arg("fmt")
+        .arg("--version")
+        .stdout(Stdio::null())
+        .status()
+        .context("Failed to run cargo")?;
+    if !rustfmt_status.success() {
+        eprintln!(
+            "{YELLOW}Please install \"rustfmt\" to format Rust, e.g. via:\n\
+            rustup component add rustfmt{YELLOW:#}"
+        );
+        return Ok(());
+    }
+    if args.all {
+        let mut formatter = Command::new("cargo");
+        formatter.arg("fmt");
+        formatter.arg("--all");
+        if args.check {
+            formatter.arg("--check");
+        }
+        run_formatter(&mut formatter, "cargo fmt")?;
+    }
+    let rust_files = files_with_extension(&args.paths, "rs")?;
+    if rust_files.is_empty() {
+        return Ok(());
+    }
+    let mut formatter = Command::new("rustfmt");
+    if args.check {
+        formatter.arg("--check");
+        formatter.arg("--files-with-diff");
+    }
+    formatter.args(rust_files);
+    run_formatter(&mut formatter, "rustfmt")
+}

crates/xtask/src/gettext.rs

@@ -0,0 +1,552 @@
+use crate::{CARGO, CommandExt, files_with_extension};
+use anyhow::{Context as _, Result, bail};
+use clap::{Args, Subcommand};
+use fish_build_helper::po_dir;
+use pcre2::bytes::Regex;
+use std::{
+    fs::OpenOptions,
+    io::{Write as _, stdout},
+    path::{Path, PathBuf},
+    process::Command,
+    thread::spawn,
+};
+
+#[derive(Args)]
+pub struct GettextArgs {
+    /// Path to the directory into which the messages from the Rust sources have been extracted.
+    /// If this is not specified, fish will be compiled with the `gettext-extract` feature to
+    /// obtain the messages.
+    #[arg(long)]
+    rust_extraction_dir: Option<PathBuf>,
+
+    #[command(subcommand)]
+    task: Task,
+}
+
+#[derive(Subcommand)]
+enum Task {
+    /// Check whether the PO files are up to date.
+    /// Prints a diff and exits non-zero if they are outdated.
+    /// Considers all our PO files by default, also allows explicitly specifying which files to
+    /// consider.
+    Check { paths: Vec<PathBuf> },
+    /// Add a PO file for a new language.
+    New {
+        /// An ISO 639-1 language identifier (ISO 639-2 if the former does not exits),
+        /// optionally followed by and underscore and an ISO 3166-1 country code to specify the variant,
+        /// e.g. `de` or `pt_BR`.
+        language: String,
+    },
+    /// Update PO files.
+    /// This will delete entries for msgids which are no longer used in the sources and introduce
+    /// new, untranslated entries for messages which do not have an entry yet.
+    /// This tool should run every time a change to the messages localized via gettext occurs,
+    /// including fish script files, where many strings are implicitly localized.
+    /// Considers all our PO files by default, also allows explicitly specifying which files to
+    /// consider.
+    Update { paths: Vec<PathBuf> },
+}
+
+fn get_po_paths<P: AsRef<Path>>(specified_paths: &[P]) -> Result<Vec<PathBuf>> {
+    let extension = "po";
+    if specified_paths.is_empty() {
+        files_with_extension([po_dir()], extension)
+    } else {
+        files_with_extension(specified_paths, extension)
+    }
+}
+
+fn update_po_file<P: AsRef<Path>, Q: AsRef<Path>>(file_to_update: P, template: Q) -> Result<()> {
+    Command::new("msgmerge")
+        .args([
+            "--no-wrap",
+            "--update",
+            "--no-fuzzy-matching",
+            "--backup=none",
+            "--quiet",
+        ])
+        .arg(file_to_update.as_ref())
+        .arg(template.as_ref())
+        .run()?;
+    let msgattrib_output_file = fish_tempfile::new_file().context("Failed to create temp file")?;
+    Command::new("msgattrib")
+        .args(["--no-wrap", "--no-obsolete"])
+        .arg("-o")
+        .arg(msgattrib_output_file.path())
+        .arg(file_to_update.as_ref())
+        .run()?;
+    crate::copy_file(msgattrib_output_file.path(), file_to_update.as_ref())?;
+    Ok(())
+}
+
+pub fn gettext(args: GettextArgs) -> Result<()> {
+    let template = match args.rust_extraction_dir {
+        Some(dir) => template::Template::new(dir)?,
+        None => {
+            let temp_dir = fish_tempfile::new_dir().context("Failed to create temp file")?;
+            Command::new(CARGO)
+                .args([
+                    "check",
+                    "--workspace",
+                    "--all-targets",
+                    "--features=gettext-extract",
+                ])
+                .env("FISH_GETTEXT_EXTRACTION_DIR", temp_dir.path())
+                .run()?;
+            template::Template::new(temp_dir.path())?
+        }
+    };
+    let mut template_file = fish_tempfile::new_file().context("Failed to create temp file")?;
+    template_file
+        .get_mut()
+        .write_all(template.serialize())
+        .with_context(|| format!("Failed to write to temp file {:?}", template_file.path()))?;
+    template_file
+        .get_mut()
+        .flush()
+        .with_context(|| format!("Failed to flush temporary file {:?}", template_file.path()))?;
+
+    match args.task {
+        Task::Check { paths } => {
+            let mut thread_handles = vec![];
+            for path in get_po_paths(&paths)? {
+                let template_path_buf = template_file.path().to_owned();
+                let handle = spawn(move || -> Result<Option<Vec<u8>>> {
+                    let tmp_copy =
+                        fish_tempfile::new_file().context("Failed to create temp file")?;
+                    crate::copy_file(&path, tmp_copy.path())?;
+                    update_po_file(tmp_copy.path(), template_path_buf)?;
+                    let diff_output = Command::new("diff")
+                        .arg("-u")
+                        .arg(&path)
+                        .arg(tmp_copy.path())
+                        .output()
+                        .context("Failed to run diff")?;
+                    if diff_output.status.success() {
+                        Ok(None)
+                    } else {
+                        Ok(Some(diff_output.stdout))
+                    }
+                });
+                thread_handles.push(handle);
+            }
+            let mut found_diff = false;
+            let mut error = None;
+            for handle in thread_handles {
+                // SAFETY: `handle.join()` only returns `Err` if the thread panicked.
+                // Our threads should not panic, and if they do, it's OK to deal with the unexpected
+                // behavior by panicking here as well.
+                match handle.join().unwrap() {
+                    Ok(None) => {}
+                    Ok(Some(diff)) => {
+                        found_diff = true;
+                        stdout()
+                            .write_all(&diff)
+                            .context("Could not write to stdout")?;
+                    }
+                    Err(e) => {
+                        error = Some(e);
+                    }
+                }
+            }
+            if let Some(e) = error {
+                return Err(e);
+            }
+            if found_diff {
+                bail!(
+                    "Not all PO files are up to date.\n\
+                     Run `cargo xtask gettext update` to bring them up to date automatically.\
+                    "
+                );
+            }
+            Ok(())
+        }
+        Task::New { language } => {
+            let language_regex = Regex::new("^[a-z]{2,3}(_[A-Z]{2})?$").unwrap();
+            if !language_regex.is_match(language.as_bytes()).unwrap() {
+                bail!(
+                    "The language name '{language}' does not match the expected format.\n\
+                    It needs to be a two-letter ISO 639-1 language code, \
+                    or a three-letter ISO 639-2 language code \
+                    if no ISO 639-1 code exists for the language.\n\
+                    Optionally, the language code can be followed be an underscore \
+                    followed by an ISO 3166-1 country code to indicate a regional variant.\n\
+                    Check the existing file names in {:?} for examples.",
+                    po_dir()
+                );
+            }
+            // TODO (MSRV>=1.91): use with_added_extension instead of with_extension
+            let po_path = po_dir().join(&language).with_extension("po");
+            let mut new_po_file = OpenOptions::new()
+                .create_new(true)
+                .write(true)
+                .open(&po_path)
+                .with_context(|| format!("Failed to create file at {po_path:?}"))?;
+            let mut header = String::new();
+            let line_prefix = "# fish-note-sections: ";
+            let lines = [
+                "Translations are divided into sections, each starting with a fish-section-* pseudo-message.",
+                "The first few sections are more important.",
+                "Ignore the tier3 sections unless you have a lot of time.",
+            ];
+            for line in lines {
+                use std::fmt::Write as _;
+                let _ = writeln!(header, "{line_prefix}{line}");
+            }
+            new_po_file
+                .write_all(header.as_bytes())
+                .with_context(|| format!("Failed to write to {po_path:?}"))?;
+            new_po_file
+                .write_all(template.serialize())
+                .with_context(|| format!("Failed to write to {po_path:?}"))?;
+            Ok(())
+        }
+        Task::Update { paths } => {
+            let mut thread_handles = vec![];
+            for path in get_po_paths(&paths)? {
+                let template_path_buf = template_file.path().to_owned();
+                let handle =
+                    spawn(move || -> Result<()> { update_po_file(path, template_path_buf) });
+                thread_handles.push(handle);
+            }
+            let mut error = None;
+            for handle in thread_handles {
+                // SAFETY: `handle.join()` only returns `Err` if the thread panicked.
+                // Our threads should not panic, and if they do, it's OK to deal with the unexpected
+                // behavior by panicking here as well.
+                if let Err(e) = handle.join().unwrap() {
+                    error = Some(e);
+                }
+            }
+            if let Some(e) = error {
+                return Err(e);
+            }
+            Ok(())
+        }
+    }
+}
+
+mod template {
+    use crate::{CommandExt as _, files_with_extension};
+    use anyhow::{Context as _, Result, bail};
+    use fish_build_helper::workspace_root;
+    use fish_common::{UnescapeFlags, unescape_string};
+    use fish_widestring::{str2wcstring, wcs2bytes};
+    use pcre2::bytes::Regex;
+    use std::{
+        collections::{HashMap, HashSet},
+        fmt::Display,
+        fs::OpenOptions,
+        io::Read as _,
+        path::Path,
+        process::Command,
+        sync::LazyLock,
+    };
+
+    // Gettext tools require this header to know which encoding is used.
+    const MINIMAL_HEADER: &str = r#"msgid ""
+msgstr "Content-Type: text/plain; charset=UTF-8\n"
+
+"#;
+
+    #[derive(PartialEq, Eq, PartialOrd, Ord, Clone, Copy, Hash)]
+    enum LocalizationTier {
+        Tier1,
+        Tier2,
+        Tier3,
+    }
+
+    impl TryFrom<&str> for LocalizationTier {
+        type Error = ();
+
+        fn try_from(value: &str) -> Result<Self, Self::Error> {
+            match value {
+                "tier1" => Ok(Self::Tier1),
+                "tier2" => Ok(Self::Tier2),
+                "tier3" => Ok(Self::Tier3),
+                _ => Err(()),
+            }
+        }
+    }
+
+    impl Display for LocalizationTier {
+        fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+            f.write_str(match self {
+                Self::Tier1 => "tier1",
+                Self::Tier2 => "tier2",
+                Self::Tier3 => "tier3",
+            })
+        }
+    }
+
+    #[derive(Default)]
+    struct FishScriptMessages {
+        explicit: HashSet<String>,
+        implicit: HashSet<String>,
+    }
+
+    pub struct Template {
+        content: Vec<u8>,
+    }
+
+    impl Template {
+        pub fn serialize(&self) -> &[u8] {
+            &self.content
+        }
+
+        /// Create a gettext template.
+        /// `rust_extraction_dir` must be the path to a directory which contains the messages
+        /// extracted from the Rust sources.
+        pub fn new<P: AsRef<Path>>(rust_extraction_dir: P) -> Result<Self> {
+            let mut template = Self {
+                content: Vec::from(MINIMAL_HEADER.as_bytes()),
+            };
+            template.add_rust_messages(rust_extraction_dir)?;
+            template.add_fish_script_messages()?;
+            // TODO: keep internal set of msgids to avoid having to run msguniq. requires parsing
+            // gettext-extraction output
+            let msguniq_output = Command::new("msguniq")
+                .args(["--no-wrap"])
+                .run_with_stdio(template.content)?;
+            Ok(Template {
+                content: msguniq_output,
+            })
+        }
+
+        /// Expects `extraction_dir` to contain only files whose content are single PO entries which can be
+        /// concatenated into a valid PO file.
+        /// If this is the case, the messages are de-duplicated and sorted by `msguniq`.
+        /// The result is appended to `template`, with a leading section marker.
+        /// On failure, the process aborts.
+        fn add_rust_messages<P: AsRef<Path>>(&mut self, extraction_dir: P) -> Result<()> {
+            let extraction_dir = extraction_dir.as_ref();
+            let mut concatenated_content = Vec::from(MINIMAL_HEADER.as_bytes());
+
+            // Concatenate the content of all files in `extraction_dir` into `concatenated_content`.
+            for entry_result in extraction_dir
+                .read_dir()
+                .with_context(|| format!("Failed to read directory {extraction_dir:?}"))?
+            {
+                let entry = entry_result
+                    .with_context(|| format!("Failed to get entry in {extraction_dir:?}"))?;
+                let entry_path = entry.path();
+                if !entry
+                    .file_type()
+                    .with_context(|| format!("Failed to get file type of {entry_path:?}"))?
+                    .is_file()
+                {
+                    bail!("Entry in {extraction_dir:?} is not a regular file");
+                }
+                let mut file = OpenOptions::new()
+                    .read(true)
+                    .open(&entry_path)
+                    .with_context(|| format!("Failed to open file {entry_path:?}"))?;
+                file.read_to_end(&mut concatenated_content)
+                    .with_context(|| format!("Failed to read file {entry_path:?}"))?;
+            }
+
+            // Get rid of duplicates and sort.
+            let msguniq_output = Command::new("msguniq")
+                .args(["--no-wrap", "--sort-output"])
+                .env("LC_ALL", "C.UTF-8")
+                .run_with_stdio(concatenated_content)?;
+            // The Header entry needs to be removed again,
+            // because it is added outside of this function.
+            let expected_prefix = MINIMAL_HEADER.as_bytes();
+            let actual_prefix = &msguniq_output[..expected_prefix.len()];
+            if expected_prefix != actual_prefix {
+                bail!(
+                    "Prefix of msguniq output does not match expected header.\nExpected bytes:\n{expected_prefix:02x?}\nActual bytes:\n{actual_prefix:02x?}"
+                );
+            }
+            self.mark_section("tier1-from-rust");
+            self.content
+                .extend_from_slice(&msguniq_output[expected_prefix.len()..]);
+            self.content.push(b'\n');
+            Ok(())
+        }
+
+        fn mark_section(&mut self, section_name: &str) {
+            self.content
+                .extend_from_slice("msgid \"fish-section-".as_bytes());
+            self.content.extend_from_slice(section_name.as_bytes());
+            self.content
+                .extend_from_slice("\"\nmsgstr \"\"\n\n".as_bytes());
+        }
+
+        fn append_messages(&mut self, msgids: &HashSet<String>) -> Result<()> {
+            let mut unescaped_msgids = HashSet::new();
+            for msgid in msgids {
+                let unescaped_wstring = unescape_string(
+                    &str2wcstring(msgid),
+                    fish_common::UnescapeStringStyle::Script(UnescapeFlags::default()),
+                )
+                .with_context(|| format!("Failed to unescape the following string:\n{msgid}"))?;
+                unescaped_msgids.insert(
+                    String::from_utf8(wcs2bytes(&unescaped_wstring))
+                        .context("Parsed msgid is not valid UTF-8")?,
+                );
+            }
+            let mut unescaped_msgids = Vec::from_iter(unescaped_msgids);
+            unescaped_msgids.sort();
+            for msgid in &unescaped_msgids {
+                self.content
+                    .extend_from_slice(format_msgid_for_po(msgid).as_bytes());
+            }
+            Ok(())
+        }
+
+        fn add_script_tier(
+            &mut self,
+            tier: LocalizationTier,
+            messages: FishScriptMessages,
+        ) -> Result<()> {
+            if !messages.explicit.is_empty() {
+                self.mark_section(&format!("{tier}-from-script-explicitly-added"));
+                self.append_messages(&messages.explicit)?;
+            }
+            if !messages.implicit.is_empty() {
+                self.mark_section(&format!("{tier}-from-script-implicitly-added"));
+                self.append_messages(&messages.implicit)?;
+            }
+            Ok(())
+        }
+
+        fn add_fish_script_messages(&mut self) -> Result<()> {
+            let share_dir = workspace_root().join("share");
+            let relevant_file_paths = files_with_extension(
+                [
+                    share_dir.join("config.fish"),
+                    share_dir.join("completions"),
+                    share_dir.join("functions"),
+                ],
+                "fish",
+            )?;
+            let mut extracted_messages = HashMap::new();
+            for path in relevant_file_paths {
+                extract_messages_from_fish_script(path, &mut extracted_messages)?;
+            }
+            let mut messages_sorted_by_tier: Vec<_> = extracted_messages.into_iter().collect();
+            messages_sorted_by_tier.sort_by_key(|(tier, _)| *tier);
+            for (tier, messages) in messages_sorted_by_tier {
+                self.add_script_tier(tier, messages)?;
+            }
+            Ok(())
+        }
+    }
+
+    fn find_localization_tier<P: AsRef<Path>>(
+        input: &str,
+        path: P,
+    ) -> Result<Option<LocalizationTier>> {
+        static L10N_ANNOTATION: LazyLock<Regex> = LazyLock::new(|| {
+            Regex::new(r"(?:^|\n)# localization: (?<annotation_value>.*)\n").unwrap()
+        });
+        if let Some(annotation) = L10N_ANNOTATION.captures(input.as_bytes()).unwrap() {
+            // SAFETY: `tier` is the name of a capture group in the regex whose captures we are looking
+            // at. The capture is done on the bytes of an UTF-8 encoded string, so the result will also
+            // be UTF-8 encoded, and the sub-slice we are looking at will start and end at codepoint
+            // boundaries.
+            let annotation_value =
+                std::str::from_utf8(annotation.name("annotation_value").unwrap().as_bytes())
+                    .unwrap();
+            if let Ok(tier) = LocalizationTier::try_from(annotation_value) {
+                return Ok(Some(tier));
+            }
+            if annotation_value.starts_with("skip") {
+                return Ok(None);
+            }
+            bail!(
+                "Unexpected localization annotation in file {:?}: {annotation_value}",
+                path.as_ref()
+            );
+        }
+        let dirname = path
+            .as_ref()
+            .parent()
+            .with_context(|| {
+                format!(
+                    "Tried to get the parent of a path which does not have a parent: {:?}",
+                    path.as_ref()
+                )
+            })?
+            .file_name()
+            .with_context(|| {
+                format!(
+                    "The parent of {:?} does not have a filename component",
+                    path.as_ref()
+                )
+            })?;
+        let command_name = path
+            .as_ref()
+            .file_stem()
+            .with_context(|| format!("The path {:?} does not have a file stem", path.as_ref()))?;
+        if dirname == "functions"
+            && command_name
+                .to_str()
+                .is_some_and(|name| name.starts_with("fish_"))
+        {
+            return Ok(Some(LocalizationTier::Tier1));
+        }
+        if dirname != "completions" {
+            bail!(
+                "Missing localization tier for function file {:?}",
+                path.as_ref()
+            );
+        }
+        // TODO (MSRV>=1.91): use with_added_extension instead of with_extension
+        let doc_path = workspace_root()
+            .join("doc_src")
+            .join("cmds")
+            .join(command_name)
+            .with_extension("rst");
+        let doc_path_exists = std::fs::exists(&doc_path)
+            .with_context(|| format!("Failed to check whether a file exists at {doc_path:?}"))?;
+        Ok(Some(if doc_path_exists {
+            LocalizationTier::Tier1
+        } else {
+            LocalizationTier::Tier3
+        }))
+    }
+
+    fn extract_messages_from_fish_script<P: AsRef<Path>>(
+        path: P,
+        extracted_messages: &mut HashMap<LocalizationTier, FishScriptMessages>,
+    ) -> Result<()> {
+        let path = path.as_ref();
+        let file_content = std::fs::read_to_string(path)
+            .with_context(|| format!("Failed to read from {path:?}"))?;
+        let Some(tier) = find_localization_tier(&file_content, path)? else {
+            return Ok(());
+        };
+        // TODO: use proper parser instead of regex
+        static EXPLICIT_MESSAGE: LazyLock<Regex> =
+            LazyLock::new(|| Regex::new(r#"\( *_ (?<message>(['"]).+?(?<!\\)\2) *\)"#).unwrap());
+        static IMPLICIT_MESSAGE: LazyLock<Regex> = LazyLock::new(|| {
+            Regex::new(r#"(?:^|\n)(?:\s|and |or )*(?:complete|function).*? (?:-d|--description) (?<message>(['"]).+?(?<!\\)\2)"#).unwrap()
+        });
+        let messages_at_tier = extracted_messages.entry(tier).or_default();
+        for message in EXPLICIT_MESSAGE.captures_iter(file_content.as_bytes()) {
+            let message =
+                std::str::from_utf8(message.unwrap().name("message").unwrap().as_bytes()).unwrap();
+            messages_at_tier.explicit.insert(message.to_owned());
+        }
+        for message in IMPLICIT_MESSAGE.captures_iter(file_content.as_bytes()) {
+            let message =
+                std::str::from_utf8(message.unwrap().name("message").unwrap().as_bytes()).unwrap();
+            messages_at_tier.implicit.insert(message.to_owned());
+        }
+        Ok(())
+    }
+
+    fn format_msgid_for_po(msgid: &str) -> String {
+        let escaped_msgid = msgid.replace("\\", "\\\\").replace("\"", "\\\"");
+        format!(
+            "\
+            msgid \"{escaped_msgid}\"\n\
+            msgstr \"\"\n\
+            \n\
+            "
+        )
+    }
+}

crates/xtask/src/lib.rs

@@ -1,28 +1,101 @@
-use std::{ffi::OsStr, process::Command};
+use std::{
+    ffi::OsStr,
+    io::Write,
+    path::{Path, PathBuf},
+    process::{Command, Stdio},
+};
+
+use anyhow::{Context, Result, bail};
+use walkdir::WalkDir;
+
+pub mod format;
+pub mod gettext;
+pub mod shellcheck;
 
 pub trait CommandExt {
-    fn run_or_panic(&mut self);
+    fn run(&mut self) -> Result<()>;
+    fn run_with_stdio(&mut self, stdin: Vec<u8>) -> Result<Vec<u8>>;
 }
 
 impl CommandExt for Command {
-    fn run_or_panic(&mut self) {
-        match self.status() {
-            Ok(exit_status) => {
-                if !exit_status.success() {
-                    panic!("Command did not run successfully: {:?}", self.get_program())
-                }
-            }
-            Err(err) => {
-                panic!("Failed to run command: {err}");
-            }
+    fn run(&mut self) -> Result<()> {
+        if !self
+            .status()
+            .with_context(|| format!("Failed to run {:?}", self.get_program()))?
+            .success()
+        {
+            bail!("Command did not run successfully: {:?}", self.get_program())
         }
+        Ok(())
+    }
+
+    fn run_with_stdio(&mut self, stdin: Vec<u8>) -> Result<Vec<u8>> {
+        let command_name = self.get_program().to_owned();
+        let mut child = self
+            .stdin(Stdio::piped())
+            .stdout(Stdio::piped())
+            .spawn()
+            .with_context(|| format!("Failed to run {command_name:?}"))?;
+        child
+            .stdin
+            .take()
+            .unwrap()
+            .write_all(&stdin)
+            .with_context(|| format!("Failed to write to stdin of {command_name:?}"))?;
+        let command_output = child
+            .wait_with_output()
+            .with_context(|| format!("Failed to read stdout of {command_name:?}"))?;
+        if !command_output.status.success() {
+            bail!("{command_name:?} failed");
+        }
+        Ok(command_output.stdout)
     }
 }
 
-pub fn cargo<I, S>(cargo_args: I)
+pub const CARGO: &str = env!("CARGO");
+
+pub fn cargo<I, S>(cargo_args: I) -> Result<()>
 where
     I: IntoIterator<Item = S>,
     S: AsRef<OsStr>,
 {
-    Command::new(env!("CARGO")).args(cargo_args).run_or_panic();
+    Command::new(CARGO).args(cargo_args).run()
+}
+
+fn get_matching_files<P: AsRef<Path>, I: IntoIterator<Item = P>, M: Fn(&Path) -> bool>(
+    all_paths: I,
+    matcher: M,
+) -> Result<Vec<PathBuf>> {
+    let mut matching_files = vec![];
+    for path in all_paths {
+        for dir_entry in WalkDir::new(path.as_ref()) {
+            let dir_entry = dir_entry
+                .with_context(|| format!("Failed to check paths at {:?}", path.as_ref()))?;
+            let path = dir_entry.path();
+            if dir_entry.file_type().is_file() && matcher(path) {
+                matching_files.push(path.to_owned());
+            }
+        }
+    }
+    Ok(matching_files)
+}
+
+fn files_with_extension<P: AsRef<Path>, I: IntoIterator<Item = P>>(
+    all_paths: I,
+    extension: &str,
+) -> Result<Vec<PathBuf>> {
+    let matcher = |p: &Path| p.extension().is_some_and(|e| e == extension);
+    get_matching_files(all_paths, matcher)
+}
+
+fn copy_file<P: AsRef<Path>, Q: AsRef<Path>>(from: P, to: Q) -> Result<()> {
+    std::fs::copy(&from, &to)
+        .with_context(|| {
+            format!(
+                "Failed to copy from {:?} to {:?}",
+                from.as_ref(),
+                to.as_ref()
+            )
+        })
+        .map(|_| ())
 }

crates/xtask/src/main.rs

@@ -1,7 +1,9 @@
-use clap::{Parser, Subcommand};
+use anyhow::{Context, Result};
+use clap::{CommandFactory, Parser, Subcommand};
+use clap_complete::CompleteEnv;
 use fish_build_helper::as_os_strs;
 use std::{path::PathBuf, process::Command};
-use xtask::{CommandExt as _, cargo};
+use xtask::{CommandExt, cargo, format::FormatArgs, gettext::GettextArgs, shellcheck::shellcheck};
 
 #[derive(Parser)]
 #[command(
@@ -18,6 +20,10 @@ struct Cli {
 enum Task {
     /// Run various checks on the repo.
     Check,
+    /// Format files or check if they are correctly formatted.
+    Format(FormatArgs),
+    /// Work on the gettext PO files.
+    Gettext(GettextArgs),
     /// Build HTML docs
     HtmlDocs {
         /// Path to a fish_indent executable. If none is specified, fish_indent will be built.
@@ -26,52 +32,85 @@ enum Task {
     },
     /// Build man pages
     ManPages,
+    /// Run ShellCheck on non-fish shell scripts
+    #[command(name = "shellcheck")]
+    ShellCheck,
 }
 
-fn main() {
+/// Only used to enable completion generation.
+/// [`clap_complete`] is not built to account for the situation we have here, where the CLI does not
+/// correspond to a top-level shell command.
+/// We work around this here by pretending that we are building a CLI for the `cargo` command, which
+/// only has the single subcommand `xtask`.
+/// These completions can then be combined with the regular cargo completions.
+#[derive(Parser)]
+#[command(name = "cargo")]
+struct FakeCargoWrapperForCompletion {
+    #[command(subcommand)]
+    xtask: FakeCliForCompletion,
+}
+
+#[derive(Subcommand)]
+enum FakeCliForCompletion {
+    /// Run fish's xtasks
+    #[command(subcommand)]
+    Xtask(Task),
+}
+
+fn main() -> Result<()> {
+    CompleteEnv::with_factory(FakeCargoWrapperForCompletion::command).complete();
+
     let cli = Cli::parse();
     match cli.task {
         Task::Check => run_checks(),
+        Task::Format(format_args) => xtask::format::format(format_args),
+        Task::Gettext(gettext_args) => xtask::gettext::gettext(gettext_args),
         Task::HtmlDocs { fish_indent } => build_html_docs(fish_indent),
         Task::ManPages => cargo(["build", "--package", "fish-build-man-pages"]),
+        Task::ShellCheck => shellcheck(),
     }
 }
 
-fn run_checks() {
+fn run_checks() -> Result<()> {
     let repo_root_dir = fish_build_helper::workspace_root();
     let check_script = repo_root_dir.join("build_tools").join("check.sh");
-    Command::new(check_script).run_or_panic();
+    Command::new(check_script).run()
 }
 
-fn build_html_docs(fish_indent: Option<PathBuf>) {
-    let fish_indent_path = fish_indent.unwrap_or_else(|| {
-        // Build fish_indent if no existing one is specified.
-        cargo([
-            "build",
-            "--bin",
-            "fish_indent",
-            "--profile",
-            "dev",
-            "--no-default-features",
-        ]);
-        fish_build_helper::fish_build_dir()
-            .join("debug")
-            .join("fish_indent")
-    });
+fn build_html_docs(fish_indent: Option<PathBuf>) -> Result<()> {
+    let fish_indent_path = match fish_indent {
+        Some(path) => path,
+        None => {
+            // Build fish_indent if no existing one is specified.
+            cargo([
+                "build",
+                "--bin",
+                "fish_indent",
+                "--profile",
+                "dev",
+                "--no-default-features",
+            ])?;
+            fish_build_helper::fish_build_dir()
+                .join("debug")
+                .join("fish_indent")
+        }
+    };
     // Set path so `sphinx-build` can find `fish_indent`.
     // Create tempdir to store symlink to fish_indent.
     // This is done to avoid adding other binaries to the PATH.
-    let tempdir = fish_tempfile::new_dir().unwrap();
+    let tempdir = fish_tempfile::new_dir().context("Failed to create tempdir")?;
     std::os::unix::fs::symlink(
-        std::fs::canonicalize(fish_indent_path).unwrap(),
+        std::fs::canonicalize(&fish_indent_path).with_context(|| {
+            format!("Failed to canonicalize path to `fish_indent`: {fish_indent_path:?}")
+        })?,
         tempdir.path().join("fish_indent"),
     )
-    .unwrap();
-    let new_path = format!(
-        "{}:{}",
-        tempdir.path().to_str().unwrap(),
-        fish_build_helper::env_var("PATH").unwrap()
-    );
+    .context("Failed to create symlink for fish_indent")?;
+    let mut new_path = tempdir.path().as_os_str().to_owned();
+    if let Some(current_path) = std::env::var_os("PATH") {
+        new_path.push(":");
+        new_path.push(current_path);
+    }
     let doc_src_dir = fish_build_helper::workspace_root().join("doc_src");
     let doctrees_dir = fish_build_helper::fish_doc_dir().join(".doctrees-html");
     let html_dir = fish_build_helper::fish_doc_dir().join("html");
@@ -91,5 +130,5 @@ fn build_html_docs(fish_indent: Option<PathBuf>) {
     Command::new(option_env!("FISH_SPHINX").unwrap_or("sphinx-build"))
         .env("PATH", new_path)
         .args(args)
-        .run_or_panic();
+        .run()
 }

crates/xtask/src/shellcheck.rs

@@ -0,0 +1,52 @@
+use anyhow::{Context, Result};
+use fish_build_helper::workspace_root;
+use ignore::Walk;
+use pcre2::bytes::Regex;
+use std::{
+    fs::File,
+    io::{BufRead, BufReader},
+    path::{Path, PathBuf},
+    process::Command,
+    sync::LazyLock,
+};
+
+use crate::CommandExt;
+
+pub fn shellcheck() -> Result<()> {
+    Command::new("shellcheck")
+        .args(files_to_check()?)
+        .current_dir(workspace_root())
+        .run()
+}
+
+fn is_shell_script<P: AsRef<Path>>(path: P) -> Result<bool> {
+    let file = File::open(&path).with_context(|| format!("Failed to open {:?}", path.as_ref()))?;
+    let mut first_line = String::new();
+    let Ok(_) = BufReader::new(file).read_line(&mut first_line) else {
+        return Ok(false);
+    };
+    static SHEBANG_REGEX: LazyLock<Regex> = LazyLock::new(|| Regex::new("^#!.*[^i]sh").unwrap());
+    Ok(SHEBANG_REGEX
+        .is_match(first_line.trim().as_bytes())
+        .unwrap())
+}
+
+fn files_to_check() -> Result<Vec<PathBuf>> {
+    let mut files = vec![];
+    for dir_entry in Walk::new(workspace_root()) {
+        let dir_entry = dir_entry.context("Error traversing workspace")?;
+        if !dir_entry
+            .file_type()
+            .with_context(|| format!("Failed to determine file type of {dir_entry:?}"))?
+            .is_file()
+        {
+            continue;
+        }
+        let path = dir_entry.into_path();
+        if !is_shell_script(&path)? {
+            continue;
+        }
+        files.push(path);
+    }
+    Ok(files)
+}

deny.toml

@@ -1,24 +1,24 @@
 [licenses]
 # We want really high confidence when inferring licenses from text
 confidence-threshold = 0.93
-unused-allowed-license = "allow" # don't warn for unused licenses in this list
+unused-allowed-license = "allow"  # don't warn for unused licenses in this list
 allow = [
-    "BSD-2-Clause",
-    "BSD-3-Clause",
-    "BSL-1.0",
-    "CC0-1.0",
-    "GPL-2.0",
-    "GPL-2.0-only",
-    "ISC",
-    "LGPL-2.0",
-    "LGPL-2.0-or-later",
-    "MIT",
-    "MPL-2.0",
-    "PSF-2.0",
-    "Unicode-DFS-2016",
-    "Unicode-3.0",
-    "WTFPL",
-    "Zlib",
+  "BSD-2-Clause",
+  "BSD-3-Clause",
+  "BSL-1.0",
+  "CC0-1.0",
+  "GPL-2.0",
+  "GPL-2.0-only",
+  "ISC",
+  "LGPL-2.0",
+  "LGPL-2.0-or-later",
+  "MIT",
+  "MPL-2.0",
+  "PSF-2.0",
+  "Unicode-DFS-2016",
+  "Unicode-3.0",
+  "WTFPL",
+  "Zlib",
 ]
 
 [sources.allow-org]

doc_src/cmds/argparse.rst

@@ -46,7 +46,7 @@ The following ``argparse`` options are available. They must appear before all *O
     In contrast, if the known option comes first (and does not take any arguments), the known option will be recognised (e.g. ``argparse --move-unknown h -- -ho`` *will* set ``$_flag_h`` to ``-h``)
 
 **-i** or **--ignore-unknown**
-    Deprecated. This is like **--move-unknown**, except that unknown options and their arguments are kept in ``$argv`` and not moved to ``$argv_opts``. Unlike **--move-unknown**, this option makes it impossible to distinguish between an unknown option and non-option argument that starts with a ``-`` (since any ``--`` seperator in ``$argv`` will be removed).
+    Deprecated. This is like **--move-unknown**, except that unknown options and their arguments are kept in ``$argv`` and not moved to ``$argv_opts``. Unlike **--move-unknown**, this option makes it impossible to distinguish between an unknown option and non-option argument that starts with a ``-`` (since any ``--`` separator in ``$argv`` will be removed).
 
 **-S** or **--strict-longopts**
     This makes the parsing of long options more strict. In particular, *without* this flag, if ``long`` is a known long option flag, ``--long`` and ``--long=<value>`` can be abbreviated as:
@@ -221,7 +221,7 @@ These variables are passed to the function as local exported variables.
 
 The script should write any error messages to stdout, not stderr. It should return a status of zero if the flag value is valid otherwise a non-zero status to indicate it is invalid.
 
-Fish ships with a ``_validate_int`` function that accepts a ``--min`` and ``--max`` flag. Let's say your command accepts a ``-m`` or ``--max`` flag and the minimum allowable value is zero and the maximum is 5. You would define the option like this: ``m/max=!_validate_int --min 0 --max 5``. The default if you call ``_validate_int`` without those flags is to check that the value is a valid integer with no limits on the min or max value allowed.
+fish ships with a ``_validate_int`` function that accepts a ``--min`` and ``--max`` flag. Let's say your command accepts a ``-m`` or ``--max`` flag and the minimum allowable value is zero and the maximum is 5. You would define the option like this: ``m/max=!_validate_int --min 0 --max 5``. The default if you call ``_validate_int`` without those flags is to check that the value is a valid integer with no limits on the min or max value allowed.
 
 Here are some examples of flag validations::
 
@@ -251,7 +251,7 @@ Some *OPTION_SPEC* examples:
 
 - ``n/name=?`` means that both ``-n`` and ``--name`` are valid. It accepts an optional value and can be used at most once. If the flag is seen then ``_flag_n`` and ``_flag_name`` will be set with the value associated with the flag if one was provided else it will be set with no values.
 
-- ``n/name=*`` is similar, but the flag can be used more than once. If the flag is seen then ``_flag_n`` and ``_flag_name`` will be set with the values associated with each occurence. Each value will be the value given to the option, or the empty string if no value was given.
+- ``n/name=*`` is similar, but the flag can be used more than once. If the flag is seen then ``_flag_n`` and ``_flag_name`` will be set with the values associated with each occurrence. Each value will be the value given to the option, or the empty string if no value was given.
 
 - ``name=+`` means that only ``--name`` is valid. It requires a value and can be used more than once. If the flag is seen then ``_flag_name`` will be set with the values associated with each occurrence.
 

doc_src/cmds/cd.rst

@@ -6,7 +6,7 @@ Synopsis
 
 .. synopsis::
 
-    cd [DIRECTORY]
+    cd [( -L | --no-dereference ) | ( -P | --dereference )] [DIRECTORY]
 
 Description
 -----------
@@ -18,14 +18,25 @@ Description
 
 ``cd`` changes the current working directory.
 
+The :envvar:`PWD` environment variable is updated with the new working directory, and the previous directory
+is added to the :ref:`directory history <directory-history>`.
+
 If *DIRECTORY* is given, it will become the new directory. If no parameter is given, the :envvar:`HOME` environment variable will be used.
 
 If *DIRECTORY* is a relative path, all the paths in the :envvar:`CDPATH` will be tried as prefixes for it, in addition to :envvar:`PWD`.
 It is recommended to keep **.** as the first element of :envvar:`CDPATH`, or :envvar:`PWD` will be tried last.
 
-Fish will also try to change directory if given a command that looks like a directory (starting with **.**, **/** or **~**, or ending with **/**), without explicitly requiring **cd**.
+The new directory name is partially resolved to remove redundant segments (``.`` or ``..``).
 
-Fish also ships a wrapper function around the builtin **cd** that understands ``cd -`` as changing to the previous directory.
+``cd`` defaults to treating symbolic links as real directories, and not resolving them to their underlying
+targets. The ``$PWD`` :ref:`special variable <variables-special>` variable will contain the path that was
+supplied. This default behaviour can be enforced with the ``-L`` or ``--no-dereference`` option.
+
+The ``-P`` or ``--dereference`` option resolves all symbolic links first. This was the default in fish versions before 3.0.0.
+
+fish will also try to change directory if given a command that looks like a directory (starting with **.**, **/** or **~**, or ending with **/**), without explicitly requiring **cd**.
+
+fish also ships a wrapper function around the builtin **cd** that understands ``cd -`` as changing to the previous directory.
 See also :doc:`prevd <prevd>`.
 This wrapper function maintains a history of the 25 most recently visited directories in the ``$dirprev`` and ``$dirnext`` global variables.
 If you make those universal variables your **cd** history is shared among all fish instances.
@@ -45,6 +56,9 @@ Examples
     cd /usr/src/fish-shell
     # changes the working directory to /usr/src/fish-shell
 
+    cd -P /tmp/link
+    # resolves /tmp/link to its target before recording the directory
+
 See Also
 --------
 

doc_src/cmds/complete.rst

@@ -81,7 +81,7 @@ The following options are available:
 **-h** or **--help**
     Displays help about using this command.
 
-Command-specific tab-completions in ``fish`` are based on the notion of options and arguments. An option is a parameter which begins with a hyphen, such as ``-h``, ``-help`` or ``--help``. Arguments are parameters that do not begin with a hyphen. Fish recognizes three styles of options, the same styles as the GNU getopt library. These styles are:
+Command-specific tab-completions in ``fish`` are based on the notion of options and arguments. An option is a parameter which begins with a hyphen, such as ``-h``, ``-help`` or ``--help``. Arguments are parameters that do not begin with a hyphen. fish recognizes three styles of options, the same styles as the GNU getopt library. These styles are:
 
 - Short options, like ``-a``. Short options are a single character long, are preceded by a single hyphen and can be grouped together (like ``-la``, which is equivalent to ``-l -a``). Option arguments may be specified by appending the option with the value (``-w32``), or, if ``--require-parameter`` is given, in the following parameter (``-w 32``).
 

doc_src/cmds/fish_breakpoint_prompt.rst

@@ -34,6 +34,6 @@ A simple prompt that is a simplified version of the default debugging prompt::
         set -l function (status current-function)
         set -l line (status current-line-number)
         set -l prompt "$function:$line >"
-        echo -ns (set_color $fish_color_status) "BP $prompt" (set_color normal) ' '
+        echo -ns (set_color $fish_color_status) "BP $prompt" (set_color --reset) ' '
     end
 

doc_src/cmds/fish_command_not_found.rst

@@ -18,7 +18,7 @@ When fish tries to execute a command and can't find it, it invokes this function
 
 It can print a message to tell you about it, and it often also checks for a missing package that would include the command.
 
-Fish ships multiple handlers for various operating systems and chooses from them when this function is loaded,
+fish ships multiple handlers for various operating systems and chooses from them when this function is loaded,
 or you can define your own.
 
 It receives the full commandline as one argument per token, so $argv[1] contains the missing command.
@@ -50,25 +50,3 @@ Or the simple default handler::
     function fish_command_not_found
         __fish_default_command_not_found_handler $argv
     end
-
-Backwards compatibility
------------------------
-
-This command was introduced in fish 3.2.0. Previous versions of fish used the "fish_command_not_found" :ref:`event <event>` instead.
-
-To define a handler that works in older versions of fish as well, define it the old way::
-
-  function __fish_command_not_found_handler --on-event fish_command_not_found
-       echo COMMAND WAS NOT FOUND MY FRIEND $argv[1]
-  end
-
-in which case fish will define a ``fish_command_not_found`` that calls it,
-or define a wrapper::
-
-  function fish_command_not_found
-       echo "G'day mate, could not find your command: $argv"
-  end
-
-  function __fish_command_not_found_handler --on-event fish_command_not_found
-       fish_command_not_found $argv
-  end

doc_src/cmds/fish_config.rst

@@ -85,10 +85,10 @@ The format looks like this:
     fish_color_command 5c5cff
 
     [unknown]
-    fish_color_normal normal
+    fish_color_normal --reset
     fish_color_autosuggestion brblack
     fish_color_cancel -r
-    fish_color_command normal
+    fish_color_command --reset
 
 The comments provide name and background color to the web config tool.
 

doc_src/cmds/fish_greeting.rst

@@ -22,6 +22,8 @@ When an interactive fish starts, it executes fish_greeting and displays its outp
 
 The default fish_greeting is a function that prints a variable of the same name (``$fish_greeting``), so you can also just change that if you just want to change the text.
 
+If :envvar:`SHELL_WELCOME` is set, it is displayed after the greeting. This is a standard environment variable that may be set by tools like systemd's ``run0`` to display session information.
+
 While you could also just put ``echo`` calls into config.fish, fish_greeting takes care of only being used in interactive shells, so it won't be used e.g. with ``scp`` (which executes a shell), which prevents some errors.
 
 Example
@@ -39,5 +41,5 @@ A simple greeting:
 
   function fish_greeting
       echo Hello friend!
-      echo The time is (set_color yellow)(date +%T)(set_color normal) and this machine is called $hostname
+      echo The time is (set_color yellow)(date +%T)(set_color --reset) and this machine is called $hostname
   end

doc_src/cmds/fish_mode_prompt.rst

@@ -62,7 +62,7 @@ Example
           set_color --bold red
           echo '?'
       end
-      set_color normal
+      set_color --reset
     end
 
 

doc_src/cmds/fish_prompt.rst

@@ -24,6 +24,8 @@ The exit status of commands within ``fish_prompt`` will not modify the value of
 
 If :envvar:`fish_transient_prompt` is set to 1, ``fish_prompt --final-rendering`` is run before executing the commandline.
 
+If :envvar:`SHELL_PROMPT_PREFIX` or :envvar:`SHELL_PROMPT_SUFFIX` are set, they are automatically prepended and appended to the left prompt. This applies to all prompts regardless of whether ``fish_prompt`` has been customized.
+
 ``fish`` ships with a number of example prompts that can be chosen with the ``fish_config`` command.
 
 
@@ -41,7 +43,7 @@ A simple prompt:
         # $USER and $hostname are set by fish, so you can just use them
         # instead of using `whoami` and `hostname`
         printf '%s@%s %s%s%s > ' $USER $hostname \
-            (set_color $fish_color_cwd) (prompt_pwd) (set_color normal)
+            (set_color $fish_color_cwd) (prompt_pwd) (set_color --reset)
     end
 
 

doc_src/cmds/fish_vi_key_bindings.rst

@@ -27,16 +27,11 @@ Further information on how to use :ref:`vi mode <vi-mode>`.
 Differences from Vim
 --------------------
 
-Fish's vi mode aims to be familiar to vim users, but there are some differences:
+fish's vi mode aims to be familiar to vim users, but there are some differences:
 
 **Word character handling**
     In vim, underscore (``_``) is treated as a keyword character by default, so word motions like ``w``, ``b``, and ``e`` treat ``foo_bar`` as a single word. In fish, underscore is treated as punctuation, so word motions stop at underscores. For example, pressing ``w`` on ``foo_bar`` in fish stops at the ``_``, while in vim it would jump past the entire identifier.
 
-**The** ``cw`` **command**
-    In vim, ``cw`` has special behavior: when the cursor is on a non-space character, it behaves like ``ce`` (change to end of word), but when the cursor is on a space, it behaves like ``dwi`` (delete word then insert).
-
-    In fish, ``cw`` always behaves like ``dwi`` - it deletes to the start of the next word (including trailing whitespace), then enters insert mode. To get vim's ``cw`` behavior in fish, use ``ce`` instead.
-
 Examples
 --------
 

doc_src/cmds/funced.rst

@@ -40,7 +40,7 @@ Run::
 
   >_ funced fish_prompt
 
-This will open up your editor, allowing you to modify the function. When you're done, save and quit. Fish will reload the function, so you should see the changes right away.
+This will open up your editor, allowing you to modify the function. When you're done, save and quit. fish will reload the function, so you should see the changes right away.
 
 When you're done, use::
 

doc_src/cmds/function.rst

@@ -32,7 +32,7 @@ The following options are available:
     If the wrapped command is the same as the function name, this will be ignored.
 
 **-e** *EVENT_NAME* or **--on-event** *EVENT_NAME*
-    Run this function when the specified named event is emitted. Fish internally generates named events, for example, when showing the prompt. Custom events can be emitted using the :doc:`emit <emit>` command.
+    Run this function when the specified named event is emitted. fish internally generates named events, for example, when showing the prompt. Custom events can be emitted using the :doc:`emit <emit>` command.
 
 **-v** *VARIABLE_NAME* or **--on-variable** *VARIABLE_NAME*
     Run this function when the variable *VARIABLE_NAME* changes value. Note that :program:`fish` makes no guarantees on any particular timing or even that the function will be run for every single ``set``. Rather it will be run when the variable has been set at least once, possibly skipping some values or being run when the variable has been set to the same value (except for universal variables set in other shells - only changes in the value will be picked up for those).
@@ -56,10 +56,12 @@ The following options are available:
 **-V** or **--inherit-variable NAME**
     Snapshots the value of the variable ``NAME`` and defines a local variable with that same name and value when the function is defined. This is similar to a closure in other languages like Python but a bit different. Note the word "snapshot" in the first sentence. If you change the value of the variable after defining the function, even if you do so in the same scope (typically another function) the new value will not be used by the function you just created using this option. See the ``function notify`` example below for how this might be used.
 
-The event handler switches (``on-event``, ``on-variable``, ``on-job-exit``, ``on-process-exit`` and ``on-signal``) cause a function to run automatically at specific events. New named events for ``--on-event`` can be fired using the :doc:`emit <emit>` builtin. Fish already generates a few events, see :ref:`event` for more.
+The event handler switches (``on-event``, ``on-variable``, ``on-job-exit``, ``on-process-exit`` and ``on-signal``) cause a function to run automatically at specific events. New named events for ``--on-event`` can be fired using the :doc:`emit <emit>` builtin. fish already generates a few events, see :ref:`event` for more.
 
 Functions names cannot be reserved words. These are elements of fish syntax or builtin commands which are essential for the operations of the shell. Current reserved words are ``[``, ``_``, ``and``, ``argparse``, ``begin``, ``break``, ``builtin``, ``case``, ``command``, ``continue``, ``else``, ``end``, ``eval``, ``exec``, ``for``, ``function``, ``if``, ``not``, ``or``, ``read``, ``return``, ``set``, ``status``, ``string``, ``switch``, ``test``, ``time``, and ``while``.
 
+Care should be taken when creating a function of the same name as an existing shell builtin or common program. If the function behaves differently, it is very common for problems to occur within fish or in scripts written by others. Consider writing an :doc:`abbreviation <abbr>` if you are wanting to replace one tool with another for interactive use.
+
 Example
 -------
 

doc_src/cmds/history.rst

@@ -109,7 +109,7 @@ You can set the ``fish_history`` variable to another name for the current shell
 
 You can change ``fish_history`` at any time (by using ``set -x fish_history "session_name"``) and it will take effect right away. If you set it to ``"default"``, it will use the default session name (which is ``"fish"``).
 
-Other shells such as bash and zsh use a variable named ``HISTFILE`` for a similar purpose. Fish uses a different name to avoid conflicts and signal that the behavior is different (session name instead of a file path). Also, if you set the var to anything other than ``fish`` or ``default`` it will inhibit importing the bash history. That's because the most common use case for this feature is to avoid leaking private or sensitive history when giving a presentation.
+Other shells such as bash and zsh use a variable named ``HISTFILE`` for a similar purpose. fish uses a different name to avoid conflicts and signal that the behavior is different (session name instead of a file path). Also, if you set the var to anything other than ``fish`` or ``default`` it will inhibit importing the bash history. That's because the most common use case for this feature is to avoid leaking private or sensitive history when giving a presentation.
 
 Notes
 -----

doc_src/cmds/if.rst

@@ -16,7 +16,7 @@ Description
 
 ``if`` will execute the command ``CONDITION``. If the condition's exit status is 0, the commands ``COMMANDS_TRUE`` will execute.  If the exit status is not 0 and :doc:`else <else>` is given, ``COMMANDS_FALSE`` will be executed.
 
-You can use :doc:`and <and>` or :doc:`or <or>` in the condition. See the second example below.
+You can use :doc:`not <not>`, :doc:`and <and>` or :doc:`or <or>` in the condition. See the second example below.
 
 The exit status of the last foreground command to exit can always be accessed using the :ref:`$status <variables-status>` variable.
 
@@ -53,7 +53,7 @@ See also
 
 ``if`` is only as useful as the command used as the condition.
 
-Fish ships a few:
+fish ships a few:
 
 - :doc:`test` can compare numbers, strings and check paths
 - :doc:`string` can perform string operations including wildcard and regular expression matches

doc_src/cmds/math.rst

@@ -218,6 +218,6 @@ Examples
 Compatibility notes
 -------------------
 
-Fish 1.x and 2.x releases relied on the ``bc`` command for handling ``math`` expressions. Starting with fish 3.0.0 fish uses the tinyexpr library and evaluates the expression without the involvement of any external commands.
+fish 1.x and 2.x releases relied on the ``bc`` command for handling ``math`` expressions. Starting with fish 3.0.0 fish uses the tinyexpr library and evaluates the expression without the involvement of any external commands.
 
 You don't need to use ``--`` before the expression, even if it begins with a minus sign which might otherwise be interpreted as an invalid option. If you do insert ``--`` before the expression, it will cause option scanning to stop just like for every other command and it won't be part of the expression.

doc_src/cmds/read.rst

@@ -62,7 +62,7 @@ The following options control the interactive mode:
     Masks characters written to the terminal, replacing them with asterisks. This is useful for reading things like passwords or other sensitive information.
 
 **-p** or **--prompt** *PROMPT_CMD*
-    Uses the output of the shell command *PROMPT_CMD* as the prompt for the interactive mode. The default prompt command is ``set_color green; echo -n read; set_color normal; echo -n "> "``
+    Uses the output of the shell command *PROMPT_CMD* as the prompt for the interactive mode. The default prompt command is ``set_color green; echo -n read; set_color --reset; echo -n "> "``
 
 **-P** or **--prompt-str** *PROMPT_STR*
     Uses the literal *PROMPT_STR* as the prompt for the interactive mode.
@@ -95,7 +95,7 @@ The following options control how much is read and how it is stored:
     Marks the end of the line with the NUL character, instead of newline. This also disables interactive mode.
 
 **-L** or **--line**
-    Reads each line into successive variables, and stops after each variable has been filled. This cannot be combined with the ``--delimiter`` option.
+    Reads each line into successive variables, and stops after each variable has been filled. This cannot be combined with the ``--null`` option, or options to control splitting like ``--delimiter``.
 
 Without the ``--line`` option, ``read`` reads a single line of input from standard input, breaks it into tokens, and then assigns one token to each variable specified in *VARIABLES*. If there are more tokens than variables, the complete remainder is assigned to the last variable.
 

doc_src/cmds/set.rst

@@ -6,16 +6,17 @@ Synopsis
 
 .. synopsis::
 
-    set
-    set (-f | --function) (-l | --local) (-g | --global) (-U | --universal) [--no-event]
-    set [-Uflg] NAME [VALUE ...]
-    set [-Uflg] NAME[[INDEX ...]] [VALUE ...]
-    set (-x | --export) (-u | --unexport) [-Uflg] NAME [VALUE ...]
-    set (-a | --append) (-p | --prepend) [-Uflg] NAME VALUE ...
-    set (-e | --erase) [-Uflg] [-xu] [NAME][[INDEX]] ...]
-    set (-q | --query) [-Uflg] [-xu] [NAME][[INDEX]] ...]
+    set [(-f | --function) (-l | --local) (-g | --global) (-U | --universal)]
+        [(-x | --export) (-u | --unexport)]
     set (-S | --show) (-L | --long) [NAME ...]
 
+    set [-Uflg] [-xu] [--no-event] NAME [VALUE ...]
+    set [-Uflg] [--no-event] NAME[[INDEX ...]] [VALUE ...]
+    set (-a | --append) (-p | --prepend) [-Uflg] [--no-event] NAME VALUE ...
+    set (-e | --erase) [-Uflg] [--no-event] NAME[[INDEX]] ...
+
+    set (-q | --query) [-Uflg] [-xu] NAME[[INDEX]] ...
+
 Description
 -----------
 
@@ -89,7 +90,7 @@ Further options:
 **-q** or **--query** *NAME*\[*INDEX*\]
     Test if the specified variable names are defined.
     If an *INDEX* is provided, check for items at that slot.
-    With a given scope (like **--global**) or attribute (like **--exported** or **--path**) check only variables that match.
+    With a given scope (like **--global**) or attribute (like **--export** or **--path**) check only variables that match.
     Does not output anything, but the shell status is set to the number of variables specified that were not defined, up to a maximum of 255.
     If no variable was given, it also returns 255.
 
@@ -219,4 +220,4 @@ This runs fish with a temporary home directory::
 
 Notes
 -----
-- Fish versions prior to 3.0 supported the syntax ``set PATH[1] PATH[4] /bin /sbin``, which worked like ``set PATH[1 4] /bin /sbin``.
+- fish versions prior to 3.0 supported the syntax ``set PATH[1] PATH[4] /bin /sbin``, which worked like ``set PATH[1 4] /bin /sbin``.

doc_src/cmds/set_color.rst

@@ -6,12 +6,17 @@ Synopsis
 
 .. synopsis::
 
-    set_color [OPTIONS] VALUE
+    set_color [OPTIONS] [VALUE]
 
 Description
 -----------
 
-``set_color`` is used to control the color and styling of text in the terminal. *VALUE* describes that styling. *VALUE* can be a reserved color name like **red** or an RGB color value given as 3 or 6 hexadecimal digits ("F27" or "FF2277"). A special keyword **normal** resets text formatting to terminal defaults.
+``set_color`` controls the color and styling of text in the terminal.
+It writes non-printing color and text style escape sequences to standard output.
+
+*VALUE* describes the styling.
+*VALUE* can be a reserved color name like **red** or an RGB color value given as 3 or 6 hexadecimal digits ("F27" or "FF2277").
+A special keyword **normal** resets text formatting to terminal defaults, however it is not recommended and the **--reset** option is preferred as it is less confusing and more future-proof.
 
 Valid colors include:
 
@@ -26,13 +31,18 @@ Hexadecimal RGB values can be in lower or uppercase.
 
 If :envvar:`fish_term24bit` is set to 0, fish will translate RGB values to the nearest color on the 256-color palette.
 If :envvar:`fish_term256` is also set to 0, fish will translate them to the 16-color palette instead.
-Fish launched as ``fish -d term_support`` will include diagnostic messages that indicate the color support mode in use.
+fish launched as ``fish -d term_support`` will include diagnostic messages that indicate the color support mode in use.
 
 If multiple colors are specified, fish prefers the first RGB one.
 However if :envvar:`fish_term256` is set to 0, fish prefers the first named color specified.
 
 The following options are available:
 
+**-f** or **--foreground** *COLOR*
+    Sets the foreground color.
+    This is equivalent to calling ``set_color COLOR`` with the exception that the keyword **normal** will only reset the foreground color to its default, instead of all colors and modes.
+    It cannot be used with *VALUE* or **--print-colors**.
+
 **-b** or **--background** *COLOR*
     Sets the background color.
 
@@ -41,6 +51,7 @@ The following options are available:
 
 **-c** or **--print-colors**
     Prints the given colors or a colored list of the 16 named colors.
+    It cannot be used with **--foreground**.
 
 **-o** or **--bold**
     Sets bold mode.
@@ -48,17 +59,21 @@ The following options are available:
 **-d** or **--dim**
     Sets dim mode.
 
-**-i** or **--italics**
-    Sets italics mode.
+**-i** or **--italics**, or **-iSTATE** or **--italics=STATE**
+    Sets italics mode. The state can be **on** (default), or **off**.
 
-**-r** or **--reverse**
-    Sets reverse mode.
+**-r** or **--reverse**, or **-iSTATE** or **--reverse=STATE**
+    Sets reverse mode. The state can be **on** (default), or **off**.
 
-**-s** or **--strikethrough**
-    Sets strikethrough mode.
+**-s** or **--strikethrough**, or **-sSTATE** or **--strikethrough=STATE**
+    Sets strikethrough mode. The state can be **on** (default), or **off**.
 
 **-u** or **--underline**, or **-uSTYLE** or **--underline=STYLE**
-    Set the underline mode; supported styles are **single** (default), **double**, **curly**, **dotted** and **dashed**.
+    Set the underline mode; supported styles are **single** (default), **double**, **curly**, **dotted**, **dashed** and **off**.
+
+**--reset**
+    Reset the text formatting to the terminal defaults before applying the new colors and modes.
+    This is equivalent to calling ``set_color normal`` except that it is possible to set the foreground color in the same call (e.g. ``set_color --reset green``)
 
 **--theme=THEME**
     Ignored.
@@ -75,10 +90,14 @@ The following options are available:
 Notes
 -----
 
-1. Using **set_color normal** will reset all colors and modes to the terminal's default.
-2. Setting the background color only affects subsequently written characters. Fish provides no way to set the background color for the entire terminal window. Configuring the window background color (and other attributes such as its opacity) has to be done using whatever mechanisms the terminal provides. Look for a config option.
-3. Some terminals use the ``--bold`` escape sequence to switch to a brighter color set rather than increasing the weight of text.
-4. ``set_color`` works by printing sequences of characters to standard output. If used in command substitution or a pipe, these characters will also be captured. This may or may not be desirable. Checking the exit status of ``isatty stdout`` before using ``set_color`` can be useful to decide not to colorize output in a script.
+1. Using ``set_color normal`` will reset all colors and modes to the terminal's default.
+2. In contrast, ``set_color --foreground normal`` will only reset the foreground color and leave all the other colors and modes unchanged.
+3. Because of the risk of confusion, ``set_color --reset`` is recommended over ``set_color normal``.
+4. Setting the background color only affects subsequently written characters. fish provides no way to set the background color for the entire terminal window. Configuring the window background color (and other attributes such as its opacity) has to be done using whatever mechanisms the terminal provides. Look for a config option.
+5. Some terminals use the ``--bold`` escape sequence to switch to a brighter color set rather than increasing the weight of text.
+6. If you use ``set_color`` in a command substitution or a pipe, these characters will also be captured.
+   This may or may not be desirable.
+   Checking the exit status of ``isatty stdout`` before using ``set_color`` can be useful to decide not to colorize output in a script.
 
 Examples
 --------

doc_src/cmds/string-join.rst

@@ -38,7 +38,7 @@ just always use ``--`` to avoid unwelcome surprises.
 ``string join0`` adds a trailing NUL. This is most useful in conjunction with tools that accept NUL-delimited input, such as ``sort -z``.
 
 Because Unix uses NUL as the string terminator, passing the output of ``string join0`` as an *argument* to a command (via a :ref:`command substitution <expand-command-substitution>`) won't actually work.
-Fish will pass the correct bytes along, but the command won't be able to tell where the argument ends.
+fish will pass the correct bytes along, but the command won't be able to tell where the argument ends.
 This is a limitation of Unix' argument passing.
 
 .. END DESCRIPTION

doc_src/cmds/ulimit.rst

@@ -102,11 +102,11 @@ The following additional options are also understood by ``ulimit``:
 
 The ``fish`` implementation of ``ulimit`` should behave identically to the implementation in bash, except for these differences:
 
-- Fish ``ulimit`` supports GNU-style long options for all switches.
+- fish ``ulimit`` supports GNU-style long options for all switches.
 
-- Fish ``ulimit`` does not support the **-p** option for getting the pipe size. The bash implementation consists of a compile-time check that empirically guesses this number by writing to a pipe and waiting for SIGPIPE. Fish does not do this because this method of determining pipe size is unreliable. Depending on bash version, there may also be further additional limits to set in bash that do not exist in fish.
+- fish ``ulimit`` does not support the **-p** option for getting the pipe size. The bash implementation consists of a compile-time check that empirically guesses this number by writing to a pipe and waiting for SIGPIPE. fish does not do this because this method of determining pipe size is unreliable. Depending on bash version, there may also be further additional limits to set in bash that do not exist in fish.
 
-- Fish ``ulimit`` does not support getting or setting multiple limits in one command, except reporting all values using the **-a** switch.
+- fish ``ulimit`` does not support getting or setting multiple limits in one command, except reporting all values using the **-a** switch.
 
 
 Example

doc_src/completions.rst

@@ -38,7 +38,7 @@ which offers yes/no in these cases::
   > myprog -o <TAB>
   > myprog --output <TAB>
 
-Fish will also offer files by default, in addition to the arguments you specified. You would either inhibit file completion for a single option::
+fish will also offer files by default, in addition to the arguments you specified. You would either inhibit file completion for a single option::
 
   complete -c myprog -s o -l output --no-files -ra "yes no"
 
@@ -142,15 +142,15 @@ Functions beginning with the string ``__fish_print_`` print a newline separated
 Where to put completions
 ------------------------
 
-Completions can be defined on the commandline or in a configuration file, but they can also be automatically loaded. Fish automatically searches through any directories in the list variable ``$fish_complete_path``, and any completions defined are automatically loaded when needed. A completion file must have a filename consisting of the name of the command to complete and the suffix ``.fish``.
+Completions can be defined on the commandline or in a configuration file, but they can also be automatically loaded. fish automatically searches through any directories in the list variable ``$fish_complete_path``, and any completions defined are automatically loaded when needed. A completion file must have a filename consisting of the name of the command to complete and the suffix ``.fish``.
 
-By default, Fish searches the following for completions, using the first available file that it finds:
+By default, fish searches the following for completions, using the first available file that it finds:
 
 - A directory for end-users to keep their own completions, usually ``~/.config/fish/completions`` (controlled by the ``XDG_CONFIG_HOME`` environment variable);
 - A directory for systems administrators to install completions for all users on the system, usually ``/etc/fish/completions``;
 - A user-specified directory for third-party vendor completions, usually ``~/.local/share/fish/vendor_completions.d`` (controlled by the ``XDG_DATA_HOME`` environment variable);
 - A directory for third-party software vendors to ship their own completions for their software, usually ``/usr/share/fish/vendor_completions.d``;
-- The completions shipped with fish, usually installed in ``/usr/share/fish/completions``; and
+- The completions shipped with fish, which are stored in the fish program and can be seen with ``status list-files``; and
 - Completions automatically generated from the operating system's manual, usually stored in ``~/.cache/fish/generated_completions`` (controlled by ``XDG_CACHE_HOME`` environment variable).
 
 These paths are controlled by parameters set at build, install, or run time, and may vary from the defaults listed above.

doc_src/conf.py

@@ -218,7 +218,7 @@ latex_engine = "xelatex"
 
 def get_command_description(path, name):
     """Return the description for a command, by parsing its synopsis line"""
-    with open(path) as opened:
+    with open(path, encoding="utf8") as opened:
         for line in opened:
             if line.startswith(name + " - "):
                 _, desc = line.split(" - ", 1)

doc_src/design.rst

@@ -5,7 +5,7 @@ This is a description of the design principles that have been used to design fis
 
 1. Everything that can be done in other shell languages should be possible to do in fish, though fish may rely on external commands in doing so.
 
-2. Fish should be user-friendly, but not at the expense of expressiveness. Most tradeoffs between power and ease of use can be avoided with careful design.
+2. fish should be user-friendly, but not at the expense of expressiveness. Most tradeoffs between power and ease of use can be avoided with careful design.
 
 3. Whenever possible without breaking the above goals, fish should follow POSIX.
 
@@ -55,9 +55,9 @@ Different configuration options are a nightmare to maintain, since the number of
 
 Examples:
 
-- Fish allows the user to set various syntax highlighting colors. This is needed because fish does not know what colors the terminal uses by default, which might make some things unreadable. The proper solution would be for text color preferences to be defined centrally by the user for all programs, and for the terminal emulator to send these color properties to fish.
+- fish allows the user to set various syntax highlighting colors. This is needed because fish does not know what colors the terminal uses by default, which might make some things unreadable. The proper solution would be for text color preferences to be defined centrally by the user for all programs, and for the terminal emulator to send these color properties to fish.
 
-- Fish does not allow you to set the number of history entries, different language substyles or any number of other common shell configuration options.
+- fish does not allow you to set the number of history entries, different language substyles or any number of other common shell configuration options.
 
 A special note on the evils of configurability is the long list of very useful features found in some shells, that are not turned on by default. Both zsh and bash support command-specific completions, but no such completions are shipped with bash by default, and they are turned off by default in zsh. Other features that zsh supports that are disabled by default include tab-completion of strings containing wildcards, a sane completion pager and a history file.