Repository navigation
refactor(picker): migrate to skim 4.8 (ratatui), drop vendored skim-tuikit - #3137
Conversation
…uikit
Replace skim 0.20.5 (tuikit backend) with skim 4.8.0 (ratatui/crossterm).
Delete the vendor/skim-tuikit patch tree and its build machinery (Taskfile
vendor-diff, flake vendorSrc/extraDummyScript). Both carried patches are moot
in 4.8: alt-<digit> key parsing is native (binds.rs routes every single char
through KeyCode::Char), and the Output::flush write_all fix is subsumed by
stdlib BufWriter::flush in the ratatui backend.
Rewrite the picker against the 4.x API. WorktreeSkimItem::display() returns
ratatui::Line. alt-r removal rides skim's reload(remove {}) token instead of
the removed downcast path plus a signal file. Preview-tab switching and the
progressive list are driven by Skim::event_sender() + Event::Render/RunPreview.
skim 4.x renders on demand (there is no 100ms tuikit heartbeat), so the
progressive `wt switch` list rendered blank during collection. Poke
Event::Render from the collect callbacks (throttled to 16ms) to restore it.
The same on-demand model surfaced four more regressions, all fixed here:
preview-tab lag, current-row highlight, Shift-Tab back-cycle (crossterm reports
Shift-Tab under three KeyEvent shapes, so bind all three), and alt-r removal
(execute-silent is fire-and-forget in 4.x and raced its reader).
Restore styled help. skim 0.20 transitively enabled clap/unstable-markdown,
which renders our `///` doc-comments as styled help; skim 4
(default-features = false) drops it, reverting help to raw text. Depend on it
explicitly to keep help output identical.
Update the integration-test harness for the 4.x picker:
- skim 4.x queries cursor position (ESC[6n) at startup in partial-height mode.
The Unix PTY harness now answers it, mirroring the existing ConPTY responder;
without a reply skim's select() times out and the picker never initializes.
- skim 4.x draws the list/preview separator one column left (col 59), so the
panel-split capture columns shifted to match.
- Regenerate the picker TUI snapshots: the match counter no longer overlaps the
preview tab header, and the HEAD column shows the full short-SHA.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
# Conflicts: # src/commands/picker/mod.rs
worktrunk-bot
left a comment
There was a problem hiding this comment.
Reviewed the skim 0.20 → 4.x migration. The mechanics — reload(remove {}) replacing the raced execute-silent+signal-file (which under skim 4.x could remove the wrong worktree on a repeat alt-r), run_skim inlining init/run to grab event_sender() for on-demand repaints, the ModeWatcher for tab-switch preview refresh, and the PTY test harness answering skim's ESC[6n cursor query — all read as carefully reasoned and well-documented. Two things below.
1. minimal-versions is failing on this PR's new deps. Not transient — it's a deterministic under-constraint introduced by ansi-to-tui. The new ansi-to-tui 8.0.1 declares simdutf8 = "^0.1", but the impl std::error::Error/Display for Utf8Error that its thiserror::Error derive depends on only landed in simdutf8 0.1.1 (changelog: "Make both Utf8Error variants implement std::error::Error" under 0.1.1). -Z minimal-versions resolves ^0.1 down to 0.1.0, so cargo check fails with Utf8Error doesn't implement std::fmt::Display (run).
The usual floor-pin fixes it — add to [target.'cfg(unix)'.dependencies]:
# minimal-versions floor: ansi-to-tui 8 declares `simdutf8 = "0.1"`, but the
# `Utf8Error: std::error::Error` impl its thiserror derive needs only lands in
# simdutf8 0.1.1. Pin the floor so `-Z minimal-versions` resolves a usable
# version. Pulled only transitively (hence the cargo-udeps ignore).
simdutf8 = { version = "0.1.1", optional = true }plus "dep:simdutf8" in the cli feature and a normal/optional entry under [package.metadata.cargo-udeps.ignore] so check-unused-dependencies stays green. Happy to push that commit if you'd like — there's enough leeway in how the floor is wired (optional vs. plain, which udeps bucket) that I left it for you rather than guessing your preference.
2. Data-safety note (not blocking): this edits the alt-r removal path (picker/mod.rs still drives delete_branch_if_safe and worktree removal via prepare_removal). The destructive primitives themselves are untouched and the change is a net safety improvement, but per the repo's data-loss-surface policy I'm not auto-approving picker-removal edits — flagging it for the record rather than rubber-stamping.
One inline doc-comment fix below.
The run_skim doc block was left attached to `struct ModeWatcher` when the watcher was inserted above the function; the real run_skim has its own doc. Addresses review feedback on #3137. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… stack The skim 0.20 -> 4.8 migration brought a TUI/image dependency tree whose crates pin tighter floors than worktrunk's manifests declared. Run the minimal-versions CI job with `-Z direct-minimal-versions` (minimize only our own direct deps; let transitive crates resolve normally) and raise each under-specified floor to the minimum the workspace builds against. Full `-Z minimal-versions` would also minimize skim's transitive TUI/image stack (ansi-to-tui's simdutf8, ratatui's instability macro, color-eyre's eyre, ratatui-image -> image/avif -> num-* and bitvec), whose crates under-declare their own floors and don't compile at the picked versions -- upstream brokenness we'd have to pin around. Direct minimization confines the check to floors we own. worktrunk direct-dep floors (largely mirroring skim 4.8's own requirements): anyhow 1.0.102, anstyle 1.0.13, log 0.4.31, tracing 0.1.35, serde 1.0.228, serde_json 1.0.145, strsim 0.11.1, terminal_size 0.4.3, unicode-width 0.2.1, once_cell 1.21.4, which 8.0.2, regex 1.12.3, ansi-to-tui 8.0.1, tokio 1.52.3, nix 0.31.3, vt100 0.16.2 (dev). signal-hook 0.4.1: 0.4.0 pins `libc =0.2.156`, which conflicts with nix 0.31.3's `libc ^0.2.186`; 0.4.1 relaxes it to `^0.2`. skim 4.8 / ratatui 0.30.1 / chrono 0.4.31 carry over from the migration: the picker is written against 4.8's ratatui 0.30.1 API; ratatui 0.30.0 is a broken release (its `#[instability::unstable]` attribute fails to compile); `DateTime::from_timestamp` (utils.rs, statusline.rs, backup.rs) landed in chrono 0.4.31. Test helpers checked by `--workspace --all-targets` get matching floors: mock-stub (serde 1.0.228, serde_json 1.0.145) and wt-perf (clap 4.6, serde_json 1.0.145, insta 1.47.2). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reconcile the live-CI-fetch + `--prs` picker work with main's skim 4.x (ratatui/crossterm) migration (#3137), which rewrote the picker's skim integration and dropped the vendored skim-tuikit tree. Key reconciliations: - Port `PrSkimItem` to skim 4.x: `display()` returns a ratatui `Line` via the shared `ansi_to_line` helper (not `skim::AnsiString`), and the `--prs` thread sends one batched `Vec<Arc<dyn SkimItem>>` (skim 4.x's channel takes batches, not single items). - `on_update` writes the live `pr_status` slot, then pokes main's `request_render` (skim 4.x repaints on demand; the 100ms heartbeat is gone). The pr tab still can't be re-queried mid-selection, so its "press alt-6 to refresh" hint stays. - Fix the Shift-Tab rotation: main added `shift-btab`/`shift-tab` bindings with the old 5-tab `tr 12345 51234`; bump them to the 6-tab `tr 123456 612345` so Shift-Tab wraps 1 -> pr (6), not 1 -> summary (5). - Drop the stale skim-0.20.5 / vendor/NOTES.md SGR-22 and heartbeat comments; keep the explicit `\x1b[0m` reset (still needed). - Regenerate the 11 picker PTY snapshots for skim 4.x rendering (border at col 59, `[N/M]` match counter) plus the CI column and pr tab. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
worktrunk 0.61.0 Created-by: HarmonybrewBot Commit-by: HarmonybrewBot Merged-by: HarmonybrewBot Description: Created by `brew bump` --- Created with `brew bump-formula-pr`.<details> <summary>release notes</summary> <pre>## Release Notes ### Improved - **`wt list` custom columns**: Each `[list.custom-columns.<Header>]` entry in user config adds a column to `wt list` (and the `wt switch` picker), rendered per row as a minijinja template over `branch`, `worktree_path`, `worktree_name`, and `vars.*`, with optional `width` and drop priority. Values expand from in-memory data only — no subprocess runs per cell — and a column that is empty on every row is dropped. `wt list --format=json` gains a `columns` map per item. The feature is experimental, so the config shape may still change. ([#3073](max-sixty/worktrunk#3073)) - **`--config-set` for inline config overrides**: A global, repeatable `--config-set <toml>` flag overrides any user-config key for a single invocation, layered above config files and `WORKTRUNK_*` env vars. The value is a real TOML fragment, so arrays and tables work natively (`wt --config-set list.full=true list`); nested tables deep-merge, and a malformed or invalid override drops the whole `--config-set` layer with an attributed warning rather than failing the command. ([#3138](max-sixty/worktrunk#3138)) - **Picker shows cached PR/MR numbers**: The `wt switch` picker skips the networked CI-status fetch, so it previously had no CI column. It now fills PR/MR numbers from the local `.git/wt/cache/ci-status/` cache populated by earlier `wt list --full` or statusline runs, with zero network access. A stale entry (TTL passed or branch head moved) keeps its number dimmed; expired entries without a number are dropped. ([#3073](max-sixty/worktrunk#3073)) - **Faster file copies on macOS**: After a reflink (`clonefile` on APFS, which already preserves mode bits), worktrunk now skips the redundant follow-up `chmod` on macOS, saving one syscall per file in `wt step copy-ignored` and every other copy path. Linux (btrfs/XFS) still sets permissions, since `FICLONE` clones data extents only and drops the execute bit. ([#3149](max-sixty/worktrunk#3149)) ### Internal - **Picker migrated to skim 4.8 (ratatui/crossterm)**: The `wt switch` picker moved off skim 0.20.5 (tuikit) to skim 4.8.0, dropping the vendored `vendor/skim-tuikit/` patch tree (both patches it carried are now native or upstream). Two cosmetic picker changes come with it: the match counter no longer overlaps the preview-tab header, and the HEAD column shows the full short-SHA. ([#3137](max-sixty/worktrunk#3137)) - **All command spawns route through one trace chokepoint**: `CommandTrace` is now the sole emitter of `[wt-trace]` command records, so a spawn path can't silently skip tracing — `git worktree add`, previously an unattributed gap, now shows up as a labeled slice in `wt-perf timeline`. ([#3134](max-sixty/worktrunk#3134)) ### Documentation - **Hook-approval skill guidance**: The bundled worktrunk skill now frames hook approvals as user consent and no longer advocates `--yes` to bypass prompts. ([#3146](max-sixty/worktrunk#3146)) ## Install worktrunk 0.61.0 ### Install prebuilt binaries via shell script ```sh curl --proto '=https' --tlsv1.2 -LsSf https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-installer.sh | sh && wt config shell install ``` ### Install prebuilt binaries via powershell script ```sh powershell -ExecutionPolicy Bypass -c "irm https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-installer.ps1 | iex"; git-wt config shell install ``` ### Install prebuilt binaries via Homebrew ```sh brew install worktrunk && wt config shell install ``` ## Download worktrunk 0.61.0 | File | Platform | Checksum | |--------|----------|----------| | [worktrunk-aarch64-apple-darwin.tar.xz](https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-aarch64-apple-darwin.tar.xz) | Apple Silicon macOS | [checksum](https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-aarch64-apple-darwin.tar.xz.sha256) | | [worktrunk-x86_64-apple-darwin.tar.xz](https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-x86_64-apple-darwin.tar.xz) | Intel macOS | [checksum](https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-x86_64-apple-darwin.tar.xz.sha256) | | [worktrunk-x86_64-pc-windows-msvc.zip](https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-x86_64-pc-windows-msvc.zip) | x64 Windows | [checksum](https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-x86_64-pc-windows-msvc.zip.sha256) | | [worktrunk-aarch64-unknown-linux-musl.tar.xz](https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-aarch64-unknown-linux-musl.tar.xz) | ARM64 MUSL Linux | [checksum](https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-aarch64-unknown-linux-musl.tar.xz.sha256) | | [worktrunk-x86_64-unknown-linux-musl.tar.xz](https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-x86_64-unknown-linux-musl.tar.xz) | x64 MUSL Linux | [checksum](https://github.com/max-sixty/worktrunk/releases/download/v0.61.0/worktrunk-x86_64-unknown-linux-musl.tar.xz.sha256) | ### Install via Cargo ```sh cargo install worktrunk && wt config shell install ``` ### Install via Winget (Windows) ```sh winget install max-sixty.worktrunk && git-wt config shell install ``` ### Install via AUR (Arch Linux) ```sh paru worktrunk-bin && wt config shell install ``` </pre> <p>View the full release notes at <a href="https://github.com/max-sixty/worktrunk/releases/tag/v0.61.0">https://github.com/max-sixty/worktrunk/releases/tag/v0.61.0</a>.</p> </details> <hr> See merge request: Harmonybrew/homebrew-core!12090
Migrates the
wt switchpicker from skim 0.20.5 (tuikit backend) to skim 4.8.0 (ratatui/crossterm), removes thevendor/skim-tuikit/patch tree we carried against the old line, and makes the full test suite green under the new backend.Why the vendor tree goes away
We vendored skim-tuikit for two patches:
alt-<digit>key parsing and anOutput::flushwrite_allfix for dropped bytes under PTY pressure. Both are moot in 4.8. Key parsing is native —binds.rsroutes every single character (digits included) throughKeyCode::Charrather than a hardcodedalt-a..alt-zarm list. The flush fix is subsumed by stdlibBufWriter::flushin the ratatui backend, which loops over partial writes correctly. So the migration carries zero vendor patches against skim, and the build machinery that kept the vendored source alive (Taskfilevendor-diff, flakevendorSrc/extraDummyScript) is deleted too.The picker rewrite
skim 4.x is a different API.
WorktreeSkimItem::display()now returnsratatui::text::Lineinstead of anAnsiString. alt-r removal rides skim'sreload(remove {})token (the selected row'soutput(), expanded into the reload command) instead of the oldas_any().downcast_refpath (which never worked across compilation units) plus a signal file. Preview-tab switching and the progressive list redraw are driven bySkim::event_sender()+Event::Render/Event::RunPreview.The regression that motivated the bulk of this
skim 4.x renders on demand. There is no 100ms tuikit heartbeat re-rendering the frame, so the progressive
wt switchlist rendered blank while collection ran in the background — the original report. The fix pokesEvent::Renderfrom the list-collect callbacks (throttled to 16ms). The same on-demand model surfaced four more regressions, all fixed here and verified head-to-head against the 0.20 picker in a terminal: preview-tab lag, current-row highlight, Shift-Tab back-cycle (crossterm reports Shift-Tab under three distinctKeyEventshapes, so all three are bound), and alt-r removal (4.xexecute-silentis fire-and-forget and raced its reader).Help text and the test harness
Two things the integration suite caught that the unit tests did not (the prior pass ran
--lib --binsonly):clap/unstable-markdown, which renders worktrunk's///doc-comments as styled help. skim 4 withdefault-features = falsedrops it, so help reverted to raw text ([experimental]→\[experimental\]). Restored by depending onunstable-markdownexplicitly, keeping help output identical and thetest_help/test_step_alias/test_docs_are_in_syncsnapshots passing unchanged.ESC[6n) at startup in partial-height mode and blocks inselect()for the reply;portable_ptyis a bare PTY and never answered, so everyswitch_pickertest failed init with "Cursor position detection timed out." The Unix harness now answers the query, mirroring the existing ConPTY responder. skim 4.x also draws the list/preview separator one column left, so the snapshot panel-split columns shifted to match. The regenerated picker snapshots show two cosmetic changes: the match counter no longer overlaps the preview tab header, and the HEAD column shows the full short-SHA instead of a truncated one.Minimal-versions CI
The skim 4.8 dep tree pins tighter floors than our manifests declared, so the nightly
minimal-versionsjob needed updating. It now runs-Z direct-minimal-versions— minimize only our own direct deps and let transitive crates resolve normally — and the manifests raise each under-specified floor to the minimum the workspace builds against (largely mirroring skim 4.8's own requirements). Full-Z minimal-versionswould instead drag in skim's transitive TUI/image stack (ansi-to-tui, ratatui'sinstabilitymacro, color-eyre, ratatui-image → image/avif), whose crates under-declare their floors and don't compile at the picked versions; direct minimization confines the check to floors we own, so no transitive pins are needed. Normal resolution (the committedCargo.lock) is unchanged. Full floor list and thesignal-hooklibc-pin detail are in theci(min-versions)commit message.Reviewing
Start at
src/commands/picker/mod.rs(therun_skimentry point, the action keybinds, andparse_reload_remove_token), thenprogressive_handler.rs(the render pokes) anditems.rs(theLine-baseddisplay()). Test-harness changes are intests/common/pty.rs(theESC[6nresponder) andtests/integration_tests/switch_picker.rs(the panel-split columns). The rest is the dependency swap, the vendor and build-machinery deletions, and regenerated snapshots.All 3980 tests pass (
cargo run -- hook pre-merge --yes), including theshell-integration-testsPTY suite that exercises the picker end-to-end across the list, previews, scroll, create/remove, and accept flows. No automated test drives a real terminal; the interactive surface was also checked by hand against the 0.20 picker.