0.0.10 cohort — progress you can ship with

Published 2026-09-23 through crates.io Trusted Publishing. The cohort is core 0.0.6, ext 0.0.8, art 0.0.8 and CLI 0.0.10; see Publication. The previous 0.0.9 cohort is also published.

Independent package versions

Package	Version	Release role
rs-rich	0.0.6	Progress time/rate/spinner columns and task model, theme stack and theme files, ~~~ strikethrough parity, Spinner/Status animation state, Syntax/JSON measurement
rs-rich-ext	0.0.8	Real error chains, fill-region layouts, explicit Overflowing wrapper
rs-rich-art	0.0.8	ANSI16 and grayscale palettes, quadrant blocks, stretch fit, size caps, brightness/contrast/gamma
rs-rich-cli	0.0.10	Multi-file debounced --watch, the new image flags, toml 1.x

Internal requirements are exact 0.0.x pins, so the four move together and cargo tree holds one rs-rich.

What changed

Each item links the pull request that carries its evidence. The 0.0.10 plan records the differences from plan.

• Progress (#6, #383 via #496): an injectable clock (upstream get_time), task start/stop/finish times, the 30-second speed window, the task-driving API, and TimeElapsed, TimeRemaining, TransferSpeed, FileSize, TotalFileSize, Spinner and TaskProgress columns. Progress::new() now uses upstream's default columns, which include time remaining. Golden: progress_time.tsv.
• Theme stack (#3, #231): Console::push_theme, pop_theme and use_theme (a guard that pops on drop), plus Theme::from_file/read/config for upstream theme files. Golden: theme_stack.tsv.
• Strikethrough (#9, #381): tilde runs pair as markdown-it pairs them, so a ~~~x~~~ b renders a ~ + struck x + ~ b. Golden: markdown_strike.tsv.
• Spinner and Status (#15, #496): the first render starts the animation, update works mid-animation, and text is markup. Goldens: live_status.tsv.
• Measurement and overflow (#149, #436): upstream __rich_measure__ for Syntax and Json, and rich_ext::layout::Overflowing with one explicit policy. Golden: measure.tsv.
• Layouts and diagnostics (#134, #146, #151, #436): layout leaves fill their region as upstream Layout does, and Diagnostic::from_error reports real error chains instead of [cycle].
• Image modes (#125, #126, #199, #437): ANSI16 and grayscale palettes with every dither, quadrant blocks (also for --diff), stretch fit, --image-max-width/height, and brightness, contrast and gamma in a fixed order. With the new options unset, 149 existing invocations were byte-identical to the previous binary.
• Watch (#139, #438): several local files, one Live region each, file events from notify with a per-file debounce and a polling fallback.
• Parity tooling (#34, #450): Table, Rule, Padding and Align differential generators and a nightly 20,000-case run. Its first findings are filed as #442–#449.
• Release hardening (#201): crates.io Trusted Publishing, Node 24 action majors, icy_sixel 0.7 and toml 1.x.

Migration

• Status::renderable returns &Spinner. A spinner built fresh for every frame now always shows its first frame; keep one spinner for the whole animation.
• Progress::new() adds a time-remaining column. Pass .columns(…) to keep the old set.
• ext layouts with height-aware leaves (such as Panel) now fill their region. Wrap a leaf in .content_height() for the old natural height.
• Exhaustive matches need the new RichError::{ThemeStack, ThemeConfig}, ImageMode::Quadrants, ImageFit::Stretch, ImageColorMode::{Ansi16, Grayscale} and ImageArtError::InvalidAdjustment variants. ImageTransforms is no longer Eq; struct literals need ..Default::default().

Release test (2026-09-23)

Run on the integrated tree (main through #231, plus this branch's release fixes), with Rust 1.98.1 stable and NO_COLOR unset.

Check	Result
python scripts/validate_release.py --tag rs-rich-cli-v0.0.10	All 15 steps pass: fmt, three Clippy configurations, 908 Rust tests (0 failed), lean CLI tests, 36 Python release tests, version and CLI-reference checks, locked build and check, staged packages, tag plan
capture_golden.py against real rich 15.0.0 (dedicated venv, no rich-cli)	All 16 fixture files regenerate byte-identically
release.py plan per tag	rs-rich-v0.0.6, rs-rich-ext-v0.0.8, rs-rich-art-v0.0.8, rs-rich-cli-v0.0.10 each select exactly their own package
release.py plan v0.0.10	Rejected: rs-rich: manifest says 0.0.6, tag says 0.0.10 (and ext, art)
crates.io API	404 for all four new versions; 200 for the published 0.0.9 cohort (core 0.0.5, ext 0.0.7, art 0.0.7, CLI 0.0.9)
Existing tags	None of the four new tags exist; every 0.0.9-cohort tag is annotated on the remote
Consumer install	cargo install from the packaged rs-rich-cli-0.0.10 source, with siblings patched only to their packaged .crate contents: release build succeeds, rich --version reports 0.0.10
Library consumer	A fresh crate pinned to =0.0.6/=0.0.8 packaged core/ext: theme stack guard and base-pop error, a ~~~x~~~ b strikethrough, Progress speed/remaining on an injected clock (checked against rich 15.0.0), Spinner start-at-first-render, Overflowing crop
CI scripts against the final optimized binary	test_cli_terminal.py (3), test_demo_pty.py (5 process checks), test_batch_v9.py, snapshot_cli.py (6 workflow snapshots), test_watch_workflows.py (12), test_live_regions_pty.py, and the differential corpus (21 of 21 match rich 15.0.0)
Docs	mkdocs build --strict passes; gen_versions.py --check and gen_cli_reference.py --check match

The patched consumer stood in for registry resolution before publication; the release runs' exact-version registry consumers have since passed (see Publication).

What the release test found and fixed

• Stale staged packages (release tooling). The first gate run failed: rs-rich-ext was verified against a core 0.0.6 that lacked fit_to_measurement. Cargo's staging registry had reused the core 0.0.6 it unpacked when #201 was packaged, and the build of it, because it treats registry versions as immutable. The same trap can also pass a tree that uses a removed API. check_packages.py now drops cached copies of staged-only versions before verifying; a real-Cargo regression fails without the fix.
• Watch retry hint. With several files and --watch-exit-on-error, the failing region still said "watch will retry after the next change" just before the watch exited. It now matches the single-file path; the PTY test asserts it.
• Demo. The guided tour now shows 0.0.10: a pushed theme, ~~~ strikethrough, a two-file watch with Live regions, plus the Progress and art sections the feature PRs added. The watch ends on its own through --watch-exit-on-error.

Publication

Independent, annotated tags on main at e604cc5 (#200). Each release run published through Trusted Publishing (environment crates-io) and then built an exact-version registry consumer. No coordinated v0.0.10 tag exists; the release script rejects it.

Tag	Release run	crates.io
rs-rich-v0.0.6	35818053978 (attempt 2)	rs-rich 0.0.6
rs-rich-ext-v0.0.8	35818015980 (attempt 2)	rs-rich-ext 0.0.8
rs-rich-art-v0.0.8	35818038760 (attempt 2)	rs-rich-art 0.0.8
rs-rich-cli-v0.0.10	35818067984 (attempt 2)	rs-rich-cli 0.0.10

Tag-order incident. All four tags were pushed within about 40 seconds, with rs-rich-ext-v0.0.8 first. Its first attempt failed the pre-publish dry run because rs-rich ^0.0.6 was not yet on crates.io; nothing was uploaded. The other runs queued behind the crates-io-release concurrency group, and GitHub keeps only one pending run per group, so the queued runs were cancelled. Recovery re-ran the tag runs one at a time in dependency order (core, then ext and art, then CLI). Manual workflow_dispatch runs started during recovery were superseded and cancelled before publishing. Every version was uploaded exactly once. Pushing tags in order, waiting for each run, avoids this; see Registry authentication.

The old CARGO_REGISTRY_TOKEN repository secret is no longer used and should be deleted.

Installed-binary screenshots

Each image is a real PTY run of the installed consumer binary (TERM=xterm-256color, truecolor); the captured bytes are replayed verbatim into xterm.js in headless Chromium. Only the caption line is added. Hashes of the raw PTY bytes, PNGs, binary and packaged crates are in provenance.json; reproduce with scripts/terminal_shots/make_cases.py --release 0.0.10.

In the ANSI16 image, the letterbox column is the terminal theme's colour 0 (grey in xterm.js), not black: ANSI16 output follows the user's theme by design. The watch capture ends with SIGINT (exit 130); its raw bytes show the first frame drawing both files, the edit repainting only status.json, and the final frame kept on exit.

The guided tour recording is also re-recorded from the 0.0.10 build.

Issue disposition

Issue	State after merge	Suggested disposition
#3 theme stack	Closed	Done
#15 parity harness	Closed	Done
#125 image colour	Closed	Done for this slice; Atkinson, perceptual distance and Braille/Sixel/GIF colour modes were deferred in the plan
#134 layout constraints	Closed	Done
#139 watch	Closed	Done
#126 image transforms	Open	#437 said "Closes", but GitHub closes only the first issue in a list. Alpha/checkerboard modes were deferred in the plan: keep open, or close and file the remainder
#199 quadrant blocks	Open	Delivered by #437: close
#146, #151 diagnostics	Open	Delivered by #436 (the chain fix, snapshots and depth tests): close
#149 overflow policies	Open	Delivered by #436 as the maintainer chose: close
#6 Progress	Open	Keep open: RenderableColumn, Live-driven Progress, track() and the pulse bar remain
#9 Markdown	Open	Keep open: inline styling in table cells remains
#34 differential fuzzing	Open	Generators and the nightly shipped (#450); keep open while #442–#449 are fixed, or close and track those
#144 art in the CLI	Open	Keep open for future renderer and backend APIs
#198 tracker	Open	Published: close

Known gaps

The nightly differential run will report #442–#449 until they are fixed: Padding, Align, Rule, Table header, printed-Text joining, tabs with full justify, emoji matching and a zero-width Panel. They are pre-existing core divergences, not regressions in this cohort. Windows is covered by CI (windows-latest), not by this local run.