A plain <Box> (no explicit flexDirection) lays out as a row — yoga's Box
default is FLEX_DIRECTION_ROW (host/yoga.ts) — so its screen-reader children
must be joined with a space, matching Ink (which hardcodes flexDirection:'row'
in Box.tsx and derives the SR separator from style).
Root cause: flexDirection is a pure yoga prop. node-ops applies it to the yoga
node but does NOT mirror it into node.props (it is not in STYLE_PROPS), so the
yoga row default was never reflected there. screen-reader.ts read
node.props["flexDirection"], which was undefined for every live-rendered box
(both default AND explicit column/row), wrongly defaulting the separator to
"\n" in all cases.
Fix: resolve the direction from the yoga node (preferring an explicit
props.flexDirection for the direct-built unit fixtures), mirroring
static-channel.ts's resolvedFlexDirection so both SR linearization paths derive
the separator identically. Root keeps undefined → "\n" (Ink's column-default
root). row-reverse/column-reverse reversal is unaffected.
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
The first exit() call now captures its value/error and initiates teardown
synchronously; subsequent exit() calls are complete no-ops, so waitUntilExit
resolves/rejects with the FIRST value rather than the last. This mirrors
Ink's handleAppExit guard (isUnmounted || isUnmounting → early return).
Previously each exit() queued a microtask that overwrote pendingExitResult/
pendingExitError before resolveExit ran, making it last-wins. An exitInitiated
flag set at the top of exit() now guards the value capture and re-resolve,
while the deferred microtask teardown (needed because exit() is called from
inside the Vue update cycle) is preserved.
Reverses the sweep-1 refutation: exit() was NOT already first-wins guarded.
Also guard unmount-in-progress (isUnmounting parity) + value→error test.
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): record sweep-4 (G32-G38) + G32 pr-open
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): recurse into nested <Transform> in <Text> squash (Ink parity, G32)
paint.ts renderTextWithInlineStyles and text-measure.ts flattenLeaves now
factor their per-child text squashing into a recursive squashTransformChild
helper that recurses GENERICALLY into transform-typed children to any nesting
depth, applying each transform with its positional sibling index and the
innerText.length > 0 guard — matching Ink squashTextNodes generic recursion
(squash-text-nodes.ts:22-39). Previously a <Transform> nested directly inside
another <Transform> (inside a <Text>) was dropped: its grandchild loop only
handled text-leaf/virtual-text/text, so a transform grandchild contributed
nothing — silent total content loss in paint and 0-width measurement (broken
layout). Paint and measure stay behaviourally identical so layout and output
agree.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): concatenate children in screen-reader <Transform> (Ink parity, G23)
The screen-reader transform branch joined children with "\n", producing
newline-separated text where Ink concatenates. In Ink a <Transform> is an
`ink-text` node, so the screen-reader path squashes it via squashTextNodes
(squash-text-nodes.ts:42), which concatenates child text with "". Join with
"" instead of "\n" to match.
The transform node's own fn is intentionally NOT applied: verified empirically
against Ink 7.0.4 that squashTextNodes only applies the internal_transform of
*child* nodes (squash-text-nodes.ts:34-39), never of the top-level node handed
to it. A <Transform> directly under a <Box> is squashed as the top-level node,
so its own transform is skipped — yielding the bare concatenated children.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G23 pr-open, reconcile G22 merged, log G23 spec correction
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): dedup screen-reader role against immediate parent only (Ink parity, G22)
Drop the `?? options.parentRole` grandparent fallback so a role-less intermediate
resets inherited parentRole to undefined for its children, matching Ink's
immediate-parent-only role dedup (render-node-to-output.ts:68-69 passes only
`node.internal_accessibility?.role`, no fallback).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G22 pr-open, reconcile G21 merged
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): pass child sibling index to nested <Transform> (Ink parity, G21)
paint.ts renderTextWithInlineStyles and text-measure.ts flattenLeaves squash
loops now pass the child's position index to child.transform instead of a
hardcoded 0, matching Ink squash-text-nodes.ts:13,38 where internal_transform
receives the plain loop counter over node.childNodes (all siblings). A nested
<Transform> that is the Nth child of a <Text> therefore gets index = N. Both
spots use the same index basis so paint and measurement agree. Refines the
earlier G06 refutation — the inline/squash path was the real gap.
Also adds the `innerText.length > 0` guard in text-measure.ts flattenLeaves so
measurement skips transforms on empty text (matching paint.ts and Ink:34), and
converts screen-reader.ts squashTextContent to use forEach-with-index so the
nested-Transform index is correct in SR mode too (was hardcoded 0).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G21 pr-open, reconcile G20 merged
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): guard writeToStdout/writeToStderr against post-teardown writes (Ink parity, G20)
Return early if teardownStarted, mirroring Ink ink.tsx:673/702, so a write
after unmount (e.g. a stray useStdout().write or console.log routed through
writeToStdout after teardown) cannot run clear()/write/restore on an
already-torn-down renderer and corrupt the restored terminal state.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G20 pr-open, reconcile G19 merged
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): reset yoga props to default on dynamic removal (Ink parity, G19)
Removed style props now reset to the yoga default (margin/padding/min/gap/flexGrow→0, flexShrink→1, flexBasis→auto, flexDirection→ROW, flexWrap→NO_WRAP, alignItems→STRETCH, alignSelf→AUTO, justifyContent→FLEX_START, position→RELATIVE) instead of keeping a stale value — matches Ink's reconciler diff + styles.ts.
The fix threads the previous prop value (prev) from patchProp into applyYogaProp so that resets only fire on genuine removals (prev is a real value, not null/undefined from Vue's initial-mount or never-set patches). RESETTABLE_PROPS is extended with all newly resettable keys.
Follow-up blocker fixes:
- marginX/marginY/paddingX/paddingY now map to Yoga.EDGE_HORIZONTAL/EDGE_VERTICAL (matching Ink styles.ts) instead of concrete EDGE_START/END/TOP/BOTTOM. They compose with the specific edges per yoga precedence, so removing an axis shorthand no longer clobbers a surviving marginLeft/etc.
- applyYogaProp and all setters now treat null the same as undefined (value == null) for the removal/reset path. Vue's host renderer passes next=null (not undefined) when a key disappears from a spread props object (e.g. Static spreads style into host props), which previously bypassed the reset and forwarded raw null into yoga (NaN/0 corruption).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G19 pr-open, reconcile G18 merged
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): restore terminal on signal exit via signal-exit (Ink parity, G18)
Previously nothing routed a process signal to teardown(): SIGINT-as-signal,
SIGTERM or SIGHUP killed the process with the cursor hidden, the alternate
screen active and raw mode on, leaving the terminal corrupted.
Mirror Ink (ink.tsx:426): register signal-exit's onExit(teardown,
{alwaysLast:false}) at interactive mount, storing the unsubscribe fn, and
call it first thing in teardown() (ink.tsx:765) so the handler is removed on
unmount()/exit() and can't leak or double-run. teardown() stays idempotent
(teardownStarted guard) so a signal-triggered teardown plus a later unmount
won't double-run, and we don't prevent the process from exiting. Only the
live interactive, non-debug mount registers — render-to-string /
non-interactive paths never touch process signal handlers; registration is
guarded against double-registration.
Uses signal-exit v4 (named onExit export; ships ESM + types, so no
@types/signal-exit needed). PTY test sends SIGINT/SIGTERM/SIGHUP to a mounted
alt-screen app and asserts the captured output ends with show-cursor
(\x1b[?25h) + leave-alt-screen (\x1b[?1049l).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Review follow-ups (3 fixes): register signal-exit whenever interactive
(drop the !debug gate so debug-but-interactive apps, which still enter the
alt-screen/hide the cursor, restore on signal — Ink ink.tsx:426); add
!teardownStarted to the registration so a spent app instance does not
re-register on a same-instance remount (the next unmount() returns early at
the teardownStarted guard before it could unsubscribe — a leak); and make
the PTY test prove the SIGNAL drove teardown (fixture never self-unmounts, so
restore bytes can only come from the signal path) with a debug-mode signal
test, an exit-anchored waitForOutput drain, and a bounded retry for the
async-flush race under saturated runners.
Review follow-ups (2 fixes): synchronous restore flush on signal — the
signal-exit teardown path now writes the restore escapes (show-cursor,
leave-alt-screen, disable-kitty) via fs.writeSync to the stdout fd so they
reach the terminal before signal-exit re-raises the signal (a buffered async
stream.write could be lost on abrupt exit); the normal unmount path keeps async
writes. Removed the config-wide retry:3 from vitest.pty.config.ts (it masked
the whole PTY suite) and scoped a retry:2 to the signal-teardown describe only,
for the residual parent-side node-pty onData read-race under a saturated runner.
* chore(parity): ledger — G18 pr-open
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): linearize screen-reader static output + drop empty-SR-frame newline (Ink parity, G17)
(a) The live static channel flushed <Static> via the 2D grid painter
(paintIsolated) even in screen-reader mode, so bordered static items leaked
box glyphs. paintStaticNode now takes an isScreenReaderEnabled flag and
linearizes fresh static children via renderScreenReaderOutput
(skipStaticElements:false) instead — matching Ink's renderer.ts:24, which
renders node.staticNode through renderNodeToScreenReaderOutput. Non-SR static
is unchanged. render.ts commit() and render-to-string.ts thread the flag.
(b) Interactive SR frames went through renderInteractiveFrame, which appends
"\n" even for empty output, leaking a spurious blank line. Ink's SR path writes
the wrapped output directly with lastOutputToRender = wrappedOutput (no appended
newline), so an empty SR frame emits zero lines. We now suppress the trailing
newline for EMPTY SR output only — matching ink.tsx:573-626 — leaving non-SR
and non-empty SR frames untouched.
Follow-up fixes (two review findings):
- renderToString in SR mode no longer DROPS <Static> output: the SR return
branch now prepends the captured/linearized static output like the non-SR
path (Ink's SR renderer returns staticOutput when node.staticNode exists,
renderer.ts:24-33).
- The SR static linearization now honors the <Static>'s resolved flexDirection
for separator + child order (read from yoga via getFlexDirection), matching
screen-reader.ts:73-82 (row/row-reverse → space, *-reverse reverses order);
the default column case is unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G17 pr-open, reconcile G16 merged
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): let per-edge borderDimColor=false override general dim (Ink parity, G16)
edgeDim now uses `?? generalDim` (nullish) instead of `|| dimAll`, so an explicit per-edge false wins — matching Ink render-border.ts:54. The five borderDimColor prop declarations in Box.ts are changed from bare `Boolean` to `{ type: Boolean, default: undefined }` so Vue does not boolean-coerce absent per-edge dim props to false (Vue only applies absent→false casting when no explicit default is provided), preserving the undefined sentinel needed for the nullish fallback.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G16 pr-open, reconcile G14 merged
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): guard against two renderers on the same stdout (Ink parity, G14)
Make skipped-mount unmount() a pure no-op: add `skippedMount` flag set in
the instance-reuse guard branch; teardown() and resolveExit() return early
when set, so no write-barrier or stdout touch reaches the owner's stream.
Strengthen test: assert app2.unmount() writes nothing to process.stdout
(catches the pre-fix empty write-barrier) and that app1 still owns the
WeakMap entry after app2.unmount() (third mount on same stdout still warns).
* chore(parity): ledger — G14 pr-open, reconcile G13 merged
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): clear+reset on terminal-width decrease during resize (Ink parity, G11)
Track lastTerminalWidth (initialized at mount via resolveSize); onResize clears
writer + resets frameState.lastOutput/lastOutputToRender when the new width is
narrower than the previous width, mirroring the logic in ink.tsx:459-474 that
prevents duplicate overlapping re-renders on terminal narrow. Width increases and
pure height changes are unaffected.
Preserve outputHeight on narrowing (matches Ink ink.tsx:462-466 which leaves
lastOutputHeight intact): zeroing it suppressed the clearTerminal path in
shouldClearTerminalForFrame for overflowing frames (hadPreviousFrame=false).
Adds overflow-path test that was RED with outputHeight=0 and GREEN with the fix.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G11 pr-open, reconcile G10 merged
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): throw a descriptive error when raw mode is unsupported (Ink parity, G10)
Previously the raw-mode acquire path silently no-opped on a stdin where raw mode
is unsupported (non-TTY / isRawModeSupported false), so using useInput on such a
stdin did nothing with no diagnostic. Ink's handleSetRawMode (App.tsx:315-327)
instead throws immediately when enabling raw mode is unsupported, with two
distinct messages (default process.stdin vs a custom stdin) both pointing at the
isRawModeSupported docs.
StdinController.acquireRawMode now throws that two-message error on
!isRawModeSupported. The unguarded useInput path surfaces it (matching Ink's
use-input.ts, which calls setRawMode(true) ungated). useFocus now guards on
isRawModeSupported before acquiring (matching Ink's use-focus.ts), so focus
degrades to a safe no-op on a non-TTY rather than throwing. releaseRawMode keeps
its no-op guard — Ink only throws when enabling.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G10 pr-open, reconcile G09 merged
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* test(runtime): assert full raw-mode error message for exact Ink parity (G10, codex)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): wrap external stdout/stderr writes in synchronized-update markers (Ink parity, G09)
writeToStdout/writeToStderr now emit bsu/esu around clear+write+restore when
shouldSynchronize, matching the render path and Ink ink.tsx:687-728. The sync
variable was already computed at mount time (render.ts:489); the external-write
functions simply lacked the wrapping. For writeToStderr, BSU/ESU go to stdout
(not stderr) because synchronized-update mode is a stdout capability — exactly
mirroring Ink's ink.tsx:717-728 behaviour.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G09 pr-open
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): draw box borders per-edge without the min-size guard (Ink parity, G05+G15)
Removed blanket `w<2||h<2` return from drawBorder; replaced it with a
`w<1||h<1` degenerate guard. Vertical sides now start at
`offsetY = top ? 1 : 0` and run for `Math.max(0, h - topRows - bottomRows)`,
matching Ink render-border.ts:133. Fixes: (G05) a 1-cell-tall box with only
side rails rendered nothing; (G15) with borderTop=false the left/right rails
were shifted one row down. Updated 4 existing snapshots that encoded the
old buggy behavior and added 3 new tests that verified red before green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G05+G15 pr-open, reconcile G03 merged, log snapshot decision
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): render linear screen-reader output in the live commit path (Ink parity, G03)
commit() previously called paint(tuiRoot) (the 2D grid painter) unconditionally
for both the non-interactive and interactive branches, so an app mounted with
isScreenReaderEnabled emitted the visual frame (box-drawing borders, padded
grid) into the live stream instead of flat linearized text. isScreenReaderEnabled
was only consulted to disable commit throttling.
Add a renderFrame(width) helper that branches on isScreenReaderEnabled: when SR
is enabled it linearizes the tree via renderScreenReaderOutput(tuiRoot,
{ skipStaticElements: true }) and wraps it with wrapAnsi(out, width,
{ trim: false, hard: true }), mirroring Ink's onRender SR branch
(ink.tsx:598-603). Both commit branches now call renderFrame() instead of
paint() directly. The non-SR path is byte-for-byte unchanged (renderFrame
returns paint(tuiRoot)). Static output continues to flush through the existing
paintStaticNode path; full SR-static linearization parity is deferred.
render-to-string.ts already used renderScreenReaderOutput and is unaffected.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G03 pr-open, reconcile G02 merged, track G17 (SR edges)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): coalesce useAnimation ticks within the render-throttle window (Ink parity, G02)
useAnimation now coalesces scheduler ticks that fall inside the current
render-throttle window and reports delta as the time since the last
RENDERED tick, so velocity-driven motion (position += speed * delta)
advances at correct wall-clock speed even when the commit throttle is
coarser than the animation interval. Previously delta was ~one scheduler
interval per committed tick, under-integrating velocity at render time.
- animation-scheduler: createAnimationScheduler(renderThrottleMs = 0)
exposes renderThrottleMs on the AnimationScheduler (no-op variant = 0).
- render.ts: derive animationRenderThrottleMs from maxFps using Ink's
Math.max(1, ceil(1000/maxFps)); 0 on debug/screen-reader/unthrottled
paths, mirroring the commit-throttle gate.
- useAnimation: tick() skips while now < nextRenderTime; on an allowed
tick delta = now - lastRenderedTime, then nextRenderTime = now + window.
Also default maxFps to 30 (Ink parity: options.maxFps ?? 30) and derive a
single renderThrottleMs that drives BOTH the commit scheduler and the
animation scheduler, so the coalescing engages on the default non-debug
path (previously it only engaged when maxFps was passed explicitly).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G02 pr-open, reconcile G01 merged, log G02 decisions
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* fix(runtime): unmount written <Static> items to match Ink (G01)
Ink's <Static> renders `items.slice(index)` and advances `index` to
`items.length` in a post-commit `useLayoutEffect`, so once an item has been
painted it is removed from the tree and its component unmounts. vue-tui kept
every Static item mounted forever: the component always mapped the full
`props.items`, and write-once was enforced only at flush time via a positional
`writtenCount` slice — the item components never tore down.
Now the <Static> component owns a `cursor` (Ink's `index`) and renders only
`items.slice(cursor)`. The renderer advances the cursor AFTER a commit has
painted the fresh items, via an `onWritten` callback registered on the host
static node — the vue-tui analogue of Ink's post-commit layout effect. This
ordering guarantees items are written before they are sliced out and unmounted,
so no item is ever lost or re-painted.
Write-once bookkeeping moved from a positional `writtenCount` to a
`writtenNodes` Set keyed by host-node identity. A single logical item expands to
several host nodes (the <Text>/<Box> plus empty text-leaf fragment anchors Vue
inserts), so a positional count mis-sliced once the cursor advanced; identity
tracking is anchor-agnostic. The shared `paintStaticNode` helper paints children
not yet in the set, records them, prunes unmounted entries, then fires
`onWritten`; render.ts, render-to-string.ts and flushStatic all use it.
Make the cursor mirror Ink fully so it can DECREASE, not just increase.
`onWritten` now SETS the cursor to items.length (was max-with-current), and a
length watch lowers it on shrink — needed because a shrink that leaves the
already-sliced children empty produces no host mutation, hence no commit/
onWritten to re-sync. Without this, [A,B] (cursor→2) → [A] → [A,C] sliced(2)=[]
and silently dropped C. paintStaticNode now always prunes and calls onWritten
(even on empty commits), painting only when there are fresh children.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* chore(parity): ledger — G01 pr-open, reconcile G12 merged
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Replace three `stdout.columns ?? 80` / `stdout.rows ?? 24` spots in render.ts with
`resolveSize(stdout).columns/rows`. The `??` guard only falls back on null/undefined,
not on 0 — so non-TTY environments where stdout reports 0 columns would collapse Yoga
layout to width 0. Ink's `getWindowSize` (utils.ts:8-23) uses a truthy guard
(`if (columns && rows)`) and a fallback chain through terminal-size → 80/24 defaults.
`resolveSize()` in useTerminalSize.ts already implements this chain; now exported and
used by the renderer. The non-TTY viewportRows → 24 branch is preserved (Ink-aligned).
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Ink derives the focus id via useMemo(() => customId ?? random, [customId])
and keys its add/remove effect on [id], so changing the id prop re-registers
the component under the new id. vue-tui captured `const id = options.id ?? …`
once at setup (and typed id as a plain string), so it never reacted.
- Widen id to MaybeRefOrGetter<string>.
- Track the current registration and re-register (unsubscribe/remove old,
subscribe/add new, re-apply active state) in a watcher keyed on the resolved
id, mirroring Ink's [id] effect. isActive handling unchanged.
Test (test-first, verified red before the fix): focus is driven purely by
focus(id) (no Tab, which would focus by position and mask the bug); changing
the id re-registers under the new value and the old id goes dead.
Also reconciles G04 -> merged (landed in #30).
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Ink's render-border.ts computes each border edge's background from
border<Edge>BackgroundColor ?? borderBackgroundColor only — it never falls
back to the Box's own backgroundColor. vue-tui's colorizeEdge had an extra
`?? bgColor` fallback, so a Box with backgroundColor but no explicit border
background painted its background onto the border glyphs too.
Drop the fallback. Background still fills the inner content area; border
glyphs are now uncolored unless an explicit border background is set.
Tests rewritten to match Ink (per maintainer's align-to-Ink policy; see
.agents/docs/parity-ledger.md Decisions log):
- add failing-first repro "Box backgroundColor does not bleed onto border
glyphs (Ink parity)"
- "wrapped text preserves backgroundColor on every content line": assert
inner rows carry bg, border rows don't (height 4->5 so text fits)
- "Box background with border fills content area": snapshot updated so
border rows have no bg
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
The test asserting process.listenerCount("beforeExit") lived in the
parallel wait-flush.test.tsx, conflicting with the branch's rule that
process-global-state tests live in *.sequential.test.* files. Move it into
leak.sequential.test.tsx (already the home for process exit/SIGINT listener-
count assertions) and extend that file's header comment. #26 (item 2).
The resize handler painted synchronously via commit() but left the commit
scheduler's pending trailing throttle timer armed. If an update was sitting
in that timer, it fired a second doCommit() right after the resize paint —
and because shouldClearTerminalForFrame clears whenever the previous frame
overflowed the viewport, the second commit emitted a duplicate clearTerminal.
Cancel the pending trailing commit before the synchronous paint; the paint
already reflects the current tree, so the pending commit is redundant.
Regression test (test-first) in throttle.sequential.test.tsx reproduces the
double-clear (2 clears) and verifies the fix (1 clear). Closes#26 (item 1).
Fifth CI run failed on ~15 resize/throttle/flush tests that pass locally. Root
cause: the GitHub runner sets CI=true, and vue-tui computes
`interactive = !isInCi && isTTY`. With CI=true, interactive is false, so the
resize listener, cursor control, and ANSI erases are never wired up — exactly
the behavior those tests assert. Locally CI is unset, so they passed (the
local-vs-CI trap, reproducible with `CI=true vp test run`).
Set env CI:"false" in both vitest configs. The PTY child helpers already force
CI=false per-spawn for this reason; the in-process suite (and the testing
harness's render()) needs the same. Verified: `CI=true vp run ci` on a fresh
checkout (no dist) now exits 0 with all 756 + 110 tests passing.
The fourth CI run failed broadly: ~15 timing-sensitive tests (throttle, flush,
resize clear-counts, rapid-input ordering) failed under sequence.concurrent on
the 4-core ubuntu runner. These assert render/commit counts driven by the
~32ms commit throttle; in-file concurrency starves them of wall-clock when many
share few cores. It passed locally only because dev machines have more cores —
the classic "works on my 12-core mac" trap.
Remove sequence.concurrent from both the main and PTY configs; keep
fileParallelism (pool: forks), which is the proven, stable win (PTY suite still
~3x faster than serial). The it.sequential/describe.sequential markers and
context-local expect become harmless no-ops under serial in-file execution.
Verified cold (no dist): vp run ci exits 0, 0 lint warnings, all tests pass.
Enable sequence.concurrent: true in vite.config.ts so the non-PTY suite runs
concurrently like the PTY suite. Stress-verified stable (8/8 at maxForks=4);
the suite drops from ~13s to ~4-5s.
Three test patterns were incompatible with concurrency; handled per cause:
- Inline snapshots (background-color, borders): the module-level `expect`
loses snapshot test context under concurrency. Fixed in place by using the
context-local `expect` (async ({ expect }) => ...), so they stay concurrent.
- Process-global state (throttle/animation-scheduler use fake timers; leak
asserts on process exit/SIGINT listener counts and live yoga nodes): a
concurrent sibling clobbers the shared global mid-test. These genuinely
require serial execution, so they move to *.sequential.test.* files with
it.sequential / describe.sequential and a header explaining why.
`vp run ready` passes.
With resize rendering now synchronous, no PTY test depends on wall-clock
timing, so concurrent execution is safe. Enable sequence.concurrent: true.
Each test already spawns its own isolated PTY subprocess (or in-process app)
with no shared state.
Verified stable: 12/12 runs green under sequence.concurrent with forks capped
to 4 (mimicking a 4-core CI runner) — the configuration that reliably flaked
before the resize fix. No it.sequential opt-outs are needed.
The resize handler routed through scheduler.schedule(), deferring the repaint
through the ~32ms commit throttle. Ink's resized() calls onRender() directly,
and a resize is a discrete viewport change that should repaint immediately —
deferring it can leave stale/overlapping content on screen for a frame.
It also made the clearTerminal-on-overflow behavior depend on wall-clock
timing: the #450 "shrink into overflow" test passed only because the throttled
resize emitted ZERO clears (its trailing timer never fired within the test's
nextTicks) and the single clear came entirely from unmount. The test asserted
the right number for the wrong reason, and the dependency on real elapsed time
made it flaky under CPU contention.
Change the resize handler to commit() directly. Now the resize itself emits the
overflow clear deterministically. Update the test to assert the clear happens
ON the resize (clearsAfterResize - clearsBeforeResize === 1) after a single
nextTick — no longer dependent on throttle timing.
The PTY suite now runs file-parallel (see vitest.pty.config.ts), so the comment
claiming "no parallelism" was wrong. The real constraint is node-pty's forks
pool, not serial execution.
The PTY suite was the CI wall-clock bottleneck, run serially via
fileParallelism:false. The original reason for serializing was a node-pty
constraint — it needs child_process.fork(), not worker_threads — but that only
dictates the pool TYPE, not single-file execution. Each test already spawns its
own isolated PTY subprocess (helpers/term.ts, run.ts: no shared ports, temp
files, or mutable globals; cwd is the read-only fixtures dir), so files
parallelize safely.
Set pool:"forks" explicitly (the real node-pty requirement) and
fileParallelism:true. Measured: the PTY suite drops ~37s -> ~13s (~3x), and the
full `vp run ci` graph drops ~41s -> ~20s cold (no task cache, no prebuilt
dist). Verified stable across 9 isolated PTY runs (incl. maxForks capped to 4
to mimic a 4-core CI runner) and 3 cold full-graph runs — 110 PTY tests pass
every time, zero flakes. testTimeout stays 15s (slowest test ~2s) to absorb CPU
contention on smaller runners.
A single top-level `vp run check:type` should fan out (-r) and fully type-check
every package, fixtures included — no separate top-level check:fixtures step.
runtime-tests' check:type now runs its main tsc then its fixtures tsc
(check:fixtures stays as a runnable sub-script). Drop the root check:fixtures
entry and its standalone step in ready.
Each package gets check:type = `tsc --noEmit` (under its own tsconfig); root
check:type = `vp run -r check:type` fans out across the workspace. Unlike vp's
tsgolint-based type-aware path, this is the standard TypeScript compiler, so it
honors each tsconfig's real project semantics.
Additive only here — vp lint still carries typeCheck; the switch-over and ready
rewiring land in the next commit. Note: testing and runtime-tests resolve
@vue-tui/runtime types from its built dist, so check:type requires a prior
build.
Replace the opaque `vp check` bundling and the ad-hoc `ready` chain (which
shelled out via `cd ... && pnpm ...`) with explicit, composable named scripts.
Root delegates; category splits live in the package that owns them:
root: check:fmt / check:lint / check:fixtures / test / build / ready
runtime-tests: test = test:integration + test:pty; check:fixtures
`ready` now composes the named scripts in dependency order (fmt, lint, build,
fixtures, test) — build runs before check:fixtures and test:pty, which need
the built dist. Drop the broken root `dev` (referenced a non-existent website
package) and the redundant per-package `check` scripts (fmt/lint already run
workspace-wide from root). Renames: pty-test -> test:pty, typecheck:fixtures
-> check:fixtures.
The root vite.config.ts lint/fmt settings cascade into every workspace
package, and a package-level lint block merges with — rather than replaces —
the root options. So the `lint.options { typeAware, typeCheck }` and the empty
`fmt: {}` repeated in runtime, testing, and runtime-tests only restated what
each package already inherits.
Remove them. runtime-tests keeps just its `ignorePatterns` (the one genuine
per-package override); typeAware/typeCheck now come from root via the merge.
Verified with `vp check`: no lint or typecheck regression, and the PTY
fixtures stay excluded from lint.
The WithChildren shim is only exercised under jsx:"react-jsx", which lives
solely in integration/pty/fixtures/tsconfig.json. Nothing in `ready` ran tsc
against that config (vp check uses jsx:"preserve" and excludes the fixtures;
pty-test only transpiles them), so a regression in the shim — children
silently rejected, or declared props silently widened away — would pass
verification unnoticed.
Add a type-only regression fixture (not a runnable PTY program; not a
*.test.tsx, so vitest never collects it) that pins both directions of the
contract: children are accepted on Box/Text/Static/Transform, and declared
props stay validated via @ts-expect-error (invalid value, wrong type, unknown
prop, and missing required props on Transform/Static).
Wire `tsc -p integration/pty/fixtures/tsconfig.json --noEmit` into `ready` via
a typecheck:fixtures script, run after build (so @vue-tui/runtime resolves
against fresh dist types) and before pty-test, so the react-jsx path is
actually enforced rather than only manually checkable.
`lint.exclude` is not a property of OxlintConfig (the correct key is
`ignorePatterns`). The invalid property failed defineConfig overload
resolution, which surfaced as a TS2769 plus a TS2321 excessive-stack-depth
error against the recursive vitest-augmented UserConfig. Using
`ignorePatterns` excludes the PTY fixtures from lint as intended and clears
both config type errors.
Bump slice-ansi@9, string-width@8, wrap-ansi@10 and add cli-truncate@6
(both the runtime dep and the pnpm catalog entry for string-width). Rewrite
wrapText truncate variants to delegate to cli-truncate, matching Ink's
wrap-text.ts: grapheme clusters (ZWJ emoji, combining marks) stay whole and
newlines are preserved. Adjust the horizontal-clip left-edge compensation in
paint.ts because slice-ansi@9 drops a straddling wide grapheme whole rather
than splitting it, so lineX must advance by the actually-dropped width.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds an early-return guard in paintNode so nodes with DISPLAY_NONE
(already set on their Yoga node) are entirely skipped during paint,
matching Ink's renderNodeToOutput behavior. Without the guard, hidden
text/borders leaked onto visible siblings at x=0.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Correctness fixes found by max-effort review of the branch diff:
- frame-writer.sync() now updates lastFrame alongside log-update's
previousOutput. Previously the two dedup layers desynced after a sync()
(the clearTerminal path), silently dropping a legitimately-changed frame
and emitting an empty BSU/ESU pair. Adds a regression test.
- scheduler: the queuePostFlushCb callback now bails if scheduled was reset
by cancel(), so a stale callback can't commit on a torn-down tree or
re-arm an uncancellable trailing timer.
- scheduler.flush() now collects multiple concurrent waiters instead of
overwriting a single resolver — fixes a hang when two waitUntilRenderFlush()
calls await the same pending commit.
- render teardown nulls mountedClear so a post-unmount app.clear() can't
write to a torn-down stream.
- test-streams getContentWrites imports bsu/esu instead of hardcoding the
escape literals.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Horizontal clipping now runs per-line AFTER transforms instead of before,
preventing Transform-widened text from escaping clip boundaries. Left-edge
clipping also uses the actual removed width to position subsequent text
correctly when a wide char straddles the boundary.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Yunfei He