fix(tui): swallow degraded mouse-burst noise so a stalled loop can't lock the composer (#35512)

* fix(tui): swallow degraded mouse-burst noise so a stalled loop can't lock the composer

When the Node event loop blocks during a heavy render/tool-call burst, stdin
stops being drained. Mode-1003 any-motion mouse reports pile up in the kernel
buffer, get partially read, and arrive as text with the `\x1b[<` prefix AND
coordinate digits chewed off across many partial reads. The existing fragment
recovery (SGR_MOUSE_FRAGMENT_RE) only handles clean `button;col;row[Mm]`
triples, so the degraded shards leak into the composer as typed text — the user
can no longer type or exit until the stall clears.

Captured leak (Windows Terminal, during tool calls):

  M6M35;220;56M6M35;218;56M169;48M;157;47M;44M20;43M79;40M78;40M0M7M35;49;41M
  48;41M;47;40M9;15;32M[I;31M5;211;26M35;211;25M7M;220;1MM0M09;25M24M23M3;22M
  M18M99;26M32MM38M63;44M47MM1;51M M4M54M

Add two recovery layers in parseTextWithSgrMouseFragments / the text-token path:

- MOUSE_BURST_NOISE_RE: whole-text fast path. If a text token is drawn only
  from the mouse-leak alphabet (`[ ] < ; I M m`, digits, spaces) AND carries
  the structural signature of mouse coordinates (>=3 M/m terminators, a digit,
  and a `;`), swallow it wholesale.
- MOUSE_BURST_RESIDUE_RE: swallows pure-noise residue in the gaps between and
  after recovered fragments, so a partially-recovered burst doesn't trail a
  chewed-up tail into the prompt.

All three constraints together preserve real prose: `Mmm MMM mmm yummy` has no
digit/`;`, `see 1;2;3M for details` has disqualifying letters, and
`1234;56;78M9;10;11M` has only two terminators — none are swallowed.

This is defense-in-depth: it stops the leak/lockout regardless of what blocks
the loop. The underlying event-loop stall during streaming is a separate,
still-open issue that needs live-turn instrumentation to root-cause.

* fix(tui): check mouse-burst noise before fragment recovery; drop test cast

Copilot review on #35512:

- MOUSE_BURST_NOISE_RE was only evaluated when parseTextWithSgrMouseFragments
  returned null. A noise blob that contains any intact `<b;c;r M` fragment makes
  fragment recovery return non-null, so the whole-text swallow never fired and
  the code emitted a pile of recovered mouse events instead of dropping the blob
  wholesale (contradicting the comment, and doing extra work mid-stall). Move the
  noise check ahead of fragment recovery so pure-noise tokens are dropped early.
  Add a regression test for a noise blob carrying intact fragments.

- Drop the unnecessary `(e as { isPasted?: boolean })` cast in the test;
  discriminated-union narrowing on `e.kind === 'key'` exposes isPasted directly.

* Potential fix for pull request finding

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* Potential fix for pull request finding

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* Potential fix for pull request finding

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* Potential fix for pull request finding

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
This commit is contained in:
brooklyn!
2026-05-30 22:27:14 -05:00
committed by GitHub
parent 355af2c20f
commit cd067ab91e
2 changed files with 81 additions and 2 deletions

View File

@ -133,4 +133,35 @@ describe('fragmented SGR mouse recovery', () => {
expect(key).toMatchObject({ kind: 'key', sequence: '1234;56;78M9;10;11M' })
})
it('swallows a fully degraded mouse-burst noise blob without leaking prompt text', () => {
// Captured from Windows Terminal during a heavy tool-call render: the event
// loop blocked past App's 50ms flush timer, so a long burst of SGR mouse
// reports (mode 1003 any-motion) arrived as text with prefixes AND
// too degraded for SGR_MOUSE_FRAGMENT_RE (1- and 2-param remnants, a
// stray focus-in `[I`), so without the whole-text noise fast path the entire
// blob types into the composer and locks the user out.
const blob =
'M6M35;220;56M6M35;218;56M169;48M;157;47M;44M20;43M79;40M78;40M0M7M35;49;41M48;41M;47;40M9;15;32M[I;31M5;211;26M35;211;25M7M;220;1MM0M09;25M24M23M3;22MM18M99;26M32MM38M63;44M47MM1;51M M4M54M'
const [events] = parseMultipleKeypresses(INITIAL_STATE, blob)
expect(events).toEqual([])
})
it('keeps plain prose that only contains scattered M and m letters', () => {
const [[key]] = parseMultipleKeypresses(INITIAL_STATE, 'Mmm MMM mmm yummy')
expect(key).toMatchObject({ kind: 'key', sequence: 'Mmm MMM mmm yummy' })
})
it('swallows noise wholesale even when it contains intact recoverable fragments', () => {
// A noise blob can carry a few intact `<b;c;r M` fragments amid the chewed
// shards. The whole-text noise check must run BEFORE fragment recovery —
// otherwise parseTextWithSgrMouseFragments returns non-null and emits a
// pile of recovered mouse events instead of dropping the blob wholesale.
const blob = '<35;159;11M;44M20;43M0M7M<35;124;26M;47;40M9;15;32M5M2M'
const [events] = parseMultipleKeypresses(INITIAL_STATE, blob)
expect(events).toEqual([])
})
})

View File

@ -65,6 +65,34 @@ const XTVERSION_RE = /^\x1bP>\|(.*?)(?:\x07|\x1b\\)$/s
const SGR_MOUSE_RE = /^\x1b\[<(\d+);(\d+);(\d+)([Mm])$/
const SGR_MOUSE_FRAGMENT_RE = /(?<!\d)(?:\[<|<)?(?:[0-9]|[1-9][0-9]|1\d{2}|2[0-4]\d|25[0-5]);\d+;\d+[Mm]/g
// Whole-text mouse-burst noise fast path. When a heavy render blocks the event
// loop past App's 50ms flush watchdog, a long burst of SGR mouse reports (mode
// 1003 any-motion / 1006 SGR) can arrive as a single text token with prefixes
// AND coordinate digits chewed off across many partial reads. The surviving
// shards (1- and 2-param remnants, stray focus-in `[I`, lone `M`/`m`
// terminators) are too degraded for SGR_MOUSE_FRAGMENT_RE, so the leftover
// tail leaks into the composer and locks the user out (they can't type or exit).
//
// If the ENTIRE text token is drawn only from the mouse-leak alphabet
// (`[ ] < ; I M m`, digits, and the stray spaces a burst can carry) AND it
// carries the structural signature of mouse coordinates — ≥3 `M`/`m`
// terminators, at least one digit, and at least one `;` separator — swallow it
// wholesale. All three constraints together preserve real prose: `Mmm MMM mmm`
// has no digit and no `;`, `see 1;2;3M for details` contains disqualifying
// letters, and `1234;56;78M9;10;11M` has only two terminators.
// eslint-disable-next-line no-control-regex
const MOUSE_BURST_NOISE_RE = /^(?=[\s\S]*\d)(?=[\s\S]*;)(?=(?:[^Mm]*[Mm]){3})[\d;<\[\]IMm \x1b]+$/
// Residual-shard variant for the gaps BETWEEN / AFTER recovered fragments
// inside parseTextWithSgrMouseFragments. A real recovery run leaves degraded
// remnants (e.g. `M6M`, `7M;220;1MM0M`, lone `;157;47M`) that are pure
// mouse-leak alphabet but too short to satisfy the ≥3-terminator whole-text
// rule. Swallow such a residue only when it is pure alphabet AND carries a
// digit AND at least one `M`/`m` — a prose gap like ` for details ` contains
// disqualifying letters and never matches.
// eslint-disable-next-line no-control-regex
const MOUSE_BURST_RESIDUE_RE = /^(?=[^\d]*\d)(?=[^Mm]*[Mm])[\d;<\[\]IMm \x1b]+$/
function createPasteKey(content: string): ParsedKey {
return {
kind: 'key',
@ -268,6 +296,16 @@ export function parseMultipleKeypresses(
} else if (token.type === 'text') {
if (inPaste) {
pasteBuffer += token.value
} else if (MOUSE_BURST_NOISE_RE.test(token.value)) {
// Fully degraded mouse-burst noise — a heavy render (e.g. a sudo /
// secret prompt repaint) blocked the event loop past App's 50ms flush
// watchdog, so a long burst of SGR mouse reports arrived as text with
// prefixes AND coordinate digits chewed off. Checked BEFORE fragment
// recovery: a noise blob can still contain a few intact `<b;c;r M`
// fragments, and parseTextWithSgrMouseFragments would then return
// non-null and emit a pile of recovered mouse events instead of
// dropping the blob wholesale. Swallow it here so it never leaks into
// the composer (and we skip the extra fragment-recovery work mid-stall).
} else {
const mouseFragments = parseTextWithSgrMouseFragments(token.value)
@ -674,7 +712,12 @@ function parseTextWithSgrMouseFragments(text: string): ParsedInput[] | null {
}
if (first.index! > cursor) {
parsed.push(parseKeypress(text.slice(cursor, first.index!)))
const gap = text.slice(cursor, first.index!)
// Skip pure mouse-leak residue between recovered fragments; only emit
// real text gaps as keypresses.
if (!MOUSE_BURST_RESIDUE_RE.test(gap)) {
parsed.push(parseKeypress(gap))
}
}
for (const match of run) {
@ -690,7 +733,12 @@ function parseTextWithSgrMouseFragments(text: string): ParsedInput[] | null {
}
if (cursor < text.length) {
parsed.push(parseKeypress(text.slice(cursor)))
const tail = text.slice(cursor)
// Swallow a pure mouse-leak residue tail (the head fragments recovered, but
// the burst trailed off into chewed-up shards). Emit only real trailing text.
if (!MOUSE_BURST_RESIDUE_RE.test(tail)) {
parsed.push(parseKeypress(tail))
}
}
return parsed