Repository navigation
ANSI printable-suppression window swallows real Tab within 50ms of Shift+Tab #5638
Description
Summary
AnsiInputProcessor's 50 ms printable-suppression window (the OnKeyboardEventParsed/ShouldSuppressFallbackKeyDown dedup of dual-reported keys) swallows a real \t keypress arriving within 50 ms of a parsed Shift+Tab (ESC[Z), because GetPrintableText () returns "\t" for both. Fast sequences like Shift+Tab then Tab lose the Tab.
Likely a regression caused by #5602 .
Found during
v2.5.0 ecosystem validation (#5630), porting tui-cs/Editor to a preview package built from #5416 head e8aba423 (2.5.0-tig-remove-cm-followup.2).
Repro / evidence
- Update to TG 2.5.0-beta.1 Editor#287: test
RawAnsi_Tab_After_ShiftTab_Reindents_Line_On_First_Keypressis skipped with this documented; re-enable after the upstream fix.
Impact
Dropped keystrokes under fast input on the ANSI driver (e.g., outdent-then-indent in an editor). Release-gating for v2.5.0 (#5630): fix or waive before Prepare Release.
Found by Claude (Fable 5) agents during the #5630 ecosystem-validation gate.
Metadata
Metadata
Assignees
Type
Projects
- Status
Activity
Harden Editor.Text CWP re-entrancy flag; add Text-hiding contract tests
[-]v2.5.0 preview: ANSI printable-suppression window swallows real Tab within 50ms of Shift+Tab[/-][+]ANSI printable-suppression window swallows real Tab within 50ms of Shift+Tab[/+]YourRobotOverlord commentedon Aug 23, 2026
I traced this through the history. The repro is valid, but #5602 is not causal: it changed only output-buffer dirty-row handling and did not touch the input pipeline.
The current 50 ms implementation came from #5618 (
b2525d126), merged six minutes after #5602. That may be the intended reference.There is a further nuance: the underlying Shift+Tab → Tab false positive predates #5618. The
OnKeyboardEventParsed/ShouldSuppressFallbackKeyDownmechanism was introduced by #4927 (e532d6101). It already behaved as follows:ESC[Zis parsed asKey.Tab.WithShift.GetPrintableText ()returns"\t"forKey.Tab.WithShift.\treaches fallback processing, matches the pending text, and is suppressed.Before #5618, that pending suppression had no expiry. #5618 narrowed it to 50 ms but retained the collision. Its coverage exercises kitty CSI-u plus legacy duplicates, not a complete legacy Shift+Tab report followed by a separate Tab keypress.
So the attribution appears to be:
The deeper problem is provenance: any parsed keyboard event can arm logic intended to deduplicate kitty plus legacy dual reports.
ESC[Zis already a complete Shift+Tab report and should not arm suppression for a following raw Tab. A robust fix should restrict arming to inputs that can legitimately be dual-reported, or preserve parser provenance; matching only text plus elapsed time cannot reliably distinguish two physical keypresses.fix(input): scope ANSI key deduplication
fix(input): require adjacent Kitty fallback
Fixes #5638. Scope ANSI key deduplication (#5644)
Update to TG 2.5.0-beta.1 (#287)