alonso.torres fcac23c52f 🐛 Keep whole-program component failure signatures stable
Clear the current tuple after final settlement and before whole-program
history checks. A skipped tail must not supply the action kind for a final
undo or redo failure. Keep the full labelled trace and the invariant/phase
in the error report; preserve immediate and deferred action attribution.

Add real passing controls and final undo/redo corruption regressions for
local, linked-library and source-editor histories. The regression failed
24 of 84 assertions before the fix. Five focused tests now pass all 121
assertions without weakening their checks. The shrink filter stays intact.
The change covers only the generator harness and its tests.

Validation: common JVM 2,015 tests / 33,268 assertions; common JS
1,951 / 31,879; frontend 1,751 / 13,976; backend 1,793 / 9,007;
exporter 16 / 40. Six properties pass 40 programs at seed 712 in
both common runtimes. Lint and format checks pass for all four modules.

AI-assisted-by: gpt-6
2026-10-30 09:50:19 +01:00
..
2026-10-30 09:48:15 +01:00
2026-09-29 09:07:34 +02:00
2026-10-29 15:43:06 +01:00
2026-10-29 15:43:06 +01:00
2026-03-11 15:24:40 +01:00