Busy look, stopped-engine tone, focus when no retry is left, board 30
deviation fixed, cleanup name check, write-budget row as R4's own, row C
built by R4 pending the owner's validation, exact results of the wave.
- CHANGELOG: the six shell failures split into the four Target crashed
of BLK-015 and the two A0 lookup timeouts (counted as BLK-015, not A/B
proven); pgTAP 014 / 021 not proven by a run without M9; the intro
names the R3 entry's branch
- owner trial 4.15: steps strictly in order (step 5 shows what two
sheets prove, the move back to the top is step 9's last action), how
to open the working copy, the trimmed text, no Claude sentence, the
re-save sentence removed
- deploy order (database first, then the application) and its
fail-closed consequence in the SQL checkpoint and SUPABASE.md
- design review, code review and security gate results, the fix wave
run, the owner decisions left open; verifier, check:all and the owner
trial still pending
- TODO: fix wave line, three follow-ups
V1-2-6_U1_CORRECTIONS §R3 as built (A7 split superseded by D2, exact
results, deviations), the SQL checkpoint applied with corrections F2 to
F4, the owner's answers and the date limits F7, M9 in §12.3, db:sandbox
migrate, the sheet list spec and its named failures, class C rows, the
design deviations, CLAUDE.md and CHANGELOG. Gates pending.
« Aller à » a finding shows its row and opens its dialog in view mode.
E2e helpers wait for the history to settle after a layer or section
pop; the layout walk waits out the write budget. Label rows for the
nine new keys.
A shown tile the search misses carries « Ne correspond pas à la
recherche » under its header (aria-describedby). The no-result notice
now needs no shown tile at all. Flip test watchers are stopped in
afterEach, so a failing assertion never leaves the setters patched.
Draft tile gets « Annuler l'ajout »; Escape cancels from non-text targets and
from a pristine text field, never while composing or already handled. Nothing
reaches the document, history untouched, focus back on the add button.
Final review fixes. « Terminer » and every panel close re-committed the
edited field against a stale value, adding a second identical history
entry so the first « Défaire » seemed to do nothing: the tile now reads
the latest applied value (tests assert the undo-length delta). A tile
leaving the search on « Terminer » stays pinned until the next view
change and keeps focus; « Ajouter la référence » resets search and sort
and focuses the new tile. Also: refused draft dropped on a read-only
flip is announced; arrow keys respect a refused tab switch; pinned
index kept across read-only; the panel announces « Références citées »;
search text inserted literally; « Aller à » names its citer; the edit
header's name starts with « Terminer. »; undo removing the focused panel
tile moves focus per P20; no-result announced once; docs corrected.
Gates: typecheck, lint, test:run (app 1510), build, e2e:ui 251;
e2e:shell 16 specs (admin and editor-equine excluded): 117 passed / 108
skipped / 1 failed (crossing:211 stuck on the MFA step-up screen before
the editor), rerun alone 11 passed; full-row guard: no pre-existing row
changed in any table.
V1-2-6_R2_LABELS.md: the 45 i18n keys R2 adds, changes or removes
(FR, provisional EN, where, status) with the ambiguities and the
pending add-draft cancel question. RUST_ARCHITECTURE_REVIEW Part II:
class C rows for the display-only reference modules (BDR-114 A).
U1_CORRECTIONS §R2 status; BROWSER_TESTS: the two new specs, five
tabs, editor-equine never run, the admin.spec crash risk;
OWNER_TRIAL §4.10 db:trial protocol for R2 and stale four-tab /
old-panel text updated; CLAUDE.md tree.
admin.spec opened the working-copy dialog on the historical sheet, whose
single slot was taken by a copy the sandbox suite leaked. The dialog checks
move to an e2e:ui page on the real component; the suite archives its own copies.