Three static modals had no entry in the fallback Esc handler, so Esc over
them fell through and closed the editor behind (or did nothing when no
editor was open — reauth and the autofill picker could only be dismissed by
mouse).
Also scopes the stray-modal catch added in a95becd to [data-dynamic]. As
written it matched ANY open .modal, so an Esc over reauth / quick-search /
the autofill picker would have removed those static nodes from the DOM for
good, breaking the feature until restart. Only the runtime-built folder
modal carries the marker.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The folder modal is built dynamically with no id in the fallback Esc
handler's modal list, so Esc fell straight through to the slideover close
and shut the editor behind it. The capture handler already bails on any
open .modal; mirror that in the fallback by closing a stray .modal before
reaching the slideover branch. Covers every dynamically-built modal, not
just this one.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Selecting an entry inserted the batch bar as a block above the grid, pushing
every row down ~60px — the list jumped under the cursor mid-click. The bar
now renders inside .content-header, to the right of "All items", and the
header reserves its height (min-height) so appearing and disappearing shifts
nothing at all.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Covers the four classes found this session (stale editor payload, editor
dismissed without the discard gate, missing withEncryptedMeta, view-blind
UI) with the surrounding cases, not just the ones reported.
Also fixes one more instance found while auditing: deleting a folder
reassigns its entries to All server-side, but an editor open on one of them
kept pointing at the deleted folder and re-created it on Save.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Tagging or moving an entry from the batch bar / kebab / drag while that same
entry sat open in the editor was lost on the next Save: the editor still held
the payload it loaded before the change and re-shipped it wholesale.
editorAbsorbChange(id, {tag|folder}) mirrors the change into soState AND
soState.original (so the dirty check doesn't fire on an edit the user didn't
type), called from all five write paths: batchAddTag, batchMoveToFolder,
addTagToEntry, moveEntryToFolder, moveEntriesToFolder.
Two bugs found next door:
- moveEntriesToFolder (multi-drag) PUT without withEncryptedMeta. The
cleartext columns are gone server-side, so it shipped fields the server
can't store and blanked title/site/username/tags at rest.
- The kebab menu offered Pin/Edit/Duplicate/Move to trash on TRASHED rows —
the table view reaches it in trash, where only Restore / Delete forever
make sense.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The click-outside handler whitelists chrome (topbar, sidebar, pagination…)
but not .batch-bar, so Delete / Clear / Move to folder / Add tag counted as
an outside click and ran requestCloseSlideOver before their own handler.
Whitelist it, plus .content-header which carries buttons of its own
(Empty trash). Those actions now decide for themselves what happens to the
editor via closeEditorForDeleted.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Batch Delete had the bug the single ✕ just lost: it never touched the open
editor, so an entry could be trashed while its editor stayed open on it,
unsaved edits and all. Same hole in batch purge, permanent delete and empty
trash.
closeEditorForDeleted(ids, alreadyConfirmed) now owns the rule for all five
call sites: close the editor only when it holds one of the deleted ids, and
show the discard prompt only when no delete modal was shown (one modal per
action; cancelling it deletes nothing).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
With askBeforeDelete off, clicking ✕ on the entry being edited trashed it
and closed the editor with no prompt whatsoever — a misclick cost the edits
silently. That path now runs the discard gate (no-op unless dirty). When
askBeforeDelete is on, its own confirm still stands alone: one modal.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Clicking the ✕ on any row called closeSlideOver() unconditionally, so an
open editor with unsaved edits on a DIFFERENT entry vanished without the
discard prompt. Close it only when it holds the entry being trashed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The note textarea showed its plaintext to anyone glancing at the screen,
unlike every other secret in the editor. An eye button on the NOTE label
masks it with -webkit-text-security (a textarea has no type="password").
The button IS the setting: it flips state.maskNotes, which is synced via
settings_json, so the choice sticks across entries and devices — no extra
Settings row for a one-bit preference.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
.entry-meta and the first order-6 action both had margin-left:auto, so the
free space was SPLIT between them — the chips landed at a different x on
every row depending on how wide the rest of the row was. .entry-meta is
always present (empty when the entry has no chips), so its auto alone puts
the actions at the far edge and right-aligns the chips against them.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The note placeholder was still flex:1, so note rows' copy button sat at the
far edge while password rows' sat at 110px. Same fixed track for both, full
label on hover.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The list-view rule set `width: 110px`, but the base .entry-pw is `flex: 1`
(basis 0), so width was ignored and the pill stretched the full row width —
dragging the copy/open buttons to the far edge. `flex: 0 0 110px` pins it,
and the actions land right after it.
Names that ellipsise in the fixed-width title column (list) and the name
cell (table) now carry a title attr, so the full name shows on hover.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A sticky box is held so its *margin* box respects `top`, so the -24px
margin-top used to cancel .content's padding pushed the header's border box
24px down from the scrollport edge — rows scrolled through the strip above
it. Move the top padding off .content and onto .content-header instead: same
spacing at rest, opaque box flush with the scrollport when stuck.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
"All items · N items" scrolled away like the pager did. It now sticks at
the scrollport top (padding-top/negative margin cancel .content's padding
so nothing shows through the gap), and the three things below it offset
by --hdr-h: batch bar, pagination, and the table header (--hdr-h + --pag-h).
Both vars are measured on .content in renderEntries.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The page buttons sat at the top of the grid and scrolled away, so paging
through a long list meant scrolling back up. Sticky at the scrollport top
with an opaque background; the table header now sticks below it via
--pag-h (measured after append, so a wrapped pager still lines up).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The gear hugged the left edge of the (width:100%) actions column because the
header th, unlike the body cells, has no col-actions class / text-align:right.
Set textAlign right on the actions header cell so the picker sits in the
top-right corner.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The show/hide-columns gear only makes sense in table view but sat in the
global toolbar (needing applyColsWrapVisibility to hide it elsewhere). Build
it inside the actions-column header cell in renderTable instead, with inline
handlers (recreated each render, so no dependence on once-wired listeners on
a static node). Removed #colsWrap from the toolbar + the now-obsolete
applyColsWrapVisibility and its callers; kept the menu id so
renderTableColsMenu still populates it, and a live-queried outside-click close.
Added .cols-wrap CSS (was an inline position:relative on the old node).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Part A — Ctrl+Shift+A (new entry from window title) is now editable like the
other 3 global hotkeys, following the same pattern end to end: state.newEntryHotkey
(synced), a capture button in Settings > Autofill, dup cross-check, reset-to-
defaults, autofillPushHotkeys, and Bridge.setAutofillHotkeys carrying ne_mods/
ne_vk. Delphi gets SetNewEntryHotkey + stores the live combo (FNewEntryHotkeyMods/
Vk) so ExecuteAutofill's post-fill re-register uses the user's combo instead of
resetting to Ctrl+Shift+A. UMainForm parses ne_* unconditionally (works with
autofill off).
Part B — new-entry from the tray pre-filled the title with "Password Manager":
GetForegroundWindow resolved to our own hidden window. Only capture the title
when the foreground window belongs to another process; otherwise pass '' (blank).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Ctrl+Shift+P (password autofill) with Ctrl+Shift physically held triggered
new-entry: the clear-field step injects a real Ctrl+A, and the held modifiers
(re-pressed by keyboard auto-repeat faster than ForceReleaseModifiers clears
them) make the system see Ctrl+Shift+A = our new-entry global hotkey. Injected
keys are only Ctrl+A + Tab, and only 'A' matches a hotkey, so unregister
NEW_ENTRY around the SendInput (try/finally re-register) — no WM_HOTKEY is
generated. Password/username go as VK_PACKET unicode, can't chord.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
viewMode is a synced setting. On launch localStorage is wiped (ephemeral port
= new origin) so state.viewMode defaults to 'cards' and applyViewMode()
highlights the cards button. loadServerSettings then set state.viewMode='table'
and re-rendered the content (table) but never re-synced the .view-btn
highlight or the column-picker — so table content showed under a cards-active
toolbar. Re-apply both in the visual-settings block, alongside theme/sidebar.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
6c-1 removed the client caller, but the server handler stayed armed: a stray
authed+CSRF'd call could still re-encrypt only 1 blob of 8 (data loss) or
rewrite a -v2 account's hash_algo (permanent lockout, 6c-4). Gutted the ~165
destructive lines and replaced the body with an auth+CSRF gate then HTTP 410.
Both risks are now closed server-side, not merely unreachable by convention.
The vestigial kdfMigration login signal is left (client ignores it, endpoint
410s) — removing it would cascade through a login-path param for no benefit.
Fixed the now-false comments pointing at migrate-kdf.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
SQLite defaults foreign_keys OFF per connection, and the schema relies on
ON DELETE CASCADE (entry_attachments, entries_password_history, recovery_keys).
FireDAC is said to default it On but that's undocumented here — set
ForeignKeys=On so a permanent delete never orphans encrypted blobs. Correct
in both cases (no-op if already On). Base check before fix: 0 orphans.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
RequireCSRF looked up csrf_token by user_id ORDER BY created_at DESC LIMIT 1
— the newest session, not the one whose Bearer token was presented. Two
consequences: created_at's 1-second resolution made two same-second sessions
a coin-flip (phantom 403s), and conceptually it accepted "any valid CSRF for
this user" rather than the one bound to this session. Now queried by
token_hash = SHA256(bearer) AND user_id. Authenticate already located this
row + checked expiry, so the expiry clause is dropped. Signature unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
CreateSession wrote FormatDateTime(IncHour(Now,24)) = local, but RequireCSRF
and CleanupExpired compare against SQLite datetime('now') = UTC. On negative
UTC offsets the CSRF window fell short of the 24h auth window: between the two,
GETs authed fine but every write got 403 "Invalid CSRF token". Now
IncHour(NowUTC,24) on write + LExpires < NowUTC in Authenticate, so all three
readers are UTC-consistent. Last FormatDateTime(...,Now) in the backend —
closes §2.2 for real. NowUTC + DateUtils already in uses.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
/migrate-kdf changed the AES key (100k->600k) but only re-encrypted
encrypted_password, leaving totp_secret/custom_fields/*_enc + attachments
under the old key = permanent data loss. Both live accounts are already at
600k so it never fired, but the path was armed and auto-triggered by the
server's login signal.
Removed the two triggers (doLogin + unlock) and the now-dead runKdfMigration.
Legacy <600k accounts upgrade safely via Change Master Password instead
(re-encrypts everything + moves to Argon2id). The Delphi endpoint is left
inert: without the client orchestrating the key swap it can't corrupt
anything, so no untested server change. Closes 6c-4 too (unreachable now).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
List view faked a table with flex+order, but the title used flex:1 so it
absorbed all free space — and since trailing content (tags/folder/actions)
varies per row, the fixed-width pw pill landed at a different x on each row.
Give the title a fixed 240px track so pw starts at the same x everywhere;
move the free-space absorption to margin-left:auto on the trailing group
(entry-meta + order-6 actions) so they still float to the far right. Copy +
globe stay in .entry-pw-row (order 4), pinned next to the pw.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Swept every user-facing surface (index.html, all js strings/toasts/dialogs,
Delphi tray/balloon/dialog text) — no French in the UI. The remaining French
is project docs (CLAUDE.md/CODE_AUDIT.md), which is the doc language, not a
mix to fix. i18n stays a future feature, not a bug.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Generator, Authenticator, TOTP generator, Vault health, Audit log, Import,
Export and Settings are now reachable from Ctrl+K. View-based tools reuse
the sidebar buttons' click handlers so their cache invalidation rides along.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Explicit opt-out for users without Quick Unlock who don't want to retype the
master password after every sleep. Default ON = exact historical behavior
(lock on WTS lock/suspend, with the documented Quick Unlock exemption —
DPAPI already gates access via the Windows account). Synced setting,
Settings > Security.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Cheatsheet open -> Ctrl+K -> Esc closed the cheatsheet UNDER the palette:
the priority chain tested cheatsheet/history before cmdPalette. Palette now
sits right after confirmModal (still the absolute top) in the one-surface-
per-keystroke chain.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Ctrl+Shift+A with Settings open created the entry panel BELOW the Settings
panel — invisible. openSlideOver now closes Settings first (root fix: covers
every editor-open path, not just the hotkey), placed after the discard guard
so cancelling keeps Settings. The inverse — opening Settings over the editor
— stays as-is by design (Esc: Settings first, editor second).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Same class as the reported import bug, other door: POST /entries always
stamped now, so an entry arriving on a device via sync lost its original
creation date (and advertised a fake edit via fresh updated_at).
HandleCreateEntry now honours body timestamps like bulk-import (absent ->
now); applyRemoteSnapshot already ships them through encryptImportEntry,
normalized to the DB format. Regular saves/duplicates send none - unchanged.
password_changed_at shares the created_at param, which is faithful: the
password is at least as old as the entry.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
vault_entries uses SQLite's space-separated UTC format everywhere, and both
sorting and sync last-write-wins compare the strings lexically — so a foreign
JSON import carrying strict-ISO 'T'/millis/Z/offset timestamps would slot in
with a different format and subtly break ordering and merge arbitration.
normalizeImportTimestamp converts any ISO-ish variant to 'YYYY-MM-DD
HH:MM:SS' UTC (bare strings treated as UTC, garbage -> '' = server stamps
now). +1 unit test (69 total).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
soSave had no re-entrancy guard: a second click while the first run awaited
encryption/POST ran the whole save again -> two POSTs, two entries. Wrapped
in a soSaving latch (same class as the sync-button guard); body moved to
soSaveInner so every early validation return releases the latch via finally.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Same Esc fall-through class: the menu's Esc handler was bubble-phase and
didn't stop the keystroke, so the slideover capture handler fired first and
popped the discard prompt while the menu also hid. Capture + stopPropagation,
gated on the menu being visible; registered before the slideover handler
(installCustomContextMenu runs at the top of init) so ordering is guaranteed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Hunted the classes behind the recent quick-search fixes across the codebase:
- Esc priority: the command palette is .cmd-palette (not .modal), so the
slideover's capture-phase Esc handler didn't see it and popped the
discard-confirm UNDER the open palette. Capture handler now also yields
to the palette.
- Esc fall-through: the fallback Esc branch closed palette + slideover +
entry modal + generator ALL on one keystroke. Now closes exactly one
surface per keystroke, topmost first.
- Modifier chord: the real-VK Tab between username and password becomes
Shift+Tab if the user holds Shift mid-fill -> focus moves backward and
the password lands in the username field. ForceReleaseModifiers before
SendVKey(VK_TAB).
- Stale placement: RestoreFromTray replayed the MinimizeToTray snapshot for
a merely taskbar-minimised window, teleporting it to the last tray-hide
position. Snapshot now replays only on a genuine tray return (captured
before Show flips Visible); iconic windows use Windows' own placement.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- The qsInput Esc handler didn't stopPropagation, so the SAME keystroke fell
through to the document-level Esc handlers which, seeing the modal now
closed, also closed (or discard-prompted) the dirty slideover behind it.
- autofill/cancel now hands the foreground back to the saved target HWND:
Esc after a fill-mode hotkey returns the user to the window they came
from instead of leaving our app focused. Locked-vault path still ends
focused on us (cancelAutofill runs before focusApp).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
wasHidden counted IsIconic too, so a window merely minimised to the taskbar
("visible" to the user) was treated as tray-origin: Esc/pick sent it to the
TRAY, vanishing from the taskbar it came from. wasHidden is now strictly
"came from the tray" (Visible=false); an iconic window gets restored by the
hotkey and simply stays open when the modal closes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Opening a fresh new entry while a dirty unsaved one was open skipped the
discard prompt: both soState.id and id are null, so soState.id !== id was
false and "switching" never triggered. Existing->anything worked (ids
differ). OR in the null/null case.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The "clears in 30s" suffix was hardcoded in ~14 toasts. Single helper
clipClearSuffix() reads clipboardClearSeconds (''=Never); prefixes concatenated
to it. Cheatsheet text made generic "(auto-clears)".
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Single choke point: copySecure overrides any positive clearAfterMs with the
clipboardClearSeconds setting (0 = user disabled). The 15 call sites keep
passing 30000 unchanged — positive just means "auto-clear this secret";
explicit 0 (username copies) still never clears. Synced setting + a select in
Settings > Security (Clipboard privacy). Win+V history exclusion is
deliberately NOT exposed — a password manager must not offer to leak into
history / cloud clipboard.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Copy a password (30s auto-clear armed), then copy something else from
another app before the timer fires: the tick emptied the clipboard anyway,
destroying the user's newer content. Guard with the Win32 clipboard sequence
number: SetText snapshots GetClipboardSequenceNumber, ClearIfOurs only
empties when it hasn't moved. Applied to the auto-clear timer AND the
clear-on-minimize path (same bug class); the explicit JS clipboard/clear
command stays unconditional (user-initiated).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Residual new-entry trigger (1 in 6): password chars go out as
KEYEVENTF_UNICODE (VK_PACKET, can't match a hotkey) — the real chord risk is
the Ctrl+A clear-field, which sends a real VK_A. If the user re-presses
Ctrl+Shift mid-sequence, that VK_A becomes physical Ctrl+Shift+A = our own
new-entry hotkey. ForceReleaseModifiers now runs inside
SendSelectAllAndDelete, at the risky instant, not just once up front.
Balloon: was gated by "Show tray notifications" (OFF for this user) — now
gated by its own synced setting "Tray alert when autofill is blocked"
(autofillFailBalloon, Settings > Autofill, default ON), carried as notify=0
on cmd://autofill/execute. ShowBalloon no longer gates internally; each
caller applies its own setting.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Root cause of "Ctrl+Shift+P opened the new-entry modal": if the user still
holds Ctrl+Shift when WaitForModifierRelease times out (1s), every password
letter is typed as a Ctrl+Shift+<letter> chord — garbage in the field AND it
fires our own global hotkeys (a password containing 'a' triggers Ctrl+Shift+A
= new entry). ForceReleaseModifiers now injects KEYUP for any still-held
modifier before typing.
Also: when the fill is blocked (elevated target) while the window is hidden
in the tray, the in-app toast is invisible — show a tray balloon instead.
ShowFirstTimeBalloon generalized into ShowBalloon(title, text, warning),
gated by the existing "Show tray notifications" setting.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Ctrl+A + Del sent before each field misbehaves on targets where Ctrl+A
isn't select-all (terminals, some remote desktops). New synced setting
(autofillClearField, Settings > Autofill) gates it: JS appends clear=0 to
cmd://autofill/execute when off; ExecuteAutofill wraps the three
SendSelectAllAndDelete calls behind AClearFirst. Absent param = ON, so
existing behavior is unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
autofillFillEntry (Ctrl+Shift+L/P path) still toasted "Password filled:"
optimistically alongside the honest UIPI failure toast. Route it through
autofillPendingToast / Bridge.onAutofillResult like the quick-search path,
and label with entryDisplayName (site is often empty -> "filled:" + nothing).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
r-zakarya