Patch 0.31.5: pin transitive security floors (tqdm, idna)

backend/requirements.txt carries only loose `>=` direct constraints and
no compiled lockfile, so osv-scanner v2 resolves the transitive graph and
pins each unpinned transitive package to its MINIMUM published version —
surfacing CVEs in ancient releases pip would never install. That fired a
phantom HIGH (tqdm 4.9.0, GHSA-r7q7-xcjw-qx8q) plus an idna finding on the
OHM Patchwatch radar, even though the live deployment runs patched versions.

Add `tqdm>=4.66.3` and `idna>=3.15` to anchor the resolver to the patched
range. osv-scanner now reports zero findings. No direct-import / schema /
API / config change; deployed installs already satisfy the floors.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

CHANGELOG.md

@@ -23,6 +23,94 @@ skip versions are the composition of each intervening adjacent
 release's steps in order — no A-to-B path is pre-computed beyond
 that.
 
+## 0.31.5 — 2026-06-03
+
+**Patch — supply-chain hygiene: pin security floors on two unbounded
+*transitive* dependencies so vulnerability scanners stay anchored to the
+patched range. No direct-import, schema, API, config, overlay, or secret
+change; deployed installs already satisfy these floors, so no rebuild or
+redeploy is required to become compliant. No upgrade steps.**
+
+`backend/requirements.txt` declares only loose (`>=`) direct constraints
+and ships no compiled lockfile. osv-scanner v2 resolves the *transitive*
+graph and, for any unpinned transitive package, pins it to its **minimum**
+published version — surfacing CVEs in ancient releases that `pip install`
+would never actually choose. Two such phantoms were firing:
+
+- **`tqdm`** (pulled by `google-generativeai`) resolved to `4.9.0` (2016),
+  flagging `GHSA-r7q7-xcjw-qx8q` (HIGH) and friends — fixed in 4.66.3.
+- **`idna`** (pulled by `httpx` / `anthropic`) resolved to `3.9.0`,
+  flagging `GHSA-65pc-fj4g-8rjx` — fixed in 3.15.
+
+Adding `tqdm>=4.66.3` and `idna>=3.15` anchors the resolver to the patched
+range. The live deployment already runs patched versions (e.g. tqdm 4.67.x);
+this only corrects what the scanner *sees*. The proper long-term fix is a
+fully-compiled lockfile — tracked separately.
+
+## 0.31.4 — 2026-06-01
+
+**Patch — bug fix + UI polish: secondary buttons that were invisible on
+light surfaces now render legibly, and the RFC view's breadcrumb action
+bar is harmonized into one coherent control group. CSS-only
+(`frontend/src/App.css`); no schema, API, config, overlay, or secret
+change — a plain frontend rebuild applies it. No upgrade steps. Shipped
+from driver session 0059.0.**
+
+`.btn-link` was authored as a *dark-header* utility — white text on a
+translucent-white fill (`rgba(255,255,255,0.15)`), the established
+on-dark pattern for the app header's "Sign out". But the same class is
+reused on **light** surfaces: the RFC breadcrumb action bar
+(`RFCView.jsx`), the PR view's diff-mode toggle and "Edit title" control
+(`PRView.jsx`), the invitations and inbox modals, and the discussion
+panel. On those near-white backgrounds the buttons were white-on-white —
+present in the DOM, fully functional, but visually invisible. The
+reported symptom: on a super-draft's header, "Metadata", "Claim
+ownership", and "Invitations" looked *missing*, while the filled CTAs
+("Start Contributing", "Graduate to RFC repo") rendered fine because
+their fill carried them.
+
+The fix is root-cause, not a per-site patch:
+
+- The base `.btn-link` rule is now a proper light-surface secondary
+  button (white fill, hairline `--c-gray-300` border, `--c-gray-700`
+  label, hover darkens both). This corrects every light-surface reuse at
+  once.
+- The original translucent-on-dark treatment is preserved for the one
+  legitimate dark-surface use via an `.app-header .btn-link` scope, so
+  the header "Sign out" is unchanged.
+- The breadcrumb action bar (`.breadcrumb-actions`) normalizes every
+  action — the discuss/contribute toggle, the filled CTAs, and the
+  secondary buttons — to one height, radius, and type scale, so the row
+  reads as a single intentional control group. The bar now `flex-wrap`s
+  instead of clipping buttons off the right edge when the set is wide.
+- The diff-mode toggle's active option now reads as clearly selected
+  (filled ink) rather than relying on a weight change alone.
+
+Smooth hover transitions and the keyboard focus ring were already
+provided globally by the v0.21.0 interaction-polish layer, so this
+change adds no new motion or focus rules — it only corrects resting-state
+color/contrast and harmonizes sizing.
+
+## 0.31.3 — 2026-05-30
+
+**Patch — admin Users tab: "Last seen" reads "Never" for unclaimed
+invites. Visual/logic only in `Admin.jsx`. A plain frontend rebuild
+applies it.**
+
+An admin-created invite row showed a real-looking "Last seen" timestamp
+identical to "Signed up," implying the invitee had visited when they
+hadn't. Cause: `users.last_seen_at` is `NOT NULL DEFAULT (datetime('now'))`
+(`migrations/001_users_and_audit.sql`) and the invite INSERT
+(`invites.py`) sets neither timestamp, so both default to the
+row-creation instant; `last_seen_at` only advances on a real
+authentication. Since an unclaimed invite has provably never
+authenticated (that unclaimed state is exactly what drives the
+"PENDING INVITE" badge), the Users tab now renders **"Never"** for the
+Last-seen cell of a pending-invite row instead of the misleading
+default. Signed-up (the invite-created date) is unchanged.
+
+Upgrade steps: none. **SHOULD** deploy as a normal code deploy.
+
 ## 0.31.2 — 2026-05-29
 
 **Patch — landing (`/`) welcome panel spacing. Visual only: CSS in

VERSION

@@ -1 +1 @@
-0.31.2
+0.31.5

backend/app/invites.py

@@ -144,10 +144,16 @@ def create_invite(
     The invitee `users` row is provisioned with:
       * `permission_state='granted'` — the admin's hand is the grant;
         the v0.8.0 self-serve `pending` queue is for the other path.
-      * `last_seen_at = NULL` — the discriminator for "invited but
-        not yet arrived" per the §16 / roadmap design. Every sign-in
-        path stamps `last_seen_at` to now, so a NULL value means the
-        invited user has not clicked through yet.
+      * `created_at` / `last_seen_at` — NOT set here, so both fall
+        through to the column default `datetime('now')` (the column is
+        `NOT NULL`; see `migrations/001_users_and_audit.sql` and the
+        longer note below). The "invited but not yet arrived" state is
+        therefore NOT carried on the user row — it is the existence of
+        an unclaimed `user_invite_tokens` row, surfaced as the listing's
+        `pending_invite` field. Consumers that want a truthful
+        last-seen MUST treat a pending-invite row as never-seen rather
+        than trusting `last_seen_at` (every real sign-in path stamps it
+        to now, but an unclaimed invite has never hit one).
       * `gitea_id = NULL`, `gitea_login = NULL` — same as a v0.7.0
         OTC-provisioned user; the OAuth identity is grandfathered if
         the user ever lands through that path.

backend/requirements.txt

@@ -9,3 +9,12 @@ google-generativeai>=0.8
 openai>=1.50
 PyYAML>=6.0
 bcrypt>=4.2
+
+# Security floors on unbounded *transitive* deps (not direct imports). osv-scanner
+# resolves an unpinned transitive package to its MINIMUM PyPI version, which surfaces
+# CVEs in ancient releases pip would never actually install. These floors keep the
+# Patchwatch radar (and any scan) anchored to the patched range.
+#   tqdm <- google-generativeai ; GHSA-r7q7-xcjw-qx8q et al., fixed in 4.66.3
+#   idna <- httpx / anthropic   ; GHSA-65pc-fj4g-8rjx,        fixed in 3.15
+tqdm>=4.66.3
+idna>=3.15

frontend/package.json

@@ -1,7 +1,7 @@
 {
   "name": "rfc-app-frontend",
   "private": true,
-  "version": "0.31.2",
+  "version": "0.31.5",
   "type": "module",
   "scripts": {
     "dev": "vite",

frontend/src/App.css

@@ -34,13 +34,31 @@
 .role-owner { background: var(--c-warning-accent); }
 .role-admin { background: var(--c-accent-strong); }
 
+/* The default surface for .btn-link is LIGHT (breadcrumb bar, PR view,
+ * modals, discussion panel, inbox). It renders as a quiet secondary
+ * button: white fill, hairline border, dark label. The dark app-header
+ * reuse ("Sign out") opts back into the translucent-on-dark treatment
+ * via the .app-header scope below. (Before v0.31.4 the base rule WAS the
+ * dark-header style, so every light-surface .btn-link was white-on-near-
+ * white and effectively invisible.) */
 .btn-link {
-  color: var(--c-white); text-decoration: none;
-  background: var(--color-on-dark-soft);
+  display: inline-flex; align-items: center;
+  color: var(--c-gray-700); text-decoration: none;
+  background: var(--c-white);
+  border: 1px solid var(--c-gray-300);
   border-radius: var(--radius-md); padding: 4px 10px;
-  font-size: var(--text-base);
+  font-size: var(--text-base); cursor: pointer;
+}
+.btn-link:hover {
+  background: var(--c-gray-50); border-color: var(--c-gray-400); color: var(--c-ink);
+}
+/* Dark header reuse: restore the original translucent-white treatment. */
+.app-header .btn-link {
+  color: var(--c-white); background: var(--color-on-dark-soft); border-color: transparent;
+}
+.app-header .btn-link:hover {
+  color: var(--c-white); background: var(--color-on-dark-hover); border-color: transparent;
 }
-.btn-link:hover { background: var(--color-on-dark-hover); }
 
 .btn-signin-header {
   color: var(--c-white); text-decoration: none;
@@ -599,7 +617,7 @@
 }
 
 .rfc-breadcrumb {
-  display: flex; align-items: center; gap: 8px;
+  display: flex; align-items: center; flex-wrap: wrap; gap: 8px;
   padding: 10px 16px;
   border-bottom: 1px solid var(--c-gray-200);
   background: var(--c-gray-50);
@@ -612,7 +630,25 @@
 }
 .breadcrumb-sep { color: var(--c-gray-300); }
 .breadcrumb-meta { color: var(--c-gray-500); font-size: var(--text-sm); }
-.breadcrumb-actions { margin-left: auto; display: flex; gap: 8px; align-items: center; }
+.breadcrumb-actions {
+  margin-left: auto;
+  display: flex; flex-wrap: wrap; justify-content: flex-end;
+  gap: 8px; align-items: center; min-width: 0;
+}
+/* Normalize every action in the bar to one height + shape so the mode
+ * toggle, the filled CTAs (Start Contributing / Open PR / Graduate) and
+ * the secondary buttons (Metadata / Claim ownership / Invitations / …)
+ * line up as a single, intentional control group. Higher specificity
+ * than the per-variant rules, so it harmonizes their size/radius/type
+ * without disturbing each variant's fill colors. */
+.breadcrumb-actions > button,
+.breadcrumb-actions > a {
+  display: inline-flex; align-items: center;
+  height: 30px; padding: 0 12px;
+  border-radius: var(--radius-md);
+  font-size: var(--text-sm); font-weight: 600;
+  white-space: nowrap;
+}
 
 .btn-mode-toggle {
   font-size: var(--text-sm); font-weight: 600;
@@ -1444,7 +1480,8 @@
   font-size: var(--text-sm);
 }
 .diff-mode-toolbar .btn-link.active {
-  font-weight: 600; color: var(--c-ink);
+  font-weight: 600; color: var(--c-white);
+  background: var(--c-ink); border-color: var(--c-ink);
 }
 .pr-diff-accent {
   margin-left: auto;

frontend/src/components/Admin.jsx

@@ -326,7 +326,14 @@ function UserRow({ user: u, busy, onChangeRole, onToggleMute, onFlipPermission }
           )}
         </td>
         <TimeCell value={u.created_at} />
-        <TimeCell value={u.last_seen_at} />
+        {/* An unclaimed admin invite has provably never authenticated, so
+            last_seen_at is just the row-creation default (it equals
+            created_at). Render the truth — "Never" — rather than a
+            timestamp that reads like a real visit. */}
+        <TimeCell
+          value={pendingInvite ? null : u.last_seen_at}
+          emptyLabel={pendingInvite ? 'Never' : '—'}
+        />
       </tr>
       {state === 'pending' && u.beta_request_reason ? (
         <tr className="user-row-reason">
@@ -345,8 +352,8 @@ function UserRow({ user: u, busy, onChangeRole, onToggleMute, onFlipPermission }
 // Render a "YYYY-MM-DD HH:MM:SS" timestamp as an intentional date-over-time
 // stack (date prominent, time quiet below) rather than letting a narrow
 // column wrap the value mid-string. Falls back to an em-dash when absent.
-function TimeCell({ value }) {
-  if (!value) return <td className="muted">—</td>
+function TimeCell({ value, emptyLabel = '—' }) {
+  if (!value) return <td className="muted">{emptyLabel}</td>
   const [date, ...rest] = String(value).split(' ')
   const time = rest.join(' ')
   return (