Gate licences in CI and publish an SBOM (#282)

The licence audit that found org.json in the APK was run by hand. Nothing stopped the next
transitive dependency arriving the same way, and "we track licences" was a claim rather than
something anyone could check.

TWO GATES, BOTH FAIL CLOSED.

scripts/license-check.js audits the server's npm tree. scripts/android-license-check.js
resolves the real releaseRuntimeClasspath — everything that can enter the APK a customer
installs — and checks it against android/licenses.json, where each entry records the licence
AND the evidence for it. A dependency nobody has recorded fails the build. That is the case
worth catching: org.json reached customers because it arrived transitively and nothing ever
asked what licence it carried.

Denied: AGPL, GPL, SSPL, Commons Clause, BUSL, and the JSON Licence. Weak copyleft (LGPL,
MPL, EPL, CDDL) is reported but does not fail — it is a judgement, and the judgement should
be made by someone who knows they are making it. Anything unrecognised fails; a package whose
licence we cannot identify is not one we ship.

⚠️ THE SERVER GATE INSTALLS --omit=dev, AND THAT IS THE POINT. A developer checkout carries
sharp, whose @img/sharp-wasm32 declares LGPL-3.0-or-later. It is a test fixture generator that
never reaches a server, but a scanner pointed at a dev tree reports LGPL and contradicts the
answer we give customers. Auditing the production install is what makes the answer defensible.

SBOM. Every release now publishes screentinker-sbom-<version>.cdx.json — CycloneDX 1.5, every
production dependency with version, purl and licence, generated from a production install. CI
uploads one on every run too. That is what turns the claim into something a customer or an
underwriter can verify themselves.

Neither script takes a dependency: a gate that needs its own supply chain audited is worth
less than one that does not.

Verified by mutation rather than assumed. Injecting GPL-3.0-or-later, AGPL-3.0, the JSON
Licence, SSPL-1.0, and a package with no licence field each fail the server gate; MIT and
LGPL pass (LGPL reported). Removing the org.json exclusion fails the Android gate by name;
dropping a group from the policy fails it as unrecorded. Both restored, both green.

Found and fixed while building it: npm ls exits non-zero for any tree problem — an extraneous
package is enough — which made the gate abort instead of auditing. It now reads the listing
either way and only aborts on genuinely empty output.

docs/licensing.md records the policy, how to run the gates, and the dev-vs-production trap.

1676/1676 pass.

.github/workflows/ci.yml

@@ -86,6 +86,39 @@ jobs:
         working-directory: android
         run: ./gradlew :app:testDebugUnitTest --no-daemon
 
+      # Every artifact that can enter the APK must have a licence on file. This runs here
+      # rather than in its own job because the Gradle cache and Android SDK are already warm.
+      - name: Licence gate (APK runtime classpath)
+        run: node scripts/android-license-check.js
+
+  licenses:
+    name: Licence gate + SBOM (production deps)
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v6
+      - uses: actions/setup-node@v6
+        with:
+          node-version: '20'
+          cache: npm
+          cache-dependency-path: server/package-lock.json
+
+      # --omit=dev on purpose, and it is the whole point of the job. A developer checkout
+      # carries sharp, whose @img/sharp-wasm32 declares LGPL-3.0-or-later; it is a test
+      # fixture generator that never reaches a server. Auditing anything other than a
+      # production install would report a licence we do not actually ship.
+      - name: Install production dependencies only
+        working-directory: server
+        run: npm ci --omit=dev
+
+      - name: Licence gate
+        run: node scripts/license-check.js --sbom sbom/screentinker-server.cdx.json
+
+      - uses: actions/upload-artifact@v4
+        with:
+          name: sbom
+          path: sbom/
+          if-no-files-found: error
+
   smoke:
     name: Boot smoke + version check
     runs-on: ubuntu-latest
@@ -109,6 +142,12 @@ jobs:
         working-directory: server
         env:
           SELF_HOSTED: 'true'
+          # Boot WITH the collector on. This block is config-gated and only the
+          # statistics-collecting deployment sets the flag, so it had never executed in CI,
+          # on alpha, or in any test - and a load-time crash inside it took production down
+          # while every check was green. Code only one deployment runs is exactly the code
+          # CI has to execute.
+          TELEMETRY_COLLECTOR: '1'
         run: |
           node server.js > "$RUNNER_TEMP/server.log" 2>&1 &
           echo $! > "$RUNNER_TEMP/server.pid"
@@ -131,6 +170,21 @@ jobs:
           test "$REPORTED" = "$EXPECTED"
           echo "OK: status ok, version $REPORTED matches VERSION"
 
+      # Booting is not enough on its own - the collector could be mounted and broken. Prove
+      # the routes it adds actually answer, so a fault inside that block fails here rather
+      # than on the single deployment that turns it on.
+      - name: Assert the collector routes answer when enabled
+        run: |
+          STATS="$(curl -sf http://localhost:3001/api/public/stats)"
+          echo "stats: $STATS"
+          test "$(echo "$STATS" | jq -r 'has("screens") and has("installs")')" = "true"
+          REPORT="$(curl -s -o /dev/null -w '%{http_code}' -X POST \
+            -H 'Content-Type: application/json' -d '{"bad":1}' \
+            http://localhost:3001/api/telemetry/report)"
+          echo "malformed report -> HTTP $REPORT"
+          test "$REPORT" = "400"
+          echo "OK: collector mounted and answering"
+
       - name: Stop server
         if: always()
         run: kill "$(cat "$RUNNER_TEMP/server.pid")" 2>/dev/null || true

.github/workflows/release.yml

@@ -85,6 +85,15 @@ jobs:
           ./scripts/build-autorun-zip.sh -o autorun.zip
           ls -la autorun.zip
 
+      # A published SBOM is what turns "we track licences" into something a customer or an
+      # underwriter can check for themselves. Built from a PRODUCTION install — a dev tree
+      # would list packages (sharp and its LGPL-bearing wasm variant) that never ship.
+      - name: Generate SBOM (production dependencies)
+        run: |
+          ( cd server && npm ci --omit=dev )
+          node scripts/license-check.js --sbom "screentinker-sbom-${{ steps.ver.outputs.version }}.cdx.json"
+          ls -la screentinker-sbom-*.cdx.json
+
       - name: Build source tarball (bundles the .wgt; the signed apk is added by scripts/finalize-release.sh)
         run: |
           OUT="screentinker-${{ steps.ver.outputs.version }}.tar.gz"
@@ -100,15 +109,42 @@ jobs:
       - name: Generate release notes
         run: |
           PREV="${{ steps.ver.outputs.prev }}"
+          VERSION="${{ steps.ver.outputs.version }}"
+
+          # Prefer the hand-written CHANGELOG section for this version.
+          #
+          # The generated list is commit SUBJECTS, which describe the work, not the release: cutting
+          # 1.9.34 produced a page reading "chore(release): v1.9.34" and one changelog commit, while
+          # the entry describing single sign-on, the native-dependency removal, the update fixes and
+          # every outside contributor sat in CHANGELOG.md and was never published. The notes on the
+          # release page are what most people actually read, so they should be the written ones.
+          #
+          # awk rather than sed: the body contains regex metacharacters and markdown that a sed range
+          # would mangle. This takes everything between `## <version>` and the next `## ` heading.
+          CHANGELOG_BODY="$(awk -v v="## $VERSION" '
+            $0 == v {found=1; next}
+            found && /^## / {exit}
+            found {print}
+          ' CHANGELOG.md)"
+
           {
             echo "## ScreenTinker ${{ steps.ver.outputs.tag }}"
             echo
-            echo "### Changes"
-            if [ -n "$PREV" ]; then
-              git log --no-merges --pretty='- %s' "${PREV}..${{ steps.ver.outputs.tag }}"
+            if [ -n "$(printf '%s' "$CHANGELOG_BODY" | tr -d '[:space:]')" ]; then
+              echo "$CHANGELOG_BODY"
             else
-              echo "_First tagged release. Most recent changes:_"
-              git log --no-merges --pretty='- %s' -n 30 "${{ steps.ver.outputs.tag }}"
+              # No entry for this version — fall back to commit subjects rather than publish a
+              # release with no notes at all. scripts/bump-version.sh already warns when the
+              # CHANGELOG has no matching heading; this is the same gap showing up downstream.
+              echo "_No CHANGELOG entry for $VERSION; listing commits instead._"
+              echo
+              echo "### Changes"
+              if [ -n "$PREV" ]; then
+                git log --no-merges --pretty='- %s' "${PREV}..${{ steps.ver.outputs.tag }}"
+              else
+                echo "_First tagged release. Most recent changes:_"
+                git log --no-merges --pretty='- %s' -n 30 "${{ steps.ver.outputs.tag }}"
+              fi
             fi
             echo
             echo "### Artifacts"
@@ -127,6 +163,7 @@ jobs:
               echo "- Docker image: \`ghcr.io/screentinker/screentinker:${{ steps.ver.outputs.version }}\` (also \`:latest\`)."
             fi
             echo "- \`ScreenTinker.apk\` - signed Android player (attached during release finalization)."
+            echo "- \`screentinker-sbom-${{ steps.ver.outputs.version }}.cdx.json\` - CycloneDX 1.5 software bill of materials for the server's production dependencies, with the licence of every component."
           } > RELEASE_NOTES.md
           cat RELEASE_NOTES.md
 
@@ -144,6 +181,7 @@ jobs:
             --notes-file RELEASE_NOTES.md \
             "${TARBALL}" \
             autorun.zip \
+            "screentinker-sbom-${{ steps.ver.outputs.version }}.cdx.json" \
             tizen/ScreenTinker.wgt
 
   docker:

.gitignore

@@ -59,3 +59,7 @@ audit/
 # Local SQLite artifacts (any extension the tooling might produce)
 *.sqlite
 *.sqlite3
+
+# Generated by scripts/license-check.js --sbom (CI publishes it as a release asset)
+sbom/
+*.cdx.json

CHANGELOG.md

@@ -1,405 +1,392 @@
 # Changelog
 
-## 1.9.34-alpha14
+## 1.9.36
 
-The update fix from alpha13, confirmed over the air. A panel that had failed every update attempt
-for hours took this one unattended, in about seventy seconds, with no prompt on screen — the first
-successful over-the-air update on that hardware since the staging bug was introduced.
+A single fix. **1.9.36 replaces 1.9.35** — see below for whether that affects you.
 
-### Added — Raspberry Pi notes in the operations runbook
-Three traps from a Pi 5 report, two of which are not Pi-specific: a piped installer cannot really
-ask you anything (the pipe is its input, so every prompt takes the default), X11 tools fail silently
-on Wayland (so screen blanking and cursor hiding can be entirely absent while appearing configured),
-and overlay filesystems protect an SD card by discarding writes — safe for a player, quietly
-destructive for a server whose database is written continuously.
+### Fixed — 1.9.35 would not start on a server collecting install statistics
 
-## 1.9.34-alpha13
+A server with install-statistics collection switched on could not start 1.9.35. It threw
+`ReferenceError: Cannot access 'db' before initialization` while loading, before it began listening,
+and a service manager configured to restart it would do so in a loop.
 
-### Fixed — a player that could not update, while downloading content perfectly well
-Updates were always written to external storage. On a player where that location exists but cannot
-be written to, the download failed the instant it began — before any data arrived — and reported
-only that it had failed to download or verify. Nothing pointed at a directory, and the same player
-was downloading and caching content without trouble the whole time, because content goes to
-internal storage.
+**Almost nobody is affected.** The fault is inside a block that only runs when a server is configured
+to *collect* statistics from other installs — not when it merely reports its own. That is a single
+deployment, not a normal install. If you have never set `TELEMETRY_COLLECTOR`, 1.9.35 runs correctly
+and this release changes nothing for you.
 
-Updates now go to the first location that genuinely accepts them, starting with the player's own
-internal storage, which is always available. External storage is still used when it works, since
-the file remains there for a manual install. Each location is tested by writing to it rather than
-by asking whether it is writable — the previous check asked, was told yes, and the write then
-failed anyway.
+The cause was a reference to the database resolved when the file loaded rather than when the request
+arrived, in code that had been moved earlier in the same release.
 
-If no location works, the player now names every one it tried and why.
+### Changed — the startup check now covers configuration only one deployment uses
 
-⚠️ **A player already stuck cannot be rescued by this release.** The broken path is how updates
-arrive, and the "Push an APK" button used it too, so a player in that state needs one update
-installed by hand. Afterwards it recovers on its own and stays fixed.
+The fault above shipped through a full test suite and every CI job green, because the affected block
+is switched on by configuration that no test set. It had never executed anywhere except the one
+server that turns it on.
 
-## 1.9.34-alpha12
+The startup smoke check now boots with that configuration enabled and confirms the routes it adds
+actually answer. Code that only one deployment runs is exactly the code an automated check has to
+exercise, and it now does.
 
-### Fixed — updates were refused as "already newer" from alpha10 onward
-The update check compared the prerelease part of a version as plain text, so "alpha11" sorted below
-"alpha8" — because "1" comes before "8". Every build from alpha10 on was therefore treated as older
-than alpha8 and alpha9, and the server refused to offer it while naming it as the latest version in
-the same reply. A player on those builds could not be moved forward at all, and nothing about the
-failure pointed at version ordering.
+### Upgrading
 
-Prerelease numbers are now compared as numbers. Ordinary precedence is unchanged: beta still comes
-after alpha, rc after beta, and a finished release still beats any prerelease of the same version.
+No migrations, no configuration changes, and no dependency changes from 1.9.35 — this release only
+alters when one value is read. Upgrading from 1.9.34 or earlier, the 1.9.35 note still applies:
+`npm ci --omit=dev` is required in both directions, which `scripts/upgrade.sh` already runs.
 
-The BrightSign host package carried the same comparison and is fixed with it — there, a wrong
-answer replaces the script that starts the player.
+## 1.9.35
 
-## 1.9.34-alpha11
+A maintenance release. Two faults where the product was working correctly and still looked broken to
+whoever was standing in front of the screen, plus the dependency advisories that could reach a running
+server.
 
-### Fixed — an update that will not install now says why
-A panel that failed to update reported only "failed to download or failed signature verification" —
-one sentence covering seven distinct causes, three of which are download failures where
-verification never runs. Every specific reason went to logcat, which an unprivileged app cannot
-read on Android 9, so in the field the message named a symptom shared by unrelated problems.
+No migrations and no configuration changes. See the upgrade note at the end of this entry.
 
-The player now reports the actual cause: the HTTP status the server returned, how many bytes
-arrived, whether the signing certificates could be read at all, whether the key genuinely differs,
-or that the download could not be written and where.
+### Fixed — a player could get stuck on an update it was never able to install
 
-### Fixed — the destination is checked before downloading
-Nothing verified that the player could actually store the file. A directory path can be returned
-and still be missing, unwritable, on a volume that has gone away, or full — and all of those
-surfaced as the same opaque failure as a network fault. The player now proves the directory is
-writable (by writing to it, not by asking) and that there is room for the APK plus the copy the
-installer stages, before it starts.
+A staged update is saved under a filename built from the version the server advertised. If a server
+advertised one version while still serving the file for an older one, the player saved the old file
+under the new name — and from then on found it, verified its signature, accepted it, and installed
+something that changed nothing. The version never moved, so the same update was offered again, and the
+player retried the same no-op until it hit its attempt limit.
 
-### Fixed — a readable APK is no longer refused on Android 9 and 10
-On those versions the archive's signing certificate comes from a legacy path that can return
-nothing, and the update was then refused with no way to distinguish that from a real mismatch. The
-player now reads the signature itself before giving up. Verification is unchanged: the certificate
-is still compared against the installed app, and anything unsigned, tampered with, or signed by a
-different key is still rejected.
+The signature check passed the whole time, correctly: the file was genuine, it was simply the wrong
+one. Worse, fixing the server did not help, because the bad file was reused before anything was
+downloaded. Recovery meant deleting the file on the device by hand.
 
-## 1.9.34-alpha10
+A staged update is now reused only when the version inside the file matches the version being
+installed, and a fresh download is checked the same way before it is applied. A server serving the
+wrong file now says so — *"server served 1.9.33 but advertised 1.9.34 — the update on the server is
+stale"* — and the file is deleted instead of kept. That makes this self-healing: once the server is
+corrected, the player recovers on its own.
 
-### Changed — install statistics report as soon as you opt in, and say when they cannot
-Turning sharing on now sends a report immediately rather than waiting for the next daily tick, and
-a failed attempt is recorded and explained. A server whose outbound traffic is filtered previously
-looked identical to one where the feature was broken: sharing on, nothing arriving, no way to tell
-which. Settings now names the address that did not answer and why, and a later success clears the
-warning.
+**Clear update cache** on the device page discards every staged update on a player. The version check
+should make it rarely necessary; it exists because a player already holding a bad file predates this
+release and cannot benefit from the check, and because the alternative is a cable and a laptop.
 
-### Added — keep your own copy of the statistics
-`TELEMETRY_EXTRA_ENDPOINT` posts the same three fields to a collector you run.
+### Fixed — directory search showed the system keyboard on top of its own
 
-It is **additional, not a redirect** — the shared report still goes to ScreenTinker, which is why
-it is named EXTRA rather than ENDPOINT, and Settings lists every destination a report goes to. It
-is also independent of the sharing switch, because it is your server posting to your host: if you
-want your own statistics and nothing sent to us, set it and leave sharing off. Each destination is
-attempted separately, so one being unreachable never stops the other.
+The directory-search widget draws its own on-screen keyboard, sized and themed to the panel and on by
+default. On Android it was never visible. The page puts the cursor in a real text field, which is the
+signal for the device to raise its system keyboard — over the bottom of the screen, exactly where the
+widget's keyboard is.
 
-## 1.9.34-alpha9
+So a wall-mounted directory showed the phone keyboard: split across the screen, with microphone, GIF,
+emoji and a settings key that opens the keyboard vendor's own interface on a kiosk. On one panel the
+only keyboard installed was voice input, so touching the search box opened a microphone. The widget's
+own keyboard had been underneath the whole time.
 
-### Added — opt-in install statistics
-ScreenTinker cannot see how widely it is deployed, because self-hosted installs are private by
-design and should stay that way. A platform administrator is now asked, once, whether this install
-will share how many screens it runs.
+When the widget draws a keyboard, it now tells the device not to raise one. Turn the built-in keyboard
+off and the system keyboard behaves as before — with nothing to cover, it is the only way left to type.
 
-The whole payload is three fields — a random instance ID, the version, and the screen count — and
-nothing else: no hostnames, addresses, organization or user names, device names, content, or
-configuration. Settings → Install statistics shows the **actual payload this server would send**,
-generated live from its own data, alongside what it last really sent and when, so the claim can be
-checked rather than taken on trust. Full detail in [docs/telemetry.md](docs/telemetry.md).
+### Changed — the dependency advisories that could reach a running server are cleared
 
-Off until enabled, and both answers are remembered — declining is permanent, so the prompt does not
-return after an update.
+Every high-severity advisory affecting a production install is resolved, including eight in the mail
+library covering SMTP command injection and header injection. The remaining advisories are in
+development-only tooling that is not installed on a server and cannot be reached from one.
 
-The random ID exists only so repeat reports from one server count as one server. It makes a report
-pseudonymous rather than anonymous, which the wording says plainly. Because sharing is opt-in, any
-total published from this is a floor — "at least N screens" — never an estimate of the install base.
+The real-time connection to players is deliberately untouched: the fix there was a patch to the message
+parser with no change to the format players speak, so nothing about an existing player's connection
+changes.
 
-## 1.9.34-alpha8
+Sending mail was previously covered only by tests that substituted the mail library for a stand-in,
+which would have stayed green through any change in the library itself. It is now also tested against
+the real one.
 
-### Fixed — an APK download with no external storage went nowhere, silently
-A panel could not install an update by OTA **or** by the dashboard's "Push an APK" button. Both
-staged the download in `getExternalFilesDir()`, which returns null when external storage is
-unavailable — not exotic on signage hardware: no emulated volume, a vendor ROM that never mounts
-one, an ejected card, storage still unmounted early in boot. A null parent silently produces a
-*relative* path, so the download was written to the process working directory, which is not
-writable. The write failed and the panel reported only "failed to download or failed signature
-verification".
+### Added — an install that collects statistics can show the total on its landing page
 
-Every signal pointed away from the cause: the HTTP request succeeds (the server records a served
-download at the exact second of each failure), the signing key is fine, and because nothing is ever
-written there is no partial file to find. It also never recovers — every attempt fails the same way.
+Where install statistics are being collected, the landing page can show how many screens have been
+deployed in total. It is an aggregate across every install that chooses to report, so it says nothing
+about any single one.
 
-Downloads now fall back to internal storage, which cannot be unmounted.
+This does nothing on a normal install: the figure is served only where collection is switched on, so a
+private server never publishes its own screen count, and the line is hidden entirely rather than
+showing a zero.
 
-⚠️ **This cannot repair a panel already affected.** The broken download path is the delivery
-mechanism, and the Push APK button shared the bug, so a panel in this state needs one manual install
-to escape it. The release protects panels that are not yet affected.
+### Changed — release notes are the written ones
 
-## 1.9.34-alpha7
+Published release notes now come from this file rather than from a list of commit subjects. The
+previous release announced itself as one commit titled "chore(release)" while the entry describing it
+sat here unread.
 
-### ⚠️ Upgrading to this build requires reinstalling dependencies
+### ⚠️ Upgrading from 1.9.34 reinstalls dependencies
 
-The two dependency changes below — dropping `sharp` and pinning `better-sqlite3` — alter
-`server/package.json`, so **`npm ci --omit=dev` is required, not optional** — in both directions.
-The runbook's rollback step marks that command "only if dependencies changed"; for this release
-they did.
+This release changes `server/package.json`, so **`npm ci --omit=dev` is required, not optional** — in
+both directions. `scripts/upgrade.sh` already runs it, and the server repairs a missed install at
+startup where it can reach the npm registry.
 
-- **Upgrading**: `scripts/upgrade.sh` already runs it. A hand-rolled deploy that skips it leaves a
-  `better-sqlite3` that no longer matches `package.json`.
-- **Rolling back past this build**: the reinstall is **mandatory**. Earlier builds import `sharp` at
-  runtime to thumbnail images, and this build removes it from the production dependencies — so
-  rolling back the code without reinstalling leaves a server whose image ingest cannot load its
-  decoder. `lib/preflight-deps.js` catches this at boot and repairs it, but do not rely on that as
-  the plan.
+Docker deployments need no action; dependencies are installed inside the image.
 
-Docker deployments need no action either way: dependencies are installed inside the image.
+## 1.9.34
 
-No migrations, no configuration changes, no player-side changes.
+Single sign-on is the headline, rebuilt rather than extended — because of a vulnerability in what
+was there before. Alongside it: the last native image dependency is gone, several players that
+could not install updates now can, and an install can optionally report how many screens it runs.
+
+No migrations and no configuration changes. See the upgrade note at the end of this entry.
+
+### Fixed — the old sign-in path could be replayed by any site you had signed into
+What shipped as "OAuth" verified almost nothing. It asked whether an **access** token was valid and
+then trusted the email address that came back, never asking the only question that matters: *who was
+this token issued for?* Any other site a user had signed into — anything holding a token with the
+right scope — could replay it against ScreenTinker and receive a session as that user. No password,
+no interaction from the victim.
+
+Identity now comes from an **ID token only**, with the signature checked against the provider's
+keys and `iss`, `aud`, `azp`, `exp` and `nonce` all verified. One flow for every provider:
+Authorization Code with PKCE, completed server-side. Google and Microsoft became ordinary entries
+rather than hand-written special cases, which is what removed the two paths that were wrong.
+
+### Added — organizations bring their own single sign-on
+Instance-wide providers stay the default and are now unlimited in number. On top of that, an
+organization can connect its own identity provider — Entra, Okta, Auth0, Keycloak, anything speaking
+OpenID Connect — configured by that organization's own admins in Settings, with no operator
+involvement and no restart.
+
+A provider may only assert addresses at domains the organization has **proved it controls**, via a
+TXT record at `_screentinker-verify.<domain>`. An unverified claim lapses after eight hours and
+releases the domain, so a typo cannot park someone else's domain indefinitely. A domain belongs to
+one organization only. Proof by delegated name (CNAME) is refused outright: it would need a wildcard
+zone we do not operate, and it would turn a subdomain takeover into an apex takeover.
+
+An organization's provider never appears publicly. The login page reveals it only after someone
+enters an address at a verified domain, so a guessed domain cannot confirm who your customers are.
+
+**Require single sign-on** is available per organization: passwords refused, other providers
+refused, the instance's own Google and Microsoft buttons refused — otherwise "requires SSO" would
+just be renaming the bypass. Turning it *off* again needs a platform administrator to approve the
+request, so one compromised org admin cannot quietly reopen password login. Break-glass for a
+platform administrator is the correct password and nothing else, and a wrong password returns the
+same refusal everyone else gets, so it cannot be used to discover whether an account exists.
+
+⚠️ **Enabling it clears the passwords** of members at verified domains. That is not reversible
+without a reset.
+
+Entra sends no `email_verified` claim, which is why a Microsoft provider is trusted on other
+grounds: an instance-wide Microsoft entry is pinned to a single directory chosen by the operator,
+and an organization's own provider is believed once it has verified a domain — the DNS proof stands
+in for the claim, since whoever controls a domain's DNS controls its mail. A provider that has
+verified nothing assumes nothing, and an explicit `email_verified: false` is refused from anyone.
+Other providers that verify addresses without saying so can opt in with
+`OIDC_<SLUG>_ASSUME_EMAIL_VERIFIED=true`.
+
+**With no SSO environment variables set, the product behaves exactly as it did before.**
+
+### Added — an existing account can move to single sign-on
+Signing in with a provider has always refused to take over an account that already has a password,
+and that refusal is right — otherwise anyone who could persuade a provider to assert your address
+would inherit your account. But the way out had never been built, so an account created with a
+password simply could not use single sign-on.
+
+**Settings → Sign-in method** now offers it, in both directions. An account has exactly **one**
+credential: linking **deletes** the password, and the confirmation says so, because a password left
+behind is a second way in that you believe you replaced. Unlinking asks for the new password first
+and applies both changes together, so the account is never left without a way in.
+
+The account being linked is the one you are **signed in as**, never whichever account matches the
+address the provider returns — that is what separates linking from the takeover the login page
+refuses. Only providers configured on this server can be linked; an organization's own provider
+cannot attach itself to an account.
+
+### Changed — the login page asks who you are before how you sign in
+The password box appears once you have entered your address and continued, rather than sitting there
+from the start. That is what lets the page check whether your organization uses single sign-on
+*before* offering you a credential, so someone whose company requires it is shown that rather than a
+password box that was always going to be refused. Correcting your address takes you back a step.
+
+The address is no longer looked up on every keystroke — it answered for half-finished domains,
+changed the form under you mid-address, and could exhaust a shared office network's lookup budget
+before anyone had tried to sign in. The instance's own provider buttons stay visible throughout, so
+the page no longer changes shape while you type.
+
+Setup instructions for both operators and organization admins are in
+[docs/sso-setup.md](docs/sso-setup.md), written from configuring real Google and Entra applications
+— including the one that catches everyone: the Microsoft tenant setting names the directory that
+*authenticates the user*, which for personal accounts is not the directory the application is
+registered in.
+
+### Changed — image processing no longer needs a native library
+Thumbnails and image measurement are now pure JavaScript, with WebAssembly decoders for webp and
+avif, running on a worker thread. Nothing in the image path is a compiled binary any more, and
+`better-sqlite3` is the only native module left.
+
+A native module needs a prebuilt binary matching both the platform and the Node version; when there
+isn't one the server fails at load with an error that reads like database corruption rather than a
+missing image library. That class of failure is gone from this half of the product.
+
+Format support is unchanged in practice: jpeg, png, gif, tiff and bmp decode directly, webp and avif
+through WebAssembly. `.heic` still produces no thumbnail — it never did, because the image library
+in use decodes AV1 but refuses HEVC.
+
+Decoding moved off the main thread deliberately. Pure JavaScript costs about a second for a
+12-megapixel photo, which in-process would stall everything else — and the thumbnail backfill walks
+an entire library at startup, which is exactly how a maintenance task turns into missed heartbeats
+and players marked offline. Thumbnailing is slower in wall-clock terms and no longer competes with
+serving requests.
+
+### Fixed — players that could not install an update
+Three separate faults, each able to strand a player on an old version.
+
+**Updates were written to external storage.** Where that location is absent, or exists but cannot be
+written to, the download failed the instant it began — before any data arrived — and reported only
+that it had failed to download or verify. The same player could be caching content perfectly well
+throughout, because content goes to internal storage. Updates now go to the first location that
+genuinely accepts them, starting with internal storage, and each candidate is tested by *writing to
+it* rather than by asking whether it is writable — the previous check asked, was told yes, and the
+write failed anyway.
+
+**Prerelease versions were ordered as text**, so a build numbered 10 or higher sorted below one
+numbered 8 or 9. A player on such a build was told it was already up to date and could not be moved
+forward, while the server named the newer build as latest in the same reply. Numbers in version
+names are now compared as numbers. The BrightSign host package carried the same comparison and is
+fixed with it — there, a wrong answer replaces the script that starts the player.
+
+**A readable update was refused on Android 9 and 10**, where a downloaded file's signing certificate
+comes from a legacy path that can return nothing. The player now reads the signature itself before
+giving up. Verification is unchanged: the certificate is still compared against the installed app,
+and anything unsigned, tampered with, or signed by a different key is still rejected.
+
+A failed update now also says which of those things went wrong, instead of one message covering
+every possible cause.
+
+⚠️ **A player already stuck cannot be rescued by this release**, because the broken path is how
+updates arrive and the "Push an APK" button used it too. Such a player needs one update installed by
+hand, after which it recovers on its own and stays fixed.
 
 ### Fixed — the Android player could leave a band down one edge of the screen
-A panel would sometimes not fill its display: content rendered into a stage sized from a window that
-no longer existed, leaving a bar the exact size of the hidden system bar. It was intermittent
-because it depended on whether the app was measured before or after the system UI was hidden — the
-same screen could come up correct after a reboot and wrong after an app restart.
-
-The stage is now measured from the current window and re-measured when focus changes, so a late
-change to the usable area is picked up instead of being baked in at startup.
+A panel would sometimes not fill its display, leaving a bar the exact size of the hidden system bar.
+It was intermittent because it depended on whether the app was measured before or after the system
+UI was hidden — the same screen could come up correct after a reboot and wrong after an app restart.
+The stage is now measured from the current window and re-measured when focus changes.
 
 Reported on an RK356x Android box, where it was compounded by an unrelated HDMI mode problem;
 pinning the output resolution fixed the corruption, and this fixes the band that remained.
 
+### Added — opt-in install statistics
+ScreenTinker cannot see how widely it is deployed, because self-hosted installs are private by
+design and should stay that way. A platform administrator is asked, once, whether this install will
+share how many screens it runs.
+
+The whole payload is three fields — a random instance ID, the version, and the screen count — and
+nothing else: no hostnames, addresses, organization or user names, device names, content or
+configuration. Settings shows the **actual payload this server would send**, generated live from its
+own data, alongside what it last really sent and when, so the claim can be checked rather than taken
+on trust. Turning it on reports immediately, and a blocked outbound connection is named along with
+the address to allow, rather than failing silently.
+
+Off until enabled, and both answers are remembered — declining is permanent, so the prompt does not
+return after an update. `TELEMETRY_EXTRA_ENDPOINT` posts the same three fields to a collector you
+run; it is **additional, not a redirect**, and independent of the sharing switch, so an operator who
+wants their own numbers and nothing sent to us can set it and leave sharing off.
+
+The random ID exists only so repeat reports from one server count as one server, which makes a
+report pseudonymous rather than anonymous — the wording says so plainly. Because sharing is opt-in,
+any total published from it is a floor, never an estimate of the install base. Full detail in
+[docs/telemetry.md](docs/telemetry.md).
+
+### Added — organizations may re-enable same-origin widgets, deliberately
+Widget isolation removed `allow-same-origin`, which also broke embedding for sites that enforce
+strict CORS. There is now an organization-level switch to put it back, behind a modal requiring a
+typed acknowledgement, with a persistent banner while it is on. It needs an organization owner or
+admin — a workspace admin is deliberately not enough — and the change is written to the activity
+log. Contributed by @ChrisChrome.
+
+The widget editor's **Preview is excluded** from that switch. Preview renders inside the dashboard
+where the admin's session token lives, so honouring the setting there would let anyone who can
+author a widget lift the session of whichever admin clicked Preview. The setting exists so
+*displays* can embed origin-strict sites; a display holds a device token, an admin's browser does
+not.
+
+### Fixed — RSS tickers ran at a speed that depended on how much news there was
+Scroll speed set a fixed total time for the whole strip to cross the screen regardless of length, so
+a feed with twenty items was dragged past in the same seconds as a feed with one — too fast to read,
+and it appeared to jump back to the start. It now holds a constant rate, so more items simply take
+proportionally longer and every item scrolls fully into and out of view. Contributed by @ChrisChrome.
+
+### Fixed — user-controlled text is escaped where it reaches the page
+An audit pass over the frontend's HTML sinks, escaping the ones that receive user-controlled data.
+Also here: dashboard banners no longer overlap the sidebar, shift the layout, or vanish when
+switching views, and the main content no longer collapses to a narrow column.
+
 ### Added — an operations runbook
-`docs/operations.md`. How to deploy, verify, and roll back an instance in both shapes it runs in
-(git + systemd, and Docker), including what to back up first, how to tell a deploy actually worked,
-and the traps that are only obvious once they have bitten you.
-
-### Changed — image processing no longer uses a native module
-`sharp` is gone. Thumbnailing and image measurement are pure JavaScript (Jimp) with WebAssembly
-codecs for webp and avif, running on a worker thread. `sharp` remains as a development dependency
-for test fixtures, so `--omit=dev` excludes it from a deployed install entirely.
-
-The motivation is that a native module needs a prebuilt binary matching both the platform and the
-Node ABI; when there isn't one, the failure arrives at load time and reads like database corruption
-rather than a missing image library. `better-sqlite3` is now the only native module left.
-
-Format support is unchanged in practice. jpeg, png, gif, tiff and bmp decode natively; webp and avif
-via WebAssembly. `.heic` still produces no thumbnail — it never did, because the `sharp` builds in
-use decode AV1 but refuse HEVC.
-
-Images are decoded on a worker thread rather than in-process. Pure-JavaScript decoding costs about a
-second for a 12-megapixel photo, which in-process would block the event loop — and the thumbnail
-backfill walks an entire library at boot, which is exactly how a maintenance task turns into missed
-heartbeats and players marked offline. Thumbnailing is slower in wall-clock terms than the native
-library was, and no longer competes with serving requests.
+[docs/operations.md](docs/operations.md): how to deploy, verify and roll back an instance in both
+shapes it runs in, what to back up first, how to upgrade Node.js safely, and the traps that are only
+obvious once they have bitten you — including three from a Raspberry Pi 5 report, two of which are
+not Pi-specific. A piped installer cannot really ask you anything, because the pipe is its input and
+every prompt takes the default. X11 tools fail silently on Wayland, so screen blanking and cursor
+hiding can be entirely absent while appearing configured. And an overlay filesystem protects an SD
+card by discarding writes — safe for a player, quietly destructive for a server whose database is
+written continuously.
 
 ### Changed — `better-sqlite3` pinned to 12.9.0
-Preparation for a future Node 22 upgrade, landed separately so the runtime move and the database
-driver move stay independently reversible.
+Preparation for a future Node.js 22 upgrade, landed separately so the runtime and the database
+driver can move independently rather than as one flag day.
 
 The pin is **exact on purpose**. 12.9.0 is the last release publishing prebuilt binaries for both
 the current and the next Node major; later 12.x releases dropped the older one while still
-advertising support for it in `engines`. A caret range would resolve to one of those and silently
-turn installation into a from-source compile. `lib/preflight-deps.js` explains this at the point
-anyone debugging the resulting failure would be reading.
+advertising support for it. A caret range would resolve to one of those and silently turn
+installation into a source build. Nothing in the query API changed.
 
-Nothing in the query API changed — every major since 9.x was bumped only to drop end-of-life Node
-and Electron versions.
+### ⚠️ Upgrading from 1.9.33 reinstalls dependencies
+This release changes `server/package.json`, so **`npm ci --omit=dev` is required, not optional** —
+in both directions.
 
-## 1.9.34-alpha6
+- **Upgrading**: `scripts/upgrade.sh` already runs it, and the server repairs a missed install at
+  startup where it can reach the npm registry.
+- **Rolling back past this release**: mandatory. Earlier builds load a native image library at
+  runtime that this release removes, so rolling back the code without reinstalling leaves a server
+  whose image ingest cannot load its decoder.
 
-### Added — a setup guide for single sign-on
-`docs/sso-setup.md`. The README explained what SSO is and listed the settings; it did not say where
-to click, which of several plausible values to use, or what a given failure means. The guide walks
-both the operator configuring Google or Microsoft for the whole instance and an organization admin
-bringing their own provider, and ends with a table of every error the sign-in can produce alongside
-what usually causes it.
+Docker deployments need no action either way; dependencies are installed inside the image.
 
-Written from configuring real Google and Entra applications rather than from the code, so the
-pitfalls in it are the ones that actually cost time — most of all that the Microsoft tenant setting
-names the directory that *authenticates the user*, which for personal accounts is not the directory
-the application is registered in.
+### Known limitations
+Deliberately unresolved, and worth knowing:
 
-## 1.9.34-alpha5
-
-### Fixed — the Link button in Settings always said "Authentication required"
-`alpha4` shipped account linking with a button that could never work. It navigated the browser
-straight at the link endpoint, and this app's session is a token held in the page rather than a
-cookie — so the request arrived with no credentials and was refused, every time.
-
-The page now asks the server for the sign-in URL first, using its session, and follows that. Nothing
-about the link itself changed.
-
-## 1.9.34-alpha4
-
-Two changes to how signing in works, both found by configuring real Google and Microsoft sign-in
-rather than by reading the code.
-
-### Added — an existing account can move to single sign-on
-Signing in with a provider has always refused to take over an account that already has a password.
-That refusal is right: otherwise anyone who could get a provider to assert your address would inherit
-your account. But the way out — sign in with your password, then link from Settings — had never been
-built, so it was a dead end. An account created with a password simply could not use single sign-on.
-
-**Settings → Sign-in method** now offers it. An account with a password can link one of the
-providers this server offers; an account on a provider can unlink back to a password.
-
-An account has exactly **one** credential. Linking **deletes** the password, and the confirmation
-says so plainly, because a password left behind is a second way in that you believe you replaced.
-Unlinking asks for the new password first and applies both changes together, so the account is never
-left, even briefly, with no way to sign in.
-
-The account being linked is the one you are **signed in as** — never whichever account matches the
-email the provider returns. That is what separates linking from the takeover the login page refuses.
-Only the providers configured on this server can be linked; an organization's own provider cannot
-attach itself to an account.
-
-### Changed — the login page asks who you are before how you sign in
-The password box now appears once you have entered your email address and continued, rather than
-sitting there from the start. That is what lets the page ask whether your organization uses single
-sign-on *before* offering you a credential: if it does, you are shown that instead of a password box
-that was going to be refused. Correcting your address takes you back a step so the answer matches
-what you actually typed.
-
-The address is no longer looked up on every keystroke. It answered for half-finished domains, changed
-the form under you mid-address, and could exhaust a shared office IP's lookup budget before anyone
-had tried to sign in.
-
-Google and Microsoft buttons now stay visible throughout, including for organizations that require
-their own provider. Such a login is still refused by the server; the page no longer changes shape
-while you type.
-
-## 1.9.34-alpha3
-
-Completes what `alpha2` half-fixed. That release made the operator's own Microsoft button work and
-left customers' Entra tenants broken, which is the wrong way round.
-
-### Fixed — a customer's own Entra tenant was refused after its domain went green
-An organization that brings its own Microsoft tenant publishes the DNS record, watches its domain
-verify, and was then refused at login with `email_unverified`. Entra sends no such claim, and an
-organization's provider was never allowed to assume one.
-
-The refusal fired *after* domain confinement had already passed, so it was not the domain check doing
-its job — it was a second check asking for something Microsoft does not emit. Requiring a claim a
-provider cannot send is not a security control; it is an outage.
-
-An organization's provider is now believed **once it has verified a domain**. The DNS proof is what
-stands in for the claim: whoever controls a domain's DNS controls its mail, which is the same trust
-that makes a verification link meaningful. A provider that has verified nothing still assumes
-nothing, domain confinement is unchanged, and an explicit `email_verified: false` is still refused
-from anyone.
-
-The assumption follows from the proof and is never a stored setting — an organization cannot switch
-it on for itself.
-
-## 1.9.34-alpha2
-
-Everything in `1.9.34-alpha1`, plus one fix without which the headline feature could not be used with
-Microsoft at all.
-
-### Fixed — Microsoft sign-in could never complete
-The login callback required the identity provider to assert `email_verified: true`. **Entra ID v2
-does not send that claim**, so a Microsoft login would authenticate correctly against the tenant and
-then be refused on the way back with `email_unverified`. Found while configuring a real Entra
-application, before a single sign-in was attempted; the SSO tests checked how the Microsoft issuer
-string is built but never put a Microsoft-shaped token through the policy.
-
-The strict check was itself a fix — an earlier version accepted an omitted claim — and it stays
-exactly as strict for a provider a **customer** configures, because such a provider is chosen by the
-party it vouches for and its bare assertion is worth nothing. What changed is recognising this as a
-question about *who was trusted* rather than about what the token contained: an instance-wide
-provider is chosen by the operator, and a Microsoft entry is pinned to a single tenant GUID, so only
-that directory can issue a token this server will accept.
-
-An explicit `email_verified: false` is still refused from anyone, and an organization's own provider
-can never make the assumption. Google is unchanged — it does send the claim.
-
-Other identity providers that verify addresses without saying so in the token can opt in with
-`OIDC_<SLUG>_ASSUME_EMAIL_VERIFIED=true`.
-
-### Fixed — documentation that would have cost you an afternoon
-`MICROSOFT_CLIENT_SECRET` is read by the server but was missing from the README table. The redirect
-URI must be registered under Entra's **Web** platform, not **SPA** — a SPA registration is rejected
-at the token endpoint, because this exchange runs server-side and sends no browser `Origin`. And the
-`email` optional claim has to be added under *Token configuration → ID*, or the token arrives with no
-address at all.
-
-## 1.9.34-alpha1
-
-**A prerelease, for the alpha instance.** It is not a production build and no display should be
-pulled onto it except deliberately: a prerelease sorts *below* its own release in semver, so a player
-that takes `1.9.34-alpha1` without being opted in will see `1.9.33` as newer and roll itself back.
-Opting a display in is what stops that.
-
-The headline is single sign-on, rebuilt from nothing — and the reason it was rebuilt rather than
-extended is a vulnerability in what was there before.
-
-### Fixed — the old sign-in path could be replayed by any site you had signed into
-What shipped as "OAuth" verified almost nothing. The Google path asked `tokeninfo` whether an
-**access** token was valid and then trusted the email address in the reply. The Microsoft path handed
-a bearer token to Graph `/me` and trusted that. Neither asked the only question that matters: *who
-was this token issued for?*
-
-So any other site a user had signed into — anything that had requested `email` or `User.Read` — held
-a token it could replay against ScreenTinker and receive a session as that user. No password, no
-interaction from the victim.
-
-Identity now comes from an **ID token only**, with signature checked against the provider's JWKS and
-`iss`, `aud`, `azp`, `exp` and `nonce` all verified. One flow for every provider: Authorization Code
-with PKCE, completed server-side. Google and Microsoft became ordinary entries rather than special
-cases, which is what removed the two hand-written paths that were wrong.
-
-### Added — organizations bring their own identity provider
-Instance-wide providers stay the default and are now unlimited in number. On top of that an
-organization may configure its own provider, but only for domains it has **proved it controls** — a
-TXT record at `_screentinker-verify.<domain>`. An unverified claim lapses after eight hours and
-releases the domain, so a typo cannot park someone else's domain indefinitely.
-
-Proof by delegated name (CNAME) is refused outright. It would have required a wildcard zone we do not
-operate, and worse, it would turn a subdomain takeover into an apex takeover.
-
-**SSO-only** is available per organization: passwords refused, other providers refused, the instance
-Google button refused. Turning it *off* again needs a platform admin to approve the request, so one
-compromised org admin cannot quietly reopen password login. Break-glass for a platform admin is the
-correct password and nothing else — and a wrong password returns the same 403 everyone else gets, so
-it cannot be used to discover whether an account exists.
-
-With no SSO environment variables set, the product behaves exactly as it did before. That was
-verified in a browser, not merely reasoned about.
-
-### Added — organizations may re-enable same-origin widgets, deliberately
-Widget isolation removed `allow-same-origin`, which also broke embedding for sites that enforce strict
-CORS. There is now an org-level switch to put it back, behind a modal that requires the operator to
-type an acknowledgement, with a persistent banner while it is on. Enabling it needs an organization
-owner or admin — a workspace admin is deliberately not enough — and the change is written to the
-activity log. Contributed by @ChrisChrome.
-
-The **widget editor's Preview is excluded** from that switch. Preview renders inside the dashboard,
-where the admin's session token lives, so honouring the setting there would have let anyone who can
-author a widget lift the session of whichever admin clicked Preview. The setting exists so *displays*
-can embed origin-strict sites; a display holds a device token, an admin's browser does not.
-
-### Fixed — RSS tickers ran at a speed that depended on how much news there was
-`scroll_speed` was wired straight into `animation-duration`, so it set a fixed total time for the
-whole strip to cross the screen regardless of length. A feed with twenty items was dragged past in
-the same seconds as a feed with one — too fast to read, and it appeared to jump back to the start.
-It now calibrates a constant pixels-per-second rate, so more items simply take proportionally longer
-and every item scrolls fully into and out of view. Contributed by @ChrisChrome.
-
-### Fixed — user-controlled text is escaped where it actually reaches HTML
-An audit pass over the frontend's HTML sinks, escaping the ones that receive user-controlled data.
-Also in this release: dashboard banners no longer overlap the sidebar, shift the layout or vanish
-when switching views, and the main content no longer collapses to a narrow column.
-
-### Known limitations in this alpha
-Deliberately not resolved yet, and worth knowing before testing against them:
-
-- Enabling SSO-only **clears the passwords** of members at verified domains. That is irreversible
+- Requiring single sign-on **clears the passwords** of members at verified domains, irreversibly
   without a reset.
-- The SSO-only removal queue is an availability dependency on the operator: if nobody approves, the
-  organization stays SSO-only.
+- Turning that requirement back off depends on a platform administrator approving the request; if
+  nobody does, the organization stays on single sign-on.
 - `landing.html` still interpolates plan names into HTML without escaping. Those values come from
   the plans table rather than from end users, so it is a loose end rather than an exposure.
-- `/api/provision` is limited to 5/min, so a twenty-display install day takes four minutes of waiting.
-  Pre-existing, unchanged by this release.
+- `/api/provision` is limited to 5 requests per minute, so a twenty-display install day involves
+  some waiting. Pre-existing and unchanged by this release.
+
+### Thanks
+This release — and a good deal of what came before it — exists because people outside the project
+reported problems and sent patches. Credit was recorded inconsistently at the time, so it is
+collected here rather than left scattered.
+
+**Code contributed**
+
+- **@ChrisChrome** — the organization-level widget sandbox toggle (#254) and the RSS ticker rate fix,
+  both in this release. Earlier: the Debian player/server install script (#137) and web player
+  auto-connect (#6).
+- **@BlazzzPlay** — eight merged pull requests across 1.9.4 to 1.9.13: server-side preview sessions
+  to work around CSP (#151), the Android hidden settings menu (#152), sending device identity on
+  reconnect before pairing (#164), the dashboard version indicator and update check (#165, #181),
+  authenticated thumbnail loading (#182), the server URL in the Add Display modal and the Releases
+  link on the APK download page (#210), and uploads respecting the current folder (#211).
+- **@a10kiloham** — boot-time thumbnail healing with ffmpeg diagnostics and packaging (#244), the
+  screenshot-request verdict toast and the reverse-proxy header pitfall it documented (#243), and a
+  configurable maximum upload size (#233).
+- **@albanobattistella** — the Italian translation, and its updates since (#2, #145, #232).
+
+**Reported**
+
+- **@carloblu74** — the Raspberry Pi 5 report behind #245, which found five defects in the installer
+  and kiosk launcher that nothing in this repository would have caught, because nothing here had ever
+  executed those scripts on a Pi. The runbook notes above come from it.
+- **@bold-media-group** — by a wide margin the largest source of field reports, across roughly fifty
+  issues: the OTA rollout and version-advertising problems, event-loop lag under long uptime, video
+  wall behaviour, Tizen playback regressions, and the content-loading failures that led to resumable
+  downloads.
+- **@Smiley-k**, **@Semetra22**, **@patrickfinardi09**, **@hapishyguy**, **@Nikhil12656**,
+  **@gittyguy92** and **@Obe-BoldMediaGroup** — bug reports and feature requests across the 1.9.x
+  line, including SMTP transport, playlist item scheduling, and the Android playlist-order fault
+  behind #234.
+
+Several of the hardest faults this year were found by someone running the product on hardware the
+project does not own. That is worth saying plainly.
 
 ## 1.9.33
 

VERSION

@@ -1 +1 @@
-1.9.34-alpha14
+1.9.36

android/app/build.gradle.kts

@@ -13,8 +13,8 @@ android {
         targetSdk = 34
         // Env-overridable so device-owner reinstalls (which require an ever-increasing
         // versionCode — downgrades are blocked) don't churn this file each build.
-        versionCode = (System.getenv("VERSION_CODE") ?: findProperty("VERSION_CODE") as String? ?: "120").toInt()
-        versionName = System.getenv("VERSION_NAME") ?: findProperty("VERSION_NAME") as String? ?: "1.9.34-alpha14"
+        versionCode = (System.getenv("VERSION_CODE") ?: findProperty("VERSION_CODE") as String? ?: "123").toInt()
+        versionName = System.getenv("VERSION_NAME") ?: findProperty("VERSION_NAME") as String? ?: "1.9.36"
     }
 
     signingConfigs {
@@ -87,8 +87,24 @@ dependencies {
     implementation("androidx.media3:media3-exoplayer:1.2.1")
     implementation("androidx.media3:media3-ui:1.2.1")
 
-    // Socket.IO client
-    implementation("io.socket:socket.io-client:2.1.0")
+    // Socket.IO client.
+    //
+    // org.json is excluded deliberately. socket.io-client pulls org.json:json:20090211
+    // transitively, and that artifact was being packaged into the APK in full — 19 classes,
+    // including CDL, XML, JSONML and its own Test class. It carries the JSON License, whose
+    // "shall be used for Good, not Evil" clause is not OSI-approved, is treated as non-free by
+    // Debian and Fedora, and is Category X at Apache. Shipping it in a commercially distributed
+    // binary is an avoidable licensing problem: it is not copyleft, but it is not a licence we
+    // want to have to explain.
+    //
+    // Nothing is lost. Android provides org.json in the platform (since API 1, and minSdk is 24),
+    // and the only classes either side actually touches are JSONObject, JSONArray and JSONTokener.
+    // The full method surface used — by socket.io/engine.io and by our own Kotlin — is
+    // get/getString/getLong/getJSONArray/getJSONObject/has/keys/length/isNull/put/NULL,
+    // the opt* family, and JSONTokener.nextValue. Every one is platform API.
+    implementation("io.socket:socket.io-client:2.1.0") {
+        exclude(group = "org.json", module = "json")
+    }
 
     // WorkManager for background downloads
     implementation("androidx.work:work-runtime-ktx:2.9.0")

android/app/src/main/java/com/remotedisplay/player/MainActivity.kt

@@ -941,6 +941,14 @@ class MainActivity : AppCompatActivity() {
                     if (::updateChecker.isInitialized) updateChecker.checkForUpdate(forced = true)
                 }
                 // #161 device-owner tooling: push + silently install an arbitrary APK from a URL.
+                // Escape hatch for a panel holding a stale/bad staged APK: drop every cached file
+                // so the next check downloads afresh. Only ever deletes caches.
+                "clear_update_cache" -> {
+                    if (::updateChecker.isInitialized) {
+                        val n = updateChecker.clearUpdateCache()
+                        Log.i("MainActivity", "clear_update_cache removed $n file(s)")
+                    }
+                }
                 "install_apk" -> {
                     val url = payload?.optString("url", "") ?: ""
                     if (url.isNotBlank() && ::updateChecker.isInitialized) {

android/app/src/main/java/com/remotedisplay/player/service/UpdateChecker.kt

@@ -415,7 +415,7 @@ class UpdateChecker(private val context: Context) {
             // #139: reuse a previously-downloaded, verified APK for this version instead of
             // re-pulling ~8.7 MB every cycle. The file also stays on disk as the artifact for a
             // manual install when silent install isn't possible.
-            if (apkFile.exists() && verifyApkSignature(apkFile)) {
+            if (apkFile.exists() && cachedApkIs(apkFile, version) && verifyApkSignature(apkFile)) {
                 Log.i(TAG, "Reusing cached verified APK: ${apkFile.absolutePath} (${apkFile.length()} bytes)")
                 handler.post { installApk(apkFile) }
                 return true
@@ -448,6 +448,17 @@ class UpdateChecker(private val context: Context) {
             // Verify the downloaded APK is our package AND signed by the same key as
             // the currently-installed app before installing. An attacker can't forge
             // our signature, so this holds even over an untrusted transport.
+            // The server advertises a version and separately serves a file; the two can drift. A
+            // stale APK behind a current version number installs as a NO-OP, so the version never
+            // changes, the update is attempted again, and the panel loops until its attempts are
+            // spent — reporting a download failure, which it is not. Say what actually happened.
+            if (!cachedApkIs(apkFile, version)) {
+                val got = apkVersionName(apkFile) ?: "unreadable"
+                lastFailure = "server served $got but advertised $version — the update on the server is stale"
+                Log.e(TAG, "Version mismatch: advertised $version, downloaded $got")
+                apkFile.delete()
+                return false
+            }
             if (!verifyApkSignature(apkFile)) {
                 // lastFailure was set precisely inside verifyApkSignature; keep it, and add the
                 // size so a truncated download is distinguishable from a genuine cert mismatch.
@@ -585,6 +596,52 @@ class UpdateChecker(private val context: Context) {
 
     // True only if the downloaded APK is this same package and shares a signing
     // certificate with the installed app. Fail-closed on any error.
+    /* The versionName inside an APK file, or null if it cannot be read. */
+    private fun apkVersionName(apkFile: File): String? = try {
+        context.packageManager.getPackageArchiveInfo(apkFile.absolutePath, 0)?.versionName
+    } catch (e: Throwable) {
+        Log.w(TAG, "Could not read version from ${apkFile.name}: ${e.message}")
+        null
+    }
+
+    /*
+     * Is this file actually the version we mean to install?
+     *
+     * The cache is keyed by FILENAME, and the filename is built from the version the server
+     * advertised — so a file called ScreenTinker-1.9.34.apk containing 1.9.33 passes a signature
+     * check (same key), gets reused on every attempt, and installs as a no-op forever. Fixing the
+     * server does not clear it; only deleting the file does. Checking the version inside makes that
+     * self-healing instead of needing a hand on the device.
+     */
+    private fun cachedApkIs(apkFile: File, version: String): Boolean {
+        val got = apkVersionName(apkFile) ?: return false
+        if (got == version) return true
+        Log.w(TAG, "Cached ${apkFile.name} contains $got, expected $version — discarding")
+        return false
+    }
+
+    /*
+     * Delete every staged APK. The escape hatch for a panel holding a bad download: it forces the
+     * next check to fetch again rather than reuse. Safe at any time — these are only ever caches,
+     * re-fetched on demand.
+     */
+    fun clearUpdateCache(): Int {
+        var n = 0
+        for (dir in listOfNotNull(
+            File(context.filesDir, "Download"),
+            context.getExternalFilesDir(Environment.DIRECTORY_DOWNLOADS),
+            File(context.cacheDir, "Download"),
+        )) {
+            val files = try { dir.listFiles() } catch (_: Throwable) { null } ?: continue
+            for (f in files) {
+                if (!f.name.endsWith(".apk")) continue
+                if (f.delete()) n++
+            }
+        }
+        report("info", "Update cache cleared ($n file(s)) — the next check will download afresh")
+        return n
+    }
+
     private fun verifyApkSignature(apkFile: File): Boolean {
         return try {
             val pm = context.packageManager

android/licenses.json

@@ -0,0 +1,45 @@
+{
+  "_comment": [
+    "Licence policy for everything on the Android release runtime classpath — i.e. everything that",
+    "can end up inside the APK customers install.",
+    "",
+    "scripts/android-license-check.js resolves the real classpath and checks it against this file.",
+    "An artifact that appears in neither 'artifacts' nor 'groups' FAILS: a new transitive dependency",
+    "must be looked at by a person before it ships, which is exactly how org.json:json:20090211 got",
+    "into the APK unnoticed in the first place.",
+    "",
+    "Record what you verified in 'evidence' — the point is to be able to answer 'how do you know?'"
+  ],
+
+  "groups": {
+    "androidx":                   { "license": "Apache-2.0", "evidence": "AndroidX / Jetpack, Apache-2.0 across the board" },
+    "com.google.android.material": { "license": "Apache-2.0", "evidence": "Material Components for Android" },
+    "com.google.code.gson":       { "license": "Apache-2.0", "evidence": "google/gson LICENSE" },
+    "com.google.crypto.tink":     { "license": "Apache-2.0", "evidence": "google/tink LICENSE" },
+    "com.google.errorprone":      { "license": "Apache-2.0", "evidence": "google/error-prone LICENSE" },
+    "com.google.guava":           { "license": "Apache-2.0", "evidence": "google/guava LICENSE" },
+    "com.google.j2objc":          { "license": "Apache-2.0", "evidence": "google/j2objc LICENSE" },
+    "com.squareup.okhttp3":       { "license": "Apache-2.0", "evidence": "square/okhttp LICENSE" },
+    "com.squareup.okio":          { "license": "Apache-2.0", "evidence": "square/okio LICENSE" },
+    "org.checkerframework":       { "license": "MIT",        "evidence": "checker-framework, MIT for the qualifiers" },
+    "org.jetbrains":              { "license": "Apache-2.0", "evidence": "JetBrains annotations" },
+    "org.jetbrains.kotlin":       { "license": "Apache-2.0", "evidence": "Kotlin stdlib" },
+    "org.jetbrains.kotlinx":      { "license": "Apache-2.0", "evidence": "kotlinx coroutines" },
+    "io.socket":                  { "license": "MIT",        "evidence": "socket.io-client-java LICENSE (MIT)" }
+  },
+
+  "artifacts": {},
+
+  "denied": {
+    "org.json:json": {
+      "why": "JSON Licence — the 'shall be used for Good, not Evil' clause. Not OSI-approved, non-free per Debian and Fedora, Apache Category X. Arrives transitively via socket.io-client and was previously packaged into the APK in full (19 classes). Excluded in app/build.gradle.kts; Android provides org.json in the platform from API 1 and minSdk is 24, so nothing is lost."
+    }
+  },
+
+  "denied_licenses": [
+    { "match": "AGPL", "why": "network copyleft" },
+    { "match": "GPL",  "why": "strong copyleft in a commercially distributed binary" },
+    { "match": "SSPL", "why": "not OSI-approved, service-scope obligations" },
+    { "match": "JSON", "why": "field-of-use restriction" }
+  ]
+}

docs/licensing.md

@@ -0,0 +1,103 @@
+# Licensing
+
+ScreenTinker is MIT. This page records how we know what our dependencies are licensed under,
+so the answer to "do you track licences?" is something you can check rather than something you
+have to take on trust.
+
+## The short answer
+
+**No GPL or AGPL anywhere in the product.** Neither the server nor the Android player links,
+bundles, or ships anything under strong or network copyleft.
+
+## Where the answer comes from
+
+Two gates run in CI on every push, and both fail closed — a dependency whose licence nobody has
+recorded fails the build rather than shipping unnoticed.
+
+| Gate | Covers | Script |
+|---|---|---|
+| Licence gate + SBOM (production deps) | the server's npm tree | `scripts/license-check.js` |
+| Licence gate (APK runtime classpath) | everything that can enter the APK | `scripts/android-license-check.js` |
+
+Run either locally:
+
+```sh
+cd server && npm ci --omit=dev && cd ..
+node scripts/license-check.js                       # server
+node scripts/android-license-check.js               # APK
+node scripts/license-check.js --sbom sbom/x.json    # also write an SBOM
+```
+
+Neither script has dependencies of its own. A gate that needs its own supply chain audited is
+worth less than one that doesn't.
+
+## ⚠️ Audit the production install, not the checkout
+
+**A licence scanner pointed at a developer checkout will report LGPL, and it will be wrong about
+what we ship.**
+
+`sharp` is a `devDependency` — a fixture generator for the image tests — and one of its platform
+binaries, `@img/sharp-wasm32`, declares `Apache-2.0 AND LGPL-3.0-or-later AND MIT`. It is never
+installed on a server: production installs with `npm ci --omit=dev`, which both the CI gate and
+`scripts/upgrade.sh` use.
+
+If someone challenges the answer with a scan of the repo, this is the discrepancy they have found.
+
+`sharp` is kept deliberately: it is the *independent* implementation used to generate fixtures for
+the pure-JavaScript image path that replaced it. Generating those fixtures with the library under
+test would mean a decode bug could produce a fixture that hides the same bug.
+
+## Policy
+
+**Allowed** — MIT, MIT-0, ISC, 0BSD, BSD-2-Clause, BSD-3-Clause, Apache-2.0, BlueOak-1.0.0,
+Unlicense, CC0-1.0, Python-2.0, WTFPL, Zlib, CC-BY-4.0.
+
+**Denied** — AGPL, GPL, SSPL, Commons Clause, BUSL, and the JSON Licence.
+
+**Reported but not failed** — LGPL, MPL, EPL, CDDL, OSL, EUPL. Weak copyleft is file- or
+library-scoped and usually fine when merely linked, but it is a judgement, and the judgement should
+be made by someone who knows they are making it.
+
+**Unrecognised — fails.** A package with no licence we can identify is not a package we ship. Where
+a dependency ships a real licence *file* but declares no `license` field, it is recorded as an
+exception in the script with the evidence that was read off disk (currently `exif-parser` and
+`thirty-two`, both MIT).
+
+### Why the JSON Licence is denied
+
+`org.json:json:20090211` arrived transitively through `socket.io-client` and was **packaged into the
+APK in full** — 19 classes, including ones nothing referenced. Its licence carries the clause *"The
+Software shall be used for Good, not Evil"*: not OSI-approved, treated as non-free by Debian and
+Fedora, and Category X at Apache. Not copyleft, but not a term to accept in a binary distributed
+commercially.
+
+It is now excluded in `android/app/build.gradle.kts`. Nothing is lost — Android has provided
+`org.json` in the platform since API 1 and `minSdk` is 24 — and `android/licenses.json` denies it by
+name so it cannot return quietly.
+
+## SBOM
+
+Every release publishes `screentinker-sbom-<version>.cdx.json`: **CycloneDX 1.5**, listing every
+production dependency with its version, package URL, and licence. Generated from a production
+install, so it describes what actually runs.
+
+CI also uploads one as a build artifact on every run.
+
+## Vendored code
+
+Anything committed under `frontend/vendor/` **ships in the release tarball** and must carry its
+licence notice as a separate file — minifiers strip headers, which is exactly when the notice has to
+be kept alongside. See `frontend/vendor/README.md`.
+
+## The GLSL transitions
+
+The 14 shaders in `shared/Transitions/` are original work. Each carries its author and licence in
+the file header, and none derives from Shadertoy, gl-transitions, glslsandbox or similar. "GL
+Transitions v1" in those headers refers to the *interface convention* — the function signature the
+renderer calls — not to borrowed code.
+
+## Limits
+
+These gates identify licences from declared metadata and recorded evidence. They are not a
+clean-room provenance review, and they do not detect code copied into the repository without
+attribution.

docs/openapi.yaml

@@ -1,7 +1,7 @@
 openapi: 3.1.0
 info:
   title: ScreenTinker Public API
-  version: 1.9.34
+  version: 1.9.36
   description: |
     Public, token-scoped REST API for ScreenTinker digital signage.
 

frontend/js/i18n/en.js

@@ -683,6 +683,8 @@ export default {
   'device.ctl.screen_on': 'Screen On',
   'device.ctl.launch_player': 'Launch Player',
   'device.ctl.force_update': 'Force Update',
+  'device.ctl.clear_update_cache': 'Clear Update Cache',
+  'device.ctl.clear_update_cache_tip': 'Delete any update file this display has already downloaded, so the next check fetches a fresh copy. Use if updates keep failing.',
   'device.ctl.shutdown': 'Shutdown',
   // Remote tab
   'device.remote.start_prompt': 'Click "Start Remote" to begin',
@@ -773,6 +775,7 @@ export default {
   'device.toast.screen_on_sent': 'Screen on command sent',
   'device.toast.launch_sent': 'Launch command sent',
   'device.toast.update_triggered': 'Update check triggered',
+  'device.toast.update_cache_cleared': 'Update cache cleared — the next check will download afresh',
   'device.toast.remote_started': 'Remote session started',
   'device.toast.command_queued': '{cmd} — device offline, will deliver on reconnect',
   'device.toast.command_undeliverable': '{cmd} — device offline and queue unavailable',

frontend/js/views/device-detail.js

@@ -501,6 +501,12 @@ async function loadDevice(deviceId, activeTab = null) {
               <path d="M21 15v4a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2v-4"/><polyline points="7 10 12 15 17 10"/><line x1="12" y1="15" x2="12" y2="3"/>
             </svg>
             ${t('device.ctl.force_update')}
+          </button>
+          <button class="btn btn-secondary btn-sm" id="clearUpdateCacheBtn" title="${t('device.ctl.clear_update_cache_tip')}">
+            <svg width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2">
+              <polyline points="3 6 5 6 21 6"/><path d="M19 6v14a2 2 0 0 1-2 2H7a2 2 0 0 1-2-2V6m3 0V4a2 2 0 0 1 2-2h4a2 2 0 0 1 2 2v2"/>
+            </svg>
+            ${t('device.ctl.clear_update_cache')}
           </button>` : ''}
           ${can('system.reboot') ? `
           <button class="btn btn-danger btn-sm" id="shutdownBtn">
@@ -1593,6 +1599,12 @@ function setupActions(device) {
     sendWithFeedback('update', 'Update', 'device.toast.update_triggered');
   });
 
+  // Drops every staged APK on the panel so the next check downloads afresh. The escape hatch for a
+  // player holding a bad download — a cached file that cannot install but is reused every attempt.
+  document.getElementById('clearUpdateCacheBtn')?.addEventListener('click', () => {
+    sendWithFeedback('clear_update_cache', 'Clear update cache', 'device.toast.update_cache_cleared');
+  });
+
   // #109: PiP overlay tester — pushes/clears an overlay via the public API (POST /api/pip).
   document.getElementById('sendPipBtn')?.addEventListener('click', async () => {
     const uri = (document.getElementById('pipUri')?.value || '').trim();

frontend/landing.html

@@ -210,6 +210,14 @@
       </a>
       <a href="#compare" class="btn btn-outline" style="padding:14px 28px;font-size:16px">See How We Compare</a>
     </div>
+
+    <!-- Live deployment count. Hidden until the number arrives, so a self-hosted instance (where
+         the endpoint does not exist) and a brand-new one (where the count is 0) show nothing at
+         all rather than an empty frame or a zero. -->
+    <p id="deployed-stat" hidden style="margin-top:28px;color:var(--muted);font-size:15px">
+      <strong id="deployed-count" style="color:var(--text,inherit);font-variant-numeric:tabular-nums"></strong>
+      screens deployed with ScreenTinker
+    </p>
   </section>
 
   <!-- Intro video -->
@@ -630,6 +638,19 @@
     // replace it on the public marketing page with a hardcoded Contact Us
     // card. Other consumers of /api/subscription/plans (billing.js,
     // settings.js, admin.js) get the full list as before.
+    /* Screens deployed. Only the deployment that collects install statistics answers this;
+       everywhere else it 404s and the line stays hidden. Failures are silent by design —
+       a marketing page must not show a broken stat, and there is nothing a visitor could
+       do about it. */
+    fetch('/api/public/stats')
+      .then(r => (r.ok ? r.json() : null))
+      .then(s => {
+        if (!s || !(s.screens > 0)) return;
+        document.getElementById('deployed-count').textContent = s.screens.toLocaleString();
+        document.getElementById('deployed-stat').hidden = false;
+      })
+      .catch(() => {});
+
     fetch('/api/subscription/plans').then(r => r.json()).then(plans => {
       const grid = document.getElementById('pricingGrid');
       const publicPlans = plans.filter(p => p.active && p.name !== 'enterprise');

frontend/vendor/README.md

@@ -4,9 +4,16 @@ Third-party libraries committed directly to the repo (not fetched from a CDN or
 from npm) so self-hosted / air-gapped instances work with no external dependency and no
 build step.
 
+**Anything added here ships in the release tarball**, so it must carry its licence notice —
+a minified bundle usually has its headers stripped, which is exactly when the notice has to
+be kept as a separate file next to it. Record the licence below and add a `<name>.LICENSE`.
+
 ## redoc.standalone.js
 - **Library:** Redoc — renders the OpenAPI reference served at `/docs`.
 - **Version:** 2.3.9
+- **Licence:** MIT — Copyright (c) 2015-present, Rebilly, Inc. Full text in
+  [`redoc.LICENSE`](redoc.LICENSE). The bundle itself carries no header (stripped by the
+  upstream minifier), which is why the notice is kept separately.
 - **Source:** https://cdn.redoc.ly/redoc/v2.3.9/bundles/redoc.standalone.js
 - **Why committed:** the API reference must render on offline instances — no CDN, no build step.
 - **Regenerate / update:**

frontend/vendor/redoc.LICENSE

@@ -0,0 +1,31 @@
+Redoc — https://github.com/Redocly/redoc
+Version vendored here: 2.3.9 (see redoc.standalone.js)
+
+The bundle in this directory is minified and its license headers were stripped upstream, so
+the notice is kept alongside it instead. MIT requires this notice to accompany the software
+wherever it is distributed, and redoc.standalone.js is included in the ScreenTinker release
+tarball.
+
+--------------------------------------------------------------------------------
+
+The MIT License (MIT)
+
+Copyright (c) 2015-present, Rebilly, Inc.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

scripts/android-license-check.js

@@ -0,0 +1,112 @@
+#!/usr/bin/env node
+'use strict';
+
+/*
+ * Licence gate for the APK.
+ *
+ *   node scripts/android-license-check.js [--sbom <path>]
+ *
+ * Resolves the real `releaseRuntimeClasspath` — every artifact that can end up inside the APK a
+ * customer installs, transitive ones included — and checks each against android/licenses.json.
+ *
+ * Fails on an artifact nobody has recorded a licence for. That is the case worth catching:
+ * org.json:json:20090211 reached customers because it arrived as a transitive dependency of
+ * socket.io-client and nothing ever asked what licence it carried.
+ */
+
+const fs = require('fs');
+const path = require('path');
+const { execFileSync } = require('child_process');
+
+const ROOT = path.join(__dirname, '..');
+const ANDROID = path.join(ROOT, 'android');
+const POLICY = JSON.parse(fs.readFileSync(path.join(ANDROID, 'licenses.json'), 'utf8'));
+const SBOM_OUT = process.argv.includes('--sbom') ? process.argv[process.argv.indexOf('--sbom') + 1] : null;
+
+function resolveClasspath() {
+  const out = execFileSync('./gradlew', ['-q', 'app:dependencies', '--configuration', 'releaseRuntimeClasspath'],
+    { cwd: ANDROID, maxBuffer: 64 * 1024 * 1024, encoding: 'utf8' });
+  const found = new Map();
+  for (const raw of out.split('\n')) {
+    // Gradle prints "group:name:requested -> resolved" when a version is upgraded; the resolved
+    // one is what ships, so prefer the right-hand side.
+    const m = raw.match(/([a-zA-Z0-9._-]+):([a-zA-Z0-9._-]+):([0-9][a-zA-Z0-9._-]*)(?:\s*->\s*([0-9][a-zA-Z0-9._-]*))?/);
+    if (!m) continue;
+    const [, group, name, requested, upgraded] = m;
+    found.set(`${group}:${name}`, { group, name, version: upgraded || requested });
+  }
+  return [...found.values()].sort((a, b) => `${a.group}:${a.name}`.localeCompare(`${b.group}:${b.name}`));
+}
+
+function licenceFor(a) {
+  const coord = `${a.group}:${a.name}`;
+  if (POLICY.denied[coord]) return { verdict: 'DENY', why: POLICY.denied[coord].why };
+  if (POLICY.artifacts[coord]) return { verdict: 'ALLOW', ...POLICY.artifacts[coord] };
+  // Longest matching group prefix wins, so a specific rule beats a broad one.
+  const groups = Object.keys(POLICY.groups)
+    .filter(g => a.group === g || a.group.startsWith(g + '.'))
+    .sort((x, y) => y.length - x.length);
+  if (groups.length) return { verdict: 'ALLOW', ...POLICY.groups[groups[0]] };
+  return { verdict: 'UNKNOWN' };
+}
+
+const artifacts = resolveClasspath();
+if (!artifacts.length) {
+  console.error('Resolved no artifacts — the gradle task did not run properly. Refusing to pass.');
+  process.exit(2);
+}
+
+const results = artifacts.map(a => ({ ...a, ...licenceFor(a) }));
+const denied = results.filter(r => r.verdict === 'DENY');
+const unknown = results.filter(r => r.verdict === 'UNKNOWN');
+
+for (const r of results.filter(r => r.verdict === 'ALLOW')) {
+  for (const d of POLICY.denied_licenses) {
+    if (new RegExp(d.match, 'i').test(r.license)) { denied.push({ ...r, why: `${r.license}: ${d.why}` }); }
+  }
+}
+
+const counts = results.reduce((m, r) => (m[r.license || '(unrecorded)'] = (m[r.license || '(unrecorded)'] || 0) + 1, m), {});
+console.log(`\nScope: ${results.length} artifacts on releaseRuntimeClasspath (everything that can enter the APK)\n`);
+Object.entries(counts).sort((a, b) => b[1] - a[1]).forEach(([l, n]) => console.log(`  ${String(n).padStart(4)}  ${l}`));
+
+if (SBOM_OUT) {
+  const sbom = {
+    bomFormat: 'CycloneDX',
+    specVersion: '1.5',
+    version: 1,
+    metadata: {
+      component: {
+        type: 'application',
+        name: 'screentinker-android-player',
+        version: fs.readFileSync(path.join(ROOT, 'VERSION'), 'utf8').trim(),
+        licenses: [{ license: { id: 'MIT' } }],
+      },
+    },
+    components: results.map(r => ({
+      type: 'library',
+      name: `${r.group}:${r.name}`,
+      version: r.version,
+      purl: `pkg:maven/${r.group}/${r.name}@${r.version}`,
+      licenses: r.license ? [{ license: { id: r.license } }] : [],
+    })),
+  };
+  fs.mkdirSync(path.dirname(SBOM_OUT), { recursive: true });
+  fs.writeFileSync(SBOM_OUT, JSON.stringify(sbom, null, 2));
+  console.log(`\nSBOM: ${SBOM_OUT} (${sbom.components.length} components, CycloneDX 1.5)`);
+}
+
+let failed = false;
+if (denied.length) {
+  failed = true;
+  console.log('\nDENIED');
+  denied.forEach(r => console.log(`  ${r.group}:${r.name}:${r.version}\n      ${r.why}`));
+}
+if (unknown.length) {
+  failed = true;
+  console.log('\nUNRECORDED — a new dependency reached the APK with no licence on file.');
+  console.log('Look it up, then add it to android/licenses.json with evidence, or exclude it.');
+  unknown.forEach(r => console.log(`  ${r.group}:${r.name}:${r.version}`));
+}
+console.log(failed ? '\nFAIL: licence policy violated.\n' : '\nOK: every artifact in the APK has a recorded, permitted licence.\n');
+process.exit(failed ? 1 : 0);

scripts/license-check.js

@@ -0,0 +1,184 @@
+#!/usr/bin/env node
+'use strict';
+
+/*
+ * Licence gate for the dependencies that actually SHIP.
+ *
+ *   node scripts/license-check.js [--sbom <path>] [--include-dev]
+ *
+ * Run from a PRODUCTION install (`npm ci --omit=dev`). That is the whole point: a developer
+ * checkout carries `sharp`, whose `@img/sharp-wasm32` declares LGPL-3.0-or-later. It is a test
+ * fixture generator, it is devDependencies-only, and it never reaches a server — but a scanner
+ * pointed at a dev tree reports LGPL and contradicts the answer we give customers. Auditing the
+ * installed production tree is what makes the answer defensible.
+ *
+ * Exits non-zero on anything denied or unresolved, so CI fails before a licence can arrive
+ * unnoticed through a transitive bump.
+ *
+ * No dependencies, deliberately — a gate that needs its own supply chain audited is worth less.
+ */
+
+const fs = require('fs');
+const path = require('path');
+const { execFileSync } = require('child_process');
+
+const args = process.argv.slice(2);
+const INCLUDE_DEV = args.includes('--include-dev');
+const SBOM_OUT = args.includes('--sbom') ? args[args.indexOf('--sbom') + 1] : null;
+const SERVER_DIR = path.join(__dirname, '..', 'server');
+
+/* ── policy ───────────────────────────────────────────────────────────────────
+ * ALLOW: permissive, no distribution obligation beyond keeping the notice.
+ * DENY:  strong/network copyleft, plus licences we will not ship for other reasons.
+ * Anything matching neither is REVIEW — it fails, and a human decides. Failing closed
+ * matters more than being clever: the risk is a licence arriving that nobody looked at.
+ */
+const ALLOW = [
+  /^MIT$/i, /^MIT-0$/i, /^ISC$/i, /^0BSD$/i, /^BSD-2-Clause$/i, /^BSD-3-Clause$/i,
+  /^Apache-2\.0$/i, /^BlueOak-1\.0\.0$/i, /^Unlicense$/i, /^CC0-1\.0$/i, /^Python-2\.0$/i,
+  /^WTFPL$/i, /^Zlib$/i, /^CC-BY-4\.0$/i,
+];
+
+const DENY = [
+  { re: /\bAGPL/i,               why: 'network copyleft — obligations trigger on serving, not distributing' },
+  { re: /\bGPL-[123]|\bGPLv[123]|(^|[^L])\bGPL\b/i, why: 'strong copyleft — links into a product we distribute commercially' },
+  { re: /\bSSPL/i,               why: 'server-side public licence — not OSI-approved, service-scope obligations' },
+  { re: /\bCommons-Clause/i,     why: 'commercial-use restriction' },
+  { re: /\bBUSL|Business Source/i, why: 'source-available, not open source' },
+  { re: /Good, not Evil|^JSON$/i,  why: 'JSON Licence — field-of-use clause, Apache Category X, non-free per Debian/Fedora' },
+];
+
+// Weak copyleft: file- or library-scoped, generally fine when merely linked, but never silently.
+const REVIEW = [/\bLGPL/i, /\bMPL/i, /\bEPL/i, /\bCDDL/i, /\bOSL/i, /\bEUPL/i, /\bCPL/i];
+
+/*
+ * Packages that ship a real licence FILE but declare no `license` field in package.json.
+ * Each entry records what was read off disk, so this is a documented finding rather than a
+ * blanket exemption. Re-verify if the version changes.
+ */
+const EXCEPTIONS = {
+  'exif-parser': { license: 'MIT', evidence: 'LICENSE.md — "The MIT License"' },
+  'thirty-two':  { license: 'MIT', evidence: 'LICENSE.txt — MIT, Copyright (c) 2011 Chris Umbel' },
+  'screentinker': { license: 'MIT', evidence: 'repository root LICENSE' },
+};
+
+function classify(id) {
+  if (!id) return { verdict: 'UNKNOWN' };
+  for (const d of DENY) if (d.re.test(id)) return { verdict: 'DENY', why: d.why };
+  // A GPL-with-exception (Classpath, linking) is not the thing we are guarding against.
+  if (/WITH .*exception/i.test(id)) return { verdict: 'REVIEW', why: 'copyleft with a linking exception' };
+  for (const r of REVIEW) if (r.test(id)) return { verdict: 'REVIEW', why: 'weak copyleft' };
+  // Composite expressions: every term must be allowed.
+  const terms = id.split(/\s+(?:OR|AND)\s+|[()]/).map(s => s.trim()).filter(Boolean);
+  if (terms.length && terms.every(t => ALLOW.some(a => a.test(t)))) return { verdict: 'ALLOW' };
+  return { verdict: 'UNKNOWN' };
+}
+
+function readLicense(dir) {
+  let pkg;
+  try { pkg = JSON.parse(fs.readFileSync(path.join(dir, 'package.json'), 'utf8')); } catch { return null; }
+  let lic = pkg.license;
+  if (lic && typeof lic === 'object') lic = lic.type;
+  if (!lic && Array.isArray(pkg.licenses)) lic = pkg.licenses.map(l => l.type || l).join(' OR ');
+  return { name: pkg.name, version: pkg.version, license: lic || null };
+}
+
+/*
+ * `npm ls` exits non-zero for any tree problem — an extraneous package, a peer-dep complaint —
+ * while still printing the full listing. Treating that as fatal would turn a routine tree quirk
+ * into an unexplained CI failure, and worse, a licence check that never actually ran. Read the
+ * output either way; a genuinely empty result is the only thing worth aborting on.
+ */
+function listInstalled() {
+  const argv = ['ls', ...(INCLUDE_DEV ? [] : ['--omit=dev']), '--all', '--parseable'];
+  const opts = { cwd: SERVER_DIR, maxBuffer: 64 * 1024 * 1024, encoding: 'utf8' };
+  try {
+    return execFileSync('npm', argv, opts);
+  } catch (e) {
+    if (e.stdout && e.stdout.trim()) return e.stdout;
+    console.error('npm ls produced no output:\n' + (e.stderr || e.message));
+    process.exit(2);
+  }
+}
+
+const dirs = listInstalled().split('\n').filter(Boolean);
+
+const pkgs = [];
+const seen = new Set();
+for (const d of dirs) {
+  const info = readLicense(d);
+  if (!info || !info.name) continue;
+  const key = `${info.name}@${info.version}`;
+  if (seen.has(key)) continue;
+  seen.add(key);
+
+  let license = info.license;
+  let note = null;
+  if (!license && EXCEPTIONS[info.name]) {
+    license = EXCEPTIONS[info.name].license;
+    note = `no license field; ${EXCEPTIONS[info.name].evidence}`;
+  }
+  pkgs.push({ ...info, license, note, ...classify(license) });
+}
+
+const denied = pkgs.filter(p => p.verdict === 'DENY');
+const review = pkgs.filter(p => p.verdict === 'REVIEW');
+const unknown = pkgs.filter(p => p.verdict === 'UNKNOWN');
+
+const counts = pkgs.reduce((m, p) => (m[p.license || '(none)'] = (m[p.license || '(none)'] || 0) + 1, m), {});
+console.log(`\nScope: ${pkgs.length} packages (${INCLUDE_DEV ? 'INCLUDING dev' : 'production only, --omit=dev'})\n`);
+Object.entries(counts).sort((a, b) => b[1] - a[1]).forEach(([l, n]) => console.log(`  ${String(n).padStart(4)}  ${l}`));
+
+if (SBOM_OUT) {
+  // CycloneDX 1.5, hand-built. A standard format customers and underwriters recognise, without
+  // taking a dependency on a generator to produce it.
+  const sbom = {
+    bomFormat: 'CycloneDX',
+    specVersion: '1.5',
+    version: 1,
+    metadata: {
+      component: {
+        type: 'application',
+        name: 'screentinker',
+        version: fs.readFileSync(path.join(__dirname, '..', 'VERSION'), 'utf8').trim(),
+        licenses: [{ license: { id: 'MIT' } }],
+      },
+      properties: [{ name: 'screentinker:scope', value: INCLUDE_DEV ? 'all' : 'production' }],
+    },
+    components: pkgs
+      .filter(p => p.name !== 'screentinker')
+      .sort((a, b) => a.name.localeCompare(b.name))
+      .map(p => ({
+        type: 'library',
+        name: p.name,
+        version: p.version,
+        purl: `pkg:npm/${p.name.replace('@', '%40')}@${p.version}`,
+        licenses: p.license ? [{ license: /[()]| OR | AND /.test(p.license) ? { name: p.license } : { id: p.license } }] : [],
+      })),
+  };
+  fs.mkdirSync(path.dirname(SBOM_OUT), { recursive: true });
+  fs.writeFileSync(SBOM_OUT, JSON.stringify(sbom, null, 2));
+  console.log(`\nSBOM: ${SBOM_OUT} (${sbom.components.length} components, CycloneDX 1.5)`);
+}
+
+let failed = false;
+if (denied.length) {
+  failed = true;
+  console.log('\nDENIED');
+  denied.forEach(p => console.log(`  ${p.name}@${p.version}  ->  ${p.license}\n      ${p.why}`));
+}
+if (unknown.length) {
+  failed = true;
+  console.log('\nUNRESOLVED — no recognised licence. Read the package, then add it to EXCEPTIONS');
+  console.log('with the evidence, or remove the dependency.');
+  unknown.forEach(p => console.log(`  ${p.name}@${p.version}  ->  ${p.license || '(no license field)'}`));
+}
+if (review.length) {
+  // Not fatal, but never silent — weak copyleft is a judgement call, and the judgement should be
+  // made by a person who knows it is being made.
+  console.log('\nREVIEW (not failing)');
+  review.forEach(p => console.log(`  ${p.name}@${p.version}  ->  ${p.license}   ${p.why}`));
+}
+
+console.log(failed ? '\nFAIL: licence policy violated.\n' : '\nOK: no denied or unresolved licences.\n');
+process.exit(failed ? 1 : 0);

server/lib/player-capabilities.js

@@ -312,6 +312,9 @@ const COMMAND_CAPABILITY = {
   launch: 'system.restart_player',
   refresh: 'system.restart_player',
   update: 'system.self_update',
+  // Clearing the staged-APK cache is part of the same self-update surface: a player that can
+  // update itself is a player that can hold a bad download and needs a way to drop it.
+  clear_update_cache: 'system.self_update',
 
   // display
   screen_on: 'display.power',

server/package-lock.json

@@ -1,12 +1,13 @@
 {
   "name": "screentinker",
-  "version": "1.9.34-alpha14",
+  "version": "1.9.36",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "screentinker",
-      "version": "1.9.34-alpha14",
+      "version": "1.9.36",
+      "license": "MIT",
       "dependencies": {
         "@azure/msal-node": "^5.2.1",
         "@jsquash/avif": "^1.3.0",
@@ -21,7 +22,7 @@
         "jimp": "^1.6.1",
         "jsonwebtoken": "^9.0.3",
         "multer": "^1.4.5-lts.1",
-        "nodemailer": "^6.9.16",
+        "nodemailer": "^9.0.5",
         "otplib": "^12.0.1",
         "qrcode": "^1.5.4",
         "socket.io": "^4.7.2",
@@ -1941,9 +1942,9 @@
       }
     },
     "node_modules/brace-expansion": {
-      "version": "2.1.2",
-      "resolved": "https://registry.npmjs.org/brace-expansion/-/brace-expansion-2.1.2.tgz",
-      "integrity": "sha512-w5JZcKgdhDOgOwm8H+KgbosopHMuGcl6qbulwjtz3SM7I7P3yW1eAjzMPLrIE+NQ9vjgANKHWeMHnrT0OXW1oA==",
+      "version": "2.1.4",
+      "resolved": "https://registry.npmjs.org/brace-expansion/-/brace-expansion-2.1.4.tgz",
+      "integrity": "sha512-hGfVzPxthbf3+2yjg/RBs60cB0FhqBS/zvdV/4wn4/BmN0bNMMHPc4V/BbFieqf1TKAGGAHnY4eSjajCl0f2Xg==",
       "license": "MIT",
       "dependencies": {
         "balanced-match": "^1.0.0"
@@ -3434,9 +3435,9 @@
       "license": "ISC"
     },
     "node_modules/ip-address": {
-      "version": "10.2.0",
-      "resolved": "https://registry.npmjs.org/ip-address/-/ip-address-10.2.0.tgz",
-      "integrity": "sha512-/+S6j4E9AHvW9SWMSEY9Xfy66O5PWvVEJ08O0y5JGyEKQpojb0K0GKpz/v5HJ/G0vi3D2sjGK78119oXZeE0qA==",
+      "version": "10.5.0",
+      "resolved": "https://registry.npmjs.org/ip-address/-/ip-address-10.5.0.tgz",
+      "integrity": "sha512-R5SnVLJmgYYvf2F2ZgwSBnelz5G4q5AxIC277GDfUaNbrZKNANcBC7RHqYYePlszf4kBolVkJauG0ZjHHFh55g==",
       "license": "MIT",
       "engines": {
         "node": ">= 12"
@@ -3544,9 +3545,9 @@
       "license": "BSD-3-Clause"
     },
     "node_modules/js-yaml": {
-      "version": "4.3.0",
-      "resolved": "https://registry.npmjs.org/js-yaml/-/js-yaml-4.3.0.tgz",
-      "integrity": "sha512-1td788aAnnZ5qs7V2QIRl1owjtYpbKt749Y3xauqQgwIIGF/xXWz1wMTEBx5O3LK3lXLVuqXPdPxj2BoFHaW9Q==",
+      "version": "4.3.1",
+      "resolved": "https://registry.npmjs.org/js-yaml/-/js-yaml-4.3.1.tgz",
+      "integrity": "sha512-CY6crGq313MX8GkwvB7tzgp99vjQxY1++5y10/BKN/GUfHqWaOGQMNZkBvqSzsZKWk/ijwHlWzzkLulsGHhjWQ==",
       "dev": true,
       "funding": [
         {
@@ -3913,9 +3914,9 @@
       "license": "MIT"
     },
     "node_modules/nodemailer": {
-      "version": "6.10.1",
-      "resolved": "https://registry.npmjs.org/nodemailer/-/nodemailer-6.10.1.tgz",
-      "integrity": "sha512-Z+iLaBGVaSjbIzQ4pX6XV41HrooLsQ10ZWPUehGmuantvzWoDVBnmsdUcOIDM1t+yPor5pDhVlDESgOMEGxhHA==",
+      "version": "9.0.5",
+      "resolved": "https://registry.npmjs.org/nodemailer/-/nodemailer-9.0.5.tgz",
+      "integrity": "sha512-wvjiKvjczmsN7U/8006JOdXubgBk2XFAbioDMbT+sM7cPs0QrhJTa6KBRX7P5REGGkDcLUz/EarWidb8G8C1jQ==",
       "license": "MIT-0",
       "engines": {
         "node": ">=6.0.0"
@@ -4931,9 +4932,9 @@
       "license": "MIT"
     },
     "node_modules/socket.io-parser": {
-      "version": "4.2.6",
-      "resolved": "https://registry.npmjs.org/socket.io-parser/-/socket.io-parser-4.2.6.tgz",
-      "integrity": "sha512-asJqbVBDsBCJx0pTqw3WfesSY0iRX+2xzWEWzrpcH7L6fLzrhyF8WPI8UaeM4YCuDfpwA/cgsdugMsmtz8EJeg==",
+      "version": "4.2.7",
+      "resolved": "https://registry.npmjs.org/socket.io-parser/-/socket.io-parser-4.2.7.tgz",
+      "integrity": "sha512-IH/iSeO9T6gz1KkFleGDWkG9N3dl4jXVYUtMhIqH10Md0ttMer8nUNWiP1DKuNrybD2xBrixLJdCC9J6ECoYkg==",
       "license": "MIT",
       "dependencies": {
         "@socket.io/component-emitter": "~3.1.0",

server/package.json

@@ -1,6 +1,7 @@
 {
   "name": "screentinker",
-  "version": "1.9.34-alpha14",
+  "version": "1.9.36",
+  "license": "MIT",
   "description": "ScreenTinker - Digital Signage Management Server",
   "main": "server.js",
   "scripts": {
@@ -23,7 +24,7 @@
     "jimp": "^1.6.1",
     "jsonwebtoken": "^9.0.3",
     "multer": "^1.4.5-lts.1",
-    "nodemailer": "^6.9.16",
+    "nodemailer": "^9.0.5",
     "otplib": "^12.0.1",
     "qrcode": "^1.5.4",
     "socket.io": "^4.7.2",

server/routes/telemetry-collector.js

@@ -0,0 +1,68 @@
+'use strict';
+
+/*
+ * Opt-in install statistics — the COLLECTOR side, plus the public aggregate the marketing
+ * page reads.
+ *
+ * Mounted only when TELEMETRY_COLLECTOR=1, so a normal self-hosted install exposes neither
+ * route. That gate is doing real work on both: the report endpoint is unauthenticated, and
+ * the aggregate would otherwise let any anonymous visitor read a private instance's screen
+ * count off its own landing page.
+ *
+ * A factory rather than a bare router so the database can be injected — which is what lets
+ * this be tested at all. It was previously inline in server.js and had no tests.
+ */
+
+const express = require('express');
+
+module.exports = function createTelemetryCollectorRouter(db) {
+  const router = express.Router();
+
+  /*
+   * Deliberately unauthenticated: a self-hosted instance has no credential with us, and
+   * issuing one would mean an enrolment handshake for what is a three-integer postcard.
+   *
+   * Upsert keyed on instance_id, so an install that reports daily occupies one row forever
+   * rather than 365 a year. Nothing here reads or stores the request IP — receiving one is
+   * unavoidable, logging it would quietly make a pseudonymous report an identifiable one.
+   */
+  router.post('/telemetry/report', express.json({ limit: '2kb' }), (req, res) => {
+    const { instance_id: id, version, screen_count: screens } = req.body || {};
+    // Validate rather than trust: this endpoint is open, so a malformed or hostile body must
+    // land as a 400, never as a row that poisons the count it exists to produce.
+    if (typeof id !== 'string' || !/^[0-9a-f-]{36}$/i.test(id)) return res.status(400).json({ error: 'bad instance_id' });
+    if (version != null && (typeof version !== 'string' || version.length > 40)) return res.status(400).json({ error: 'bad version' });
+    if (!Number.isInteger(screens) || screens < 0 || screens > 100000) return res.status(400).json({ error: 'bad screen_count' });
+    db.prepare(`INSERT INTO telemetry_reports (instance_id, version, screen_count, first_seen, last_seen)
+                VALUES (?, ?, ?, strftime('%s','now'), strftime('%s','now'))
+                ON CONFLICT(instance_id) DO UPDATE SET
+                  version = excluded.version, screen_count = excluded.screen_count, last_seen = excluded.last_seen`)
+      .run(id, version || null, screens);
+    res.json({ ok: true });
+  });
+
+  /*
+   * The aggregate, for the marketing page. An aggregate across every install that reports,
+   * so it discloses nothing about any one of them.
+   *
+   * Cached, because this sits on a public landing page and the number moves in hours, not
+   * milliseconds. A scraper hitting it in a loop costs one query per interval, not per request.
+   */
+  const STATS_TTL_MS = 5 * 60 * 1000;
+  let statsCache = { at: 0, body: null };
+
+  router.get('/public/stats', (req, res) => {
+    const now = Date.now();
+    if (!statsCache.body || now - statsCache.at > STATS_TTL_MS) {
+      const row = db.prepare(
+        'SELECT COUNT(*) AS installs, COALESCE(SUM(screen_count), 0) AS screens FROM telemetry_reports'
+      ).get();
+      statsCache = { at: now, body: { screens: row.screens, installs: row.installs } };
+    }
+    // Public and cacheable, but never for long by a shared cache: the number is meant to climb.
+    res.set('Cache-Control', 'public, max-age=300');
+    res.json(statsCache.body);
+  });
+
+  return router;
+};

server/routes/widgets.js

@@ -1206,6 +1206,14 @@ function renderDirectorySearch(c) {
 
   // ----- on-screen keyboard (drives the same filter path as typing) -----
   if (cfg.show_onscreen_keyboard) {
+    /* Tell the platform not to raise ITS keyboard for this field. We autofocus a real
+       <input>, which on Android is the signal to throw the system IME over the bottom of
+       the screen - directly on top of the keyboard we draw below, so a directory panel
+       showed Google's keyboard (mic, GIF and emoji keys included) and never showed its
+       own. The buttons write input.value directly, so suppressing the platform keyboard
+       costs nothing here. Ignored by browsers that don't know inputmode, which is the
+       right fallback: a desktop preview keeps behaving exactly as before. */
+    input.setAttribute('inputmode', 'none');
     var kb = document.getElementById('keyboard');
     function press(ch) { input.value += ch; try { input.focus(); } catch(e){} onInput(); }
     ['1234567890','qwertyuiop','asdfghjkl','zxcvbnm'].forEach(function(r){

server/server.js

@@ -978,31 +978,19 @@ app.get('/api/version', (req, res) => {
 app.use('/api/status', require('./routes/status'));
 
 /*
- * Opt-in install statistics — COLLECTOR side. Inert unless TELEMETRY_COLLECTOR=1, so a normal
- * self-hosted install never exposes this at all; only the deployment that gathers the numbers
- * turns it on. Deliberately unauthenticated: a self-hosted instance has no credential with us,
- * and issuing one would mean an enrolment handshake for what is a three-integer postcard.
- *
- * Upsert keyed on instance_id, so a install that reports daily occupies one row forever rather
- * than 365 a year. Nothing here reads or stores the request IP — receiving one is unavoidable,
- * logging it would quietly make a pseudonymous report an identifiable one.
+ * Opt-in install statistics — the COLLECTOR side, plus the public aggregate the marketing
+ * page reads. Both live in routes/telemetry-collector.js; both are mounted only when
+ * TELEMETRY_COLLECTOR=1, so a normal self-hosted install exposes neither.
  */
 if (process.env.TELEMETRY_COLLECTOR === '1') {
-  app.post('/api/telemetry/report', express.json({ limit: '2kb' }), (req, res) => {
-    const { instance_id: id, version, screen_count: screens } = req.body || {};
-    // Validate rather than trust: this endpoint is open, so a malformed or hostile body must
-    // land as a 400, never as a row that poisons the count it exists to produce.
-    if (typeof id !== 'string' || !/^[0-9a-f-]{36}$/i.test(id)) return res.status(400).json({ error: 'bad instance_id' });
-    if (version != null && (typeof version !== 'string' || version.length > 40)) return res.status(400).json({ error: 'bad version' });
-    if (!Number.isInteger(screens) || screens < 0 || screens > 100000) return res.status(400).json({ error: 'bad screen_count' });
-    db.prepare(`INSERT INTO telemetry_reports (instance_id, version, screen_count, first_seen, last_seen)
-                VALUES (?, ?, ?, strftime('%s','now'), strftime('%s','now'))
-                ON CONFLICT(instance_id) DO UPDATE SET
-                  version = excluded.version, screen_count = excluded.screen_count, last_seen = excluded.last_seen`)
-      .run(id, version || null, screens);
-    res.json({ ok: true });
-  });
-  console.log('[telemetry] collector enabled at POST /api/telemetry/report');
+  /* `require('./db/database').db`, not the module-scope `db` — that binding is declared far
+     below this line, so naming it here throws "Cannot access 'db' before initialization" at
+     load and the process never starts. The inline handler this replaced only touched `db`
+     inside a request callback, which runs long after the binding exists; passing it to a
+     factory made the reference eager. Every neighbouring call site in this region resolves
+     the same lazy way. */
+  app.use('/api', require('./routes/telemetry-collector')(require('./db/database').db));
+  console.log('[telemetry] collector enabled at POST /api/telemetry/report (+ GET /api/public/stats)');
 }
 
 // #146 BILLING: Usage Report on its OWN route (NOT part of /api/status — billing is revenue

server/test/directory-search.test.js

@@ -59,6 +59,24 @@ test('show_onscreen_keyboard flag is carried into the page config', async () =>
   assert.ok(html.includes('"show_onscreen_keyboard":false'), 'keyboard flag inlined (page hides the keyboard when false)');
 });
 
+test('the built-in keyboard suppresses the platform one', async () => {
+  // The page autofocuses a real <input>, which on Android raises the system IME over the
+  // bottom of the screen - covering the keyboard this widget draws itself. A panel showed
+  // Gboard, complete with a mic key, and never showed its own keyboard.
+  seed('search_kb_on', 'directory-search', { source_widget_id: 'board1', show_onscreen_keyboard: true });
+  const { html } = await fetchRender('search_kb_on');
+  assert.ok(/setAttribute\(\s*'inputmode'\s*,\s*'none'\s*\)/.test(html),
+    'page tells the platform not to raise its own keyboard');
+
+  // Only when we are drawing one. With the built-in keyboard off there is nothing to cover,
+  // and the platform keyboard is the only way left to type.
+  seed('search_kb_off2', 'directory-search', { source_widget_id: 'board1', show_onscreen_keyboard: false });
+  const off = await fetchRender('search_kb_off2');
+  assert.ok(off.html.includes('"show_onscreen_keyboard":false'), 'flag inlined as false');
+  assert.ok(!/inputmode="none"/.test(off.html),
+    'the input is not statically marked inputmode=none - suppression is gated on the flag at runtime');
+});
+
 test('missing source -> friendly fallback page, not a 500', async () => {
   seed('search_missing', 'directory-search', { source_widget_id: 'does-not-exist' });
   const { status, html } = await fetchRender('search_missing');

server/test/email-nodemailer-live.test.js

@@ -0,0 +1,94 @@
+// The rest of the email tests mock nodemailer through require.cache, which proves our code
+// calls sendMail correctly but says nothing about whether nodemailer still ACCEPTS what we
+// hand it. That gap is exactly where a dependency bump breaks sending: the suite stays green
+// while mail stops leaving the building.
+//
+// This drives the real library against a throwaway SMTP server on a loopback port, using the
+// option and message shapes services/email.js actually builds. No network, no credentials.
+const test = require('node:test');
+const assert = require('node:assert');
+const net = require('node:net');
+const nodemailer = require('nodemailer');
+const { buildSmtpMessage } = require('../services/email');
+
+// A minimal SMTP server that answers enough of the protocol to accept one message and
+// records the conversation, so assertions can be made about what went over the wire.
+function startFakeSmtp() {
+  const seen = { commands: [], data: '' };
+  return new Promise((resolve) => {
+    const srv = net.createServer((sock) => {
+      let inData = false;
+      sock.write('220 fake.local ESMTP\r\n');
+      sock.on('data', (buf) => {
+        const chunk = buf.toString();
+        if (inData) {
+          seen.data += chunk;
+          if (/\r\n\.\r\n/.test(seen.data)) { inData = false; sock.write('250 OK queued\r\n'); }
+          return;
+        }
+        for (const line of chunk.split('\r\n').filter(Boolean)) {
+          seen.commands.push(line);
+          const cmd = line.split(' ')[0].toUpperCase();
+          if (cmd === 'EHLO' || cmd === 'HELO') sock.write('250-fake.local\r\n250 SIZE 10485760\r\n');
+          else if (cmd === 'MAIL' || cmd === 'RCPT') sock.write('250 OK\r\n');
+          else if (cmd === 'DATA') { inData = true; sock.write('354 send it\r\n'); }
+          else if (cmd === 'QUIT') { sock.write('221 bye\r\n'); sock.end(); }
+          else sock.write('250 OK\r\n');
+        }
+      });
+      sock.on('error', () => {});   // a client hanging up mid-conversation is not a test failure
+    });
+    srv.listen(0, '127.0.0.1', () => resolve({ srv, port: srv.address().port, seen }));
+  });
+}
+
+test('the installed nodemailer accepts the options and messages we build', async () => {
+  const { srv, port, seen } = await startFakeSmtp();
+  try {
+    // The shape getSmtpTransporter() constructs.
+    const transporter = nodemailer.createTransport({
+      host: '127.0.0.1', port, secure: false, tls: { rejectUnauthorized: false },
+    });
+
+    // The object `from` form, used when a display name overrides the configured one.
+    await transporter.sendMail({
+      from: { name: 'ScreenTinker', address: 'noreply@example.com' },
+      to: 'user@x.com',
+      subject: '[ScreenTinker] Hello',
+      html: '<p>hi there</p>',
+      text: 'hi there',
+    });
+
+    assert.ok(seen.commands.some(c => /^EHLO/i.test(c)), 'greets the server');
+    assert.ok(seen.commands.some(c => /^MAIL FROM:<noreply@example\.com>/i.test(c)),
+      'envelope sender is the configured address, not the display name');
+    assert.ok(seen.commands.some(c => /^RCPT TO:<user@x\.com>/i.test(c)), 'envelope recipient');
+    assert.match(seen.data, /Subject: \[ScreenTinker\] Hello/, 'subject and its prefix survive encoding');
+    assert.match(seen.data, /From: ScreenTinker <noreply@example\.com>/, 'display-name form still renders');
+    assert.match(seen.data, /Content-Type: multipart\/alternative/, 'text and html sent as alternatives');
+
+    // The string `from` form, used when there is no override.
+    await transporter.sendMail({
+      from: 'ScreenTinker <noreply@example.com>', to: 'user@x.com', subject: 'Welcome', html: '<p>x</p>',
+    });
+    transporter.close();
+  } finally {
+    srv.close();
+  }
+});
+
+test('buildSmtpMessage output is something the installed nodemailer can send', async () => {
+  const { srv, port, seen } = await startFakeSmtp();
+  try {
+    const transporter = nodemailer.createTransport({
+      host: '127.0.0.1', port, secure: false, tls: { rejectUnauthorized: false },
+    });
+    // Built by our own code rather than hand-written here, so the two cannot drift apart.
+    await transporter.sendMail(buildSmtpMessage('to@x.com', 'Subj', 'plain', '<p>rich</p>', 'Sender Name'));
+    assert.ok(seen.commands.some(c => /^RCPT TO:<to@x\.com>/i.test(c)), 'recipient reached the wire');
+    assert.match(seen.data, /Subject: Subj/, 'subject reached the wire');
+    transporter.close();
+  } finally {
+    srv.close();
+  }
+});

server/test/telemetry-collector.test.js

@@ -0,0 +1,94 @@
+'use strict';
+
+// The collector and the public aggregate it feeds. Both were previously inline in server.js
+// with no test at all — the endpoint that decides what a public marketing page claims, and the
+// unauthenticated one anyone on the internet can POST to.
+
+const test = require('node:test');
+const assert = require('node:assert/strict');
+const express = require('express');
+const Database = require('better-sqlite3');
+
+const db = new Database(':memory:');
+db.exec(`CREATE TABLE telemetry_reports (
+  instance_id TEXT PRIMARY KEY,
+  version TEXT,
+  screen_count INTEGER NOT NULL,
+  first_seen INTEGER NOT NULL,
+  last_seen INTEGER NOT NULL
+);`);
+
+const app = express();
+app.use('/api', require('../routes/telemetry-collector')(db));
+const server = app.listen(0);
+let base;
+
+test.before(async () => {
+  await new Promise(r => (server.listening ? r() : server.once('listening', r)));
+  base = `http://127.0.0.1:${server.address().port}`;
+});
+test.after(() => server.close());
+
+const uuid = (n) => `0000000${n}-0000-4000-8000-00000000000${n}`.slice(0, 36).padEnd(36, '0');
+const report = (body) => fetch(`${base}/api/telemetry/report`, {
+  method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify(body),
+});
+
+test('a well-formed report is accepted and stored', async () => {
+  const res = await report({ instance_id: uuid(1), version: '1.9.34', screen_count: 12 });
+  assert.equal(res.status, 200);
+  assert.deepEqual(await res.json(), { ok: true });
+  const row = db.prepare('SELECT * FROM telemetry_reports WHERE instance_id = ?').get(uuid(1));
+  assert.equal(row.screen_count, 12);
+  assert.equal(row.version, '1.9.34');
+});
+
+test('reporting again updates the row rather than adding one', async () => {
+  // An install reports daily. If this ever inserted instead of updating, one install would
+  // occupy 365 rows a year and the public figure would count it 365 times.
+  await report({ instance_id: uuid(1), version: '1.9.35', screen_count: 20 });
+  const n = db.prepare('SELECT COUNT(*) AS c FROM telemetry_reports WHERE instance_id = ?').get(uuid(1)).c;
+  assert.equal(n, 1, 'still a single row for this install');
+  const row = db.prepare('SELECT * FROM telemetry_reports WHERE instance_id = ?').get(uuid(1));
+  assert.equal(row.screen_count, 20, 'count is the latest reported');
+  assert.equal(row.version, '1.9.35');
+});
+
+test('hostile or malformed bodies are refused, not stored', async () => {
+  const before = db.prepare('SELECT COUNT(*) AS c FROM telemetry_reports').get().c;
+  const bad = [
+    { instance_id: 'not-a-uuid', screen_count: 1 },
+    { instance_id: uuid(2) },                                    // no count
+    { instance_id: uuid(2), screen_count: -1 },
+    { instance_id: uuid(2), screen_count: 1e9 },                 // absurd, would skew the total
+    { instance_id: uuid(2), screen_count: 1.5 },                 // not an integer
+    { instance_id: uuid(2), screen_count: 1, version: 'v'.repeat(41) },
+    {},
+  ];
+  for (const body of bad) {
+    const res = await report(body);
+    assert.equal(res.status, 400, `expected 400 for ${JSON.stringify(body)}`);
+  }
+  assert.equal(db.prepare('SELECT COUNT(*) AS c FROM telemetry_reports').get().c, before,
+    'nothing rejected reached the table');
+});
+
+test('the public aggregate sums screens across installs and names no one', async () => {
+  db.prepare(`INSERT INTO telemetry_reports (instance_id, version, screen_count, first_seen, last_seen)
+              VALUES (?,?,?,?,?)`).run(uuid(3), '1.9.34', 480, 1, 1);
+  const res = await fetch(`${base}/api/public/stats`);
+  assert.equal(res.status, 200);
+  const body = await res.json();
+  assert.equal(body.screens, 500, '20 + 480');
+  assert.equal(body.installs, 2);
+  // The whole payload — no instance ids, no versions, nothing per-install.
+  assert.deepEqual(Object.keys(body).sort(), ['installs', 'screens']);
+  assert.match(res.headers.get('cache-control') || '', /max-age=300/);
+});
+
+test('the aggregate is cached, so a scraper cannot turn page views into queries', async () => {
+  db.prepare(`INSERT INTO telemetry_reports (instance_id, version, screen_count, first_seen, last_seen)
+              VALUES (?,?,?,?,?)`).run(uuid(4), '1.9.34', 999, 1, 1);
+  const body = await (await fetch(`${base}/api/public/stats`)).json();
+  assert.equal(body.screens, 500, 'still the cached figure, not 1499');
+});

tizen/config.xml

@@ -1,6 +1,6 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <widget xmlns="http://www.w3.org/ns/widgets" xmlns:tizen="http://tizen.org/ns/widgets"
-        id="http://screentinker.com/player" version="1.9.34" viewmodes="maximized">
+        id="http://screentinker.com/player" version="1.9.36" viewmodes="maximized">
     <tizen:application id="ScrnTinkr1.ScreenTinker" package="ScrnTinkr1" required_version="2.4"/>
     <tizen:profile name="tv"/>
     <name>ScreenTinker</name>