fix(player): let the background-audio toggle govern backgrounding again

Locking the screen kept a video's audio playing whether or not the background-audio button was on. Reported as "audio only mode is always active even if not selected". The button (UR-040) was built for the WebView <video> path, where losing visibility kills the decode: it chose between handing off to a native audio stream and letting playback stop. Native video then became the default renderer (DR-188), and on that path playback runs through ExoPlayer inside a MediaSessionService -- a foreground media service whose entire purpose is to keep playing while the app is hidden. Nothing stopped it, and nothing in the codebase paused on background. So the button governed a handoff that no longer had a gap to bridge. There was no interruption to paper over, and a user who never touched it got background playback anyway. The gating made it self-concealing: MainActivity.onStop only dispatched 'jellytau-background' when backgroundAudioEnabled was already true. The one notification that the app had gone away was itself conditional on the setting, so with the button OFF nothing could react even in principle. onStop and onStart now fire unconditionally and carry the two facts only the activity knows -- whether the toggle is armed, and whether Android put the window into picture-in-picture. What to do about it is decided in Rust (player/background_policy.rs), because it depends on whether the item has a picture to lose: video + toggle off -> Pause video + toggle on -> HandOffToAudio music, either -> KeepPlaying (no picture to give up) picture-in-picture -> KeepPlaying (the window is still on screen) It takes no renderer parameter on purpose. Two renderers with two behaviours and one toggle reaching only one of them is what produced the defect; a rule that cannot see the renderer cannot reproduce it. Two failure modes are deliberate. A decision call that fails leaves playback alone rather than risking silence mid-listen. An event with no detail -- older Kotlin against newer JS -- reads as "armed, not PiP", degrading to the previous behaviour instead of pausing unexpectedly. Foregrounding resumes only what backgrounding paused: a video the user paused themselves before locking stays paused. Written test-first per CLAUDE.md. The stub encoded today's behaviour (nothing ever pauses) and failed exactly as reported -- `left: KeepPlaying, right: Pause` -- before the rule was implemented. Verified on a device, R8-minified, both directions: [player_background_action] video=true armed=false pip=false -> Pause [player_background_action] video=true armed=true pip=false -> HandOffToAudio UR-040 / DR-224 / UT-211.
fix(ci): give the builder image what linuxdeploy needs for the AppImage
2026-08-22 10:09:46 +02:00 · 2026-08-22 02:52:32 +02:00 · 2026-08-21 23:14:16 +02:00 · 2026-08-21 22:32:04 +02:00 · 2026-08-21 20:30:54 +00:00 · 2026-08-21 22:30:28 +02:00
33 changed files with 1830 additions and 470 deletions
@@ -28,7 +28,7 @@ jobs:
    if: "!startsWith(github.event.head_commit.message, 'chore(release)')"
    runs-on: linux/amd64
    container:
-      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08
+      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08.1

    steps:
      - name: Checkout repository
@@ -89,6 +89,15 @@ jobs:
      # sit on master until somebody cut a tag. These three steps are what make
      # those configs load-bearing. All are project deps installed by
      # `bun install`; nothing is fetched at job time.
+      # Cheap tripwire for a class of defect this repo kept hitting: tooling on
+      # a rarely-taken path. scripts/build-android.sh ran `npm install` on its
+      # clean-build branch -- in a bun project, ignoring bun.lock and
+      # re-resolving the tree, which is how the Tauri plugin crate/package
+      # versions drifted apart and broke a release build. It survived because
+      # clean builds are rare.
+      - name: Check build tooling
+        run: bash scripts/check-tooling.sh
+
      - name: Check formatting
        run: bun run format:check

@@ -107,6 +116,28 @@ jobs:
          bunx svelte-kit sync
          bun run check

+      # Tauri refuses to build when a plugin's Rust crate and npm package are on
+      # different minor versions. Nothing here runs `tauri build` -- that only
+      # happens on a tag -- so a mismatch introduced on master stayed invisible
+      # until the release build, which is where it was found: v0.10.0 prep hit
+      # `tauri-plugin-log (v2.8.0) : @tauri-apps/plugin-log (v2.9.0)`. `cargo
+      # check`, clippy, the tests and svelte-check had all passed.
+      #
+      # `tauri info` performs the same comparison the bundler does, without a
+      # build. Grepping its output is crude, but the alternative is discovering
+      # this at tag time again.
+      - name: Check Tauri plugin versions match
+        run: |
+          set -e
+          if bunx tauri info 2>&1 | tee /tmp/tauri-info.txt | grep -q "version mismatched"; then
+            echo "::error::A Tauri plugin's Rust crate and npm package versions disagree."
+            echo "::error::The release build will refuse to start. Align them in"
+            echo "::error::src-tauri/Cargo.toml and package.json (both are pinned exactly)."
+            grep -A6 "version mismatched" /tmp/tauri-info.txt || true
+            exit 1
+          fi
+          echo "✅ Tauri plugin crate/package versions agree."
+
      # Coverage rather than a bare `bun run test`: same suite, plus the
      # thresholds in vitest.config.ts, so a large untested module or a deleted
      # test fails here instead of being noticed months later.
@@ -156,7 +187,7 @@ jobs:
    runs-on: linux/amd64
    needs: test
    container:
-      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08
+      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08.1
    env:
      ANDROID_HOME: /opt/android-sdk
      ANDROID_SDK_ROOT: /opt/android-sdk
@@ -225,7 +256,7 @@ jobs:
    name: Supply Chain
    runs-on: linux/amd64
    container:
-      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08
+      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08.1

    steps:
      - name: Checkout repository
@@ -21,7 +21,7 @@ jobs:
    name: Run Tests
    runs-on: linux/amd64
    container:
-      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08
+      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08.1
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
@@ -94,7 +94,7 @@ jobs:
    runs-on: linux/amd64
    needs: test
    container:
-      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08
+      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08.1
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
@@ -146,9 +146,29 @@ jobs:
      # If TAURI_SIGNING_PRIVATE_KEY is ever absent the build fails loudly rather
      # than quietly shipping an unsigned release that no client will accept --
      # which is the behaviour we want.
+      # Same hazard as the Windows job: the bundle directory is never cleaned by
+      # cargo and the runner reuses src-tauri/target, while the copy step below
+      # globs bundle/deb/*.deb and friends. Windows is where this actually bit
+      # (v0.8.2 shipped thirteen stale installers), but only because Linux
+      # packaging is newer -- the glob is identical. Remove the directory so a
+      # stale artifact cannot exist to be copied.
+      - name: Clear previous bundle output
+        run: rm -rf src-tauri/target/release/bundle
+
      - name: Build for Linux
        run: bun run tauri build
        env:
+          # linuxdeploy's bundled `strip` cannot parse the `.relr.dyn` section
+          # modern toolchains emit, and fails on every bundled library:
+          #   strip: libzstd.so.1: unknown type [0x13] section `.relr.dyn'
+          #   failed to bundle project `failed to run linuxdeploy`
+          # Ubuntu 23.10+ links with -z pack-relative-relocs by default, so this
+          # image hits it. Skipping strip is linuxdeploy's documented escape
+          # hatch; the cost is a larger AppImage. Found by building the target
+          # locally before tagging -- nothing in CI builds the app, so a release
+          # would have been the first time anyone discovered the AppImage target
+          # does not work.
+          NO_STRIP: "true"
          TAURI_SIGNING_PRIVATE_KEY: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY }}
          TAURI_SIGNING_PRIVATE_KEY_PASSWORD: ${{ secrets.TAURI_SIGNING_PRIVATE_KEY_PASSWORD }}

@@ -170,12 +190,15 @@ jobs:
          # Without nullglob an unmatched pattern stays literal, so test each
          # candidate instead. Same POSIX-only rule as traceability-check.yml.
          #
-          # The .AppImage.tar.gz + .sig pair is what the updater downloads and
-          # verifies; the plain .AppImage is what a human downloads. Both ship.
+          # Tauri v2 signs the .AppImage ITSELF and writes <name>.AppImage.sig
+          # beside it -- there is no .AppImage.tar.gz unless
+          # bundle.createUpdaterArtifacts is set to "v1Compatible". The updater
+          # downloads the same AppImage a human does and verifies that .sig, so
+          # both files must ship or the manifest points at a signature nobody
+          # can fetch.
          for bundle in \
            src-tauri/target/release/bundle/appimage/*.AppImage \
-            src-tauri/target/release/bundle/appimage/*.AppImage.tar.gz \
-            src-tauri/target/release/bundle/appimage/*.AppImage.tar.gz.sig \
+            src-tauri/target/release/bundle/appimage/*.AppImage.sig \
            src-tauri/target/release/bundle/deb/*.deb \
            src-tauri/target/release/bundle/rpm/*.rpm; do
            [ -e "$bundle" ] || continue
@@ -212,7 +235,7 @@ jobs:
    # baked into the builder image. No toolchain installs here — the image has
    # cargo-xwin, clang/clang-cl, lld, llvm, nsis and the msvc target.
    container:
-      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08
+      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08.1
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
@@ -285,7 +308,7 @@ jobs:
    runs-on: linux/amd64
    needs: test
    container:
-      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08
+      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08.1
    env:
      ANDROID_HOME: /opt/android-sdk
      ANDROID_SDK_ROOT: /opt/android-sdk
@@ -357,8 +380,12 @@ jobs:
          keyPassword=${{ secrets.ANDROID_KEY_PASSWORD }}
          EOF

+      # `--apk` is a boolean flag, not `--apk true`. tauri-cli took a value here
+      # until 2.10; from 2.11 the stray `true` is parsed as a positional and the
+      # command fails with "unexpected argument 'true' found" before building.
+      # This line and scripts/build-android.sh must agree.
      - name: Build signed Android APK
-        run: bun run tauri android build --apk true --target aarch64
+        run: bun run tauri android build --apk --target aarch64

      - name: Collect & verify signed APK
        run: |
@@ -384,7 +411,7 @@ jobs:
    needs: [build-linux, build-windows, build-android]
    if: startsWith(github.ref, 'refs/tags/v')
    container:
-      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08
+      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08.1
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
@@ -413,6 +440,19 @@ jobs:
          name: jellytau-android
          path: artifacts/android/

+      # Runs before the SBOM, the checksums and the upload -- everything
+      # downstream describes this set of files, so a stale artifact must be
+      # caught before it gets hashed into SHA256SUMS and published as though it
+      # belonged to this release.
+      #
+      # See the script for the eight months of releases that shipped their
+      # predecessors' Windows installers.
+      - name: Verify artifacts belong to this release
+        run: |
+          ./scripts/check-release-artifacts.sh \
+            "${{ steps.tag_name.outputs.VERSION }}" \
+            artifacts/linux artifacts/windows artifacts/android
+
      # Software Bill of Materials, one per half of the app. Without it there is
      # no answer to "does this release contain <vulnerable crate>?" other than
      # rebuilding the tag and re-resolving it. cargo-cyclonedx is in the builder
@@ -462,8 +502,13 @@ jobs:
          APPIMAGE_URL=""
          NSIS_URL=""

-          for f in artifacts/linux/*.AppImage.tar.gz; do
+          # Tauri v2 signs the AppImage itself; <name>.AppImage.sig sits beside
+          # it. Verified against a real signed build before tagging -- the
+          # v1-style .AppImage.tar.gz is never produced with
+          # createUpdaterArtifacts: true.
+          for f in artifacts/linux/*.AppImage; do
            [ -e "$f" ] || continue
+            case "$f" in *.sig) continue;; esac
            APPIMAGE_URL="${BASE}/$(basename "$f")"
            [ -e "$f.sig" ] && APPIMAGE_SIG="$(cat "$f.sig")"
          done
@@ -482,9 +527,11 @@ jobs:
            exit 1
          fi

-          # Release notes for the update prompt come from the traceability graph,
-          # same source as the release body.
-          NOTES="$(bun run release:notes 2>/dev/null | head -c 4000 || echo "See the release page for details.")"
+          # What the in-app update prompt shows. Same reviewed source as the
+          # release body -- the CHANGELOG section for this version, not the
+          # traceability draft.
+          NOTES="$(awk -v ver="## $VERSION" '$0==ver{f=1;next} /^## /{if(f)exit} f' CHANGELOG.md | head -c 4000)"
+          [ -n "$NOTES" ] || NOTES="See the release page for details."

          jq -n \
            --arg version "$PLAIN" \
@@ -549,24 +596,37 @@ jobs:
          # release rather than shipping and failing for users.
          sha256sum -c SHA256SUMS

-      # Release notes come from the traceability graph, not from a hardcoded
-      # heredoc. scripts/release-notes.ts resolves the commit range's changed
-      # files to their TRACES ids and then to requirement descriptions, grouping
-      # UR into Features and DR/IR into Improvements -- which is what CLAUDE.md
-      # has asked for all along, while this workflow pasted a fixed block of
-      # install instructions and a line saying "see CHANGELOG.md for detailed
-      # changes". It also linked "GitHub Issues" on a Gitea-hosted project.
+      # The published body is the hand-written CHANGELOG.md section for this
+      # version. `bun run release:notes` is printed into the job log as a
+      # drafting aid, but is NOT published: CLAUDE.md is explicit that its
+      # output is "a reviewed draft, not a final changelog", and publishing it
+      # unreviewed proved the point -- a range containing a repo-wide prettier
+      # sweep resolved to nearly the whole requirement matrix and produced notes
+      # claiming one release had added the entire application.
+      #
+      # A missing CHANGELOG section fails the release. A release whose notes say
+      # nothing is worse than one that waits for a maintainer to write two
+      # sentences, and the checklist already requires that entry.
      - name: Prepare release notes
        id: release_notes
        run: |
          set -e
          VERSION="${{ steps.tag_name.outputs.VERSION }}"
+
+          echo "📋 Traceability draft (for reference; not published):"
+          bun run release:notes 2>/dev/null || echo "(could not derive a draft)"
+          echo ""
+
+          # The section between this version's heading and the next one.
+          CHANGES=$(awk -v ver="## $VERSION" '$0==ver{f=1;next} /^## /{if(f)exit} f' CHANGELOG.md)
+          if [ -z "$(echo "$CHANGES" | tr -d '[:space:]')" ]; then
+            echo "::error::CHANGELOG.md has no '## $VERSION' section."
+            echo "::error::Add the entry for this version and re-tag; see docs/release-checklist.md."
+            exit 1
+          fi
+
          {
-            echo "## JellyTau $VERSION"
-            echo ""
-            # A generated summary of what actually changed; falls back to a
-            # pointer rather than failing the release if the range is odd.
-            bun run release:notes 2>/dev/null || echo "See the commit log for changes in this release."
+            echo "$CHANGES"
            echo ""
            echo "### Downloads"
            echo ""
@@ -578,8 +638,8 @@ jobs:
            echo "| Windows | \`*-setup.exe\` (NSIS). Unsigned — SmartScreen may warn on first run. |"
            echo "| Android | \`*.apk\` sideload, or \`*.aab\` for Play Console |"
            echo ""
-            echo "Desktop builds update themselves from here on: JellyTau checks this"
-            echo "release feed and can install a new version in place."
+            echo "Desktop builds check for updates from here and can install a new"
+            echo "version in place, verifying its signature first."
            echo ""
            echo "### Verifying your download"
            echo ""
@@ -599,6 +659,7 @@ jobs:
            echo "---"
            echo "Report a problem: ${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}/issues"
          } > release_notes.md
+
          echo "📝 Release notes:"
          cat release_notes.md

@@ -21,7 +21,7 @@ jobs:
    name: Build & publish docs to gitea-pages
    runs-on: linux/amd64
    container:
-      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08
+      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08.1

    steps:
      - name: Checkout code
@@ -17,7 +17,7 @@ jobs:
    runs-on: linux/amd64
    name: Check Requirement Traces
    container:
-      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08
+      image: gitea.tourolle.paris/dtourolle/jellytau-builder:2026.08.1

    steps:
      - name: Checkout repository
@@ -9,6 +9,125 @@ generated trace matrix lives in [docs/traceability.md](docs/traceability.md).
 For how long each fixed defect had been shipping before it was found, see
 [docs/defect-windows.md](docs/defect-windows.md).

+## v0.10.1
+
+A single fix, for something that had been quietly overriding a choice you made.
+
+### 🐛 Fixes
+
+- **Locking the screen no longer keeps playing a video's audio unless you asked
+  it to.** The player has a background-audio button: turn it on and the sound
+  carries on when you lock the screen or leave the app, turn it off and playback
+  stops. It stopped working when video moved to the native renderer — which plays
+  through a media service designed to keep going while the app is hidden — and
+  nothing was left to stop it. So the audio continued whether the button was on
+  or off, and there was no way to make it behave otherwise. The button governs it
+  again: with it off, locking the screen pauses the video and unlocking resumes
+  where you were; with it on, the audio continues as before. Music is untouched —
+  it keeps playing when backgrounded, as a music player should — and a video in a
+  picture-in-picture window keeps playing too, because the window is still on
+  screen. If you had already paused before locking, it stays paused.
+  (UR-040 → DR-224)
+
+## v0.10.0
+
+Two things you can see, and a great deal of work on how this project builds and
+ships itself. The app can now update itself, and it can tell you what it did
+when something goes wrong — both of which existed as gaps rather than as bugs,
+which is why they lasted so long.
+
+### ✨ Changes
+
+- **JellyTau can update itself.** Anyone who installed an AppImage or ran the
+  Windows installer was frozen on that version permanently: nothing in the app
+  ever mentioned that a newer one existed, and the release page was the only
+  announcement. Settings → Updates now checks, shows what changed, and installs
+  and restarts on request. Each download is verified against JellyTau's signing
+  key before anything is installed, so a substituted file is refused rather than
+  run. Android is deliberately not wired to this — an app may not replace its own
+  APK, that is the system installer's job — and is given a link to the releases
+  page instead of a button that would fail. (UR-077 → DR-217)
+
+- **You can export a diagnostics bundle.** Until now the app forgot everything it
+  had done the moment it closed. Logs went to standard output, which nobody sees
+  when launching from a desktop icon, and on Android went nowhere at all — so the
+  backend was invisible on the platform where the hardest playback bugs live. A
+  crash left nothing behind. Logs are now kept in a size-capped file that
+  survives a restart, a crash is recorded before the app dies, and Settings →
+  Diagnostics exports the lot as one file to attach to a bug report. Access
+  tokens and passwords are stripped before anything is written to disk, not
+  merely before it is exported. Nothing is transmitted anywhere; you attach the
+  file yourself. (UR-078 → DR-218)
+
+- **Linux gets an AppImage again.** The release notes have advertised one for
+  months while the build never produced it — the packaging step looked for the
+  file, found nothing, and said nothing. (DR-217)
+
+### 🐛 Fixes
+
+- **Releases no longer ship every Windows installer ever built.** Every release
+  from v0.1.0 to v0.8.2 carried its predecessors': sixteen installers on v0.8.2,
+  thirteen of them stale, and a download list on v0.5.0 reaching back to 0.1.0.
+  The build directory is never cleaned and the build machine reuses it, so each
+  release collected whatever was left behind. It went unnoticed for eight months
+  because nothing looked wrong — the files were real and the page merely looked
+  busy. The stale files have been removed from the published releases, the build
+  now clears that directory first, and a check refuses to publish a release
+  containing an artifact from a different version. (DR-220)
+
+- **Release notes now say what changed.** All 35 previous releases published the
+  same block of generic install instructions, whose "What's New" section was a
+  link to a file that does not resolve from a release page. Every release page
+  now carries its own entry from this changelog, and the past ones have been
+  filled in. (DR-219)
+
+### 🔒 Security and supply chain
+
+- **Dependencies are now checked against a vulnerability database on every
+  build.** They never had been. The first run found eight vulnerabilities and one
+  unsoundness in the Rust dependency graph — all of them fixed by an update
+  nobody had a reason to run. Licences are checked against an allow-list too, so
+  nothing gets redistributed inside a release that does not permit it.
+  (DR-216)
+
+- **Every release publishes checksums and a bill of materials.** `SHA256SUMS`
+  lets you verify a download (`sha256sum -c SHA256SUMS`); the SBOM lists what
+  went into the build, so "does this release contain <vulnerable library>?" has
+  an answer that is not "rebuild it and find out". (DR-216)
+
+- **Builds are reproducible again.** Every CI job named a container image tag
+  that was rewritten in place, so rebuilding an old release did not necessarily
+  rebuild the same thing. Jobs now pin an immutable tag. The one dependency that
+  comes from a git branch rather than a package registry is pinned to an exact
+  revision, closing a path by which new upstream code could arrive unreviewed in
+  a library linked into the player. (DR-216)
+
+### 🧹 Under the hood
+
+- Formatting, linting and type-checking now run in CI. All three were configured
+  and enforced by nothing: 199 files did not match the project's own formatter, a
+  type error could sit on the main branch until somebody cut a release, and the
+  test-coverage command had been broken for months by a dependency mismatch.
+  Coverage now has a floor that only moves up. (DR-215)
+
+- The traceability matrix counts requirements implemented by configuration.
+  Several carried the necessary annotations and were being counted as uncovered
+  because the extraction tool only read source files. (DR-215)
+
+- The project now has a security policy, contribution guide, code of conduct,
+  issue and pull-request templates, and an operations document covering the
+  builder image, the release secrets, and what losing the signing key would mean.
+
+- The app framework moved from Tauri 2.9.5 to 2.11.5. Nothing about this is
+  visible in use, but it is worth recording that it did not go quietly: the
+  windowing layer beneath Tauri quietly stopped publishing the Android JavaVM
+  and application handle that this app's credential storage had been reading for
+  its whole life. Nothing here had changed; a side effect several dependencies
+  down had simply gone away, and the app aborted on launch on every Android
+  device. JellyTau now sets that handle itself rather than relying on someone
+  else to do it. Caught by installing on a real tablet before release — no test
+  suite runs the app. (UR-012 → DR-223)
+
 ## v0.9.1

 A one-line fix to the home screen, released on its own because it is the kind of
@@ -141,6 +141,17 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
    lld \
    llvm \
    nsis \
+    # AppImage bundling. linuxdeploy embeds xdg-open into the AppImage and
+    # aborts the whole bundle if it is missing:
+    #   failed to bundle project: xdg-open binary not found
+    # It is present on most desktop distros, which is why the AppImage built on
+    # a developer machine and failed here. desktop-file-utils and zsync are the
+    # other two linuxdeploy commonly wants (desktop-file-validate, and zsync for
+    # delta updates), added together so a missing one does not cost another
+    # image rebuild and another failed release build.
+    xdg-utils \
+    desktop-file-utils \
+    zsync \
    && rm -rf /var/lib/apt/lists/* \
    # Ubuntu's clang package ships clang but NOT the clang-cl alias that cc-rs
    # invokes for MSVC targets. clang-cl is the same binary in MSVC-compat mode,
@@ -5,12 +5,12 @@
    "": {
      "name": "jellytau",
      "dependencies": {
-        "@tauri-apps/api": "^2",
-        "@tauri-apps/plugin-log": "^2.9.0",
-        "@tauri-apps/plugin-opener": "^2",
+        "@tauri-apps/api": "^2.11.1",
+        "@tauri-apps/plugin-log": "2.9.0",
+        "@tauri-apps/plugin-opener": "^2.5.4",
        "@tauri-apps/plugin-os": "^2.3.2",
        "@tauri-apps/plugin-process": "^2.3.1",
-        "@tauri-apps/plugin-updater": "^2.10.1",
+        "@tauri-apps/plugin-updater": "2.10.1",
        "hls.js": "^1.6.15",
        "svelte-dnd-action": "^0.9.69",
      },
@@ -20,7 +20,7 @@
        "@sveltejs/kit": "^2.9.0",
        "@sveltejs/vite-plugin-svelte": "^6.2.4",
        "@tailwindcss/vite": "^4.1.18",
-        "@tauri-apps/cli": "^2",
+        "@tauri-apps/cli": "^2.11.4",
        "@testing-library/svelte": "^5.3.1",
        "@vitest/coverage-v8": "^4.0.18",
        "@vitest/ui": "^4.0.16",
@@ -255,35 +255,35 @@

    "@tailwindcss/vite": ["@tailwindcss/vite@4.1.18", "", { "dependencies": { "@tailwindcss/node": "4.1.18", "@tailwindcss/oxide": "4.1.18", "tailwindcss": "4.1.18" }, "peerDependencies": { "vite": "^5.2.0 || ^6 || ^7" } }, "sha512-jVA+/UpKL1vRLg6Hkao5jldawNmRo7mQYrZtNHMIVpLfLhDml5nMRUo/8MwoX2vNXvnaXNNMedrMfMugAVX1nA=="],

-    "@tauri-apps/api": ["@tauri-apps/api@2.9.1", "", {}, "sha512-IGlhP6EivjXHepbBic618GOmiWe4URJiIeZFlB7x3czM0yDHHYviH1Xvoiv4FefdkQtn6v7TuwWCRfOGdnVUGw=="],
+    "@tauri-apps/api": ["@tauri-apps/api@2.11.1", "", {}, "sha512-M2FPuYND2m+wh5hfW9ZpSdxMPdEJovPBWwoHJmwUpysTYNHaOkVFN419m/K0LIgjb/7KU2vBgsUepJWugQCvAA=="],

-    "@tauri-apps/cli": ["@tauri-apps/cli@2.9.6", "", { "optionalDependencies": { "@tauri-apps/cli-darwin-arm64": "2.9.6", "@tauri-apps/cli-darwin-x64": "2.9.6", "@tauri-apps/cli-linux-arm-gnueabihf": "2.9.6", "@tauri-apps/cli-linux-arm64-gnu": "2.9.6", "@tauri-apps/cli-linux-arm64-musl": "2.9.6", "@tauri-apps/cli-linux-riscv64-gnu": "2.9.6", "@tauri-apps/cli-linux-x64-gnu": "2.9.6", "@tauri-apps/cli-linux-x64-musl": "2.9.6", "@tauri-apps/cli-win32-arm64-msvc": "2.9.6", "@tauri-apps/cli-win32-ia32-msvc": "2.9.6", "@tauri-apps/cli-win32-x64-msvc": "2.9.6" }, "bin": { "tauri": "tauri.js" } }, "sha512-3xDdXL5omQ3sPfBfdC8fCtDKcnyV7OqyzQgfyT5P3+zY6lcPqIYKQBvUasNvppi21RSdfhy44ttvJmftb0PCDw=="],
+    "@tauri-apps/cli": ["@tauri-apps/cli@2.11.4", "", { "optionalDependencies": { "@tauri-apps/cli-darwin-arm64": "2.11.4", "@tauri-apps/cli-darwin-x64": "2.11.4", "@tauri-apps/cli-linux-arm-gnueabihf": "2.11.4", "@tauri-apps/cli-linux-arm64-gnu": "2.11.4", "@tauri-apps/cli-linux-arm64-musl": "2.11.4", "@tauri-apps/cli-linux-riscv64-gnu": "2.11.4", "@tauri-apps/cli-linux-x64-gnu": "2.11.4", "@tauri-apps/cli-linux-x64-musl": "2.11.4", "@tauri-apps/cli-win32-arm64-msvc": "2.11.4", "@tauri-apps/cli-win32-ia32-msvc": "2.11.4", "@tauri-apps/cli-win32-x64-msvc": "2.11.4" }, "bin": { "tauri": "tauri.js" } }, "sha512-R8xGtMpwyetawSqm9kYOuMmEqkhUbvcUy8n0aNXIxollKBLESUu5f4Fx+64hgASYm1H+jSWq6jCW6zqTnH6hqQ=="],

-    "@tauri-apps/cli-darwin-arm64": ["@tauri-apps/cli-darwin-arm64@2.9.6", "", { "os": "darwin", "cpu": "arm64" }, "sha512-gf5no6N9FCk1qMrti4lfwP77JHP5haASZgVbBgpZG7BUepB3fhiLCXGUK8LvuOjP36HivXewjg72LTnPDScnQQ=="],
+    "@tauri-apps/cli-darwin-arm64": ["@tauri-apps/cli-darwin-arm64@2.11.4", "", { "os": "darwin", "cpu": "arm64" }, "sha512-1ryOF3ZhpZ/nemHV5zVwBQBz9jDGKmKPvWPADOhc83ig0P4bMc2iER4NbC6r9sjeIZ6RVQ4g3RZIYvezhcl4TQ=="],

-    "@tauri-apps/cli-darwin-x64": ["@tauri-apps/cli-darwin-x64@2.9.6", "", { "os": "darwin", "cpu": "x64" }, "sha512-oWh74WmqbERwwrwcueJyY6HYhgCksUc6NT7WKeXyrlY/FPmNgdyQAgcLuTSkhRFuQ6zh4Np1HZpOqCTpeZBDcw=="],
+    "@tauri-apps/cli-darwin-x64": ["@tauri-apps/cli-darwin-x64@2.11.4", "", { "os": "darwin", "cpu": "x64" }, "sha512-uFsGQAAfuyz1k/yGLmkWfkBlgKAqZfxqlHmLWx81QU27RJWfmbNHCIq8T8w1e+VClleIuZUjpHWfoE4E3DLo3A=="],

-    "@tauri-apps/cli-linux-arm-gnueabihf": ["@tauri-apps/cli-linux-arm-gnueabihf@2.9.6", "", { "os": "linux", "cpu": "arm" }, "sha512-/zde3bFroFsNXOHN204DC2qUxAcAanUjVXXSdEGmhwMUZeAQalNj5cz2Qli2elsRjKN/hVbZOJj0gQ5zaYUjSg=="],
+    "@tauri-apps/cli-linux-arm-gnueabihf": ["@tauri-apps/cli-linux-arm-gnueabihf@2.11.4", "", { "os": "linux", "cpu": "arm" }, "sha512-IaHZn5CdBL21oUmjiVOS1ctw6Ip1O0pjp70FwOWmYz1myWe0SY96ZIj2FYf7pT0m8bI2h/hrs5ZbEXXh44/MkQ=="],

-    "@tauri-apps/cli-linux-arm64-gnu": ["@tauri-apps/cli-linux-arm64-gnu@2.9.6", "", { "os": "linux", "cpu": "arm64" }, "sha512-pvbljdhp9VOo4RnID5ywSxgBs7qiylTPlK56cTk7InR3kYSTJKYMqv/4Q/4rGo/mG8cVppesKIeBMH42fw6wjg=="],
+    "@tauri-apps/cli-linux-arm64-gnu": ["@tauri-apps/cli-linux-arm64-gnu@2.11.4", "", { "os": "linux", "cpu": "arm64" }, "sha512-N41/ukTRVe6XSuUTESuFdGeOW2i7k62tK+6gHK5Kd5/q5RPvvi19GaWAVPPb9u95HSGmTChSolBfzynUsssFaA=="],

-    "@tauri-apps/cli-linux-arm64-musl": ["@tauri-apps/cli-linux-arm64-musl@2.9.6", "", { "os": "linux", "cpu": "arm64" }, "sha512-02TKUndpodXBCR0oP//6dZWGYcc22Upf2eP27NvC6z0DIqvkBBFziQUcvi2n6SrwTRL0yGgQjkm9K5NIn8s6jw=="],
+    "@tauri-apps/cli-linux-arm64-musl": ["@tauri-apps/cli-linux-arm64-musl@2.11.4", "", { "os": "linux", "cpu": "arm64" }, "sha512-v277UnT/fB64xAfSroL5N3Km3tLmvATWqJJw/wRI+g6o+HkeD0slyE7gOhNs1MbjE41R7bQOTxMVoL3aomUJmw=="],

-    "@tauri-apps/cli-linux-riscv64-gnu": ["@tauri-apps/cli-linux-riscv64-gnu@2.9.6", "", { "os": "linux", "cpu": "none" }, "sha512-fmp1hnulbqzl1GkXl4aTX9fV+ubHw2LqlLH1PE3BxZ11EQk+l/TmiEongjnxF0ie4kV8DQfDNJ1KGiIdWe1GvQ=="],
+    "@tauri-apps/cli-linux-riscv64-gnu": ["@tauri-apps/cli-linux-riscv64-gnu@2.11.4", "", { "os": "linux", "cpu": "none" }, "sha512-qqgNkQ2u1yZHxjhxsZaxUtRDW8dIqIYm33rx/mzwQv0SfY9x1B+iraj8vWeFiXjjSVVhEMepXSOts1TqPzvXNQ=="],

-    "@tauri-apps/cli-linux-x64-gnu": ["@tauri-apps/cli-linux-x64-gnu@2.9.6", "", { "os": "linux", "cpu": "x64" }, "sha512-vY0le8ad2KaV1PJr+jCd8fUF9VOjwwQP/uBuTJvhvKTloEwxYA/kAjKK9OpIslGA9m/zcnSo74czI6bBrm2sYA=="],
+    "@tauri-apps/cli-linux-x64-gnu": ["@tauri-apps/cli-linux-x64-gnu@2.11.4", "", { "os": "linux", "cpu": "x64" }, "sha512-2VRNWl84FOH0m2giiDkO2h0QXlcMJeX+zJDpI5kDIQAx6s+geF3v48F4DXfJez4GS/FdoDGnPnw1C2iYGbQ7bQ=="],

-    "@tauri-apps/cli-linux-x64-musl": ["@tauri-apps/cli-linux-x64-musl@2.9.6", "", { "os": "linux", "cpu": "x64" }, "sha512-TOEuB8YCFZTWVDzsO2yW0+zGcoMiPPwcUgdnW1ODnmgfwccpnihDRoks+ABT1e3fHb1ol8QQWsHSCovb3o2ENQ=="],
+    "@tauri-apps/cli-linux-x64-musl": ["@tauri-apps/cli-linux-x64-musl@2.11.4", "", { "os": "linux", "cpu": "x64" }, "sha512-o9GyhYor/nc7xarmwDE3ka2szuW3uuZzXjHWh64Q8YX5AtSgxdQkFWzrY4O8KiGtVNvFBI14H3Q49Qj5TOIP/A=="],

-    "@tauri-apps/cli-win32-arm64-msvc": ["@tauri-apps/cli-win32-arm64-msvc@2.9.6", "", { "os": "win32", "cpu": "arm64" }, "sha512-ujmDGMRc4qRLAnj8nNG26Rlz9klJ0I0jmZs2BPpmNNf0gM/rcVHhqbEkAaHPTBVIrtUdf7bGvQAD2pyIiUrBHQ=="],
+    "@tauri-apps/cli-win32-arm64-msvc": ["@tauri-apps/cli-win32-arm64-msvc@2.11.4", "", { "os": "win32", "cpu": "arm64" }, "sha512-ld5Ehb598m0VkYyylRPNeCFsBe/km0jxis6KgMpl3IGY6I/i1RwQXO05I1AsXUXO2WC6AvB/Lw4qTf/asiuEiQ=="],

-    "@tauri-apps/cli-win32-ia32-msvc": ["@tauri-apps/cli-win32-ia32-msvc@2.9.6", "", { "os": "win32", "cpu": "ia32" }, "sha512-S4pT0yAJgFX8QRCyKA1iKjZ9Q/oPjCZf66A/VlG5Yw54Nnr88J1uBpmenINbXxzyhduWrIXBaUbEY1K80ZbpMg=="],
+    "@tauri-apps/cli-win32-ia32-msvc": ["@tauri-apps/cli-win32-ia32-msvc@2.11.4", "", { "os": "win32", "cpu": "ia32" }, "sha512-12Hxi0XX/H5VFxO/bGgHkFWhml9VMgEOu9CidjeCeTNQ1l6fpUlbiGgSP7CLI3PFtW9/FfbeHieZ+kyWK5H7CA=="],

-    "@tauri-apps/cli-win32-x64-msvc": ["@tauri-apps/cli-win32-x64-msvc@2.9.6", "", { "os": "win32", "cpu": "x64" }, "sha512-ldWuWSSkWbKOPjQMJoYVj9wLHcOniv7diyI5UAJ4XsBdtaFB0pKHQsqw/ItUma0VXGC7vB4E9fZjivmxur60aw=="],
+    "@tauri-apps/cli-win32-x64-msvc": ["@tauri-apps/cli-win32-x64-msvc@2.11.4", "", { "os": "win32", "cpu": "x64" }, "sha512-+vDiqBIU5dMISg/wNvX3sF+ZHfgJGJ5T0AcO+EHNXV9GGAG+P5fzodlDXD3QdKCRgZxMoCm5PPvj3BqLNjBthw=="],

    "@tauri-apps/plugin-log": ["@tauri-apps/plugin-log@2.9.0", "", { "dependencies": { "@tauri-apps/api": "^2.11.0" } }, "sha512-Ql8okrnsguk0eDq1GvRfttFV5KaeW/7vcao6bdbkXCRJ1+2sWE15ZJvJVEKVANrOKy1mRngqC3IFIAP+wP5qSw=="],

-    "@tauri-apps/plugin-opener": ["@tauri-apps/plugin-opener@2.5.2", "", { "dependencies": { "@tauri-apps/api": "^2.8.0" } }, "sha512-ei/yRRoCklWHImwpCcDK3VhNXx+QXM9793aQ64YxpqVF0BDuuIlXhZgiAkc15wnPVav+IbkYhmDJIv5R326Mew=="],
+    "@tauri-apps/plugin-opener": ["@tauri-apps/plugin-opener@2.5.4", "", { "dependencies": { "@tauri-apps/api": "^2.11.0" } }, "sha512-1HnPkb+AmgO29HBazm4uPLKB+r7zzcTBW1d0fyYp1uP+jwtpoiNDGKMMzz58SFp49nOIrxdE3aUJtT57lfO9CQ=="],

    "@tauri-apps/plugin-os": ["@tauri-apps/plugin-os@2.3.2", "", { "dependencies": { "@tauri-apps/api": "^2.8.0" } }, "sha512-n+nXWeuSeF9wcEsSPmRnBEGrRgOy6jjkSU+UVCOV8YUGKb2erhDOxis7IqRXiRVHhY8XMKks00BJ0OAdkpf6+A=="],

@@ -761,9 +761,9 @@

    "@tailwindcss/oxide-wasm32-wasi/tslib": ["tslib@2.8.1", "", { "bundled": true }, "sha512-oJFu94HQb+KVduSUQL7wnpmqnfmLsOA/nAh6b6EH0wCEoK0/mPeXU6c3wKDV83MkOuHPRHtSXKKU99IBazS/2w=="],

-    "@tauri-apps/plugin-log/@tauri-apps/api": ["@tauri-apps/api@2.11.1", "", {}, "sha512-M2FPuYND2m+wh5hfW9ZpSdxMPdEJovPBWwoHJmwUpysTYNHaOkVFN419m/K0LIgjb/7KU2vBgsUepJWugQCvAA=="],
+    "@tauri-apps/plugin-os/@tauri-apps/api": ["@tauri-apps/api@2.9.1", "", {}, "sha512-IGlhP6EivjXHepbBic618GOmiWe4URJiIeZFlB7x3czM0yDHHYviH1Xvoiv4FefdkQtn6v7TuwWCRfOGdnVUGw=="],

-    "@tauri-apps/plugin-updater/@tauri-apps/api": ["@tauri-apps/api@2.11.1", "", {}, "sha512-M2FPuYND2m+wh5hfW9ZpSdxMPdEJovPBWwoHJmwUpysTYNHaOkVFN419m/K0LIgjb/7KU2vBgsUepJWugQCvAA=="],
+    "@tauri-apps/plugin-process/@tauri-apps/api": ["@tauri-apps/api@2.9.1", "", {}, "sha512-IGlhP6EivjXHepbBic618GOmiWe4URJiIeZFlB7x3czM0yDHHYviH1Xvoiv4FefdkQtn6v7TuwWCRfOGdnVUGw=="],

    "@testing-library/dom/aria-query": ["aria-query@5.3.0", "", { "dependencies": { "dequal": "^2.0.3" } }, "sha512-b0P0sZPKtyu8HkeRAfCq0IfURZK+SuwMjY1UXGBU27wpAiTwQAIlq56IbIO+ytk/JjS1fMR14ee5WBBfKi5J6A=="],

@@ -61,6 +61,11 @@ filled. Keep a couple of dated tags live and prune the rest.

 The order matters — CI breaks if the workflow lands before the image exists.

+A caveat learned the hard way: the *trailing* layer is only fast for `cargo
+install` tools. Adding an **apt** package invalidates the packaging layer, which
+sits above the `cargo-xwin`/`cargo-deny` installs, so those recompile too — a
+~20 minute rebuild rather than ~2.
+
 ```bash
 # 1. Edit Dockerfile.builder. Put new tools in the TRAILING layer: it exists so
 #    a tool change is a ~2 min rebuild instead of ~15.
@@ -80,6 +85,48 @@ docker run --rm gitea.tourolle.paris/dtourolle/jellytau-builder:2026.09 \
 toolchain inside the job — a toolchain install in CI. Bump both, rebuild, push,
 then merge.

+## Tauri plugin versions are pinned in pairs
+
+Every Tauri plugin exists twice: a Rust crate in `src-tauri/Cargo.toml` and an
+npm package in `package.json`. **The Tauri CLI refuses to build when the two are
+on different minor versions** — not a warning, a hard stop before compilation.
+
+Both sides are therefore pinned *exactly* (`"2.8.0"`, not `"^2.8.0"`). A caret
+range is what let them drift apart in the first place: `bun add` took the latest
+npm package while cargo held an older crate, and nothing noticed until a release
+build refused to start.
+
+Nothing in `build-and-test.yml` runs `tauri build` — that happens only on a tag —
+so this class of breakage used to be invisible until release day. The
+`Check Tauri plugin versions match` step runs `tauri info`, which performs the
+same comparison without building.
+
+To upgrade a plugin, move **both** sides together and re-run that step. Expect
+the Rust side to be the constraint: a newer plugin crate may pull a large
+transitive upgrade (bumping `tauri-plugin-log` to 2.9.0 also moved `wry`,
+`wasm-bindgen`, `web-sys` and `webkit2gtk`), which touches the webview and
+therefore video playback. That is a change to make deliberately, with a full
+build and a playback check — not one to slip into a release.
+
+## AppImage needs more than the Rust toolchain
+
+`linuxdeploy` (which Tauri downloads at build time to assemble the AppImage)
+shells out to distro tools that a minimal server image does not have. It aborts
+the whole bundle on the first one missing:
+
+```
+failed to bundle project: xdg-open binary not found
+```
+
+The image therefore carries `xdg-utils`, `desktop-file-utils` and `zsync`. This
+is a class of failure that **cannot be caught by building locally**: a developer
+machine is a desktop and has all three, so the AppImage builds there and fails in
+CI. It cost one release build to find.
+
+Tauri's AppImage bundler also downloads `linuxdeploy`, `AppRun` and two plugin
+scripts from GitHub during the build. That is Tauri's behaviour, not ours, but it
+means an AppImage build depends on GitHub being reachable from the runner.
+
 ## Secrets

 Managed with the `tea` CLI (`tea actions secrets list`) or the repo settings UI.
@@ -410,6 +410,12 @@ Internal architecture, components, and application logic.
 | DR-216 | Dependencies are gated on known vulnerabilities and on licence compatibility, and the build graph is pinned to what is actually shipped. The project had no scanning of any kind: nothing checked the ~500-crate Rust graph or the JS packages against an advisory feed, and nothing checked that everything redistributed inside an MIT-licensed bundle permits it. The first run found eight vulnerabilities and one unsoundness — `bytes`, four in `rustls-webpki`, `time`, two in `quick-xml`, `rand` — every one closed by a `cargo update` nobody had reason to run. `cargo deny` (src-tauri/deny.toml) now runs in CI over advisories, licences, bans and sources. Two structural fixes matter as much as the gate: the graph is scoped to the targets actually shipped, so an advisory against an Apple-only path is correctly absent rather than ignored by ID; and the one git dependency (`libmpv`) is pinned by revision instead of by branch, since a branch means any `cargo update` silently substitutes new upstream code in the one dependency that is unsigned and links a C library into the player. Licence findings are recorded rather than waved through — `libmpv`/`libmpv-sys` are LGPL-2.1, which the app satisfies by dynamic linking, and that carries obligations (keep the linkage dynamic; ship libmpv's licence text with any bundle carrying the .so) | Tooling | - | Done |
 | DR-217 | In-app update, desktop only, over a manifest we control. `tauri-plugin-updater` and `tauri-plugin-process` are compiled for everything except Android/iOS — spelled as a target-triple cfg rather than `cfg(desktop)`, which Cargo does not evaluate in a `[target.'cfg(…)']` table and which therefore drops the dependency silently, surfacing much later as "Permission updater:default not found". The release workflow signs updater artifacts with a minisign key held in Gitea secrets and publishes `latest.json` to a dedicated `updater` branch, read over Gitea's raw-file URL: this instance serves `/releases/download/<tag>/<asset>` but returns 404 for `/releases/latest/download/<asset>`, so there is no stable latest-release URL to point at, and the docs branch is force-pushed by publish-docs.yml so it cannot host the manifest either. Bundle targets gain `appimage`, which the release notes had been advertising for months while `tauri.conf.json` never built it — the artifact step globbed for `*.AppImage`, found nothing, and said nothing | Tooling | UR-077 | Done |
 | DR-218 | Persistent, redacted logging and a diagnostics export. `tauri-plugin-log` replaces the `env_logger` stdout-only init, giving a rotating 5 MB file, a webview target in dev, and — the single largest gain — logcat on Android, where `env_logger`'s stdout went nowhere. **Redaction runs in the log formatter, not at export**: a credential in a file on the device is already a disclosure, so stripping it on the way out would be too late; the exporter redacts a second time to cover files written by older builds. `api_key`/`X-Emby-Token`/`Authorization`/`"AccessToken"`/`Token="…"` all reduce to `[REDACTED]` while host, item ids and filenames are deliberately kept — a bundle scrubbed of those is one nobody can debug from. The server URL is reduced to scheme and host, dropping any embedded `user:pass@`. The panic hook chains to the previous hook rather than replacing it, because `utils/lock.rs` installs a silencing hook around tests that provoke poisoned locks on purpose. The chosen level persists to disk and is re-applied at startup, since reproducing a bug usually means restarting into it. The frontend facade keeps its untouched `console.*` pass-through (DR-204) and additionally forwards a stringified copy at info and above, so one file holds both halves of the app in order — which is what makes a race between them legible after the fact | Tooling | UR-078 | Done |
+| DR-219 | Release notes are the reviewed CHANGELOG entry, not a generated draft. Every release from v0.0.1 to v0.9.1 published the same ~1,050 bytes of generic install instructions whose "What's New" section said "See CHANGELOG.md" — a link that does not resolve from a release page. Thirty-five releases, byte-identical, telling a reader nothing about what changed. The workflow now publishes the `## <version>` section of CHANGELOG.md and fails the release if that section is absent, since notes that say nothing are worse than a build that waits for two sentences. `release:notes` is printed into the job log as a drafting aid but is deliberately *not* published: CLAUDE.md calls its output "a reviewed draft, not a final changelog", and publishing it unreviewed proved why — a range containing a repo-wide formatting sweep resolved to nearly the entire requirement matrix and produced notes claiming one release had added the whole application. The script now skips cosmetic commits (`chore(format)`, `chore(deps)`, `style`) when deriving a range's files, and says how many it skipped rather than silently reporting a smaller set | Tooling | - | Done |
+| DR-220 | A release ships only its own artifacts. `src-tauri/target/*/release/bundle/` is not versioned, cargo never cleans it, and the CI runner reuses the target directory — so the copy step's `bundle/**/*-setup.exe` glob collected every installer ever built there. Every release from v0.1.0 to v0.8.2 shipped its predecessors': sixteen Windows installers on v0.8.2, thirteen of them stale, and a download list on v0.5.0 reaching back to 0.1.0. It went unnoticed for eight months because there was nothing to notice — the upload loop reported success, the files were real, and the page looked busy rather than wrong. It stopped only when an unrelated cache change wiped the runner's target dir, leaving the defect dormant rather than fixed. Both desktop builds now clear the bundle directory first, so a stale file cannot exist to be copied — filtering the copy by version would have hidden it instead. `scripts/check-release-artifacts.sh` is the backstop for the next route nobody predicts: it runs before the SBOM, the checksums and the upload, and refuses to publish when any artifact's embedded version disagrees with the tag | Tooling | - | Done |
+| DR-221 | The release path is exercised before a tag exists. Nothing in `build-and-test.yml` runs `tauri build` — only a tag does — so a whole class of breakage was invisible until release day, and two instances of it were sitting on master at once. Tauri refuses to build when a plugin's Rust crate and npm package differ by minor version, which the updater and logging work had introduced (`tauri-plugin-log 2.8.0` against `@tauri-apps/plugin-log 2.9.0`) while `cargo check`, clippy, the tests and `svelte-check` all passed; both sides are now pinned exactly rather than by caret, since a caret is what let them separate, and CI runs `tauri info` to compare them without building. The AppImage target had never once been built: linuxdeploy carries a `strip` too old to parse the `.relr.dyn` section modern toolchains emit, so bundling failed on every library — and Ubuntu 23.10+ links with `-z pack-relative-relocs` by default, so the builder image fails the same way a modern Arch host does. `NO_STRIP=true` is linuxdeploy's documented escape hatch; the cost is a larger, unstripped bundle. Both were found by building the target locally before tagging rather than by publishing a release that could not build | Tooling | - | Done |
+| DR-222 | Build tooling matches the package manager the project declares. `scripts/build-android.sh` ran `npm install` on its clean-build path — in a bun project, where `packageManager` says bun and `bun.lock` is the committed lockfile. npm ignores that lockfile, re-resolves the whole tree from package.json, and writes a `package-lock.json` that `.gitignore` then hides. That is not a style preference: the JS halves of the Tauri plugins are pinned exactly against Cargo.lock because the CLI refuses to build when a plugin's crate and package differ by minor version, and a silent re-resolve is precisely how they drift apart. It survived because clean builds are rare — the shape shared by nearly every defect found preparing v0.10.0, where the code running on every commit was healthy and the code running on a release, a tag or a clean build had no guard at all. `scripts/check-tooling.sh` fails on any npm/yarn/pnpm invocation or foreign lockfile | Tooling | - | Done |
+| DR-223 | The Android JavaVM and Application are published into `ndk_context` by this crate, not by a transitive dependency. Seven call sites (five in credentials.rs, two in lib.rs) read that process-global to reach JNI, and nothing here ever set it — `tao` did, three levels below anything this project names in Cargo.toml. tao 0.35.3 moved those pointers into a private struct and stopped publishing them, so the Tauri 2.11 upgrade made the first credential read abort the process on every launch: `PANIC ... android context was not initialized`. Our code had not changed; an undocumented side effect of the windowing layer had gone. The invariant is now owned here rather than assumed: `JNI_OnLoad` captures the JavaVM as the shared library loads, and the Application is resolved lazily via `ActivityThread.currentApplication()` and pinned as a global reference for the process lifetime — the Application rather than the Activity, since that is what `SecureStorage.initialize()` immediately reduces its argument to. Failure degrades to the encrypted-file credential path and is logged, rather than aborting. Found only by installing on a device: nothing in CI runs the app | Security | UR-012 | Done |
+| DR-224 | Backgrounding the app obeys the background-audio toggle on every renderer. The toggle (UR-040) was built for the WebView `<video>` path, where losing visibility kills the decode: it chose between handing off to a native audio stream and letting playback stop. Native video then became the default renderer (DR-188), and on that path playback runs through ExoPlayer inside a `MediaSessionService` — a foreground media service whose purpose is to keep playing while the app is hidden. Nothing paused it and nothing in the codebase paused on background, so locking the screen kept the audio going whether or not the toggle was on: the toggle governed a handoff that no longer had a gap to bridge, and users got background playback they never asked for. The decision now lives in Rust (`player/background_policy.rs`) and both renderers obey it: a video with the toggle off pauses, with the toggle on hands off to audio, music is never paused by backgrounding, and picture-in-picture keeps playing because the window is still on screen (UR-041). It takes no renderer parameter on purpose — the split between the two paths is what produced the defect | Player | UR-040 | Done |
 | DR-198 | The webview runs under a real Content-Security-Policy, and the asset protocol is scoped to the one directory it still serves. `csp` was `null`, which disables CSP entirely: any script that reached the web layer — through a future `{@html}`, a dependency, or a devtools paste — would have inherited the whole IPC surface, and with it the user's session. `script-src 'self'` (Tauri injects a nonce for SvelteKit's inline bootstrap script at build time, so no `'unsafe-inline'` is needed) plus `object-src`/`frame-src 'none'` and `base-uri 'self'` is the part that is genuinely restrictive. `img-src`/`media-src`/`connect-src` cannot be: the Jellyfin origin is typed in by the user at run time and is commonly plain `http` on a LAN, so they allow `http:`/`https:` — a wide grant for *data*, but one that still bars `file:`, `filesystem:` and scripting schemes, and leaves `script-src` untouched. `style-src` keeps `'unsafe-inline'` because Svelte compiles `style="…"` attributes (including `app.html`'s `display: contents` wrapper) into markup; this is safe only while no `<style>` element survives into `index.html`, since a nonce there would make Tauri's injection outrank — and therefore void — `'unsafe-inline'`. `worker-src blob:` and `media-src blob:` are hls.js: it demuxes in a worker built from a blob and attaches MSE through `URL.createObjectURL`. `asset:` and `http://asset.localhost` are the same protocol under the two naming schemes `convertFileSrc` emits (custom scheme on Linux/macOS, `http` host on Windows/Android); `ipc:`/`http://ipc.localhost` is the invoke transport, which would otherwise be blocked by `connect-src`. A run-time CSP naming the server origin exactly was rejected: Tauri computes the header from immutable config when it serves the HTML, so it would mean rebuilding config and reloading the webview on every server change, for a policy the user can already point anywhere. The asset-protocol scope narrows from `$APPDATA/**` to `$APPDATA/thumbnails/**` — since DR-137 moved downloaded media to the loopback server, `imageCache` is the only `convertFileSrc` caller left, so the database and the encrypted-token fallback file no longer sit inside the grant | Security | UR-012, UR-071 | Done |

 ---
@@ -709,6 +715,8 @@ Internal architecture, components, and application logic.
 | UT-207 | The hero banner's rotation timer restarts from the moment of a manual change: a swipe 5.5s into a 6s interval waits a further 6s instead of firing the leftover 500ms, repeated restarts never stack timers, and `stop()` ends rotation | DR-038 | Done |
 | UT-208 | The update decision: each numeric version field is compared in order, the installed version is not offered to itself, a leading `v` is tolerated because that is how the tags are written, a pre-release sorts below the release of the same number so 0.9.2-rc1 is not offered to somebody on 0.9.2, a missing patch field reads as zero rather than NaN, mobile reports link-only while desktop reports install, and absent release notes normalise to null rather than undefined | DR-217 | Done |
 | UT-209 | Redaction and forwarding. Rust: every credential shape reduces to `[REDACTED]` while the host, username and neighbouring parameters survive; redaction is idempotent, leaves ordinary lines alone, does not fire on the word "token" in prose, and does not panic on multi-byte input; a server URL keeps only scheme and host and drops an embedded `user:pass@`; an unparseable level falls back to info rather than failing at startup. Frontend: info and above forward while debug does not, a message the level filter suppressed is not forwarded, a throwing forwarder neither propagates nor prevents the console write, and an `Error` renders as name and message rather than the `{}` that `JSON.stringify` produces | DR-218 | Done |
+| UT-210 | Cosmetic-commit detection for release notes: a `chore(format)`, `chore(deps)` or `style` subject is skipped when deriving a range's changed files, while `fix`, `feat`, `ci`, `docs`, a bare `chore:` and `chore(release):` are kept; and the word "format" appearing later in a subject ("fix(duration): format times over 24 hours") does not make a real fix look cosmetic | DR-219 | Done |
+| UT-211 | The background decision: a video with the toggle off pauses (the reported defect, where the media service kept playing regardless), a video with it on hands off to audio, music keeps playing whatever the toggle says because it has no picture to lose, picture-in-picture keeps playing in every combination since the window is still visible, and the answer does not vary by renderer | DR-224 | Done |

 ### Integration Tests

@@ -28,7 +28,7 @@ know how something *works*, read

 **Next free requirement ids** (always re-check
 [requirements.md](../requirements.md) before allocating): **UR-079**,
-**IR-033**, **DR-219**. Three specs below suggested ids that have since been
+**IR-033**, **DR-225**. Three specs below suggested ids that have since been
 taken by other work; each carries a ⚠️ note at the top.

 ## Partially implemented
@@ -1,6 +1,6 @@
 {
  "name": "jellytau",
-  "version": "0.9.1",
+  "version": "0.10.1",
  "description": "A cross-platform Jellyfin client built with Tauri, SvelteKit and Rust.",
  "author": "Duncan Tourolle <duncan@tourolle.paris>",
  "license": "MIT",
@@ -29,6 +29,7 @@
    "format:check": "prettier --check .",
    "check:boundary": "bash scripts/check-frontend-boundary.sh",
    "check:links": "bash scripts/check-doc-links.sh",
+    "check:tooling": "bash scripts/check-tooling.sh",
    "hooks:install": "./scripts/install-hooks.sh",
    "android:build": "./scripts/build-android.sh",
    "android:build:release": "./scripts/build-android.sh release",
@@ -55,12 +56,12 @@
    "release:notes": "bun run scripts/release-notes.ts"
  },
  "dependencies": {
-    "@tauri-apps/api": "^2",
-    "@tauri-apps/plugin-log": "^2.9.0",
-    "@tauri-apps/plugin-opener": "^2",
+    "@tauri-apps/api": "^2.11.1",
+    "@tauri-apps/plugin-log": "2.9.0",
+    "@tauri-apps/plugin-opener": "^2.5.4",
    "@tauri-apps/plugin-os": "^2.3.2",
    "@tauri-apps/plugin-process": "^2.3.1",
-    "@tauri-apps/plugin-updater": "^2.10.1",
+    "@tauri-apps/plugin-updater": "2.10.1",
    "hls.js": "^1.6.15",
    "svelte-dnd-action": "^0.9.69"
  },
@@ -70,7 +71,7 @@
    "@sveltejs/kit": "^2.9.0",
    "@sveltejs/vite-plugin-svelte": "^6.2.4",
    "@tailwindcss/vite": "^4.1.18",
-    "@tauri-apps/cli": "^2",
+    "@tauri-apps/cli": "^2.11.4",
    "@testing-library/svelte": "^5.3.1",
    "@vitest/coverage-v8": "^4.0.18",
    "@vitest/ui": "^4.0.16",
@@ -8,7 +8,7 @@
 # tarball/VCS URL and drop the local-copy prepare() step.

 pkgname=jellytau
-pkgver=0.9.1
+pkgver=0.10.1
 pkgrel=1
 pkgdesc="A cross-platform Jellyfin client"
 arch=('x86_64')
@@ -82,7 +82,17 @@ fi
 if [ "$CLEAN" = "1" ]; then
    echo "🧹 Clearing build caches (clean build)..."
    rm -rf node_modules/.vite dist .svelte-kit .next build target src-tauri/target 2>/dev/null || true
-    npm install > /dev/null 2>&1
+    # `bun install`, NOT `npm install`. This is a bun project (see packageManager
+    # in package.json) and bun.lock is the lockfile that is committed; npm
+    # ignores it, re-resolves the tree from package.json alone, and writes a
+    # package-lock.json that .gitignore then hides.
+    #
+    # That is not cosmetic. The Tauri CLI refuses to build when a plugin's Rust
+    # crate and npm package differ by minor version, so the JS side is pinned
+    # exactly to match Cargo.lock; a re-resolve is precisely how those halves
+    # drift apart again. A clean build must not be able to change what gets
+    # installed.
+    bun install > /dev/null 2>&1
 fi

 # Step 1: Sync Android source files
@@ -94,22 +104,28 @@ echo "🎨 Building frontend..."
 bun run build

 # Step 2: Build Android APK
+# `--apk` is a boolean flag, NOT `--apk true`.
+#
+# tauri-cli took a value here until 2.10; from 2.11 it is a plain flag and the
+# stray `true` is parsed as a positional argument, failing with
+# "error: unexpected argument 'true' found" before the build starts. Found by
+# deploying to a device after the Tauri 2.9.5 -> 2.11.5 upgrade.
 if [ "$BUILD_TYPE" = "release" ] && [ "$SIDE_BY_SIDE" = "1" ]; then
    # A release build in the debug slot: R8 still runs, but the applicationId is
    # suffixed and the debug keystore signs it (read by build.gradle.kts from
    # JT_SIDE_BY_SIDE), so the real key is not needed and it replaces any other
    # .debug install cleanly. Deliberately does NOT write keystore.properties.
    echo "📦 Building side-by-side release APK (com.dtourolle.jellytau.debug)..."
-    JT_SIDE_BY_SIDE=1 bun run tauri android build --apk true "${TARGET_ARGS[@]}"
+    JT_SIDE_BY_SIDE=1 bun run tauri android build --apk "${TARGET_ARGS[@]}"
 elif [ "$BUILD_TYPE" = "release" ]; then
    # Configure release signing from .env (single source of truth). Must run
    # after sync-android-sources.sh, since gen/android is (re)generated there.
    ./scripts/write-keystore-properties.sh
    echo "📦 Building release APK..."
-    bun run tauri android build --apk true "${TARGET_ARGS[@]}"
+    bun run tauri android build --apk "${TARGET_ARGS[@]}"
 else
    echo "📦 Building debug APK..."
-    bun run tauri android build --apk true --debug "${TARGET_ARGS[@]}"
+    bun run tauri android build --apk --debug "${TARGET_ARGS[@]}"
 fi

 echo ""
@@ -26,7 +26,25 @@ bun run build

 # --bundles overrides tauri.conf.json bundle.targets so this script controls
 # exactly which Linux formats are produced (never NSIS here).
-bun run tauri build --bundles "$BUNDLES"
+# TRACES: | DR-221
+#
+# 🔴 NO_STRIP=true is required for the AppImage bundle.
+#
+# linuxdeploy (which Tauri downloads and runs to build the AppImage) carries its
+# own `strip`, and that copy is too old to parse the `.relr.dyn` section modern
+# toolchains emit for RELR relocations. It fails on essentially every bundled
+# library:
+#
+#   strip: libzstd.so.1: unknown type [0x13] section `.relr.dyn'
+#   failed to bundle project `failed to run linuxdeploy-x86_64.AppImage`
+#
+# Ubuntu 23.10+ links with -z pack-relative-relocs by default, so the CI builder
+# image hits this exactly as a modern Arch host does. Skipping the strip step is
+# linuxdeploy's own documented escape hatch; the cost is an unstripped, larger
+# AppImage (~153 MB for a build that bundles libmpv and its ffmpeg stack).
+#
+# Remove this only after confirming a linuxdeploy release that understands RELR.
+NO_STRIP=true bun run tauri build --bundles "$BUNDLES"

 BUNDLE_ROOT="src-tauri/target/release/bundle"
 echo ""
@@ -46,6 +46,28 @@ bun run build
 # from tauri.conf.json (bundle.targets includes "nsis"), which is not subject to
 # that CLI validation — the bundler then picks nsis once it knows the target is
 # Windows.
+# TRACES: | DR-221
+#
+# 🔴 Clear the bundle output before building.
+#
+# The bundle directory is not versioned and is never cleaned by cargo, and the
+# CI runner reuses src-tauri/target between builds. The copy step below globs
+# `bundle/**/*-setup.exe`, so every stale installer left there was picked up and
+# attached to the release: v0.8.2 shipped sixteen Windows installers, thirteen
+# of them from earlier versions, and v0.5.0 offered users a download list going
+# back to 0.1.0. Every release from v0.1.0 to v0.8.2 did this. It stopped only
+# because an unrelated change wiped the runner's target dir, so it is dormant
+# rather than fixed.
+#
+# Filtering the copy by version would hide it; removing the directory means a
+# stale file cannot exist to be copied. scripts/check-release-artifacts.sh is
+# the backstop if some other path reintroduces one.
+BUNDLE_DIR="src-tauri/target/$TARGET/release/bundle"
+if [[ -d "$BUNDLE_DIR" ]]; then
+  echo "🧹 Clearing previous bundle output at $BUNDLE_DIR"
+  rm -rf "$BUNDLE_DIR"
+fi
+
 if [[ "$WIN_BUNDLES" == "none" ]]; then
  bun run tauri build --runner cargo-xwin --target "$TARGET" --no-bundle
 else
@@ -0,0 +1,103 @@
+#!/usr/bin/env bash
+# Refuse to publish a release whose artifacts are not all from this release.
+#
+# TRACES: | DR-220
+#
+#   ./scripts/check-release-artifacts.sh <version> <dir> [<dir>...]
+#
+# e.g.
+#   ./scripts/check-release-artifacts.sh v0.9.2 artifacts/linux artifacts/windows
+#
+# ## The defect this exists for
+#
+# Every JellyTau release from v0.1.0 to v0.8.2 shipped every Windows installer
+# ever built. `src-tauri/target/*/release/bundle/` is not versioned, cargo never
+# cleans it, and the CI runner reuses the target directory between builds — so
+# the copy step's `bundle/**/*-setup.exe` glob collected the whole history. By
+# v0.8.2 that was sixteen installers, thirteen of them stale. v0.5.0 offered
+# users a download list going back to 0.1.0.
+#
+# Nobody noticed for eight months. There was nothing to notice with: the upload
+# loop reported success, the assets were real files, and the release page looked
+# busy rather than wrong.
+#
+# The builds now clear the bundle directory first, which removes the cause. This
+# is the backstop for the next thing that reintroduces a stale file by a route
+# nobody predicted — a cached directory, a restored artifact, a hand-copied fix.
+#
+# ## What it checks
+#
+# Every file whose name embeds a semantic version must embed *this* version.
+# Files with no version in the name (jellytau-release.apk, jellytau.exe,
+# SHA256SUMS, latest.json) are accepted: they are produced fresh each build and
+# have no version to disagree with.
+
+set -euo pipefail
+
+if [ "$#" -lt 2 ]; then
+  echo "usage: $0 <version> <dir> [<dir>...]" >&2
+  exit 2
+fi
+
+VERSION_RAW="$1"
+shift
+# Accept the tag form (v0.9.2) or the bare form (0.9.2).
+VERSION="${VERSION_RAW#v}"
+
+echo "🔎 Checking release artifacts are all version ${VERSION}…"
+
+FOUND=0
+STALE=0
+UNVERSIONED=0
+
+for dir in "$@"; do
+  if [ ! -d "$dir" ]; then
+    echo "   (no $dir — skipping)"
+    continue
+  fi
+
+  # -print0/read -d '' so a filename with a space cannot split into two.
+  while IFS= read -r -d '' file; do
+    name="$(basename "$file")"
+    FOUND=$((FOUND + 1))
+
+    # First x.y.z in the filename, if any.
+    embedded="$(printf '%s' "$name" | grep -oE '[0-9]+\.[0-9]+\.[0-9]+' | head -1 || true)"
+
+    if [ -z "$embedded" ]; then
+      UNVERSIONED=$((UNVERSIONED + 1))
+      continue
+    fi
+
+    if [ "$embedded" != "$VERSION" ]; then
+      echo "   ❌ $name carries version $embedded"
+      STALE=$((STALE + 1))
+    fi
+  done < <(find "$dir" -type f -print0)
+done
+
+echo ""
+echo "   $FOUND file(s) checked; $UNVERSIONED carry no version in the name."
+
+if [ "$FOUND" -eq 0 ]; then
+  echo "❌ No artifacts found at all. A release with no files is a failed build," >&2
+  echo "   not an empty one." >&2
+  exit 1
+fi
+
+if [ "$STALE" -gt 0 ]; then
+  echo ""
+  echo "❌ $STALE artifact(s) belong to a different version than ${VERSION}." >&2
+  echo "" >&2
+  echo "   This is how every release from v0.1.0 to v0.8.2 came to ship its" >&2
+  echo "   predecessors' Windows installers: src-tauri/target/*/release/bundle/" >&2
+  echo "   is never cleaned and the runner reuses it, so a glob picks up" >&2
+  echo "   whatever was left behind." >&2
+  echo "" >&2
+  echo "   The builds clear that directory first, so seeing this means a stale" >&2
+  echo "   file arrived by some other route. Find it before publishing — do not" >&2
+  echo "   delete the file and re-run." >&2
+  exit 1
+fi
+
+echo "✅ Every versioned artifact is ${VERSION}."
@@ -0,0 +1,73 @@
+#!/usr/bin/env bash
+# Refuse tooling that contradicts what this project actually uses.
+#
+# TRACES: | DR-222
+#
+#   ./scripts/check-tooling.sh
+#
+# ## Why
+#
+# This is a bun project: `packageManager` in package.json says so, bun.lock is
+# the committed lockfile, and .gitignore hides the other package managers'
+# lockfiles precisely so they cannot be committed by accident.
+#
+# scripts/build-android.sh nonetheless ran `npm install` on its clean-build
+# path. npm ignores bun.lock, re-resolves the whole tree from package.json, and
+# writes a package-lock.json that .gitignore then hides from view.
+#
+# That is not a style preference. The Tauri CLI refuses to build when a plugin's
+# Rust crate and npm package differ by minor version, so the JS side is pinned
+# exactly against Cargo.lock -- and a silent re-resolve is exactly how those
+# halves drift apart again. The drift already cost one release build.
+#
+# It survived because the clean-build path runs rarely. That is the shape of
+# nearly every defect found while preparing v0.10.0: the code that runs on every
+# commit was fine, and the code that runs on a release, a clean build or a tag
+# had no guard at all.
+
+set -uo pipefail
+
+REPO_ROOT="$(git rev-parse --show-toplevel)"
+cd "$REPO_ROOT" || exit 1
+
+FAILED=0
+
+echo "🔎 Checking build tooling is consistent with packageManager…"
+
+# Only the *invocations* matter. A comment explaining why npm is wrong, or a
+# .gitignore entry naming package-lock.json, is not a violation -- so match a
+# command at the start of a line or after a shell separator.
+PATTERN='(^|[;&|(]|&&|\|\||\bthen |\bdo |[[:space:]]{4,})(npm|yarn|pnpm)[[:space:]]+(install|ci|add|run|exec)\b'
+
+MATCHES="$(grep -rInE "$PATTERN" \
+  --include='*.sh' --include='*.yml' --include='*.yaml' \
+  scripts/ .gitea/ 2>/dev/null | grep -v '^\s*#' || true)"
+
+if [ -n "$MATCHES" ]; then
+  echo "❌ A non-bun package manager is invoked:"
+  echo "$MATCHES" | sed 's/^/   /'
+  echo ""
+  echo "   This project uses bun (packageManager in package.json, bun.lock"
+  echo "   committed). npm/yarn/pnpm ignore that lockfile and re-resolve the"
+  echo "   dependency tree, which is how the Tauri plugin crate/package"
+  echo "   versions drifted apart and broke a release build."
+  echo ""
+  echo "   Use: bun install / bun run / bunx"
+  FAILED=1
+fi
+
+# A lockfile from another manager should never exist here; .gitignore hides
+# them, so one can sit in a working tree unnoticed and change what installs.
+for stray in package-lock.json yarn.lock pnpm-lock.yaml; do
+  if [ -f "$stray" ]; then
+    echo "❌ $stray exists. Another package manager has run here."
+    echo "   Delete it and run: bun install"
+    FAILED=1
+  fi
+done
+
+if [ "$FAILED" -eq 0 ]; then
+  echo "✅ Only bun is used, and no foreign lockfile is present."
+fi
+
+exit "$FAILED"
@@ -49,6 +49,11 @@ describe("isTracedSourceFile", () => {
    expect(isTracedSourceFile("src-tauri/deny.toml")).toBe(true);
    expect(isTracedSourceFile("src-tauri/rust-toolchain.toml")).toBe(true);
    expect(isTracedSourceFile("scripts/hooks/pre-commit")).toBe(true);
+    // Shell tooling is listed individually, not globbed: most scripts/*.sh
+    // implement nothing, and adding one should be a decision.
+    expect(isTracedSourceFile("scripts/check-release-artifacts.sh")).toBe(true);
+    expect(isTracedSourceFile("scripts/build-desktop-linux.sh")).toBe(true);
+    expect(isTracedSourceFile("scripts/logcat.sh")).toBe(false);
  });

  it("does not scan CI workflows, whose comments discuss TRACES in prose", () => {
@@ -95,6 +95,14 @@ const TOOLING_FILES = new Set([
  "scripts/hooks/pre-commit",
  "src-tauri/deny.toml",
  "src-tauri/rust-toolchain.toml",
+  // Shell tooling that implements a requirement. Named individually rather than
+  // globbing scripts/*.sh: most of these scripts implement nothing, and the
+  // point of the list is that adding a file is a decision.
+  "scripts/install-hooks.sh",
+  "scripts/check-release-artifacts.sh",
+  "scripts/build-desktop-linux.sh",
+  "scripts/build-windows-cross.sh",
+  "scripts/restore-ownership.sh",
 ]);

 /** Directory names that never contain hand-written traced source. */
@@ -0,0 +1,58 @@
+/**
+ * Tests for release-note derivation.
+ *
+ * TRACES: | DR-219 | UT-210
+ *
+ * The bug these were written against: `bun run release:notes v0.9.1..HEAD`
+ * listed *every user requirement in the project* as a feature of the release.
+ * The range contained a repo-wide `prettier --write` sweep, so `git diff
+ * --name-only` reported 199 files, their TRACES comments resolved to nearly the
+ * whole matrix, and the result claimed one release had added the entire
+ * application.
+ *
+ * That mattered more than it looked: build-release.yml now generates the
+ * published release body from this script, so the noise would have shipped.
+ */
+
+import { describe, it, expect } from "vitest";
+import { isCosmeticCommit } from "./release-notes";
+
+describe("isCosmeticCommit", () => {
+  it("treats a formatting sweep as cosmetic", () => {
+    // The actual commit that triggered this.
+    expect(isCosmeticCommit("chore(format): run prettier over src/ and scripts/")).toBe(true);
+    expect(isCosmeticCommit("style: reindent the player module")).toBe(true);
+    expect(isCosmeticCommit("style(player): reindent")).toBe(true);
+  });
+
+  it("treats a lockfile-only dependency bump as cosmetic", () => {
+    // Touches package.json/bun.lock, which carry no TRACES, but a `chore(deps)`
+    // that also edits source would still be caught by that source file.
+    expect(isCosmeticCommit("chore(deps): bump vitest to 4.1.11")).toBe(true);
+  });
+
+  it("does NOT treat ordinary work as cosmetic", () => {
+    expect(isCosmeticCommit("fix(player): restart the hero banner timer")).toBe(false);
+    expect(isCosmeticCommit("feat(updater): in-app update on desktop")).toBe(false);
+    expect(isCosmeticCommit("ci: make the frontend gates real")).toBe(false);
+    expect(isCosmeticCommit("docs: add SECURITY.md")).toBe(false);
+  });
+
+  it("does not mistake a chore that is not formatting for a formatting one", () => {
+    // `chore(release)` bumps versions and must still be attributable; a bare
+    // `chore:` could be anything, so it is NOT skipped by default.
+    expect(isCosmeticCommit("chore(release): v0.9.2")).toBe(false);
+    expect(isCosmeticCommit("chore: tidy up the queue helper")).toBe(false);
+  });
+
+  it("is not fooled by the word format appearing later in a subject", () => {
+    // A real fix to formatting *code* is not a cosmetic commit.
+    expect(isCosmeticCommit("fix(duration): format times over 24 hours correctly")).toBe(false);
+    expect(isCosmeticCommit("feat: add a format picker to settings")).toBe(false);
+  });
+
+  it("handles an empty or malformed subject without throwing", () => {
+    expect(isCosmeticCommit("")).toBe(false);
+    expect(isCosmeticCommit("   ")).toBe(false);
+  });
+});
@@ -54,11 +54,86 @@ function loadRequirementDescriptions(): Map<string, string> {
  return map;
 }

+/**
+ * Commit subjects whose changes carry no requirement meaning.
+ *
+ * `chore(format)` / `style` rewrite files without changing behaviour;
+ * `chore(deps)` moves lockfiles. Anything else — including a bare `chore:` and
+ * `chore(release):` — is assumed to mean something and is kept.
+ *
+ * Anchored at the start of the subject on purpose: "fix(duration): format times
+ * over 24 hours" is a real fix to formatting *code*, not a formatting commit.
+ */
+const COSMETIC_SUBJECT = /^(chore\(format\)|chore\(deps\)|style)(\([^)]*\))?\s*:/i;
+
+/**
+ * Does this commit subject describe a change with no requirement meaning?
+ *
+ * Exported for scripts/release-notes.test.ts.
+ *
+ * TRACES: | DR-219
+ */
+export function isCosmeticCommit(subject: string): boolean {
+  return COSMETIC_SUBJECT.test(subject.trim());
+}
+
+/**
+ * Files the range changed, excluding those touched only by cosmetic commits.
+ *
+ * Why not a plain `git diff --name-only <range>`: that is what this did, and a
+ * single repo-wide `prettier --write` inside the range made it report 199 files
+ * whose TRACES comments resolved to nearly the entire requirement matrix. The
+ * generated notes for v0.9.2 claimed the release had added the whole
+ * application — and build-release.yml publishes this output, so the noise would
+ * have shipped.
+ *
+ * Walking commit by commit and skipping the cosmetic ones keeps a file that a
+ * sweep *and* a real change both touched: it is still listed by the real
+ * commit. Only files touched exclusively by cosmetic commits drop out, which is
+ * exactly the intent.
+ *
+ * Merge commits produce no output from `git diff-tree` without `-m`, and are
+ * skipped deliberately: everything they merge is already in the range as its
+ * own commit, so including them would double-count.
+ */
 function changedFiles(range: string): string[] {
-  const cmd = range ? `git diff --name-only ${range}` : "git ls-files"; // untagged repo: describe everything currently traced
-  return sh(cmd)
-    .split("\n")
-    .filter((f) => f && existsSync(f));
+  // Untagged repo: describe everything currently traced.
+  if (!range) {
+    return sh("git ls-files")
+      .split("\n")
+      .filter((f) => f && existsSync(f));
+  }
+
+  // NUL between hash and subject so a subject containing anything at all is safe.
+  const log = sh(`git log --no-merges --format=%H%x00%s ${range}`);
+  if (!log) return [];
+
+  const files = new Set<string>();
+  let skipped = 0;
+
+  for (const line of log.split("\n")) {
+    const [sha, ...subjectParts] = line.split("\u0000");
+    const subject = subjectParts.join("\u0000");
+    if (!sha) continue;
+
+    if (isCosmeticCommit(subject)) {
+      skipped++;
+      continue;
+    }
+
+    for (const f of sh(`git diff-tree --no-commit-id --name-only -r ${sha}`).split("\n")) {
+      if (f && existsSync(f)) files.add(f);
+    }
+  }
+
+  if (skipped > 0) {
+    // Say what was dropped rather than silently reporting a smaller set.
+    console.error(
+      `ℹ️  Skipped ${skipped} cosmetic commit(s) (formatting/deps) when deriving notes.`,
+    );
+  }
+
+  return [...files];
 }

 /** Collect requirement IDs referenced by TRACES comments in the given files. */
@@ -132,4 +207,7 @@ function main() {
  console.log(out.join("\n"));
 }

-main();
+// Guarded so this module stays importable from release-notes.test.ts.
+if (import.meta.main) {
+  main();
+}
@@ -1,6 +1,6 @@
 [package]
 name = "jellytau"
-version = "0.9.1"
+version = "0.10.1"
 description = "A cross-platform Jellyfin client"
 authors = ["Duncan Tourolle <duncan@tourolle.paris>"]
 license = "MIT"
@@ -164,17 +164,33 @@ class MainActivity : TauriActivity() {
   */
  override fun onStop() {
    super.onStop()
-    if (backgroundAudioEnabled) {
-      dispatchWebEvent("jellytau-background")
+    // Fires unconditionally now. It used to be gated on backgroundAudioEnabled,
+    // which meant the frontend was never told the app had gone away unless the
+    // toggle was already on -- so with the toggle OFF nothing could react, and
+    // on the native path ExoPlayer's media service simply kept playing. That is
+    // the whole defect: the toggle appeared to do nothing because the only
+    // notification of backgrounding was itself gated on the toggle (DR-224).
+    //
+    // What to DO about it is decided in Rust (player_background_action); this
+    // only reports the two facts the activity alone knows.
+    val inPip = if (android.os.Build.VERSION.SDK_INT >= android.os.Build.VERSION_CODES.N) {
+      isInPictureInPictureMode
+    } else {
+      false
    }
+    dispatchWebEvent(
+      "jellytau-background",
+      "{\"backgroundAudioArmed\": $backgroundAudioEnabled, \"inPictureInPicture\": $inPip}"
+    )
  }

  /** The app is visible again — tell the frontend to resume WebView video. */
  override fun onStart() {
    super.onStart()
-    if (backgroundAudioEnabled) {
-      dispatchWebEvent("jellytau-foreground")
-    }
+    // Also unconditional, for the same reason: a video paused on background has
+    // to be told it is visible again, and that pause happens with the toggle
+    // OFF. The frontend ignores this when it has nothing to restore.
+    dispatchWebEvent("jellytau-foreground")
  }

  /**
@@ -182,14 +198,14 @@ class MainActivity : TauriActivity() {
   * evaluateJavascript pattern already used to unmute video elements. Posted to
   * the WebView thread; safe no-op if the WebView isn't found yet.
   */
-  private fun dispatchWebEvent(name: String) {
+  private fun dispatchWebEvent(name: String, detailJson: String = "null") {
    val webView = mediaWebView ?: run {
      android.util.Log.w("MainActivity", "dispatchWebEvent('$name'): no WebView")
      return
    }
    webView.post {
      webView.evaluateJavascript(
-        "window.dispatchEvent(new CustomEvent('$name'));",
+        "window.dispatchEvent(new CustomEvent('$name', { detail: $detailJson }));",
        null
      )
      android.util.Log.d("MainActivity", "Dispatched web event: $name")
@@ -0,0 +1,167 @@
+//! Publishes the Android JavaVM and application Context into `ndk_context`.
+//!
+//! TRACES: UR-012 | DR-223
+//!
+//! # Why this exists
+//!
+//! Seven places in this crate (five in `credentials.rs`, two in `lib.rs`) reach
+//! the JNI environment through [`ndk_context::android_context`], which reads a
+//! process-global pair of pointers: the `JavaVM` and a `Context` jobject.
+//!
+//! Nothing here ever set that global. `tao` did — the windowing layer beneath
+//! `wry`, several dependencies below anything this project names. tao 0.34.5
+//! called `ndk_context::initialize_android_context(...)` while starting the
+//! Android activity, and our code simply read what it had left behind.
+//!
+//! **tao 0.35.3 stopped.** It keeps the same two pointers in a private
+//! `AndroidContext` struct of its own and no longer publishes them. The moment
+//! that landed (via the Tauri 2.9.5 → 2.11.5 upgrade), the first credential
+//! read on Android aborted the process:
+//!
+//! ```text
+//! PANIC at ndk-context/src/lib.rs:72: android context was not initialized
+//!   8: ndk_context::android_context
+//!   9: jellytau_lib::run::{{closure}}
+//! ```
+//!
+//! Not a crash in our code, and not a change to our code: an undocumented side
+//! effect of a transitive dependency disappeared. The lesson worth keeping is
+//! that relying on *someone else* to populate a global is a dependency you
+//! cannot see in `Cargo.toml` and will not be told about when it breaks.
+//!
+//! # Why restore the global rather than rewrite the call sites
+//!
+//! Threading a VM and Context handle through seven call sites — including the
+//! credential path — is a larger and riskier change than owning the invariant
+//! those call sites already depend on. This module makes the assumption true
+//! instead of removing it, and the seven callers are untouched.
+//!
+//! # How
+//!
+//! `JNI_OnLoad` gives us the `JavaVM` the instant the shared library loads,
+//! which is the earliest and most reliable moment available — nothing in the app
+//! can run before it. It does *not* give us a Context, so that is resolved
+//! lazily on first use via `ActivityThread.currentApplication()`, by which point
+//! the Application object certainly exists.
+//!
+//! The Context published is the **Application**, not the Activity. That is what
+//! the consumers want anyway (`SecureStorage.initialize()` immediately calls
+//! `context.applicationContext`), and it cannot outlive its own lifetime the way
+//! a retained Activity reference would.
+
+use std::ffi::c_void;
+use std::sync::atomic::{AtomicPtr, Ordering};
+use std::sync::OnceLock;
+
+use jni::objects::GlobalRef;
+use jni::sys::{jint, JNI_VERSION_1_6};
+use jni::JavaVM;
+
+/// The `JavaVM`, captured at library load.
+static JAVA_VM: AtomicPtr<c_void> = AtomicPtr::new(std::ptr::null_mut());
+
+/// A global reference to the Application, kept alive for the process lifetime.
+///
+/// `ndk_context` stores a bare pointer and does not own the reference, so the
+/// `GlobalRef` must outlive every read. A local reference would be freed the
+/// moment the frame that created it returned, leaving a dangling jobject that
+/// only misbehaves later.
+static APP_CONTEXT: OnceLock<GlobalRef> = OnceLock::new();
+
+/// Whether the `ndk_context` global has been populated.
+static PUBLISHED: OnceLock<bool> = OnceLock::new();
+
+/// Called by the Android runtime when `libjellytau_lib.so` is loaded.
+///
+/// Verified that neither tao, wry nor tauri defines `JNI_OnLoad` in this
+/// library, so there is nothing to collide with. Returning the JNI version is
+/// mandatory — returning 0 makes `System.loadLibrary` fail.
+///
+/// TRACES: UR-012 | DR-223
+#[no_mangle]
+pub extern "system" fn JNI_OnLoad(vm: JavaVM, _reserved: *mut c_void) -> jint {
+    JAVA_VM.store(vm.get_java_vm_pointer().cast(), Ordering::SeqCst);
+    // Deliberately no logging here: the logger is not installed this early.
+    JNI_VERSION_1_6
+}
+
+/// Make [`ndk_context::android_context`] safe to call.
+///
+/// Idempotent and cheap after the first success. Returns an error rather than
+/// panicking: a failure here means credentials fall back to the encrypted-file
+/// path, which is a degraded mode the app already supports — far better than
+/// aborting the process, which is what the missing global did.
+///
+/// TRACES: UR-012 | DR-223
+pub fn ensure_initialized() -> Result<(), String> {
+    if PUBLISHED.get().is_some() {
+        return Ok(());
+    }
+
+    let vm_ptr = JAVA_VM.load(Ordering::SeqCst);
+    if vm_ptr.is_null() {
+        return Err(
+            "JNI_OnLoad has not run: no JavaVM captured. The library was loaded in an \
+             unexpected way, or JNI_OnLoad was stripped from the shared object."
+                .to_string(),
+        );
+    }
+
+    let vm = unsafe { JavaVM::from_raw(vm_ptr.cast()) }
+        .map_err(|e| format!("failed to adopt the JavaVM pointer: {e}"))?;
+
+    let mut env = vm
+        .attach_current_thread()
+        .map_err(|e| format!("failed to attach the current thread to the JVM: {e}"))?;
+
+    // ActivityThread.currentApplication() is the standard way to reach the
+    // Application from native code without being handed a Context. It is a
+    // hidden-but-stable API; it returns null only before the Application is
+    // constructed, which cannot be the case by the time anything here runs.
+    let activity_thread = env
+        .find_class("android/app/ActivityThread")
+        .map_err(|e| format!("android.app.ActivityThread not found: {e}"))?;
+
+    let application = env
+        .call_static_method(
+            activity_thread,
+            "currentApplication",
+            "()Landroid/app/Application;",
+            &[],
+        )
+        .map_err(|e| format!("ActivityThread.currentApplication() failed: {e}"))?
+        .l()
+        .map_err(|e| format!("currentApplication() did not return an object: {e}"))?;
+
+    if application.is_null() {
+        return Err(
+            "ActivityThread.currentApplication() returned null — the Application has not \
+             been created yet."
+                .to_string(),
+        );
+    }
+
+    let global = env
+        .new_global_ref(&application)
+        .map_err(|e| format!("failed to pin the Application as a global reference: {e}"))?;
+
+    // Store first, publish second: `ndk_context` will hold a bare pointer into
+    // this reference, so it must already be owned somewhere permanent.
+    let stored = APP_CONTEXT.get_or_init(|| global);
+    let context_ptr = stored.as_obj().as_raw().cast::<c_void>();
+
+    unsafe {
+        ndk_context::initialize_android_context(vm_ptr, context_ptr);
+    }
+    let _ = PUBLISHED.set(true);
+
+    log::info!("[INIT] Android JavaVM and Application published to ndk_context");
+    Ok(())
+}
+
+/// Whether the global has been published, for callers that want to degrade
+/// rather than attempt a JNI call.
+#[allow(dead_code)]
+pub fn is_initialized() -> bool {
+    PUBLISHED.get().is_some()
+}
@@ -840,6 +840,46 @@ pub async fn player_enter_background_audio(
 /// untouched — if it fired while backgrounded, playback is already stopped and
 /// this simply reports the last position.
 ///
+/// What playback should do now that the app is no longer visible.
+///
+/// The caller supplies only what it alone knows -- whether the per-player
+/// toggle is armed, and whether Android put the window into picture-in-picture.
+/// Everything else (what is playing, and therefore whether there is a picture to
+/// lose) is read here, because it is domain state.
+///
+/// The rule itself is in `player::background_policy`; this command is the wire.
+/// Returning `KeepPlaying` for an empty queue is deliberate: with nothing
+/// playing there is nothing to pause, and an error would make the frontend
+/// handle a case that is not a failure.
+///
+/// TRACES: UR-040, UR-041 | DR-224 | UT-211
+#[tauri::command]
+#[specta::specta]
+pub async fn player_background_action(
+    player: State<'_, PlayerStateWrapper>,
+    background_audio_armed: bool,
+    in_picture_in_picture: bool,
+) -> Result<crate::player::background_policy::BackgroundAction, String> {
+    use crate::player::background_policy::{background_action, is_video_media, BackgroundAction};
+
+    let is_video = {
+        let controller = player.0.lock().await;
+        let queue_arc = controller.queue();
+        let queue = queue_arc.lock().map_err(|e| e.to_string())?;
+        match queue.current() {
+            Some(item) => is_video_media(item.media_type),
+            None => return Ok(BackgroundAction::KeepPlaying),
+        }
+    };
+
+    let action = background_action(is_video, background_audio_armed, in_picture_in_picture);
+    info!(
+        "[player_background_action] video={} armed={} pip={} -> {:?}",
+        is_video, background_audio_armed, in_picture_in_picture, action
+    );
+    Ok(action)
+}
+
 /// TRACES: UR-040 | DR-052 | UT-061, IT-013
 #[tauri::command]
 #[specta::specta]
@@ -1,3 +1,5 @@
+#[cfg(target_os = "android")]
+mod android_context;
 mod auth;
 mod commands;
 mod connectivity;
@@ -119,6 +121,7 @@ use commands::{
    player_add_to_queue,
    player_add_track_by_id,
    player_add_tracks_by_ids,
+    player_background_action,
    player_cancel_autoplay_countdown,
    player_cancel_sleep_timer,
    // Jellyfin reporting commands
@@ -614,6 +617,14 @@ fn create_player_backend(
    {
        info!("Android platform detected - initializing ExoPlayer backend");

+        // Same precondition as the credential path: ndk_context must be
+        // populated before it is read, and nothing outside this crate populates
+        // it any more. Idempotent, so it does not matter which of the two runs
+        // first. TRACES: UR-012 | DR-223
+        if let Err(e) = crate::android_context::ensure_initialized() {
+            log::error!("[INIT] Android context unavailable for the player: {e}");
+        }
+
        // Get the Android context via ndk-context
        let ctx = ndk_context::android_context();

@@ -732,6 +743,7 @@ fn specta_builder() -> Builder<tauri::Wry> {
        .commands(tauri_specta::collect_commands![
            // Player commands
            player_play_item,
+            player_background_action,
            player_enter_background_audio,
            player_exit_background_audio,
            player_play_queue,
@@ -1270,6 +1282,21 @@ pub fn run() {
            // On Android, initialize SecureStorage BEFORE creating CredentialStore
            #[cfg(target_os = "android")]
            {
+                // Publish the JavaVM and Application into ndk_context first.
+                //
+                // Everything below reads that global. tao used to populate it
+                // and stopped doing so in 0.35 (Tauri 2.11), at which point the
+                // first read here aborted the process on launch. See
+                // android_context.rs. A failure is logged rather than fatal:
+                // credentials then fall back to the encrypted-file path, which
+                // is a supported degraded mode -- unlike aborting.
+                //
+                // TRACES: UR-012 | DR-223
+                if let Err(e) = crate::android_context::ensure_initialized() {
+                    log::error!("[INIT] Android context unavailable: {e}");
+                    log::error!("[INIT] Secure credential storage will fall back to the encrypted file.");
+                }
+
                info!("[INIT] Initializing Android SecureStorage for credentials...");
                let ctx = ndk_context::android_context();
                let vm = unsafe { jni::JavaVM::from_raw(ctx.vm().cast()) };
@@ -0,0 +1,155 @@
+//! What playback should do when the app stops being visible.
+//!
+//! TRACES: UR-040 | DR-224 | UT-211
+//!
+//! # The defect this exists for
+//!
+//! The per-player background-audio toggle (UR-040) was built for the WebView
+//! `<video>` path, where backgrounding the app kills the decode: the toggle
+//! decided whether to *hand off* to a native audio stream or let playback die.
+//!
+//! Native video then became the default renderer (DR-188). On that path playback
+//! runs through ExoPlayer inside a `MediaSessionService` — a foreground media
+//! service whose entire purpose is to keep playing when the app is not visible.
+//! Nothing stops it, and nothing in the codebase paused playback on background.
+//!
+//! So locking the screen kept the audio playing **whether or not the toggle was
+//! on**. The toggle governed a handoff that no longer had anything to hand off
+//! *from*: there was no gap in playback to bridge. A user who had never touched
+//! it got background audio anyway, which is the bug as reported.
+//!
+//! # Why the decision lives in Rust
+//!
+//! It depends on what the item *is* (a video keeps its picture; music has none
+//! to lose) and on a user setting — domain questions, not presentation ones, and
+//! the answer must be identical for both renderers. The frontend and the Android
+//! activity carry it out; neither decides it. Putting the rule in either would
+//! have reproduced exactly the split that caused this: two renderers, two
+//! behaviours, one toggle that only reached one of them.
+
+use crate::player::media::MediaType;
+
+/// What the player should do when the app is backgrounded.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize, serde::Deserialize, specta::Type)]
+#[serde(rename_all = "camelCase")]
+pub enum BackgroundAction {
+    /// Carry on. Music, and video the user explicitly asked to keep hearing
+    /// while it is in a picture-in-picture window.
+    KeepPlaying,
+    /// Swap the video stream for an audio-only one and keep playing.
+    HandOffToAudio,
+    /// Stop making sound. The user did not ask for background playback.
+    Pause,
+}
+
+/// Decide what backgrounding should do.
+///
+/// * `is_video` — whether the current item has a picture to lose. Music is
+///   never paused by backgrounding; that is what a music player is for.
+/// * `background_audio_armed` — the per-player toggle (UR-040).
+/// * `in_picture_in_picture` — the app is not "gone", it is in a floating
+///   window and still visible. Pausing here would break PiP (UR-041).
+///
+/// TRACES: UR-040, UR-041 | DR-224 | UT-211
+pub fn background_action(
+    is_video: bool,
+    background_audio_armed: bool,
+    in_picture_in_picture: bool,
+) -> BackgroundAction {
+    // PiP first: the window is still on screen, so this is not backgrounding in
+    // any sense the user would recognise.
+    if in_picture_in_picture {
+        return BackgroundAction::KeepPlaying;
+    }
+
+    // Music has no picture to lose; a music player that stopped when the screen
+    // locked would be broken in an obvious way.
+    if !is_video {
+        return BackgroundAction::KeepPlaying;
+    }
+
+    if background_audio_armed {
+        BackgroundAction::HandOffToAudio
+    } else {
+        BackgroundAction::Pause
+    }
+}
+
+/// Whether a media type has a picture that backgrounding would throw away.
+///
+/// TRACES: UR-040 | DR-224
+pub fn is_video_media(media_type: MediaType) -> bool {
+    matches!(media_type, MediaType::Video)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn video_without_the_toggle_pauses() {
+        // THE REPORTED BUG. Native video runs in a foreground media service that
+        // keeps playing when the app is hidden, and nothing paused it -- so
+        // locking the screen gave background audio to a user who never asked
+        // for it.
+        assert_eq!(
+            background_action(true, false, false),
+            BackgroundAction::Pause
+        );
+    }
+
+    #[test]
+    fn video_with_the_toggle_hands_off_to_audio() {
+        assert_eq!(
+            background_action(true, true, false),
+            BackgroundAction::HandOffToAudio
+        );
+    }
+
+    #[test]
+    fn music_always_keeps_playing() {
+        // Backgrounding a music player and having it stop would be absurd. The
+        // toggle is irrelevant here: there is no picture to give up.
+        assert_eq!(
+            background_action(false, false, false),
+            BackgroundAction::KeepPlaying
+        );
+        assert_eq!(
+            background_action(false, true, false),
+            BackgroundAction::KeepPlaying
+        );
+    }
+
+    #[test]
+    fn picture_in_picture_is_not_backgrounding() {
+        // The video is in a floating window and still on screen. Pausing would
+        // break PiP (UR-041), which is a separate feature reached through
+        // onUserLeaveHint rather than onStop.
+        assert_eq!(
+            background_action(true, false, true),
+            BackgroundAction::KeepPlaying
+        );
+        assert_eq!(
+            background_action(true, true, true),
+            BackgroundAction::KeepPlaying
+        );
+    }
+
+    #[test]
+    fn the_rule_does_not_depend_on_the_renderer() {
+        // There is deliberately no renderer parameter. The WebView path and the
+        // native path must answer identically -- the split between them is what
+        // produced the defect, because the toggle only ever reached one.
+        for armed in [true, false] {
+            let once = background_action(true, armed, false);
+            let again = background_action(true, armed, false);
+            assert_eq!(once, again);
+        }
+    }
+
+    #[test]
+    fn only_video_counts_as_having_a_picture() {
+        assert!(is_video_media(MediaType::Video));
+        assert!(!is_video_media(MediaType::Audio));
+    }
+}
@@ -4,6 +4,7 @@
 //         DR-001, DR-004, DR-005, DR-009, DR-028, DR-029, DR-047
 pub mod autoplay;
 pub mod backend;
+pub mod background_policy;
 pub mod events;
 pub mod media;
 pub mod queue;
@@ -1,7 +1,7 @@
 {
  "$schema": "https://schema.tauri.app/config/2",
  "productName": "JellyTau",
-  "version": "0.9.1",
+  "version": "0.10.1",
  "identifier": "com.dtourolle.jellytau",
  "build": {
    "beforeDevCommand": "bun run dev",
@@ -44,6 +44,7 @@
  },
  "bundle": {
    "active": true,
+    "createUpdaterArtifacts": true,
    "targets": [
      "deb",
      "rpm",
@@ -19,6 +19,31 @@ export const commands = {
 async playerPlayItem(item: PlayItemRequest) : Promise<PlayerStatus> {
    return await TAURI_INVOKE("player_play_item", { item });
 },
+/**
+ * Exit background-audio mode: stop the native audio player and return its final
+ * position so the frontend can reload the WebView `<video>` there (UR-040).
+ * 
+ * Returns the position in seconds. The sleep timer is intentionally left
+ * untouched — if it fired while backgrounded, playback is already stopped and
+ * this simply reports the last position.
+ * 
+ * What playback should do now that the app is no longer visible.
+ * 
+ * The caller supplies only what it alone knows -- whether the per-player
+ * toggle is armed, and whether Android put the window into picture-in-picture.
+ * Everything else (what is playing, and therefore whether there is a picture to
+ * lose) is read here, because it is domain state.
+ * 
+ * The rule itself is in `player::background_policy`; this command is the wire.
+ * Returning `KeepPlaying` for an empty queue is deliberate: with nothing
+ * playing there is nothing to pause, and an error would make the frontend
+ * handle a case that is not a failure.
+ * 
+ * TRACES: UR-040, UR-041 | DR-224 | UT-211
+ */
+async playerBackgroundAction(backgroundAudioArmed: boolean, inPictureInPicture: boolean) : Promise<BackgroundAction> {
+    return await TAURI_INVOKE("player_background_action", { backgroundAudioArmed, inPictureInPicture });
+},
 /**
 * Enter background-audio mode: hand playback of the currently-watched video off
 * to the native ExoPlayer *audio* path so the audio keeps playing while the app
@@ -41,13 +66,6 @@ async playerEnterBackgroundAudio(item: PlayItemRequest, positionSeconds: number)
    return await TAURI_INVOKE("player_enter_background_audio", { item, positionSeconds });
 },
 /**
- * Exit background-audio mode: stop the native audio player and return its final
- * position so the frontend can reload the WebView `<video>` there (UR-040).
- * 
- * Returns the position in seconds. The sleep timer is intentionally left
- * untouched — if it fired while backgrounded, playback is already stopped and
- * this simply reports the last position.
- * 
 * TRACES: UR-040 | DR-052 | UT-061, IT-013
 */
 async playerExitBackgroundAudio() : Promise<number> {
@@ -1976,6 +1994,23 @@ countdownSeconds: number;
 * Maximum number of episodes to auto-play consecutively (0 = unlimited)
 */
 maxEpisodes?: number }
+/**
+ * What the player should do when the app is backgrounded.
+ */
+export type BackgroundAction = 
+/**
+ * Carry on. Music, and video the user explicitly asked to keep hearing
+ * while it is in a picture-in-picture window.
+ */
+"keepPlaying" | 
+/**
+ * Swap the video stream for an audio-only one and keep playing.
+ */
+"handOffToAudio" | 
+/**
+ * Stop making sound. The user did not ask for background playback.
+ */
+"pause"
 /**
 * Smart caching configuration
 */
@@ -4,7 +4,7 @@
  import { get } from "svelte/store";
  import { goto } from "$app/navigation";
  import { commands } from "$lib/api/bindings";
-  import type { JRayActor, StreamingQuality } from "$lib/api/bindings";
+  import type { JRayActor, StreamingQuality, BackgroundAction } from "$lib/api/bindings";
  import { listen } from "@tauri-apps/api/event";
  import Hls from "hls.js";
  import type { MediaItem } from "$lib/api/types";
@@ -63,6 +63,7 @@
  import {
    setBackgroundAudioEnabled,
    subscribeAppBackgrounded,
+    type BackgroundSignal,
    subscribeAppForegrounded,
  } from "$lib/utils/backgroundAudio";
  import { platform } from "@tauri-apps/plugin-os";
@@ -825,7 +826,7 @@
    // flip the component into HTML5 mode). Unsubscribers go into nativeUnlisteners
    // so onDestroy tears them down.
    if (backgroundAudioSupported) {
-      nativeUnlisteners.push(subscribeAppBackgrounded(enterBackgroundAudioHandoff));
+      nativeUnlisteners.push(subscribeAppBackgrounded(onAppBackgrounded));
      nativeUnlisteners.push(subscribeAppForegrounded(exitBackgroundAudioHandoff));
    }

@@ -1722,6 +1723,9 @@
  const backgroundAudioSupported = platform() === "android";
  let backgroundAudioOn = $state(false); // v1: default OFF each session
  let handoffState: BackgroundAudioState = { ...initialHandoffState };
+  // Set when backgrounding paused playback, so foregrounding resumes only
+  // what we stopped -- never something the user paused themselves.
+  let pausedByBackgrounding = false;

  function toggleBackgroundAudio() {
    backgroundAudioOn = !backgroundAudioOn;
@@ -1737,6 +1741,43 @@

  // App went to background/locked while background-audio is armed: hand off to
  // native audio and stop the WebView video decode.
+  async function onAppBackgrounded(signal: BackgroundSignal) {
+    // What to do is a domain decision, not a presentation one: it depends on
+    // whether the item has a picture to lose, which is Rust's to know. This used
+    // to be decided implicitly by Kotlin gating the event on the toggle, which
+    // is why the native path -- whose media service keeps playing regardless --
+    // ignored the toggle entirely (DR-224).
+    let action: BackgroundAction;
+    try {
+      action = await commands.playerBackgroundAction(
+        signal.backgroundAudioArmed,
+        signal.inPictureInPicture,
+      );
+    } catch (e) {
+      // Never leave playback in an undefined state because a decision call
+      // failed. Continuing is the old behaviour and the safer default: it
+      // cannot silently stop something the user is listening to.
+      log.warn("Background action lookup failed; leaving playback alone:", e);
+      return;
+    }
+
+    log.debug("Background action:", action);
+    switch (action) {
+      case "keepPlaying":
+        return;
+      case "pause":
+        // The user did not ask for background playback. Remember that WE paused
+        // it, so returning to the foreground can resume rather than leaving a
+        // video mysteriously stopped.
+        pausedByBackgrounding = isPlaying;
+        if (isPlaying) await playerController.pause();
+        return;
+      case "handOffToAudio":
+        await enterBackgroundAudioHandoff();
+        return;
+    }
+  }
+
  async function enterBackgroundAudioHandoff() {
    if (!shouldEnterBackgroundAudio(backgroundAudioOn, handoffState)) return;
    // `currentTime` is the component's authoritative ABSOLUTE position (the RAF
@@ -1797,6 +1838,20 @@
  // App returned to foreground: stop native audio, reload the WebView <video> at
  // the position native reached, and restore play/pause.
  async function exitBackgroundAudioHandoff() {
+    // Resume what backgrounding paused, before the handoff check: the pause path
+    // and the handoff path are mutually exclusive, and this one leaves no
+    // handoff state to unwind. Only resumes if WE paused it -- a user who
+    // paused before locking the screen stays paused.
+    if (pausedByBackgrounding) {
+      pausedByBackgrounding = false;
+      try {
+        await playerController.play();
+      } catch (e) {
+        log.warn("Failed to resume after backgrounding:", e);
+      }
+      return;
+    }
+
    if (!shouldExitBackgroundAudio(handoffState)) return;
    // Read the native player's state BEFORE exiting — the exit stops it. If the
    // user hit pause on the lockscreen while backgrounded, that pause must
@@ -66,10 +66,31 @@ export function setBackgroundAudioEnabled(enabled: boolean): boolean {
 * Returns an unsubscribe function. No-op where unsupported (the event never
 * fires on non-Android platforms).
 */
-export function subscribeAppBackgrounded(handler: () => void): () => void {
+export interface BackgroundSignal {
+  /** Whether the per-player background-audio toggle was armed (UR-040). */
+  backgroundAudioArmed: boolean;
+  /** Whether Android put the window into picture-in-picture (UR-041). */
+  inPictureInPicture: boolean;
+}
+
+/**
+ * Older builds dispatched this event with no detail, and only when the toggle
+ * was already armed. Treat a missing detail as "armed, not PiP" so a mismatched
+ * pair degrades to the previous behaviour rather than pausing unexpectedly.
+ */
+function readSignal(event: Event): BackgroundSignal {
+  const detail = (event as CustomEvent).detail as Partial<BackgroundSignal> | null | undefined;
+  return {
+    backgroundAudioArmed: detail?.backgroundAudioArmed ?? true,
+    inPictureInPicture: detail?.inPictureInPicture ?? false,
+  };
+}
+
+export function subscribeAppBackgrounded(handler: (signal: BackgroundSignal) => void): () => void {
  if (typeof window === "undefined") return () => {};
-  window.addEventListener("jellytau-background", handler);
-  return () => window.removeEventListener("jellytau-background", handler);
+  const listener = (event: Event) => handler(readSignal(event));
+  window.addEventListener("jellytau-background", listener);
+  return () => window.removeEventListener("jellytau-background", listener);
 }

 /** Subscribe to the native "app foregrounded" signal. Returns an unsubscribe fn. */