Skip to content

appimage: Ship the browser source (CEF) in the AppImage #1865

Open

Description

@zackees

Problem. The AppImage builds with -DENABLE_BROWSER=OFF (build-aux/appimage/README.md, "Left out"). Users who move a scene collection from upstream OBS or the Flatpak lose every Browser Source and get no explanation (#1720). The only workaround is a window capture of a real browser.

Blocked by #1864 (CEF upgrade) and the Linux sandbox part of #1092. We must not ship an EOL Chromium with the sandbox off in a new package.

Work.

  1. Bundle libcef.so, resources, locales and obs-browser-page in the AppDir, then check them against hostlib's host-vs-bundled rules (GL/GBM/DRM must come from the host, appimage: Bundled GL/GBM/DRM/VA/Wayland libraries override the host driver stack #625).
  2. Sandbox inside an AppImage: no setuid chrome-sandbox is possible from a FUSE mount, so use the user-namespace sandbox. Where a host blocks it (Ubuntu 24.04 apparmor_restrict_unprivileged_userns, see bug: browser: CEF 7871 (Chromium 150.0.7871.252, end of life, two in-the-wild V8 CVEs) runs with no_sandbox on every platform #1092), apply bug: browser: CEF 7871 (Chromium 150.0.7871.252, end of life, two in-the-wild V8 CVEs) runs with no_sandbox on every platform #1092's fallback design, with a log line and a UI notice.
  3. Until this ships, make the gap visible. On load, a browser source whose plugin is missing shows "Browser sources aren't included in this AppImage build yet" in the source's properties or the missing-sources flow, so users don't lose their overlays silently or file a bug about them.

Testing. Extend build-aux/appimage/test-appimage.sh (required for packaging changes): a saved profile with a browser source pointing at a local file:// page, checking that the source registers and obs-browser-page starts from the AppImage. It runs only where AppImage tests already run, so it adds no PR CI time.

Acceptance. Browser sources from an imported scene collection render in the AppImage, the sandbox is on or the documented fallback is used, test-appimage.sh covers it, and README and features.md are updated. Item 3 can land first, on its own.

Refs #1720, #1092, #1864.

Activity

  1. zackees commented on Oct 23, 2026

    @zackees
    MemberAuthor

    Claiming: #1865 item 3 only (the notice). Items 1-2, shipping obs-browser in the AppImage, stay blocked on #1864 and the #1092 Linux sandbox work.

    1. RED: a Qt-free frontend/utility/missing-source-types helper (unique missing type ids with counts, plus a hint per type), stubbed, with a small test_missing_source_types cmocka binary (no libobs, milliseconds).
    2. GREEN: after LoadData() loads the sources, enumerate them; any type obs_source_get_display_name(id) doesn't know is listed in one "Missing Source Types" warning. It names browser sources ("this build does not include browser sources yet", with the Window Capture workaround), VLC, and any other plugin type, and writes one log line per type.

    Decisions: safe mode only logs (it leaves plugins out on purpose); the warning shows on each collection load, like Missing Files; the sources and their settings are kept untouched.

  2. added
    claimedAn agent or person has claimed this; check the claim comment before starting
    on Oct 23, 2026
  3. zackees commented on Oct 23, 2026

    @zackees
    MemberAuthor

    Item 3 merged: #1883 (0deb926). A scene collection with sources from a plugin this build lacks now gets a "Missing Source Types" warning (browser sources: not in this build yet, with the Window Capture workaround; VLC; any other plugin type), plus a log line per type. Items 1-2 (obs-browser in the AppImage) stay open, blocked on #1864 (CEF upgrade, decision pending there) and the #1092 Linux sandbox.

  4. moved this from In progress to Triage in obs-rust Multiviewon Oct 25, 2026
  5. removed
    claimedAn agent or person has claimed this; check the claim comment before starting
    on Oct 25, 2026
  6. zackees commented on Oct 25, 2026

    @zackees
    MemberAuthor

    Triage: not Ready yet. It's blocked by #1092's Linux sandbox part (the issue's own rule: no EOL Chromium and no sandbox-off in a new package). The CEF side for Linux x86_64 is done (#1948), and item 3 merged (#1883, the Missing Source Types warning). Once #1092's Linux sandbox work lands (obs-browser#3 is merged; the obs-studio side and the fallback warning #2238 are queued), items 1-2 are ready, with acceptance through test-appimage.sh loading a saved profile with a browser source. P2, size:M.

  7. zackees commented on Oct 26, 2026

    @zackees
    MemberAuthor

    Triage: still blocked by #1864 (CEF upgrade) and #1092 (sandbox), both open, so this isn't /ready as a whole. Item 3 (the "Browser sources aren't included in this AppImage build yet" notice) has no blocker and can land first, as the issue says. A worker can file it as its own sub-issue.

  8. zackees commented on Oct 26, 2026

    @zackees
    MemberAuthor

    Item 3 is already on master, from #1883 (0deb9260e). After a scene collection loads, OBSBasic::LoadData() lists the source types no loaded module provides (frontend/utility/missing-source-types.h). For browser_source it shows the notice that browser sources aren't in this build. No sub-issue is needed. What's left here is items 1 and 2, which are still blocked by #1864 and #1092.

  9. zackees commented on Oct 27, 2026

    @zackees
    MemberAuthor

    Coordinator triage (2026-10-27). Still waiting on its own declared blockers:

    So items 1-2 cannot be Ready yet: bundling an EOL Chromium with the sandbox off is the one thing the issue itself forbids. Both links are still current, so nothing needs rewriting in the body.

    One slice here is independent of both blockers and is Ready on its own: item 3, "until this ships, make the gap visible" (a browser source whose plugin is missing shows "Browser sources aren't included in this AppImage build yet"). Acceptance is testable today - a saved profile with a browser source against an AppImage build, checking the message instead of a silent drop - so it can be split out and claimed without waiting for CEF. Say the word (or just split it) and I will file it as a sub-issue of this one with item 3's text and acceptance.

    This stays out of ready until #1864 and #1092 land; it will leave the triage list then.

  10. zackees commented on Oct 27, 2026

    @zackees
    MemberAuthor

    Coordinator (2026-10-27): I linked the two blockers this issue already declares in its body as GitHub issue dependencies, so it leaves the triage list and comes back on its own when they close, instead of being re-triaged every tick.

    issueDependenciesSummary.blockedBy is 2 for this issue, so uv run -m workflow.coordinator no longer lists it under triage — the list hides any issue with an open blocker and returns it when the last one closes. Nothing else about the work changed: items 1-2 stay blocked on those two, and item 3 (the "browser sources aren't in this build" notice) is already on master from #1883.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    P2Normal prioritysize:MAbout a day

    Type

    No type

    Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions