GRI-2014: render a real slideshow instead of a blank one #192

Merged
vivy merged 1 commit from fix/GRI-2014-pptx-slideshow-render into dev 2026-11-14 13:19:11 +00:00
Member

Summary

  • soffice --show left the Impress slideshow permanently blank on this
    host (no /dev/dri): the presentation window never got its first paint.
    Loading the document normally, waiting for its window, then sending a
    synthetic F5 over XTEST (with OpenGL/Skia disabled in the profile)
    renders real content instead.
  • New crates/server/src/bin/pptx_video_presenter.rs (pure-Rust x11rb,
    no new host package) drives that keypress and tracks the slideshow via
    the EWMH _NET_WM_STATE_FULLSCREEN hint, since soffice no longer
    self-closes on its own when a show ends now that nothing passes --show.
    render_to_mp4 shells out to it the same way it shells out to
    soffice/ffmpeg/Xvfb/gs — new PptxVideoConfig::presenter_path
    (GRI-1884).
  • Existing unit and acceptance tests updated to fake the presenter's exit
    code (0 = finished, 2 = still running at the timeout) the same way they
    already fake every other tool; none of this needs a real X server to run
    in CI.

Full root-cause writeup and a fidelity report (3 real decks, including one
with a transition and one with both a transition and an entrance
animation, run through the real pipeline on this host) are on forge #189.
That comment also flags two things not fixed here: the Docker image has
none of these binaries (pre-existing, from #190), and a slideshow does not
end on its own even when fully auto-timed, so render_to_mp4's existing
"discard on timeout" contract means most real uploads still fall back to
slide images in practice — a product/acceptance-criteria decision, not a
bug this PR introduces.

Test plan

  • cargo fmt --all --check
  • cargo clippy --workspace --all-targets --locked -- -D warnings
  • cargo test -p server --lib pptx_video (unit tests, fake tools)
  • cargo test -p server --test pptx_video_acceptance_integration
    (SIGNAGE_TEST_DATABASE_URL set)
  • cargo deny check bans licenses sources
  • Manually verified on this host with real soffice/Xvfb/ffmpeg:
    plain deck, a deck with an auto-advancing fade transition, and a
    deck with both a transition and a fade entrance animation — all
    render real, correct content (see forge #189 for frames/detail).
## Summary - `soffice --show` left the Impress slideshow permanently blank on this host (no `/dev/dri`): the presentation window never got its first paint. Loading the document normally, waiting for its window, then sending a synthetic F5 over XTEST (with OpenGL/Skia disabled in the profile) renders real content instead. - New `crates/server/src/bin/pptx_video_presenter.rs` (pure-Rust `x11rb`, no new host package) drives that keypress and tracks the slideshow via the EWMH `_NET_WM_STATE_FULLSCREEN` hint, since `soffice` no longer self-closes on its own when a show ends now that nothing passes `--show`. `render_to_mp4` shells out to it the same way it shells out to `soffice`/`ffmpeg`/`Xvfb`/`gs` — new `PptxVideoConfig::presenter_path` (GRI-1884). - Existing unit and acceptance tests updated to fake the presenter's exit code (0 = finished, 2 = still running at the timeout) the same way they already fake every other tool; none of this needs a real X server to run in CI. Full root-cause writeup and a fidelity report (3 real decks, including one with a transition and one with both a transition and an entrance animation, run through the real pipeline on this host) are on forge #189. That comment also flags two things not fixed here: the Docker image has none of these binaries (pre-existing, from #190), and a slideshow does not end on its own even when fully auto-timed, so `render_to_mp4`'s existing "discard on timeout" contract means most real uploads still fall back to slide images in practice — a product/acceptance-criteria decision, not a bug this PR introduces. ## Test plan - [x] `cargo fmt --all --check` - [x] `cargo clippy --workspace --all-targets --locked -- -D warnings` - [x] `cargo test -p server --lib pptx_video` (unit tests, fake tools) - [x] `cargo test -p server --test pptx_video_acceptance_integration` (`SIGNAGE_TEST_DATABASE_URL` set) - [x] `cargo deny check bans licenses sources` - [x] Manually verified on this host with real `soffice`/`Xvfb`/`ffmpeg`: plain deck, a deck with an auto-advancing fade transition, and a deck with both a transition and a fade entrance animation — all render real, correct content (see forge #189 for frames/detail).
GRI-2014: render a real slideshow instead of a blank one
Some checks failed
CI / the pre-push hook (pull_request) Successful in 1m5s
CI / cargo fmt (pull_request) Successful in 1m9s
CI / cargo clippy (pull_request) Successful in 2m32s
CI / cargo deny (pull_request) Successful in 1m36s
CI / cargo deny advisories (pull_request) Failing after 1m56s
CI / repository boundary (pull_request) Successful in 1m17s
CI / frontend-web (pull_request) Successful in 2m18s
CI / cargo test (pull_request) Successful in 5m19s
CI / Notify gate failed (pull_request) Has been skipped
CI / CI (pull_request) Successful in 18s
8ea3b8c251
soffice --show left the slideshow permanently blank under this host's
Xvfb (no /dev/dri access): every real deck fell back to slide images,
with no way to tell a failing render from a working one. Loading the
document normally and sending a synthetic F5 over XTEST once it has
loaded, with OpenGL disabled, renders real content. Since soffice no
longer exits on its own when a show ends, a new pptx_video_presenter
binary drives the keypress and tracks the EWMH fullscreen hint to know
when the show starts and ends, in place of waiting on the process.
vivy merged commit 2a91fc3d59 into dev 2026-11-14 13:19:11 +00:00
vivy deleted branch fix/GRI-2014-pptx-slideshow-render 2026-11-14 13:19:12 +00:00
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
Grimoirix/grimoire-of-digital-signage!192
No description provided.