screentinker/server
ScreenTinker e606cc83d1 Screenshots: prove pixels arrived instead of assuming the draw worked
A BrightSign emitted BLANK screenshots and logged "Screenshot sent". With hwz
enabled the video decodes onto a hardware plane outside the browser compositor
— BrightSign's docs say the HTML/JS layer "doesn't see the pixels" — so
drawImage(video) produces a fully TRANSPARENT image and throws nothing.
Chromium 87, which this XT245 reports, fails the same way.

Both capture paths set captured/drawn = true purely because drawMediaFit() had
not thrown. So the dashboard showed a dead screen while the panel played
perfectly, and the zone path painted a black rectangle in place of the labelled
placeholder drawZonePlaceholder() exists to guarantee ("never a transparent
hole"). Success reported, nothing done.

isMediaReadable() does not catch this. It answers "am I ALLOWED to read this"
(same-origin / CORS), which is a different question from "did any pixels
arrive".

videoFrameIsCapturable() probes a 16x16 scratch canvas before committing to a
full-size draw. ALPHA is the discriminator, not colour: a scratch canvas starts
transparent and a real decoded frame writes alpha=255 even when the frame is
pure black, so a legitimate fade-to-black still reads as captured while
"nothing arrived" does not. A tainted canvas counts as captured, because
tainting only happens once cross-origin pixels have actually been drawn.

Probing BEFORE the draw matters twice: it avoids a wasted full-size drawImage on
every frame of a 1fps stream, and in the zone path it stops a black rectangle
being painted underneath the placeholder.

When a video is on screen but unreadable the status card now says so, because
that card is also what shows for "no content" — without the line an operator
would reasonably conclude the screen was blank.

Not gated on BrightSign: the same silent failure exists for any stalled decoder
or engine that declines to hand back frames.

10 tests, 964 pass.
2026-08-05 10:01:03 -05:00
..
config fix(logging): gate CF-Connecting-IP on a Cloudflare peer, not any trusted proxy 2026-07-26 10:23:04 -05:00
db Keep a workspace on schedules that outlive their device group 2026-07-30 20:53:21 -05:00
lib Finish the BrightSign port: native sync, offline fallback, multicast guard 2026-08-04 21:43:42 -05:00
middleware Merge branch 'fix/recovery-grants' into release/auth-campaign 2026-07-26 16:22:59 -05:00
player Screenshots: prove pixels arrived instead of assuming the draw worked 2026-08-05 10:01:03 -05:00
routes Make a content-only schedule actually put that content on the screen 2026-07-30 22:26:58 -05:00
scripts fix(content+android): rotation-aware media — portrait upright on dashboard AND player (#170) (#172) 2026-07-12 22:05:11 -05:00
services Make a recurring schedule respect its start and end dates 2026-07-30 22:22:59 -05:00
test Screenshots: prove pixels arrived instead of assuming the draw worked 2026-08-05 10:01:03 -05:00
ws Stamp the authenticated device on relayed playback progress 2026-07-30 22:41:06 -05:00
.gitignore feat(email): Microsoft Graph send + alert spam protection + preferences UI 2026-05-12 18:16:40 -05:00
config.js Add an operator override for self-update on MDM-managed panels 2026-07-28 23:07:30 -05:00
package-lock.json chore(release): v1.9.29-rc2 2026-08-05 00:06:20 -05:00
package.json chore(release): v1.9.29-rc2 2026-08-05 00:06:20 -05:00
server.js Finish the BrightSign port: native sync, offline fallback, multicast guard 2026-08-04 21:43:42 -05:00
smoke-ui.js Keep the smoke test out of npm test, and update the lockfile 2026-07-28 20:34:34 -05:00
version.js chore(version): single-source VERSION, env-configurable data paths, bump tooling 2026-06-10 12:56:03 -05:00