The two pixel_sim cards paint a black canvas in both galleries

[bdbarnett]

Anyone who opens dino or pixel_sim_demos from either gallery gets a black
rectangle and no way to tell it apart from a page that failed — both cards
report themselves healthy while drawing nothing. They are the only two cards
that go through utils/pixel_sim, and they are blank on both runtimes, so
this is the simulated LED matrix, not the Pyodide or WebAssembly host.

Measured on the published site, 2026-09-22, headless Chromium from
pydevices-examples/.venv, 30 s after network-idle:

card	page	data-runtime-state	#display_canvas
dino	pyscript/pyodide.html?modules=dino&deps=pydevices-pygraphics	ready	480x320, non-black pixels 0
dino	gallery/micropython.html?modules=dino	ready	480x320, non-black pixels 0
pixel_sim_demos	pyscript/pyodide.html?modules=pixel_sim_demos&deps=pydevices-palettes,pydevices-pygraphics	ready	480x320, non-black pixels 0
pixel_sim_demos	gallery/micropython.html?modules=pixel_sim_demos	ready	480x320, non-black pixels 0

No page error, no failed request, no exception in the on-page console. The log
ends exactly where success would end:

Initializing PSDisplay...
PSDisplay: initialized.
PSDisplay.rotation():  Setting rotation to 90
done setting rotation
Initializing SimPixelDisplay...
SimPixelDisplay: initialized.

So the rotation to 480x320 happens, SimPixelDisplay constructs, and then
nothing reaches the surface. dino ends on app.every(_tick, period=FRAME_MS, async_=app.timer_async) followed by one game.draw(); worth checking whether
that first draw reaches the canvas at all and whether the periodic tick ever
fires under either host, before looking at the game logic.

The other 54 Pyodide cards paint (#134), so nothing about the shared loader
path is implicated.

[bdbarnett]

Fix is in #138, awaiting merge; CI is green. Cause: 15f166a dropped display_drv.show() from both examples, but a PixelDisplay has no needs_refresh, so appdev.App never flushes the LED grid to the host canvas.

Ran .venv/bin/python pixcheck.py http://127.0.0.1:<port>/gallery/micropython.html?modules=<card> (headless Chromium, local tree with the deployed /vendor and /assets): non-black pixels 0 on main, 3325 for dino and 25600 for pixel_sim_demos on the branch. Both animate, and Space starts dino (score climbs, cacti scroll). The Pyodide card reads examples from main, so that half can only be confirmed after the merge.

[bdbarnett]

Fix merged in #138: the show() call that flushes the LED grid to the canvas (removed in 15f166a on 09-09) is back in both examples. The MicroPython gallery, served locally in headless Chromium: 0 non-black pixels on main before the fix; 3325 (dino) and 25600 (pixel_sim_demos) with it, both animating. Leaving this open until the deployed galleries, the Pyodide one included, are checked after the site rebuilds.

[bdbarnett]

Verified on the deployed site after #138, using each gallery's own card links in headless Chromium and counting non-black canvas pixels after 25 s and 27 s. MicroPython gallery: pixel_sim_demos 128475→132400, dino 131100→131325. Pyodide gallery: pixel_sim_demos 132425→130000, dino 131325, holding still until Space as expected. Both galleries paint, and pixel_sim animates in both.