mirror of
https://github.com/Klipper4FlashForge/firmware.git
synced 2026-10-21 21:02:16 -05:00
MonstrofilandClaude Opus 5
71dde065b9
* The sysfs-geometry tests and comments never caught up to the hardcode The splash-screen fix (6e43e4c,22525f7) made geometry default to the known Creator5Pro panel instead of a sysfs probe -- deliberate: sysfs reports the wrong geometry on real hardware, measured, not a test artefact. It left two things behind it. test_ffscreen.py's whole "-- geometry --" section, plus every other test that goes through the panel() fixture, asserts the sysfs-driven contract that no longer exists -- panel() never passes geometry explicitly, so every one of those tests now exercises the hardcoded default regardless of what fake sysfs data it set up, and 13 of them failed outright. Skipped rather than rewritten or deleted: the assertions are still true of the code path they test, that code path is just no longer the default one, and deleting the tests would make a future revival of sysfs-driven geometry (once the report is understood) invisible again. ffscreen.py's own header still argued FOR reading sysfs at runtime ("the two models do not have to agree on it") and for treating an explicit geometry as the exceptional case -- exactly backwards from what the code now does. Corrected to match. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Skip was disallowed where it mattered most: the printer-sim job runs REQUIRE_PRINTER_SIM=1 with no ALLOW_SKIP, so pytest.mark.skip turns into a hard failure there -- the previous fix (bd2d9df) traded 13 failures for a skip-triggered failure in a different job. The sysfs probe path in ffscreen.py is real code, just no longer what any caller reaches for by default. panel() now passes geometry=None explicitly so these tests exercise that path directly instead of skipping it -- real assertions against real (if currently unused) code, which both jobs can run to completion. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>