mirror of
https://github.com/Klipper4FlashForge/firmware.git
synced 2026-10-21 12:52:16 -05:00
MonstrofilandClaude Opus 5
f095075985
chelper/__init__.py and c_helper.so are one unit: the first declares the cffi
cdefs, the second is what they describe. FlashForge's 1.9.7 declares
void extruder_set_pressure_advance(struct stepper_kinematics *sk
, double pressure_advance, double smooth_time);
and ours takes a fourth argument. A stock flash restores c_helper.so -- it is
in the chelper.tar their software component carries -- and restores nothing
else in that directory. Their klippy.py, toolhead.py and mcu.py come back too;
gcode.py, bed_mesh.py and chelper/__init__.py do not, because their overlay is
thirty files against the three hundred and twelve in the factory tree.
So a printer that took v20260827 or v20260827b, whose software component
carried our klippy, comes back from a stock flash with our cdefs against their
library and their extruder.py calling it their way: "expects 4 arguments, got
3", at the first extruder, on a machine that has already been flashed back to
stock and has nothing left to flash.
$MODDIR/prog/stock-chelper is FlashForge's directory, off a 1.9.7 factory
/usr/prog, and anvil-link-prog.sh restores it over ours. The whole directory
rather than the two files that disagree: their __init__.py is upstream's,
check_build_code and all, so a .c newer than c_helper.so sends klippy to
mips-linux-gnu-gcc, which no printer has. c_helper.so is touched last for the
same reason.
ONLY UP TO 1.9.7. That is the firmware this copy came off; a newer FlashForge
release may pair a different __init__.py with a different library, and
overwriting it would be this bug pointed the other way. An unreadable version
is left alone too. The version is theirs to state -- no package of ours ships
a software component, so /usr/prog/PROGRAM/software still names FlashForge's.
It rides in anvil-klipper rather than anvil-core, which qa/static caught: a
MIPS c_helper.so in a package declaring Architecture: all installs on anything.
That recipe gets a postinst of its own for the same reason anvil-klipper-config
has one -- anvil-core's pass runs before this package's payload has landed.
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>