You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
make qr-onhw-smoke (the on-target datatypes.library QR-decode test, tests/gui/qr-onhw.sh / tests/gui/qr_onhw.c) currently fails on this
machine — found while verifying #110 (vendored quirc update).
Symptom: serial output shows DATATYPES=yes then LOAD=0 W=86 H=86
(the image loaded fine via qrimage_load_gray), then nothing — no DEC=,
no URI=, no END. No visible crash: a manual boot with --screenshot-after shows a normal 2.Workbench:> shell prompt, as if qr-onhw exited (or was never entered) rather than hanging or Guru'ing.
Raising SECS just produces more DATATYPES=/LOAD= boot cycles (3 at
SECS=400), suggesting a reboot loop rather than a slow decode.
Not the documented WB-HDD disk-degradation gotcha
(.claude/skills/copperline-testing) — du -sh on AMIAUTH_WB_HDD shows
~51 MB, within the normal few-tens-of-MB range, not bloated.
No CI baseline to compare against — qr-onhw-smoke isn't wired into .github/workflows/ci.yml at all; it's local-only tooling against a local
Amiberry HDD/ROM fixture (tests/gui/.env). So this may be specific to
this machine's current environment state rather than a real regression in
the codebase — worth confirming on another machine/fresh clone before
assuming it's a code bug.
To reproduce:
make qr-onhw-docker && make qr-onhw-smoke
Next steps to try:
Confirm whether it reproduces on a fresh AMIAUTH_WB_HDD clone (rule out
accumulated state in the long-lived local fixture).
Check whether qr-onhw even reaches main() — a boot-time crash before
the first DATATYPES= print would look identical to this from serial
alone; try a screenshot earlier in the boot, or check if the binary even
copies/executes (Startup-Sequence patching, SYS: assign timing).
If it does reach main() and exits after LOAD=, the qrimage_free/the
call into qr_decode_gray right after the LOAD= print is the next
place to instrument.
make qr-onhw-smoke(the on-target datatypes.library QR-decode test,tests/gui/qr-onhw.sh/tests/gui/qr_onhw.c) currently fails on thismachine — found while verifying #110 (vendored quirc update).
Symptom: serial output shows
DATATYPES=yesthenLOAD=0 W=86 H=86(the image loaded fine via
qrimage_load_gray), then nothing — noDEC=,no
URI=, noEND. No visible crash: a manual boot with--screenshot-aftershows a normal2.Workbench:>shell prompt, as ifqr-onhwexited (or was never entered) rather than hanging or Guru'ing.Raising
SECSjust produces moreDATATYPES=/LOAD=boot cycles (3 atSECS=400), suggesting a reboot loop rather than a slow decode.
Ruled out:
pre-update quirc (
git stashthe four changed files, rebuild, rerun):fails identically either way.
(
.claude/skills/copperline-testing) —du -shonAMIAUTH_WB_HDDshows~51 MB, within the normal few-tens-of-MB range, not bloated.
qr-onhw-smokeisn't wired into.github/workflows/ci.ymlat all; it's local-only tooling against a localAmiberry HDD/ROM fixture (
tests/gui/.env). So this may be specific tothis machine's current environment state rather than a real regression in
the codebase — worth confirming on another machine/fresh clone before
assuming it's a code bug.
To reproduce:
make qr-onhw-docker && make qr-onhw-smoke
Next steps to try:
AMIAUTH_WB_HDDclone (rule outaccumulated state in the long-lived local fixture).
qr-onhweven reachesmain()— a boot-time crash beforethe first
DATATYPES=print would look identical to this from serialalone; try a screenshot earlier in the boot, or check if the binary even
copies/executes (
Startup-Sequencepatching,SYS:assign timing).main()and exits afterLOAD=, theqrimage_free/thecall into
qr_decode_grayright after theLOAD=print is the nextplace to instrument.