src/qr/ vendors quirc.h, quirc_internal.h, quirc.c, decode.c,
identify.c, version_db.c from sidick/quirc
(a maintained fork of the original dlbeer/quirc,
tracked here specifically because upstream isn't active — see
THIRDPARTY.md). The fork has had fixes since we last synced,
but it's not clear whether any land in the parts AmiAuth actually exercises.
Note: unlike qrcodegen (pinned to an exact upstream commit, see
THIRDPARTY.md), there's no recorded commit hash or fetch script for our
quirc copy — it looks like a manual snapshot. Worth pinning a commit here
too once this is resolved, so future syncs are a diff against a known point
rather than "compare two working trees."
To check: diff our vendored copy against the fork's current HEAD,
focusing on whether any fix touches the actual call surface AmiAuth uses
(src/qr/qr.c is the only caller):
quirc_new / quirc_resize / quirc_destroy
quirc_begin / quirc_end (drives identify.c's geometry detection —
the float-heavy part QUIRC_FLOAT_TYPE=float exists for, see docs/ or
the Makefile's QUIRC_SRCS comment)
quirc_count / quirc_extract
quirc_decode / quirc_flip (uses version_db.c's tables)
If a relevant fix exists: pull it in, keep the ISC license header intact,
re-run make qr-onhw-smoke (the on-target QR-decode test) to confirm
nothing regressed on real/emulated 68k. If nothing relevant changed: note
that here and close, but still worth pinning a commit for next time.
src/qr/vendorsquirc.h,quirc_internal.h,quirc.c,decode.c,identify.c,version_db.cfrom sidick/quirc(a maintained fork of the original dlbeer/quirc,
tracked here specifically because upstream isn't active — see
THIRDPARTY.md). The fork has had fixes since we last synced,
but it's not clear whether any land in the parts AmiAuth actually exercises.
Note: unlike
qrcodegen(pinned to an exact upstream commit, seeTHIRDPARTY.md), there's no recorded commit hash or fetch script for ourquirc copy — it looks like a manual snapshot. Worth pinning a commit here
too once this is resolved, so future syncs are a diff against a known point
rather than "compare two working trees."
To check: diff our vendored copy against the fork's current
HEAD,focusing on whether any fix touches the actual call surface AmiAuth uses
(
src/qr/qr.cis the only caller):quirc_new/quirc_resize/quirc_destroyquirc_begin/quirc_end(drivesidentify.c's geometry detection —the float-heavy part
QUIRC_FLOAT_TYPE=floatexists for, seedocs/orthe Makefile's
QUIRC_SRCScomment)quirc_count/quirc_extractquirc_decode/quirc_flip(usesversion_db.c's tables)If a relevant fix exists: pull it in, keep the ISC license header intact,
re-run
make qr-onhw-smoke(the on-target QR-decode test) to confirmnothing regressed on real/emulated 68k. If nothing relevant changed: note
that here and close, but still worth pinning a commit for next time.