Continuous 16-slot inventory now shows a persistent, in-place presence table: one row
per UID seen this session (kept as rows, never removed), redrawn in place each scan with
present / gone (Ns ago) status and a present/total count.
Also keep the RF field up across scans in continuous mode (drop it once on exit) instead
of re-initialising every round -- faster, and avoids per-scan 'Setting ISODEP' churn when
APDU logging is enabled. Fix the reader help text (the polling loop exits on Enter, not
the pm3 button).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The 16-slot inventory reused the single-slot read timeout (ISO15693_READER_TIMEOUT,
~1558us) for every slot, so each empty slot waited the full window. That timeout only
needs to catch the START of a response (a SOF) within t1 (~320us)+margin -- once a SOF
is seen it stops applying and the full response is still received. Add a dedicated
ISO15693_INVENTORY_SLOT_TIMEOUT (~755us) for the inventory slot 0 + EOF slots, leaving
the single-slot and WRITE timeouts untouched.
Verified on-device via hf 15 list (carrier-period timestamps): per-slot ~1.0ms vs
~1.8ms, ~1.8x faster rounds, no tags missed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add ISO 15693 16-slot inventory with anti-collision tree-walking, reachable via 'hf 15 reader --all'; single-tag reading stays the default. The firmware steps the 16 slots with EOF markers and flags 'signal but no decode' slots as collision hints; the client walks the UID mask tree, validates CRC per slot to confirm real collisions, and dedups discovered tags. Single-tag reading (default) and 16-slot inventory (--all) both run under hf 15 reader.
All four call sites already pass `cmd` as a Python list, so Python's default of `shell=False` means there is no active command-injection risk. This change makes the security intent explicit, prevents accidental future regression if a call site is later changed to pass a string, and improves code clarity.
No functional behaviour is changed.