The tool downloaded only the first event, which on UM20147 is a 4,796 B
histogram -- so it could not reach the 72,560 B event that is the whole point of
the uint32 chunk-offset change.
--event takes a key in hex, 'first', 'largest' or 'all'. 'largest' is the one
worth running on an unfamiliar unit, since it is what exercises offsets past
64 KB; the output flags any event over 64 KB as carrying into params[1], and
prints the chunk count so a 71-chunk walk is visible as such.
Selected events are downloaded from inside iter_events(), so the chain cursor
sits where THOR's would -- non-selected events are walked past without a
download, which the browse walk already attests is fine.
Also: connect() returning nothing now prints a hint rather than just a bare
timeout. /dev/ttyACM* numbering shifts whenever another CDC-ACM device is
plugged in, and this cost a debugging session when the tool was aimed at a
TMI-Seismo MCU instead of the Micromate. `ls -l /dev/serial/by-id/` names each
device and is stable, so the hint says so.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ru8Lg9HkkYvX9VWWo65SmL
serversdownandClaude Opus 5
0cfec2f098