Bug Fixes
-
lager adc/lager dacfailed on every named channel. The migrated dispatchers’ channel resolver only accepted integers, but adc/dac nets are saved with named pins (AIN0,CH0,DAC0) on LabJack T7 and MCC USB-202 — so every read/write on those nets failed with “Invalid channel pin”. Named pins now pass through to the drivers, which already parse them. Hardware-verified on both instrument families. -
lager supply <net> setfailed with “Unknown action: set_mode” for every supply model, and--ocp/--ovponsupply voltage/supply currentwere silently discarded — the command reported success but protection limits never reached the instrument. Both work now, with hardware-limit validation; a protections-only call (e.g.voltage --ovp 6with no value) applies the protection.clear-ocp/clear-ovpno longer 502 on EA PSB supplies. -
set_model('discharge')on the Keithley 2281S raised an error in 0.32.0, breaking HIL flows that select discharge battery simulation over SCPI. Discharge is the instrument’s always-available idle default, not a stored model — the request is now treated as satisfied, and strict empty-slot detection for numbered slots is unchanged. -
A Joulescope JS220 could be lost until a container restart after a
lager pythonclaim handoff. The open path now retries transient post-handoff failures with a short backoff, and a wedged USB context recovers with an automatic ~2s service respawn instead of requiring manual intervention. - Hardware errors printed a raw Python dict containing the full box-side traceback, and an internal proxy failure printed a literally empty “Hardware error: ”. Driver errors now surface as their one-line message (the traceback stays in box logs), and connection failures name their cause.
- A slow box-side operation was misreported as “cannot reach box”. The CLI now distinguishes a read timeout (box reachable, operation still running) from a genuine connection failure, and USB commands get a 30s first-contact budget for slow hub discovery.