Skip to main content

​
Upgrade Notes

  • Update your boxes. Most of the fixes in this release run on the box. Run lager update --box <box-name> after you upgrade the CLI.
  • lager wifi status exits 1 when the box has no wireless interface. It used to print “Interface detection failed” and exit 0. A script that treats that exit as success must now handle the failure.
  • lager wifi status and lager wifi access-points run on the box host over SSH. They used to run inside the container, which cannot see the host’s wireless interface. access-points now scans the box’s own wireless interface by default, not wlan0.
  • lager usb on a YKUSH hub fails when a port does not switch. The hub acknowledges a switch that it ignores, and enable, disable, toggle and cycle reported success. The driver now reads the port back. A script that relied on the false success now gets an error.

​
Features

  • Yepkit YKUSH3 hubs are supported. lager instruments and lager nets add-all detect the YKUSH3 (04d8:f11b). Its three USB ports are usb nets for lager usb. Its three GPIO pins are gpio nets for lager gpi and lager gpo. The hub’s switchable 5V output is not a net.
  • The IPEVO V4K PRO document camera (1778:d00e) is detected as a webcam. lager instruments lists it, and lager nets add <name> webcam ... accepts it.
  • A gated box’s gateway can see which command each request belongs to. Requests that carry a lager login token also send the command’s name, such as supply voltage, with the CLI version and an id for the command. The name never includes arguments. A box that is not behind a gateway receives none of these headers.

​
Bug Fixes

  • The /scope page shows the scope’s real state.
    • A Rigol that loses only its capture card is no longer called unplugged. The page says that the live view stopped, and it says when the card is back.
    • The page says when the box closes its connection after a long idle.
    • The run badge shows a dash after a disconnect. It used to keep its last state, such as TRIG’D, until the next frame.
    • The page says when a scope net has no channel nets, and it gives the lager nets add command for each channel.
    • The stream request no longer returns paths on the box when a scope’s daemon fails.
  • The /scope page draws a Rigol’s trace where the scope’s own screen does.
    • display zoom now magnifies the trace and shows the zoom bar.
    • A channel’s hardware offset now moves its trace. The volts read off the trace do not change.
    • The box checks for a Rigol’s side menu again a minute after a check finds none. The menu no longer stays over the live view.
  • A stop typed in the /scope page’s console stops the measurement polling. start and start single in the console now follow the same rule as their buttons.
  • lager scope <net> cursor move-a and move-b move from where the cursor is. On a Rigol, each call started from the center of the grid. After set-a --x -200, move-a --x 50 now puts A at -150, not +50.
  • lager scope <net> status says “running” for a Rigol that the daemon opens while it runs. The command and the scope page said “stopped” after every daemon restart, power cycle or replug until you pressed Start.
  • lager scope <net> measure prints a reading without float noise. A Rigol’s duty cycle now prints as 49.987 %, not 49.986999999999995 %. A Rigol’s refusal of an averaging count no longer names 65536 captures, which the CLI does not accept.
  • A Rigol MSO5000 bus decoder is refused before anything is sent. Creating a UART, I2C, SPI, CAN or FlexRay decoder failed partway through the setup with “Function not found”. It now raises an error that says the scope does not decode buses through lager.
  • The scope daemon’s log says when the capture card stopped. Requests that wait at that moment are no longer told that the oscilloscope stopped answering.
  • The scope daemon notices a PicoScope unplugged while it was idle. The next command failed with a bare driver error. The daemon now closes the unit when a 2000-series call fails and no PicoScope is on the bus. It opens the unit again when it returns.
  • A PicoScope plugged back in gets its settings back. Channel enables, volts/div, coupling, probe, timebase and trigger returned to the driver’s defaults, so a page that pressed Start drew nothing. The same unit, by serial, now keeps its settings. A different unit starts from its defaults.
  • lager debug flash and lager debug erase no longer call an unreachable box too old. With --erase-start or --erase-size, a /health timeout told you to update the box. The command now retries twice, then says that the box or its Stout gateway did not answer.
  • lager diagnose on a USB hub net reports an unreachable endpoint as unreachable. It used to say that the box does not serve /diagnose/usbhub, and it told you to update the box.
  • lager usb says that the box service restarts, not “cannot reach box”. When a hub does not open after it re-enumerates, the box service exits to recover. It now answers first, and it tells you to retry in a few seconds.
  • Python API calls on a multi-channel supply net act on that net’s channel. On a Rigol DP800 or Keysight E36000, ovp(), ocp(), state(), clear_ovp() and similar calls through Net.get() sometimes acted on another channel. lager supply was not affected.
  • lager supply can set OVP and OCP back to a Rigol DP800’s factory limits. A DP800 ships with protection limits up to 110% of the channel’s rating, and the CLI refused any limit above the rating. The limit is now 110%. Voltage and current setpoints stay within the rating.
  • wait_for_level() on a LabJack T7 GPIO net counts the stream setup against its timeout. A call ran about 30 ms past its timeout. The call still stops the stream before it raises TimeoutError, which can take about 0.6 s.
  • A command stopped with kill releases its box lock. A killed background command, such as lager debug <net> gdbserver, used to hold its lock until the lock’s TTL ran out.
  • The box’s authorized_keys sync survives the end of an install or update. The key-sync poller now runs in its own session. A signal to the caller’s process group no longer stops it.
  • The box image carries the license notices of uldaq and the LabJack Exodriver. They are at /usr/share/licenses/lager/vendor/.

​
Known Limitations

  • The YKUSH3 GPIO pins are not fully verified on hardware. Their voltage level, and whether a written level is held, are not yet measured.
  • Two IPEVO V4K PRO cameras on one box share an address. The camera reports no serial.
  • I2C and SPI per-call overrides stay on the driver. An overrides value such as frequency_hz also applies to later calls on the same net. Set the value that you want on each call.
  • A failed resistance set on an EA solar net does not raise an error. The call returns the resistance that the instrument still has. Compare the returned value with the value you set.

​
Installation

To install this version:
To upgrade from a previous version:
Then update your boxes:

​
Resources

View Release on PyPI