search

m-labs / artiq

A leading-edge control system for quantum information experiments
https://m-labs.hk/artiq
GNU Lesser General Public License v3.0
425 stars 196 forks source link

lda: HidError and buffer mixup #65

Open jordens opened 9 years ago

jordens commented 9 years ago 
======================================================================
ERROR: test_attenuation (artiq.test.lda.TestLda) (i=8.0)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/rjordens/src/artiq-testbench/miniconda/envs/artiq/lib/python3.4/site-packages/artiq/test/lda.py", line 16, in test_attenuation
    self.assertEqual(i, self.cont.get_attenuation())
  File "/home/rjordens/src/artiq-testbench/miniconda/envs/artiq/lib/python3.4/site-packages/artiq/devices/lda/driver.py", line 205, in get_attenuation
    return (ord(self.get(0x0d, 1))/4.) * dB
  File "/home/rjordens/src/artiq-testbench/miniconda/envs/artiq/lib/python3.4/site-packages/artiq/devices/lda/driver.py", line 186, in get
    self.write(command, length)
  File "/home/rjordens/src/artiq-testbench/miniconda/envs/artiq/lib/python3.4/site-packages/artiq/devices/lda/driver.py", line 155, in write
    len(buf)))
  File "/home/rjordens/src/artiq-testbench/miniconda/envs/artiq/lib/python3.4/site-packages/artiq/devices/lda/driver.py", line 141, in _check_error
    raise HidError("{}: {}".format(ret, err))
artiq.devices.lda.driver.HidError: -1: None

======================================================================
FAIL: test_attenuation (artiq.test.lda.TestLda) (i=8.5)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/rjordens/src/artiq-testbench/miniconda/envs/artiq/lib/python3.4/site-packages/artiq/test/lda.py", line 16, in test_attenuation
    self.assertEqual(i, self.cont.get_attenuation())
AssertionError: 8.5 != 8.0

======================================================================
FAIL: test_attenuation (artiq.test.lda.TestLda) (i=9.0)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/rjordens/src/artiq-testbench/miniconda/envs/artiq/lib/python3.4/site-packages/artiq/test/lda.py", line 16, in test_attenuation
    self.assertEqual(i, self.cont.get_attenuation())
AssertionError: 9.0 != 8.5

[etc]

17:28 < rjo> ysionneau: lda/hidapi looks like it breaks intermittently and then has some queuing mixup. any ideas? http://people.phys.ethz.ch/~robertjo/artiq-ci/artiq-ci_20150705114415-nist_rj1-5db52b1_newassy.log 17:31 < ysionneau> rjo: are those tests running with real hw? 17:35 < ysionneau> ah it seems so 17:35 < ysionneau> I'll have a look tomorrow morning 17:36 < ysionneau> I guess you can open an issue with the first failing test with the Hid exception 17:43 < rjo> yes. 17:44 < rjo> i suspect there are two issues (the error and the queuing mess-up) and both are in hiapi.

jordens commented 9 years ago

This bug is probably either in the LDA hardware, the USB stack, or hidapi. @dhslichter You tested and signed of on the LDA driver. Is a workaround (closing the device, reopening, and trying again) good enough?

dhslichter commented 9 years ago

This workaround seems suitable to me. The LDA is a "slow" hardware device (so a little extra time to close and reopen is OK from time to time), and has a non-volatile memory so the function of the actual hardware won't be affected by this process.

jordens commented 9 years ago

Keep in mind that this bug might mean that the device is in a corrupted state and has arbitrary attenuation, at least for a while.

dhslichter commented 9 years ago

Well, if this is truly due to issues with hidapi we don't have much choice, right?

jordens commented 9 years ago

Hidapi is open source.

dhslichter commented 9 years ago

Sure, but it represents some work and then the hidapi maintainers have to want to integrate the suggested changes, because it seems like a bad idea to have our own artiq-friendly version of hidapi that needs to be maintained. Comments from others? Obviously it would be nice to have this bug fixed, but I don't know how much work is entailed in tracking it down and writing a fix for hidapi. Certainly I will not be able to do this.

jordens commented 9 years ago

Somebody could test whether the bug and/or the workaround have any consequences along the lines I described.

sbourdeauducq commented 5 years ago

it seems like a bad idea to have our own artiq-friendly version of hidapi that needs to be maintained.

Since the upstream hidapi seems to be pretty much abandoned, there isn't much of an alternative if we want this bug fixed.