Connection problems with python in Ubuntu

aribes

Hi, 

I have modified some example to try to reconnect (a while True and a try/catch block) if there is an exception (e.g. timeout). 

I am using Ubuntu 16.04 with the latest library and sometimes I get this warning. 

No handlers could be found for logger "pygatt.backends.gatttool.gatttool"

What can it be due to?

Eric

• What python version are you using?
  
• Do you see a simliar "no handlers" warning if using pybluez?
• Please post the full code that I can run as is on my Linux machine

aribes

Python 2.7.12 on Ubuntu 16.04

The code I am using is the sensorfusion.py example. I have only modified the connection part by wrapping in a while True and a try/catch block, so I keep trying to connect if a timeout exception is raised.

If using pybluez, I can connect but I cannot set up sensor fusion (error is "Can't set data processor!"). 

Sometimes I get a segmentation fault when setting up sensor fusion when I use pygatt.

aribes

The code is here https://pastebin.com/XWMF1bcj

Eric

You will need to either reset the board or call "mbl_mw_metawearboard_tear_down" if you cannot create the data processor.  Ideally this should be done after the script is finished collecting data.

You can also remove the "set_sample_delay" calls and only stream one sensor fusion output which will remove the need to teardown the board.

aribes

I am already calling the disconnect method (python api) which calls the mbl_mw_metawearboard_tear_down, so I guess that is ok.

How do you execute issue a soft reset in code? I guess that the process should be: detect wrong behaviour -> issue soft reset -> wait X seconds (how long should X be?) -> reconnect.

I don't quite understand your last sentence, the "remove set_sample_delay calls" one. 

Eric

Call mbl_mw_debug_reset to reboot the board.  It could take 10+ seconds for the bluez stack to detect that connection has been lost but the actual number can vary from platform to platform.

What i mean is don't make calls to "set_sample_delay"

aribes

I am using pygatt, because with pybluez I get the "cannot create data processor" error message, no matter that I reconnect or reset the board (from metawear app).

aribes

I think resetting the board should not be needed, because if I terminate the app and start again right after that, then I can connect again. 

So it seems something related to python having some library internal state which is interfering, as is the only thing that changes when I ctrl-c the app and start again.

aribes

I tried to call init_libmetawear if I get 5 timeouts in a row, but it doesn't work. Sometimes, after 2-3 timeouts, it connects to the board, but other times the program gets stuck in endless timeouts, and the only solution is ctrl-c the app and run again.

Any ideas on how could I avoid doing that? 

Eric

Serialize the API state after a successful initialization.  Then, if the same board is detected in the script, deserialize the state before initializing the API.

https://mbientlab.com/cppdocs/latest/advanced_features.html#serialization

aribes

I don't quite understand what you mean by serializing the API state. 

I see that the mbl_mw_metawearboard_initialize is called in the connect() function (python API). When does the deserialization occur? Before or after connecting to the board? 

Should I save a state per board as a cache and use that before connecting?

Eric

See commits 544c758 and fcb1d61 on the pymetawear develop branch.

Akah

In the C++ documentation the function "libmetawear.mbl_mw_metawearboard_initialize" has three parameters and in the test.py has two parameters.
When I'm trying to execute the test.py test.py example (using Python SDK), I'm getting error message below:

root@debian:/home/user/Downloads/MetaWear-SDK-Python# python test.py XX:XX:XX:XX:XX:XX
error 1537656591.703993: Error on line: 296 (src/blestatemachine.cc): Operation now in progress
Connected
Traceback (most recent call last):
  File "test.py", line 17, in <module>
    libmetawear.mbl_mw_metawearboard_initialize(device.board, init_handler)
TypeError: this function takes at least 3 arguments (2 given)

Why the function has only two parameters here?

Thanks

Eric

Your post has nothing to do with the thread topic. Post your unrelated question(s) in their own threads, do not hijack existing threads.