Fix bug in `DataHandler` for new protocol

[bplimley]

While a backlog of data is being flushed to the server in DataHandler.regular_send(), there is potential for a network error (especially if the server is receiving heavy traffic). Currently this network error is caught in DataHandler.main() but not handled correctly (cpm is saved to queue, rather than trash[0] etc.)

[tybtab]

Hi, Brian I'm not sure what the problem is for this specific issue. If a cpm fails to send, it will go to queue (send_to_memory method). If main undergoes a regular_send (network is up) then the queue is left-popped to a temporary trash variable which is used to send the cpm popped from the queue.

[bplimley]

Sorry for the delayed response. Let me just describe it more specifically with an example.

• DataHandler.main() is called with cpm=2.
  • DataHandler.regular_send() happens with cpm=2.
  • sender.send_cpm_new() happens with cpm=2.
  • There is an existing queue of data, so DataHandler.regular_send() starts flushing the queue with the while loop.
  • One of the sender.send_cpm_new() fails with a network error on a measurement with cpm=0.
• The exception is caught in DataHandler.main(). That's good but self.send_to_memory(cpm, cpm_err) is sending cpm=2 to memory, not cpm=0!

Maybe you want to use DataHandler.main() in the while loop. I haven't thought it through though.

There's also another bug here, let me make a new issue for it.

[tybtab]

I made some changes. When it tries to send entries in the queue now, it undergoes a try and except for each entry in the queue. If it fails, the entry is placed back in the queue and datahandler stops iterating through the queue. I also added some arguments to some of the methods in datahandler in order to implement this.

[bplimley]

@tybtab FYI, I tested this on my device and it looks like it works fine.