Proxy Incorrectly Handling Opened RX Port With No New Data

[kb1lqd]

Summary

Recent proxy.py updates incorrectly handle an open RX port that has no new data as it returns False when no data is to be returned.

Problem Explanation

False is returned when no data in available from an open port and proxy uses this as normal data. Telemetry and other single data item programs just ignore but multi-packet protocols such as Hermes messenger hang and fail to receive packets.

Environment

Software

Master software

Hardware

REV D1

Supporting Information

The image below shows test code additions that fix the problem by checking for False and disregarding the returned item as data.

[kb1lqc]

Slightly confused on this ticket, what should Proxy be doing?

[kb1lqd]

I placed a print statement in the hermes message script state machine for RX to print all fragment frames being parsed on a new reception. I only see the start packet arrive...

Printing out the TX path as the packets are created shows packets are being created and the START, DATA, STOP frame identifiers are correct:

Something is clearly not letting RX retrieve all the packets. Is it the updates for the CPU reduction?
#157

[kb1lqd]

I placed a print showing the count of port 3 items in proxy when sending a message through Hermes. the correct count of 3 was observed:

But the Hermes receiver got nothing (unedited):

Printing queue size in Hermes RX queuelen function shows only occasionally a single frame gets retrieved...

[kb1lqd]

Printing GetDict[] dictionary keys in proxy is showing that port 3 is not appending dictionary items!

[kb1lqd]

Using the proxy log functionality I can see 3 packets getting into proxy:

They are decoding correctly manually:

Oh damn.... @kb1lqc @reillyeon is this a bug due to the global getDicts but now there are multiple threads reading a single global item after I updated Proxy to split up each Faraday unit into it's own thread?

[kb1lqc]

@kb1lqd well this almost sounds like nominal Proxy operation. As designed it is only intended to store a queue and when anything reads the data it will pop it off. So if you append to a queue and some other thread pops it off then that is technically nominal.

Is there really a second thread reading from port 3?

[kb1lqd]

I'm saying the data is not going into the queue. The queue for port 3 gets removed or deleted... data comes into proxy but gets lost trying to append to the getdict queue for port 3.

…

On Thu, Apr 20, 2017 at 12:59 AM Bryce Salmi ***@***.***> wrote: @kb1lqd <https://github.com/kb1lqd> well this almost sounds like nominal Proxy operation. As designed it is only intended to store a queue and when anything reads the data it will pop it off. So if you append to a queue and some other thread pops it off then that is technically nominal. Is there really a second thread reading from port 3? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#162 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAUsyNONEjdGI0EidQ7cdJ_8QLaEA2JJks5rxxBTgaJpZM4M-vad> .

[kb1lqc]

This is interesting sing deque is supposed to be thread-safe:

https://docs.python.org/2/library/collections.html#deque-objects

A good example is "burning a candle from both ends":
https://pymotw.com/2/collections/deque.html#consuming

[kb1lqd]

We are storing the dequeue's in a dictionary... not thread safe.

…

On Thu, Apr 20, 2017 at 11:45 PM Bryce Salmi ***@***.***> wrote: This is interesting sing dequeu is supposed to be thread-safe: https://docs.python.org/2/library/collections.html#deque-objects A good example is "burning a candle from both ends": https://pymotw.com/2/collections/deque.html#consuming — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#162 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAUsyNJKI-ZB436jZHXMgIrPIkU0Yh-9ks5ryFB8gaJpZM4M-vad> .

[kb1lqc]

OK yeah that is correct and obvious in retrospect... Stackexchange explains it well:
https://stackoverflow.com/questions/3358770/python-dictionary-is-thread-safe

Looks like we will need a new approach to this. Any way we can quickly test out the theory with minimal impact? Maybe hack out the dictionaries so only dequeus are used and assume only one station ever?