Ticket28061 priority

[juga0]

Travis error comes from new flake8 version, not from changes in this branch

[pastly]

I don't think this does anything to solve the problem.

rp.best_priority() still returns a generator. That generator is started stored in relays. Then you iterate over the generator (just like before) but now there's a debug line logged about how long it took to iterate through all the relays we were told to measure.

Is my understanding of how the code works wrong? Assuming my understanding is correct ...

The actual fix would:

• go through all the the relays the generator gives us
• when we run out, wait until there are no more pending results <-- new thing
• then go back to the beginning.

while True:
    for target in rp.best_priority():
        # As before, tell a thread to measure the target
        # and store the async_result in pending_results
        if len(pending_results) >= max_pending_results:
            # As before, sleep and trim the list of pending results
    while len(pending_results) > 0:
        # NEW: sleep and trim the list of pending results
    # Go back to the top, now knowing all workers are idle 

[juga0]

i didn't see this comment before.
I got to think that using iterators would make the method being called for each iteration, (and i forgot to convert to list in my changes).
The new thing you propose might make sense, but yet, while that condition is not reached before the last iteration?.
And how is that i didn't get results close to each other this time?: https://trac.torproject.org/projects/tor/ticket/28061#comment:22

[juga0]

I realized that change make sense. I made that change and left it running for a day. There're no relays measured within minutes.
There're still relays being measured within one hour, but that's other issue.
Thanks for the suggestion.
Rebased also to master.