Repository navigation
Attempt to fix SAMP testing issues - #2067
Conversation
|
I've found a couple of bugs, so this might work better (fingers crossed). I'll run Travis a few times to see what happens. If everything works fine, I may actually set the default pool size to 1 since people interested in high performance can always increase it, but I don't really think it's necessary to set it to 20 by default. |
|
I'm going to re-run the Travis tests several times to see how many times it:
I think the pool size was indeed the issue - somehow having a pool of proxies resulted in a non-deterministic order of communications. The question is whether one can better enforce the order of communications, or whether we should de-implement the pool (or leave it as an optional and potentially unreliable feature). One option would be to limit the number of proxy servers to one per client. |
|
Yeah -- I'm fine with merging this now (just to get the test suite into a little better shape) and deal with the pool size later. |
Attempt to fix SAMP testing issues
2 participants
(Related to #2064)
This pull request is an attempt to make the SAMP tests deterministic. To start with, I noticed issues with garbage collection in Python 3 due to the presence of the
__del__method (which are unneeded).However, I still see the garbage collection issues when running in parallel - not sure what is going on there. Maybe @mdboom or @embray would have ideas?
I also see a lot of warnings like:
Though I haven't quite figured out yet what is going on there.
As far as the non-deterministic tests go, I wonder whether it has something to do with the use of a pool of servers as opposed to a single server. If we still see the failures here I'll try forcing a pool size of 1.