Memory leak on snmp timeout with twisted High Level API

[mcfletch]

So we're seeing a memory leak in a polling operation where if the pysnmp getCmd (twisted hlapi) times out we leak RAM. I've attached a (rather poorly behaved, kill -9 to stop) reproduction script that leaks RAM at about 1MB/6s or so by issuing 3000 timing-out requests in a loop.

It's entirely possible that there's a problem in my usage, but it looks pretty close to documented usage. My guess is that there's some state in the engine or the udptransport that isn't getting cleared.

To be clear, the real application doesn't issue 3000 requests at once, but it shows the same leak pattern where any timed-out operation seems to leak memory.

testleak.py.txt

[lextm]

Recent releases no longer support twisted, so this can be closed.