Category archives: Snaking
Discussions of programming, particularly of programming Python
Burned too many times (Can't get happy about finding a leak...)
Written by
on
in
Snaking.
Have just discovered another leak, which is the one that was showing up with the "raise an error on send" test. Somehow holding a reference to a Deferred object which had methods registered from the holding object is producing the leak.
That so shouldn't be happening. Going to have to do some reworking of the ...
I give up (How the heck can *that* be leaking?)
Written by
on
in
Snaking.
Somehow a test case which consists of creating a client, calling getTable( ) on it, and having it immediately raise a socket error when it goes to send is leaking. There must be something going wonky at the Twisted level. Maybe a deferred/callLater object getting left on the stack for every call? Maybe a demon? ...
Take a step away from the computer (That's it, nice and easy...)
Written by
on
in
Snaking.
Frustrating (I'm gnashing my teeth at this point) to spend literally hours trying to whittle down a memory leak only to get close and find that it disappears. Mirages of memory leaks plague me. Now that I'm doing a gc run after each test case, and with a few other tweaks to try and filter ...
I believe this bug is driving me nutty (But at least I have a date for the staff Christmas party...)
Written by
on
in
Snaking,
Vindaloo.
Okay, so I have the bug isolated... sort-of. I have two test cases which differ solely by 1 call. In the one case, there's a memory leak (beyond the garbage created by the bisect oid store). In the other, there's none.
The problem is, I can't see any way for the leak to fall out ...
Leaks in the test-suite (Bisect OID storage seems to create unreachable tuples...)
Written by
on
in
Snaking.
The most obvious leak in the test suite turns out to be limited to the bisect-based OID storage module. Nothing to do with the real-world system :( . Haven't been able to figure out how the garbage gets produced, the tuples are just part of a big list stored inside the Agent object. The Agent ...
Oh, yes, the memory-leak safari continues... (Just in case anyone cares...)
Written by
on
in
Snaking.
Tracking down the memory leak is not going well. So far each time I think I've isolated a test case that produces a leak it eventually stabilises at some memory level after an hour or two of operation.
There's something funky going on that I'm thinking may be a reference loop that's requiring the cyclic ...
Doesn't look like it's in any condition to update to using v4 (PySNMP is hard enough to use with full releases...)
Written by
on
in
Snaking.
After a few attempts to get everything working with PySNMP v4a1, think I'm going to have to shelve that work and move back to PySNMP 3.x. Whatever fixes may have shown up, the amount of rewriting it's requiring is just too much to consider in the middle of trying to track down this memory leak. ...
Seem to have fallen behind in tracking PySNMP (Apparently there's a version 4 now...)
Written by
on
in
Snaking.
PySNMP (which is where I'm trying to track down the memory leak at the moment) appears to have a new version in development (in CVS). Really should have been paying attention to what changed before I hit "update" in the CVS module. Looks like some pretty serious refactoring.
Also appears to have a MIB parser ...
Need to put together a PyCon presentation proposal (or three) ("Functional embarrassment to exploding decorators, the creation myth of descriptors")
Written by
on
in
Snaking.
Also would be nice to have PyOpenGL-as-ctypes as a presentation, but that requires getting enough time to actually finish (heck, start) the project.
Maybe should do something for BasicProperty, PyTable and wxOO, just a quicky "save yourself some work" thing. Maybe build a wxPython UI for Cinemon just to be perverse :) .
Hmm, ...
Anyone know a decent tool for finding memory leaks? (Even being able to reliably and automatically determine *if* there's a leak would be good...)
Written by
on
in
Snaking.
Pinning down memory leaks in large-ish systems with large numbers of libraries is a bit of a pain. I wind up using the silly Windows task manager (yes, I'm still using Windows, despite having fixed Linux (sort-of) a while ago). Sure, I don't do this very often, but would be nice to have tools when ...
Mike C. Fletcher, the VRPlumber, is a philosopher of design turned computer programmer based in Toronto, Canada. He works primarily in the Python programming language, and maintains a large number of Open Source software packages. You can reach him at: mcfletch@vrplumber.com.