> http://trac.edgewall.org/ contains at least one example of a reference > leak. It's holding up the release of 0.11 for a while. *scnr*
All my investigations on possible memory leaks in Trac have only confirmed that Python does _not_, I repeat, it does *NOT* leak any memory in Trac. Instead, what appears as a leak is an unfortunate side effect of the typical malloc implementation which prevents malloc from returning memory to the system. The memory hasn't leaked, and is indeed available for further allocations by trac. > The problem is also covered by the docs at > http://docs.python.org/dev/library/sys.html#sys.exc_info Ah, but that's not a *reference* leak. If Python (or an extension module) contains a reference leak, that's a bug. A reference leak is a leak where the reference counter is increased without ever being decreased (i.e. without the application having a chance to ever decrease it correctly). In this case, it's just a cyclic reference, which will get released whenever the garbage collector runs next, so it's not a memory leak. Regards, Martin -- http://mail.python.org/mailman/listinfo/python-list