D
dbeavon
Guest
Yes, I know about that. Our memory leaks go deeper than that. Typically the memory appears to be consumed by r-code in the heap, based on what I can tell when flailing around with windbg. We trim ABL sessions entirely so that there is *nothing* of ours that remains in the _mproapsv processes (only Progress stuff). This is especially true in the evenings when things are idle and there are long periods when _mproapsv has no ABL sessions. Aside from the CLR bridge we don't use memory outside of the normal AVM. I am quite sure we aren't use weird MEMPTR or something that might out-live any given ABL session. There is an extremely small leak within the managed memory of the CLR (related to bridge components), but that is not the leak I care about. The leak that concerns me is in unmanaged memory.
Continue reading...
Continue reading...