On Tue, 1 Jul 2003, [ISO-8859-1] Andr� Malo wrote: > ID 2 ... hmm, am I blind or don't we have an abort_fn registered in our > pools? Why? That would solve a lot of problems, IMHO. If there's an > abort_fn registered, all IDs referencing to this one are invalid.
There is such a feature in the pools code, but as far as I'm aware we don't use it. We intentionally do not check for OOM conditions because, though we've had many heated debates about this, we've always arrived at the consensus that if you hit OOM, your box is hosed anyway and virtually any effort you make to correct it other than dying will just make things worse. Not to mention the fact that several platforms (including linux iirc) have optimistic memory allocation, so even if you /are/ out of memory, you probably won't know it, so the checks would be useless anyhow. --Cliff
