On Nov 14, 2007 3:26 PM, Ruediger Pluem <[EMAIL PROTECTED]> wrote: > Hm. Currently this works fine, but in r591488, you create a subpool of > st->pool and > save it in l->pool and I understood that it was your intention that this > subpool > will be cleared sometime in the future and at this point of time > uldap_connection_cleanup > seems to be important again. As I see no harm in registering the cleanup for > this pool > and as the person who introduces clearing this subpool will have forgotten > about this cleanup > I would think that reverting would be the best here.
I was afraid that leaving cleanup registered (on a pool that isn't cleaned up) would be the more misleading option. It's not just a matter of picking when to run the cleanup -- the cleanup function itself (and the prior allocations) do not accomodate fully removing a backend connection. The details of the allocation were largely unimportant when the cleanup was intended to be run at child exit (even though it wasn't running at all) because this all storage allocated only in the child I'm glad to revert && comment until such a time that it can be worked out more fully (and yank backport) -- thanks for the review. -- Eric Covener [EMAIL PROTECTED]
