On Sat, 2008-04-19 at 17:24 +0300, Tsantilas Christos wrote: > I think we have two different problems here, the client_side code and > the comm_close in async calls.
You are correct. General async call bugs and general comm API problems have been mentioned on this thread as well. Most of these problems are related, of course, but it is rather important to understand which problem the person is talking about when discussing a specific solution :-). > If I am not wrong the bug which triggered this discussion is the bug2309 > which is squid3.0 bug where AsyncCalls does not exists. If that is indeed the case, it is a good point! > The problem here > is the client side code which the true is that it needs clean up. It is > well known to everyone tried to fix bugs or add code in this part of > squid3.. Yes, I hope everybody who worked with the client-side code is in agreement that it "needs a clean up". I believe the question opening this thread was "when to do it" (v3.1 or v3.2) not "whether to do it". It is important to note that I can hit bug2309-like assertions in trunk as well. The exact causes may be a little different, but all seem to revolve around comm_close and transaction abort handling (before and after async calls). Thank you, Alex.