Its near the bottom of worker_thread() After the h_command_loop is called we come back and test a flag in the context (something like async_waiting) and then call the async function.
Don't see why that would break off the top of my head. What I have noticed though is my maillog now grows to fill the entire disk due to messages about your CtdlClientGetLine() being called with wrong args or bad fd. My system has no contexts at this time so it shouldn't get called at all. The new context scheduling is almost done and it will take care of all that. I'm merging your changes (via HEAD) into my branch all the time so that switch should be painless. Maybe we should disable the messages for the moment. > Wed Feb 24 2010 02:32:46 EST from dothebart @ Uncensored > > > >me did: > >* catch the -1 in two places > >* add a checker loop whether more comands are waiting in the buffered IO >buffer > >don't know whether this broke the async stuff, which lines do that? maybe >checking whether IO is waiting is not enough? whats the condition for async >io? > > > >
