On Tue, 2014-03-18 at 12:11 +0100, David Henningsson wrote: > On 03/18/2014 09:23 AM, Tanu Kaskinen wrote: > > module-zeroconf-publish assumes that if it creates two defer events, > > then the one that was created first will be dispatched first. That's > > a fair assumption in my opinion, so let's fix the ordering in > > pa_mainloop. > > Are you sure there are no modules assuming the current behaviour, and > thus will be broken with this patch? I don't know any good way to check that, so I'm not 100% sure. However, I find it very unlikely that someone would expect events to be processed in inverse order. That said, I don't think this is the right fix for the zeroconf-publish crash. I was going to suggest that we'd document it in pa_mainloop_api that defer events must be dispatched in creation order, but then I realized we have no control over existing external mainloop implementations, so we can't safely add that new promise to the API. I'll fix the bug in a different way (by removing all defer event ordering assumptions from zeroconf-publish). -- Tanu
