BTW, the change suggested is by the original poster Pau Ceano. Actually I am the submitter of patch 491. For me, the assert in tcp_connector.cpp only gets a chance to show up after applying patch 491.
I don't think it is an issue of exhaustion as signaler.cpp creates purely local sockets, its address 4-tuples wouldn't overlap with those created in tcp_connector.cpp which include a remote component. On Tue, Jan 22, 2013 at 8:35 AM, Ian Barber <[email protected]> wrote: > > > On Mon, Jan 21, 2013 at 4:32 PM, KIU Shueng Chuan <[email protected]>wrote: > >> Ian, patch 491 fixes an assertion in signaler.cpp. >> >> This thread is about another assertion in tcp_connector.cpp. >> >> Both of them are due to re-use of sockets remaining in TIME_WAIT state. >> >> The first one is a Windows implementation issue. >> The second one could possibly occur on other platforms. (I have only used >> it enough on WinXP to have triggered it on WinXP though) >> >> > Yes, but the patch in 491 should reduce the amount of exhaustion, which > should lesson the chance of hitting the connection one (as the reconnect is > a less common event). If you're still hitting the problem, in the absence > of opinons of other windows users I'd be inclined to make a PR for your > change and see what the impact is like. > > Ian > > _______________________________________________ > zeromq-dev mailing list > [email protected] > http://lists.zeromq.org/mailman/listinfo/zeromq-dev > >
_______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
