> On March 16, 2017, 3:27 p.m., Andrew Stitcher wrote: > > I've looked this over, and I can't see anything obviously wrong. But > > there's a lot of detailed churn in this change, and I'm not so familiar > > with libuv, so I'm not confident enough to say "ship it". But I agree with > > the overall direction.
That'll do :) Churn was hard to avoid since the previous code assumed socket==work item. To have multi-socket listeners I needed to make the distinction and rename to keep it clear which touched everything unfortunately. We'll sort out bugs as we find them... - Alan ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/57662/#review169163 ----------------------------------------------------------- On March 15, 2017, 8:35 p.m., Alan Conway wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/57662/ > ----------------------------------------------------------- > > (Updated March 15, 2017, 8:35 p.m.) > > > Review request for qpid, Andrew Stitcher and Cliff Jansen. > > > Bugs: PROTON-1438 > https://issues.apache.org/jira/browse/PROTON-1438 > > > Repository: qpid-proton-git > > > Description > ------- > > (also available on https://github.com/alanconway/qpid-proton/tree/ipv4-ipv6) > > Improved listening behavior for pn_proactor_listen to allow selective > listening by protocol (ipv4/v6) or portable "listen to everything". > > Host can be a host name, IPV4 or IPV6 literal, or the empty string/NULL > (treated the same). The empty string listens on all local addresses. A host > name listens on all addresses associated with the name. An IPV6 literal > address (or wildcard '[::]') listens only for IPV6. An IPV4 literal address > (or wildcard '0.0.0.0') listens only for IPV4. > > - pn_proactor_listen may listen on more than one socket for ipv6/v4 or for > DNS names with multiple address records. > > - the 'backlog' applies to *each* socket > > - an error on any socket will close all the sockets of the listener, > PN_LISTERN_CLOSE event indicates all sockets are closed and provides the > error that triggered the close. > > PROTON-1438: added pn_event_condition convenience function > > A convenience function to make it easier to write generic error handling > functions for events. > > /* > * If the event context object has a condition and the condition is set > * return it, otherwise return NULL. > * If the event context object has remote and local conditions, > * try the remote condition first, then the local. > */ > PN_EXTERN struct pn_condition_t *pn_event_condition(pn_event_t *event); > > PROTON-1438: c libuv proactor - separate pn_connection_t on free > > Separate pn_connection_t from proactor internal data structures when the > proactor frees the connection - in case the application is using refcounting > to pin the connection. > > > Diffs > ----- > > proton-c/include/proton/cid.h 2d688964cbf2cf17850ee2f4f93eb6c85c7eb1a8 > proton-c/include/proton/event.h 6f93cd4d33c121d9265649814f7770a99d61cd45 > proton-c/include/proton/listener.h 2038c064a8a4fd621f584fd02db93a6e4593b708 > proton-c/include/proton/proactor.h 43b8ccb85525631818b8393c5e7f9b07ae0b348d > proton-c/src/core/engine.c 8c2aeb0d27a06384b7b1dc3488d9bb122dda7bd7 > proton-c/src/proactor/libuv.c 102fcdd8a30d2dd57d9545552bcfd695a251a66d > proton-c/src/tests/proactor.c beba46e84c75fb36677576b645fd2f39bb238827 > proton-c/src/tests/test_tools.h 9fe679c4ce99f8bd8446343ea180ad10a9a60c7c > > > Diff: https://reviews.apache.org/r/57662/diff/1/ > > > Testing > ------- > > Passes proton tests, including new proactor.c tests for ipv4/v6 listening and > connecting > > > Thanks, > > Alan Conway > >