Join GitHub today
GitHub is home to over 31 million developers working together to host and review code, manage projects, and build software together.Sign up
call destructors in broker shutdown path #1005
As discussed in #25, instance shutdown is far from clean. It's a bit better than before, now that we have the rc1, rc2, and rc3 run levels (modules loaded in rc1 are unloaded in rc3), but the "normal" path is to let the shutdown timer (started at the beginning of rc3) expire and call exit from its reactor callback.
Ideally we'd like to get to the point where all reactor watchers are stopped after their subsystems are torn down and the reactor would exit naturally, with the timeout acting as a fallback rather than the main path. That will require some design work to ensure the overlay is torn down in an orderly fashion without deadlocking distributed services, and is not attempted here.
This PR takes a step in the right direction by replacing the
I think we could close #25 if this is merged and open new issues for the remaining problems.
Also this lays some groundwork for #974 (convert to latest czmq) as zsys gets sad if any sockets are open in its
@@ Coverage Diff @@ ## master #1005 +/- ## ========================================== + Coverage 76.74% 77.32% +0.58% ========================================== Files 151 151 Lines 25724 25754 +30 ========================================== + Hits 19741 19915 +174 + Misses 5983 5839 -144