----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/71781/#review218744 -----------------------------------------------------------
Ship it! Ideally we would have a regression test, but I won't push on that :P src/master/master.cpp Lines 11483-11489 (original), 11483-11491 (patched) <https://reviews.apache.org/r/71781/#comment306618> Looks good, I guess we could also just send the event before moving into the completed set to avoid the copy and the cognitive overhead of the extra variable? - Benjamin Mahler On Nov. 21, 2019, 9:42 p.m., Benjamin Bannier wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/71781/ > ----------------------------------------------------------- > > (Updated Nov. 21, 2019, 9:42 p.m.) > > > Review request for mesos, Andrei Sekretenko and Benjamin Mahler. > > > Repository: mesos > > > Description > ------- > > Moving a framework to `frameworks.completed` transfers ownership of the > Framework to that container and e.g., using the original framework > pointer after that could fail (the framework could e.g., be already be > rotated out). > > This patch fixes the master to not use the potentially cleaned up > framework pointer anymore. > > > Diffs > ----- > > src/master/master.cpp 14b90a5e276df055bb8a570331f27cab200c9869 > > > Diff: https://reviews.apache.org/r/71781/diff/1/ > > > Testing > ------- > > `make check` > > > Thanks, > > Benjamin Bannier > >
