> On May 4, 2018, 2:45 p.m., Benjamin Bannier wrote: > > src/sched/sched.cpp > > Lines 1349 (patched) > > <https://reviews.apache.org/r/66938/diff/2/?file=2016633#file2016633line1349> > > > > Do we really need to `return` here? It seems just dropping this > > particular operation would be enough (in addition to calling `error` with > > all its side-effects). I am especially wondering about the tracking of > > operations. > > > > (With a `CHECK` the expected behavior would be simpler, not saying we > > should prefer it).
I switched back to aborting the process directly rather than calling error() - while this means we must eliminate the relevant test, it seems to make a bit more sense for the framework developer I think. It also matches what we do in a similar case, where the framework attempts to acknowledge a status update explicitly after implicit acknowledgement has been enabled. - Greg ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/66938/#review202446 ----------------------------------------------------------- On May 4, 2018, 11:20 p.m., Greg Mann wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/66938/ > ----------------------------------------------------------- > > (Updated May 4, 2018, 11:20 p.m.) > > > Review request for mesos, Benjamin Bannier, Gaston Kleiman, Jan Schlicht, and > Vinod Kone. > > > Repository: mesos > > > Description > ------- > > Since the 'SchedulerDriver' does not support operation status updates, > this patch adds a check to the driver which will abort the scheduler > if the 'id' field is set in an offer operation. > > > Diffs > ----- > > src/sched/sched.cpp 620a3b26d8bf3487b6ce922b2280be7da291df06 > src/tests/scheduler_tests.cpp 749420a23bc1a3873bd4d5aee56e78cff79bb1af > > > Diff: https://reviews.apache.org/r/66938/diff/3/ > > > Testing > ------- > > make check > > > Thanks, > > Greg Mann > >
