> On Feb. 24, 2014, 8:10 p.m., Adam B wrote: > > 3rdparty/libprocess/include/process/collect.hpp, line 340 > > <https://reviews.apache.org/r/18311/diff/4/?file=502382#file502382line340> > > > > In your implementation of _await(), you CHECK !isPending and > > !isDiscarded, but here you're binding to _await onAny result of the future. > > Why?
Yes, that is a proper bug. > On Feb. 24, 2014, 8:10 p.m., Adam B wrote: > > 3rdparty/libprocess/include/process/collect.hpp, line 299 > > <https://reviews.apache.org/r/18311/diff/4/?file=502382#file502382line299> > > > > What if future.isPending/isDiscarded/isFailed? Should this really be > > onAny()? Yes, the await-future is not implying that all/any wrapped futures are ready but only that they reached a terminal state - aka are complete (discarded, failed, ready) - just like await(list). > On Feb. 24, 2014, 8:10 p.m., Adam B wrote: > > 3rdparty/libprocess/include/process/collect.hpp, line 328 > > <https://reviews.apache.org/r/18311/diff/4/?file=502382#file502382line328> > > > > You set the return type on the .then lambda for Future<Future<T>>, but > > don't need to set it here? Why is it needed in one but not the other? Implicit type casting. - Till ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/18311/#review35318 ----------------------------------------------------------- On Feb. 27, 2014, 9:11 p.m., Till Toenshoff wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/18311/ > ----------------------------------------------------------- > > (Updated Feb. 27, 2014, 9:11 p.m.) > > > Review request for mesos and Benjamin Hindman. > > > Repository: mesos-git > > > Description > ------- > > Currently the Process::await implementation on list<Future>, by the nature of > std::list, expects equally typed futures. This new override implements await > for a tuple of futures, hence allows awaiting differently typed futures in a > single call. > > There also is a new override that allows await on a single Future, a > convenience approach for allowing a Process based await on a single Future > without forcing the user to render a list or tuple out of that. > > A C++11 and a boost-based implementation have been added. > > This patch also includes tests on those new overrides. > > > Diffs > ----- > > 3rdparty/libprocess/include/process/collect.hpp 2a73bc9 > 3rdparty/libprocess/src/tests/process_tests.cpp e899aed > > Diff: https://reviews.apache.org/r/18311/diff/ > > > Testing > ------- > > make check (clang c++11, gcc) > > > Thanks, > > Till Toenshoff > >