----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/7001/#review11357 -----------------------------------------------------------
This definitely needs to get done, so I'm stoked you took it on! However, this is the kind of thing I think merits some discussion (before unnecessary work is done ... sorry). In particular, I had created 'Try<void>' some time ago for this exact reason, but didn't use it after thinking we might want to use 'Try<Nothing>' instead. Here were the pros I saw to using 'Try<void>': + It captures the "void" return type well. ;) + We can eliminate 'Try<void>::get' so that people can't even attempt to get something that doesn't exist (although, a 'get' on a Try<Nothing> returns an object that you can't really do anything with, so it's very harmless). However, there were also cons: - The 'Try<void>' implementation is mostly duplicated code. - You have to do 'return Try<void>::some();' which doesn't read as nice as it could (at least, not as nice as 'return Nothing();'). - To do the same thing for Result and Future will require lots of duplicated code, which is at least a non-starter for Future and thus we'll probably always be using Future<Nothing> for asynchronous cases (and it seems much cleaner to be consistent). For these reasons, I was slightly more inclined towards 'Try<Nothing>'. Naturally, I'd love to hear others thoughts! - Benjamin Hindman On Sept. 11, 2012, 5:05 p.m., Ben Mahler wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/7001/ > ----------------------------------------------------------- > > (Updated Sept. 11, 2012, 5:05 p.m.) > > > Review request for mesos, Benjamin Hindman, Vinod Kone, and Jie Yu. > > > Description > ------- > > We unnecessarily have Try<bool>s all over the place, these are tri-state: > {error, some:true, some:false}. It seems most cases, we never use > {some:false} in the function or the caller. > > So, this restores some sanity to use two-state Try<void>s: {error, some} > > > Diffs > ----- > > src/linux/cgroups.hpp 1a3cdc2 > src/linux/cgroups.cpp 53d611f > src/linux/fs.hpp 31a6100 > src/linux/fs.cpp 744aea6 > src/logging/logging.cpp d6d31ec > src/slave/cgroups_isolation_module.hpp 00255b5 > src/slave/cgroups_isolation_module.cpp 8a121e0 > src/slave/gc.hpp 3760d09 > src/slave/gc.cpp 5212a41 > src/slave/process_based_isolation_module.cpp c0576bd > src/slave/slave.cpp 4ea1db1 > src/tests/cgroups_tests.cpp fbaa046 > src/tests/configurator_tests.cpp 8baed76 > src/tests/files_tests.cpp 6ef2004 > src/tests/stout_tests.cpp f690fac > src/tests/zookeeper_server.hpp 4f34910 > src/webui/webui.cpp d4f2ab9 > third_party/libprocess/include/stout/os.hpp 602db1f > third_party/libprocess/src/process.cpp 2d2b56c > > Diff: https://reviews.apache.org/r/7001/diff/ > > > Testing > ------- > > osx 10.7 gcc 4.2.1 > redhat Red Hat 4.1.2-48 gcc 4.1.2 > > make > make check > > note that SampleFrameworks.PythonFramework is consistently failing on red > hat, unrelated to this change > > > Thanks, > > Ben Mahler > >
