> On Jan. 18, 2017, 1:50 a.m., Joseph Wu wrote: > > src/launcher/default_executor.cpp, line 1055 > > <https://reviews.apache.org/r/55311/diff/1/?file=1599651#file1599651line1055> > > > > It actually feels like `process::Winsock winsock;` is more appropriate > > here, as we want the socket stack to be cleaned up afterwards as well. > > Either than, or you should add a `process::finalize(true)` befow.
Tearing down winsock before libprocess is terminated will cause ugly failures as the test framework disposes itself. Let's do a call to `process::finalize(true)`. - Alex ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/55311/#review162013 ----------------------------------------------------------- On Jan. 8, 2017, 6:28 a.m., Alex Clemmer wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/55311/ > ----------------------------------------------------------- > > (Updated Jan. 8, 2017, 6:28 a.m.) > > > Review request for mesos, Andrew Schwartzmeyer, Daniel Pravat, and Joseph Wu. > > > Repository: mesos > > > Description > ------- > > The default executor's `main` currently depends on `UPID`, which > networking libraries to do things like retrieve the address of the > current host. On Windows, this means that it is required to initialize > the winsock stack to be initialized before `UPID` is used. > > To resolve this issue, we add a call to `process::initialize` at the > beginning of the `main`, which will cause the stack to be initialized > correctly. > > > Diffs > ----- > > src/launcher/default_executor.cpp fd9963996c82461b60888397989e309d51b60aaa > > Diff: https://reviews.apache.org/r/55311/diff/ > > > Testing > ------- > > > Thanks, > > Alex Clemmer > >
