> On Dec. 16, 2016, 10:22 a.m., Alex Clemmer wrote: > > While not ideal, in the event that we were wrong about whether this symlink > > is required, it seems likely that it would at least make visible, obvious > > errors, instead of subtle ones. But, can we please create an issue to track > > our divestment from symlinks here, and tag it with the `microsoft` label so > > I can track it on our dashboard?
Note: I don't think this is the right thing to do, but it does show how little we actually rely on symlinks (on Windows, and only on Windows). I'm working on a short design doc for removing the need for symlinks on Windows (tests) entirely. In non-test environments, we'll still want symlinks (especially for the long-path support). - Joseph ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/54807/#review159470 ----------------------------------------------------------- On Dec. 15, 2016, 10:43 p.m., Joseph Wu wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/54807/ > ----------------------------------------------------------- > > (Updated Dec. 15, 2016, 10:43 p.m.) > > > Review request for mesos, Andrew Schwartzmeyer, Alex Clemmer, and Michael > Park. > > > Repository: mesos > > > Description > ------- > > This fixes the 500 or so tests that are enabled on Windows builds. > > With the exception of some parts of the agent recovery pathway, > creating symlinks is not necessary for the agent to function. > The current set of tests will pass with or without the ability to > create symlinks. In the case of the ASF CI, we will never have the > ability to create symlinks on Windows. > > > Diffs > ----- > > src/slave/paths.cpp 8792cee43d94e7b0bbd7b80aebbe501236244621 > > Diff: https://reviews.apache.org/r/54807/diff/ > > > Testing > ------- > > .\support\windows-build.bat (Win10) > > > Thanks, > > Joseph Wu > >
