----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/47821/#review134840 -----------------------------------------------------------
Thanks for proposing this workaround for the SASL on Windows challenge. I must confess that I am not really *happy* with this but given our constraints, I believe the way you reduced my original patch is indeed least painful ;). So it looks good for me, barring the needed comment rewording. cmake/CompilationConfigure.cmake (lines 132 - 139) <https://reviews.apache.org/r/47821/#comment199772> Could you please reword this (and the other) comment to reflect the fact that we do have a pluggable authentication layer? I did add some clues for such reword into the linked JIRA issue. Given that I am not a native speaker, I guess you can do a much better job than I ever could for such updated comment :) Please note that we commonly have the "link" to the JIRA issue as trailing sentence; "... See MESOS-XXXX." - Till Toenshoff On May 25, 2016, 6:45 p.m., Alex Clemmer wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/47821/ > ----------------------------------------------------------- > > (Updated May 25, 2016, 6:45 p.m.) > > > Review request for mesos, Daniel Pravat, Artem Harutyunyan, Joris Van > Remoortere, and Michael Park. > > > Bugs: MESOS-5450 > https://issues.apache.org/jira/browse/MESOS-5450 > > > Repository: mesos > > > Description > ------- > > SASL is currently a hard dependency for Mesos. Per MESOS-5450, we expect > that some platforms will not support SASL (namely Windows), so this > commit will begin the first step in a several-step process of removing > it as a hard dependency. > > For this first step, this commit will shed SASL dependency in libmesos > only for Windows builds. We do this by: > > (1) Adding a preprocessor symbol, `HAS_AUTHENTICATION` that wraps the > SASL-dependent code, so that we can conditionally choose not to > compile it. > (2) Defining `HAS_AUTHENTICATION` on all Unix builds, and leaving it > undefined on all Windows builds. > (3) Logging an error and exiting the master if the user passes in > flags that depend on SASL, such as `--authenticate`. > > Notably, what we do *not* do is: > > (1) Shed SASL dependency in the tests. > > The impact of this is that, on Windows, relevant libmesos tests will > either not compile, or not pass. Naturally, as tracked by MESOS-5450, > the second phase of this series will be to shed the test dependency as > well. > > > Diffs > ----- > > cmake/CompilationConfigure.cmake 5c7833ceaed556cc4ffb650996e918c1a542c5f0 > src/Makefile.am 447bc2ab511ad173d3d911be10992be0974f4584 > src/master/master.cpp 35b428b0f7dee5954514d8860cfc498271ccf267 > src/sched/sched.cpp 4693d0dc09afc3ddbbf34e166579b6a6d71c3e38 > > Diff: https://reviews.apache.org/r/47821/diff/ > > > Testing > ------- > > > Thanks, > > Alex Clemmer > >
