> On Sept. 14, 2016, 1:59 p.m., Benjamin Bannier wrote: > > src/common/resources.cpp, line 1745 > > <https://reviews.apache.org/r/51517/diff/2/?file=1495047#file1495047line1745> > > > > It would be great if you could avoid using `default` when switching > > over enum values. See https://issues.apache.org/jira/browse/MESOS-3754 for > > some background. > > Anindya Sinha wrote: > Done at both call sites. > > Benjamin Bannier wrote: > Sorry for being unclear, but I think since these enums are just glorified > names for integers you still need to handle unknown values. What about a > pattern like this: > > ostream& operator<<(ostream& stream, const > Resource::DiskInfo::Source& source) > { > switch (source.type()) { > case Resource::DiskInfo::Source::MOUNT: > return stream << "MOUNT:" + source.mount().root(); > case Resource::DiskInfo::Source::PATH: > return stream << "PATH:" + source.path().root(); > } > > UNREACHABLE(); > } > > Jiang Yan Xu wrote: > This style LGTM. Just to clarify, "since these enums are just glorified > names" you mean because we haven't turned on `-Wswitch`? In the proposed new > style we use `UNREACHABLE();` for readability and not for "unknown values" > right? > > Anindya Sinha wrote: > Updated. > > Benjamin Bannier wrote: > Thanks for accommodating these last minute requests Anindya. > > In general above pattern could still trigger `UNREACHABLE` at runtime when > > a) a caller forces an uncovered integer into `Type`, e.g., with > `source.set_type(static_cast<Resource::DiskInfo::Source::Type>(42));`, or > b) when new values are added to the `Type` enum, but code switching over > values is not recompiled. > > It looks like neither issue is likely for this particular code, but the > suggested pattern should be a pretty safe general solution.
OK then `-Wswitch` can only help to some extent I guess. In Mesos code itself we would certainly discourage code that would introduce the two above cases. I guess it's valuable to mention these in MESOS-3754? In cases where you can't directly return from switch cases it seems we cannot completely eliminate the need for ``` default: UNREACHABLE(); ``` when considering the above cases, no? Of course none of these concerns are applicable to this patch. Just thought you should share these points on MESOS-3754 too. :) - Jiang Yan ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/51517/#review148967 ----------------------------------------------------------- On Sept. 14, 2016, 3:32 p.m., Anindya Sinha wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/51517/ > ----------------------------------------------------------- > > (Updated Sept. 14, 2016, 3:32 p.m.) > > > Review request for mesos and Jiang Yan Xu. > > > Bugs: MESOS-6060 > https://issues.apache.org/jira/browse/MESOS-6060 > > > Repository: mesos > > > Description > ------- > > We also log the disk type (MOUNT or PATH) for each persistent disk as > well as the root path for these disks while logging resources. Note > that this is logged only when source is present, otherwise there is > no additional content logged. > > > Diffs > ----- > > src/common/resources.cpp f841e91b81975a887e64a76a709708ed7946025f > src/tests/resources_tests.cpp f627862870279e5a778aa83e1ddcaa88c27ba1cb > src/v1/resources.cpp 4410a8f95ef0ebe6ec7aee38cd47f6f83f863867 > > Diff: https://reviews.apache.org/r/51517/diff/ > > > Testing > ------- > > All tests passed. > > > Thanks, > > Anindya Sinha > >
