> On Nov. 2, 2015, 3:52 p.m., Neil Conway wrote: > > How about backward compatibility? Adding a note to docs/upgrades.md seems a > > good idea, at the very least. Are we pretty confident that no one else is > > looking at this data, and/or we're happy to break anyone that is?
Good point about backward compatibility. However, if the `data` field contains random bytes that might result into the entire json becoming "unrenderable", then it's broken anyways. Also note that we don't expose `TaskInfo::data` and `TaskStatus::data` fields over state.json either. We can use the dev/user mailing lists to get a feel for it. - Kapil ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/39611/#review104789 ----------------------------------------------------------- On Nov. 2, 2015, 3:21 p.m., Joseph Wu wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/39611/ > ----------------------------------------------------------- > > (Updated Nov. 2, 2015, 3:21 p.m.) > > > Review request for mesos, Ben Mahler and Artem Harutyunyan. > > > Bugs: MESOS-3771 > https://issues.apache.org/jira/browse/MESOS-3771 > > > Repository: mesos > > > Description > ------- > > We do not encode or otherwise preprocess the raw binary in the ExecutorInfo > or TaskInfo `data` fields before outputting them to JSON via the state > endpoints. This means that the outputted JSON may be invalid (i.e. non-UTF8) > if `data` was filled in with arbitrary bytes. > > The `data` fields are being removed because the state endpoints should not be > writing binary data in the first place. > > > Diffs > ----- > > src/common/http.cpp f56d8a178e1f61d91adfeaad32b47718adbc4379 > src/slave/http.cpp d6df97fb64077fb051dd78fe35bf04e8be0331ff > > Diff: https://reviews.apache.org/r/39611/diff/ > > > Testing > ------- > > `make check` > > Apparently, none of our tests check these fields. > > (Discovered via manual tests) Our JSON parser (PicoJson) is not restrictive > enough to catch the invalid-JSON that results. It will happily parse the > binary blobs (roundtripping successfully). > > > Thanks, > > Joseph Wu > >