-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/39611/#review104789
-----------------------------------------------------------


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?

- Neil Conway


On Nov. 2, 2015, 8:21 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, 8:21 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
> 
>

Reply via email to