Create an EQC statem test for the DVV style riak_object #807
Conversation
Since the put_merge code was copied into two tests, both of which only manipulate riak_object, it seemed a refactor made sense. Looking at the put merge code, it only manipulates internal state of riak_object, so I think it is the correct place for the code. It also smooths the way for a general refactor of riak_object into a CRDT and riak_kv into riak_crdt (mwah ha ha ha ha!)
ghost
assigned 
| FrontierClock = vclock:increment(Actor, Timestamp, MergedClock), | ||
| {ok, Dot} = vclock:get_entry(Actor, FrontierClock), | ||
| %% Assign an event to the new value | ||
| DottedPutObject = assign_dot(NewObject, Dot, dvv_enabled()), |
There was a problem hiding this comment.
Didn't you just remove this assign_dot function from the module?
There was a problem hiding this comment.
Removed the exported 'assign_dot/2', kept this private one.
There was a problem hiding this comment.
Yes, but it's missing from the diff above.
| ]. | ||
|
|
||
| %% Like when a user does a put without first fetching (blank vv) @see | ||
| %% next_state/3 for the return tuple being prepended to vnode_data |
There was a problem hiding this comment.
note to me, when I address Dave's review comments, fix this too.
|
this test seems fine. It passed against 100k EQC runs. |
|
I noticed src/riak_kv_mutator.erl is still in your branch, can you rebase against develop? |
|
I addressed all comments except that desire to simplify the fake preflist generation. I really should address that, but I want the code this test tests into the build on Friday. Honestly I'll simplify the test in a later PR, since it will need to change to address that weird forwarding siblings issue. I rebased, and now when I try and push git tells me I need to pull and merge. Do I force push? Some other thing? Please can you help? |
|
I'd say create a new rebased branch + PR, close this one but reference it from the new PR. |
|
Closing and opening new PR for rebased branch. |
|
Link to new PR: #835 |
When making test I had to cut and paste the
riak_kv_vnode:put_mergeinto the test. The same code is also cut and pasted intoec_eqc. Seems an obvious candidate for a refactor. Andriak_objectseemed like the location sinceput_mergetinkered withriak_objectstate.