plug bg_manager_eqc into eunit and address potential race

[jrwest]

previously this eqc was not run as part of the eunit suite (#485). When plugging it in a found a failure due to a race-condition in the test:

1> bg_manager_eqc:run_check().
Starting Quviq QuickCheck version 1.29.1
   (compiled at {{2013,6,27},{9,46,39}})
Licence for Basho reserved until {{2014,5,29},{12,7,3}}
Failed!
[{set,{var,2},{call,bg_manager_eqc,set_concurrency_limit,[d,1,false]}},
 {set,{var,5},{call,bg_manager_eqc,start_process,[]}},
 {set,{var,7},{call,bg_manager_eqc,get_lock,[d,{var,5},[]]}},
 {set,{var,11},{call,bg_manager_eqc,stop_process,[{var,5}]}},
 {set,{var,13},{call,bg_manager_eqc,all_resources,[all]}}]


Final State:
---------------
alive = true
bypassed = false
enabled = true
procs = [{<0.222.0>,not_running}]
limits = [{d,1}]
locks = [{d,[{#Ref<0.0.0.262>,<0.222.0>,[],released}]}]
counts = []
tokens = []
---------------


background_mgr tables:
---------------
[{enabled,true},{bypassed,false},{{info,d},{resource_info,lock,1,true}}]
---------------
[{{given,d},{resource_entry,d,lock,<0.222.0>,[],#Ref<0.0.0.262>}}]
---------------


bg_manager monitors:
---------------
{monitors,[]}
---------------

bg_manager_eqc:set_concurrency_limit(d, 1, false) -> undefined
bg_manager_eqc:start_process() -> <0.222.0>
bg_manager_eqc:get_lock(d, <0.222.0>, []) -> #Ref<0.0.0.262>
bg_manager_eqc:stop_process(<0.222.0>) -> ok
bg_manager_eqc:all_resources(all) ->
  [#bg_stat_live{
     resource = d, type = lock, owner = {<0.222.0>, []} }]

Reason: {postcondition, {1, '/=', 0}}
false

This PR addresses both those issues.

see comment in changeset for more details

[jrwest]

not sure why the build is failing. a test was skipped because there should be 370 w/ this change. The bg_manager_eqc test did run.

[reiddraper]

@jrwest here's the failure, eunit's reporting is just stupid. This should be fixed in the merge commit:

module 'hashtree'
  hashtree: snapshot_test...[0.870 s] ok
  hashtree: delta_test...[0.028 s] ok
  hashtree: sha_test_...[9.125 s] ok
  hashtree: eqc_test_...*timed out*
  undefined

[jrwest]

bah, easy to miss. thanks.

[reiddraper]

I'll grab review now.

[reiddraper]

Can you elaborate on what local messaging semantics this relies on?

[jrwest]

"local" may be ancillary. its relying on ordering of messages between two processes A and B (the test process and the bg manager). Its not bullet proof because technically the 2 messages (enabled and the next command) may still arrive before the monitor message but its a lot less likely now -- i was reproducing it every 1k tests and now can't in 100k regularly.

[reiddraper]

Cool, thanks.

[jrwest]

build failed again for unrelated reason: #596

[reiddraper]

@jrwest so, shall we get #596 fixed, and the once that is merged, the bors merge-commit should pass here?

[jrwest]

@borshop merge