OTP 17 compatibility

[shino]

This PR adds OTP 17 compatibility.

• depends on OTP 17 compatibility riak_cs_multibag#30.
• includes changes of Add OTP 17 compatibility shino/xmerl#3 which has been merged.

Tested with

• Erlang/OTP R16B02 basho patch 5 (a little old)
• Erlang/OTP 17.4 (EQC version 1.33.2 is used)

---

Note:

In 17.4, os:cmd/* treats input as a list of unicode code points and
it outputs results in a list of unicode code points too.
To avoid this, user administration interface in rtcs* is changed
to use erlcloud instead of curl (and fight with IO/unicode/etc.)

[kuenishi]

Why don't you use riakc_obj:metadata() here?

[kuenishi]

Please ping a pre6 tag to basho/riak_cs_multibag#30 and update this branch.

[kuenishi]

Ah, the same unit test failing again!! Looks like eunit parallelism has changed?

$ ./rebar eunit skip_deps=true suites=riak_cs_gc_manager_eqc
Building riak-cs-ee
==> rel (eunit)
==> riak_cs (eunit)
....................................................................................................(x10)..................
..................................................................................(x100)...................................
..........(x10)..(x1).........Failed! Reason: 
{'EXIT',{{badmatch,{error,{already_started,<0.7643.0>}}},
         [{riak_cs_gc_manager_eqc,'-prop_set_interval/0-fun-1-',1,
                                  [{file,"test/riak_cs_gc_manager_eqc.erl"},
                                   {line,94}]}]}}
After 5630 tests.
infinity
riak_cs_gc_manager_eqc:80: eqc_test_...*failed*
in function riak_cs_gc_manager_eqc:'-eqc_test_/0-fun-5-'/1 (test/riak_cs_gc_manager_eqc.erl, line 80)
**error:{assertEqual_failed,[{module,riak_cs_gc_manager_eqc},
                     {line,80},
                     {expression,"eqc : quickcheck ( eqc : testing_time ( 10 , ? QC_OUT ( prop_set_interval ( ) ) ) )"},
                     {expected,true},
                     {value,false}]}
  output:<<"Starting Quviq QuickCheck version 1.33.1

[shino]

Please ping a pre6 tag to basho/riak_cs_multibag#30 and update this branch.

Sure thing 😄

[shino]

Ah, the same unit test failing again!! Looks like eunit parallelism has changed?

• What OTP version? 16 or 17 or both?
• Particular on this branch?

[kuenishi]

Parameterized types like dict:dict(ref(), timer()) and dict:dict(timer(), ref()) ?

[shino]

What a big love for types ❗

[kuenishi]

What OTP version? 16 or 17 or both?

17.5.6.2, today's HEAD of basho/otp, basho-otp-17 branch. Not tested in R16 but I'm sure it passes ...

Particular on this branch?

I think so.

[kuenishi]

Great work; we still have several points to address. Let's work on them.

[kuenishi]

The checksum of my OTP17 is 57799e53416c915a202909 .

[kuenishi]

I'm sorry, I cannot reproduce above error after I removed .eunit directory. Current error is:

Warning: eqc_fsm header included, but no commands found in module riak_cs_list_objects_fsm_eqc
Compiled src/riak_cs_access.erl
test/riak_cs_list_objects_fsm_eqc.erl:0: variable '_' is unbound
test/riak_cs_list_objects_fsm_eqc.erl:0: variable '_' is unbound
test/riak_cs_list_objects_fsm_eqc.erl:0: variable '_' is unbound
ERROR: eunit failed while processing /Users/kuenishi/cs-2.1/riak_cs: rebar_abort
make: *** [test] Error 1

[kuenishi]

Fmm, weird. As I updated quick check then failed to reproduce. Please ignore my ado about errors.

[shino]

np!

[shino]

Addressed comments. Also fixed the error of riak_cs_gc_manager_eqc which could be reproduced with OTP 17 and EQC 1.36.1. Ready for review again.

[kuenishi]

@borshop merge