Try to stabilize rooms for rosters

[aleklisi]

This PR addresses #

Proposed changes include:

• describe the functionality changes
• describe new or updated tests
• describe changes to the documentation

[codecov]

Codecov Report

Merging #2319 into gdpr-retrieve-mam will decrease coverage by 13.98%.
The diff coverage is 100%.

@@                  Coverage Diff                   @@
##           gdpr-retrieve-mam    #2319       +/-   ##
======================================================
- Coverage              76.74%   62.76%   -13.99%     
======================================================
  Files                    334      334               
  Lines                  29154    29152        -2     
======================================================
- Hits                   22374    18297     -4077     
- Misses                  6780    10855     +4075

Impacted Files	Coverage Δ
src/mod_roster.erl	75.98% <100%> (-4.12%)	⬇️
src/mod_auth_token_rdbms.erl	0% <0%> (-100%)	⬇️
src/mod_aws_sns.erl	0% <0%> (-100%)	⬇️
src/inbox/mod_inbox_rdbms_mysql.erl	0% <0%> (-100%)	⬇️
src/mam/mam_message_compressed_eterm.erl	0% <0%> (-100%)	⬇️
src/inbox/mod_inbox_muclight.erl	0% <0%> (-100%)	⬇️
src/inbox/mod_inbox_rdbms_pgsql.erl	0% <0%> (-100%)	⬇️
src/mam/mam_jid.erl	0% <0%> (-100%)	⬇️
src/mam/mam_jid_mini.erl	0% <0%> (-100%)	⬇️
src/inbox/mod_inbox_one2one.erl	0% <0%> (-100%)	⬇️
... and 127 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 070f079...e161e3c. Read the comment docs.

[DenysGonchar]

it doesn't matter how much time we wait for the change in the roster. room doesn't appear at all. to repeate the issue we have to run 3 suites one after another:

• gdpr
• mam
• muc_light

to reproduce locally run:
test-runner.sh --preset riak_mnesia --skip-small-tests --skip-cover gdpr mam muc_light

[DenysGonchar]

to be more specific problem is caused by the failure in end_per_group/2 function for riak_timed_yz_buckets_muc_light group in the mam suite:

=== Reason: {badrpc,timeout}
  in function  escalus_rpc:call_with_cookie_match/6
     called as escalus_rpc:call_with_cookie_match(mongooseim@localhost,
                                                  mod_muc_light_db_backend,
                                                  force_clear,[],5000,
                                                  ejabberd)
  in call from mam_SUITE:end_state/3 (mam_SUITE.erl, line 874)
  in call from mam_SUITE:end_per_group/2 (mam_SUITE.erl, line 660)
  in call from test_server:ts_tc/3 (test_server.erl, line 1539)
  in call from test_server:run_test_case_eval1/6 (test_server.erl, line 1140)
  in call from test_server:run_test_case_eval/9 (test_server.erl, line 987)

crash in the end_state/3 blocks execution of end_modules/3 function, and so prevents correct stop of the mod_muc_light module:

  657: end_per_group(Group, Config) ->
  658:     C = configuration(Group),
  659:     B = basic_group(Group),
  660:     Config1 = end_state(C, B, Config),
  661:     Config2 = end_modules(C, B, Config1),
  662:     escalus_fresh:clean(),
  663:     delete_users(Config2).
  ...
  823: end_modules(C, muc_light, Config) ->
  824:     end_modules(C, generic, Config),
  825:     dynamic_modules:stop(host(), mod_muc_light),
  826:     Config;
  827: end_modules(_, _, Config) ->
  828:     [stop_module(host(), M) || M <- mam_modules()],
  829:     Config.

test command:

test-runner.sh --preset riak_mnesia --skip-small-tests --skip-cover gdpr mam:riak_timed_yz_buckets_muc_light muc_light

---

the root cause

gdpr suite doesn't remove muc_light rooms gracefully, so having it executed before mam suite results in longer execution of mod_muc_light_db_backend:force_clear/0