Fix slow MAM muc_prefs_cases tests #4286
Conversation
This has a huge impact on the testing times for these cases. Since the room is reused across `muc_run_prefs_case` calls, there is no need to resend presence every time - the users are room participants all the time. The issue is that when (re)entering, Alice would wait for 4 stanzas - 3 presences and 1 room subject message. However, only in the beginning, as a new user would she receive the room subject - subsequent waits for 4 stanzas failed silently after only 3 presences, but only after waiting the 5s timeout. This has happend 6 times per testcase, and there were 3 tescases using this mechanism, so there is 3*6*5=90s less pointless waiting now.
gustawlippa
changed the title
|
elasticsearch_and_cassandra_26 / elasticsearch_and_cassandra_mnesia / 41863b7 small_tests_25 / small_tests / 41863b7 small_tests_26 / small_tests / 41863b7 small_tests_26_arm64 / small_tests / 41863b7 ldap_mnesia_25 / ldap_mnesia / 41863b7 dynamic_domains_mysql_redis_26 / mysql_redis / 41863b7 ldap_mnesia_26 / ldap_mnesia / 41863b7 dynamic_domains_pgsql_mnesia_25 / pgsql_mnesia / 41863b7 dynamic_domains_pgsql_mnesia_26 / pgsql_mnesia / 41863b7 pgsql_cets_26 / pgsql_cets / 41863b7 dynamic_domains_mssql_mnesia_26 / odbc_mssql_mnesia / 41863b7 pgsql_mnesia_25 / pgsql_mnesia / 41863b7 internal_mnesia_26 / internal_mnesia / 41863b7 pgsql_mnesia_26 / pgsql_mnesia / 41863b7 mysql_redis_26 / mysql_redis / 41863b7 mssql_mnesia_26 / odbc_mssql_mnesia / 41863b7 |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## feature/instrument #4286 +/- ##
======================================================
- Coverage 84.70% 84.67% -0.04%
======================================================
Files 556 556
Lines 33890 33890
======================================================
- Hits 28708 28697 -11
- Misses 5182 5193 +11 ☔ View full report in Codecov by Sentry. |
This PR fixes three testcase added recently. I'm having issues with this case in my other PR, and noticed that it's surprisingly slow.
Time before fix:
Time after fix:
Explanation:
Since the room is reused across
muc_run_prefs_casecalls, there is no need to resend presence every time - the users are room participants all the time. The issue is that when (re)entering, Alice would wait for 4 stanzas - 3 presences and 1 room subject message. However, only in the beginning, as a new user would she receive the room subject - subsequent waits for 4 stanzas failed silently after only 3 presences, but only after waiting the 5s timeout. This has happend 6 times per testcase, and there were 3 tescases using this mechanism, so there is 3 * 6 * 5=90s less pointless waiting now.I think maybe Escalus should fail loudly when less then N stanzas arrive when waiting for N stanzas.