Avoid inconsistent coverage of drain callback in streaming code

[Martchus]

• Ensure code section mentioned in https://progress.opensuse.org/issues/167803 is consistently covered
• See further commit messages for other improvements; this PR will generally improve the coverage of streaming-related code

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.96%. Comparing base (957f9c2) to head (c12ea6e).
Report is 5 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #6016      +/-   ##
==========================================
+ Coverage   98.93%   98.96%   +0.03%     
==========================================
  Files         396      396              
  Lines       39341    39345       +4     
==========================================
+ Hits        38923    38939      +16     
+ Misses        418      406      -12     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[kalikiana]

I guess the finish/tick bits are what makes it reliable. Good catch!

[Martchus]

I guess the finish/tick bits are what makes it reliable.

No, the code section mentioned in the ticket was simply not covered at all by this unit test. It was only covered by the fullstack test and that was not reliable. Finishing the controller is for extending coverage even a little bit further (see codecov report).

The one-tick bits are so that all events are handled before leaving the subtest. Otherwise they'd be handled in the next subtest where certain objects have already been destroyed leading to Perl warnings. This would have been a problem after the unit test is now executing more of the event processing code.