1320: fix deletion of reused files

[kkoehn]

unsolved problem of deleting reused files, when parent gets deleted

see problem description in #1320

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 94.77%. Comparing base (972ee15) to head (d27e358).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1696   +/-   ##
=======================================
  Coverage   94.77%   94.77%           
=======================================
  Files         130      130           
  Lines        3330     3330           
=======================================
  Hits         3156     3156           
  Misses        174      174           

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

[MrSerth]

I've spent further time debugging and looking for a solution. Finally, I wasn't able to do so with the current architecture and, hence, need to suggest a major architectural change:

Could you try to perform the necessary changes directly by updating and saving the changes (either with the @task.update method or by first assigning new values and then using @task.save)?

Of course, this means the service will return an altered object, persisted object already and not a modified version (that needs to be save). However, the big advantage is that we can more fine granular control which object gets saved how and don't need to wrap around the Rails magic.

Then, you should wrap the execute block into a database transaction. This allows you to perform the changes (i.e., on files) and they "look" persisted in this transaction even if they are not. You may then also use something like @task.reload to get an updated version of the record and move over to the next parts (i.e., tests, model solutions). If you only want to reload associations, you can do so with @task.association(:tests).reload (to force, use @task.association(:tests).reload(true)).

I am rather confident that we will be able to navigate around the current issue with this approach, even if it changes the API. Hence, I would suggest to follow with this approach.