Currently we have about 10 core services and based on SonarQube all of them have code coverage under 78%. After analysis, I discovered some problems which are listed below.

Problems

• lack of integration tests on UI
• lack of edge cases in tests which is generally the cause of regression
• adding tests only for 'happy path'

• a lot of duplicates (situation when we have a few tests for the method and in the most of them we have the same 'when... then')

• code duplicates between services (there are some classes which are duplicated in many services, but tests are in one service e.g. Message, PageableUtil)
• UI services coverages are not calculated on SonarQube, so we cannot monitor them

• incorrect tests which pass - caused by the lack of TDD

• adding contract tests rarely

Proposed solutions

• add SonarQube plugin to Intelij to be able to observe code coverage
• use Selenium for UI integration tests
• use TDD to improve the value of added tests
• add contract test (with proper 'history') to ACC
• add tests for edge cases, exceptions
• add tests for message keys
• add tests for importers/exporters