Bug Classification For Regression Tests

A set of simple criteria can be used to determine whether a bug is a candidate for a regression test often as part of the certification process

1. Test can be done on one isolated node
   • installation of rpms, configuration with yaim, valid proxy available if needed
2. Test needs more nodes
   1. a server (also included: BDII) is set up and needs to be tested via a client API/CLI from another machine (more realistic)
   2. an API needs to be tested and needs a server
      1. server needs special setup (version etc.)
      2. server can be standard
   3. several nodes are involved (FTS)
3. Test is difficult
   • web interface tests, security bugs (no info available), complicated workflow, problems that only occur after running the service for a long time etc

An SA3 investigation showed that around 50% of bugs are in the first category above.

-- OliverKeeble - 22-Feb-2010