Pathfinder development last week was focused on internal testing and bug fixes. The first pass revealed some regressions that we needed to deal with prior to a public beta. On the bright side, the added flexibility of our new locally quickest technique made it possible to fix some known bugs from previous releases. This week we will run through our test set again and (if all goes well) post a public beta shortly after.
Internal Pathfinder Testing
We have two test sets that are designed to detect any bugs or results changes that have crept in as a side effect of our code changes – regression tests and our V&V problem set. For the sort of internal testing we do during development, we use the regression tests. We usually only run the V&V set prior to a release (when we update the V&V manual).
Most of the regression tests are carefully-designed worst-case scenarios for Pathfinder’s movement algorithms. For each test, we run it against three Pathfinder modes: SFPE, SFPE+collisions, and steering. We run this set fairly frequently during development to check for regressions.
The image above shows the completed test matrix after we switched from shortest path to locally quickest path and added multiple (but not all) exit doors. As you can see from the red, we had a busy week.
Now that we’ve fixed some door crossing-related issues and improved the performance in certain cases, we’re optimistic that the first beta will be along shortly. However, first we need to re-run the tests.