Change in CQ numbers that then gets reverted again

[dhendriks]

Two days ago, in the Eclipse ESCET project, we had CQ numbers changing in the Dash License Check tool's output: https://gitlab.eclipse.org/eclipse/escet/escet/-/issues/1135. Then today, it changed back again: https://gitlab.eclipse.org/eclipse/escet/escet/-/issues/1139. I can keep updating the output to match what the tool provides, but is there any explanation for these changes? Will they keep randomly over and over again?

[waynebeaton]

The backend pulls data from a couple of different places. One of the sources of data is an old repository on Gerrit. The reference to CQ23806 is for version Apache Ant 1.10.12 in the Gerrit repository. The reference to CQ20836 is for Apache Ant 1.10.7 in the IPLab repository.

When searching for a matching version, the query will favour the closest service release. So, when asking for version 1.10.14, I would expect to match version 1.10.12.

According to the logs, the Gerrit repository is sometimes not accessible (likely due to planned brownouts) and so that 1.10.12 record is omitted from the search and we default back to the older version.

It's long past time to remove the Gerrit reference (the repository is no longer active). I'll move the data out of it and into IPLab. That should stabilise this.

[waynebeaton]

FWIW, that the CQ numbers are numerical anagrams (20836/23806) is just an interesting coincidence.

[waynebeaton]

@dhendriks I've pushed the updates. This aspect of the data source should be stable now.

FWIW, the tool should not create new issues for other service releases of Apache Ant 1.10. It is possible (however unlikely) that somebody could force a review of, say, version 1.10.14, thereby adding that version into the dataset and impacting your results. In this unlikely case, the tool would be working as designed as it is assumed that matching an exact (or closer) version will provide better results. Again, this scenario is unlikely.