[jwrosewell]

Problem

ci/run-performance-tests.ps1 runs the performance example without a data file. The example then reads 51DEGREES_DD_PATH, and when that isn't set it searches the folders from the repository root for 51Degrees-LiteV4.1.hash (exampleUtils.js).

In the nightly build, ci/fetch-assets.ps1 creates a link with that name under fiftyone.devicedetection.onpremise/tests/ that points to the TAC file. The search takes the last folder in name order first, so it reaches tests/ before the real Lite file in device-detection-cxx/device-detection-data/. The graph is therefore measured with the TAC file at present, but only because of the folder order and a link named after the Lite file.

Change

• When it exists, passes --datafile fiftyone.devicedetection.onpremise/device-detection-cxx/device-detection-data/Enterprise-HashV41.hash, the link ci/fetch-assets.ps1 makes to the TAC file.
• Without a license key the TAC file isn't downloaded, so the script writes a warning and runs the example with its default data file, as before.

Verification

• The edited script passes a PowerShell 7 parse check.
• The argument block was run locally against a stand-in for the example. With the file present it passed --datafile <path> --jsonoutput results_Test.json, and without it only --jsonoutput results_Test.json.
• As the nightly build already reaches the TAC file, the graph's figures shouldn't move when this is merged. A step in the graph after merging would mean the old run was using a different file.