Job slots in use by kind of job, and vCPUs in use going by the workers' CPU use, against the fleet's capacity (one slot per vCPU, plus extra precompile slots). Sampled every 5 minutes; hover the chart to read an earlier sample.
Reports published by this farm, newest first. Open one, or link a report directly as ?run=<run id>. The daily evaluation of julia master has its own page.
Total test time of passing packages relative to the newest daily, so 1.1 means 10% slower than today. Same versions compares only the package versions the newest daily tested; all versions also chains in versions tested on earlier dailies.
Paste this on the pull request to re-evaluate every package that regressed, against the same baseline.
Packages that failed on the primary build but passed on the baseline — grouped by why they failed, most severe first.
Changes in the other direction, and failures shared with the baseline.
Packages skipped on the primary build, grouped by why. They count as neither passes nor failures.
Every evaluated package — type two or more letters to see how any of them fared.