Nightly Omega testing does no baseline comparison

[xylar]

Nightly Omega testing reports under the build name Baseline_<compiler> on CDash but no baseline comparison happens. cron-scripts/tasks/polaris_cdash/launch_polaris_ctest.sh calls polaris suite -c ocean -t omega_nightly --model omega -w <dir> -p <omega build> with no -b, so every task runs without a baseline and polaris_cdash.py finds no POLARIS BASELINE: line to check. The nightly tests therefore catch crashes and failed validation within a run, but not a change in answers.

Making this real needs a decision about where a baseline comes from, because maintaining one by hand across five machines and eight compiler configurations would cost more than it is worth. Baselines in Polaris are generated from main rather than curated, so the question is how a blessed set gets produced and propagated automatically.

What to decide

• What blesses a baseline. A label on a pull request is one option: labelling it would generate baseline output from that merge commit and publish it to the machine's shared area.
• Where blessed baselines live per machine, and how long they are kept.
• Whether the nightly run compares against the most recent blessed baseline or against a fixed release.
• What happens on the machines where nightly Polaris tests do not currently run at all (pm-cpu, pm-gpu, aurora — see Cron job installation Status #631), since a baseline there has nothing to compare.

Meanwhile

The build name should stop claiming to be a baseline comparison, which is part of #645.

---

Posted by Claude Code on @xylar's behalf. The testing, analysis and wording
above are AI-authored; please check them accordingly.