Skip to content

History / Test references and report generation

Revisions

  • Drop the cross-version /SP timing claim; state what -sp is and does Published results predate PerformanceLogging, so a discontinuity note on the report page is noise for the people reading them. The flag keeps its row in the table and a short section: /SP exists since GeoDMS 20.16.0, is off, and when on measures and reports per-operation cost and footprint (plus the memory diagnostics on the same gate) into the log. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

    @MaartenHilferink MaartenHilferink committed Aug 23, 2026
  • Drop the CalcCache from the docs: it was retired in 2022 The timings section explained warm runs by a cache that no longer exists (same misreading GeoDMS #1189 cleaned out of the engine docs). What the suite actually reuses is the per-round LocalDataDir root, and by default that is wiped at round start -- so a default round is cold, and only -resume (or -report-only) keeps it. The location tables named the old shared C:\LocalData\regression layout rather than runs\<run_id>. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

    @MaartenHilferink MaartenHilferink committed Aug 23, 2026
  • Document the full.py flags the table had drifted from, and the /SP window -sp (2026-08-23), -resume and -no-gui were never listed; -MT1..3 either. Point readers at `full.py -h` as the authoritative list. /SP was on every command line until 2026-08-23, but only bites from 20.15.0 where PerformanceLogging was introduced, so only 20.15.0+ folders measured before that date carry the overhead -- nothing published (through v20.12.0) does; in practice the unpublished 20.16.0/20.17.0 folders on OVSRV05. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

    @MaartenHilferink MaartenHilferink committed Aug 23, 2026
  • Add operational wiki: Linux-on-Windows, Claude runbook, references & report generation Capture hard-won insights from the regression-suite work: - Running-Linux-tests-on-Windows: WSL invocation, swap/reboot, per-test wsl --shutdown, drvfs write limits, GUI Qt 6.4-vs-6.10 blocker - Running-tests-with-Claude: detach long runs (session reaper), watchdog caveats, -tests substring trap, PowerShell pitfalls, integrity-first (don't mask regressions) - Test-references-and-report-generation: cross-version check before rebaselining, version-specific refs, Make_/ReadOnly guard, hollow-OK/false-pass, report rules Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

    @eoudejans eoudejans committed Jun 16, 2026