Nightly Jenkins runs collect Windows host metadata and pass it to benchmark.py parse via --machine-info. The dashboard System info block reads from data/run_environment.csv.

powershell -File scripts/collect_machine_info.ps1 -OutputPath machine_info.json
python scripts/benchmark.py parse <dir> `
  --run-id nightly-<commit>-<build> `
  --channel nightly `
  --commit-hash <hash> `
  --date <iso-date> `
  --machine-info machine_info.json

See scripts/machine_info.example.json for the JSON shape. Wired in status-app/scripts/push_benchmark.sh.

Channels

Nightly, pull-request, and release results are stored separately and never mixed:

nightly → data/
PR → data/desktop/pr/<number>/
release → data/desktop/releases/<series>/ (for example 2.39)

The manual Jenkins job (status-app/e2e/manual-benchmark-windows) publishes PR and RC runs. It does not promote a final release into the nightly baseline.

Promote a final release baseline

After a fully passed final build (for example 2.39.0, not an RC) is published, promote that run once by hand. Look up run_id in data/desktop/releases/2.39/runs.csv, then:

python scripts/benchmark.py promote-baseline --release-data-dir data/desktop/releases/2.39 --run-id <run_id>
python scripts/benchmark.py graphs --data-dir data --output-dir docs/desktop/nightly --channel nightly --baseline-dir data/desktop/baselines

An RC cannot be promoted. Each release series can be promoted only once.

Adding new tests

How to add a new performance test

Edit ./scripts/tests_config.toml:

[[tests]]
test_id = "test_my_feature"
display_name = "My Feature Loading Time Performance"
graph_filename = "my_feature_loading_time.png"
pattern = "test_my_feature"
ylabel = "Load Time (s)"

Add the test to a scenario page (or create a new one):

[[pages]]
slug = "my-feature"
title = "My Feature"
description = "Short description shown on the dashboard."
test_ids = ["test_my_feature"]

Android benchmarks

See docs/android/README.md for mobile navigation response time charts.

Raw CSV history: data/