# needs-benchmarks decision — before/after, against four observed PRs
# 2026-08-12T10:38:01Z  branch jongsun/ci/benchmark-paths-force-run
# logic extracted verbatim from get-requirements.yml set-needs-benchmarks

ok   #45443 benchmark mocks, builds reused        before=false after=true  (benchmark-owned files changed)
ok   #45444 gated-metrics + test, builds reused   before=false after=true  (benchmark-owned files changed)
ok   #45445 workflow + script, builds reused      before=false after=true  (benchmark-owned files changed)
ok   #45455 workflow + script, own build          before=true  after=true  (fresh builds)
ok   CONTROL app-only change, builds reused       before=false after=false (builds from base branch)
ok   CONTROL app-only change, own build           before=true  after=true  (fresh builds)

6/6 match expectation

## Observed values these cases encode (from the live PRs, 2026-08-12):
  #45443 builds-from-run=31579703000 (a main run)  run-benchmarks=skipped
  #45444 builds-from-run=31579703000 (a main run)  run-benchmarks=skipped
  #45445 builds-from-run=31579703000 (a main run)  run-benchmarks=skipped
  #45455 builds-from-run=31585190838 (its own)     run-benchmarks=ran