Skip to content

ci: raise the PERF-01 budget 200 -> 400ms for the 79-pattern library #198

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
hermanngeorge15 merged 2 commits into from
Oct 21, 2026
Merged

ci: raise the PERF-01 budget 200 -> 400ms for the 79-pattern library #198

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2a57746
ci: raise the PERF-01 budget 200 -> 400ms for the 79-pattern library …
hermanngeorge15 Oct 20, 2026
a498a11
ci: rename the perf gate step to match the budget it enforces
hermanngeorge15 Oct 21, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 22 additions & 4 deletions .github/workflows/ci.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -166,7 +166,7 @@ jobs:
# the build. Measured at ~5ms of scan time locally and well under 100ms
# end-to-end on a hosted runner; the regressed build this defends against
# was 806ms on hardware faster than CI.
- name: Performance gate — 500 files under 200ms
- name: Performance gate — 500 files under 400ms
run: |
set -euo pipefail
corpus="$(mktemp -d)"
Expand Down Expand Up @@ -197,8 +197,26 @@ jobs:
if [ "$elapsed" -lt "$best" ]; then best=$elapsed; fi
done

echo "500-file scan: ${best}ms (budget 200ms)"
if [ "$best" -ge 200 ]; then
echo "::error::PERF-01 regression: 500 files took ${best}ms, budget is 200ms (issue #29)."
# Budget raised 200 -> 400ms on 2026-10-20. The 200ms figure was set in
# 20c88d1 (2026-08-21) against a 48-pattern library; v0.2.0 ships 79, a
# 65% increase in per-file matching work, and the budget was never
# revisited. It went unnoticed because the job was being cancelled at
# 30 minutes before ever reaching this step — raising that ceiling to 60
# exposed a gate that had already been failing.
#
# Measured: 213ms on a hosted runner (#188), and 172ms best-of-5 locally
# on arm64 with individual runs spanning 172-375ms. At 200ms the gate was
# nondeterministic rather than wrong: four Dependabot PRs on 2026-10-16
# that changed no Rust at all split two-pass/two-fail (#193 #195 vs
# #194 #196), and `main` itself was red from 2026-10-10.
#
# 400ms keeps the gate load-bearing. The regression it exists to catch is
# pattern compilation moving back inside the per-file loop, which cost
# 806ms at 48 patterns and would be ~1.3s at 79 — still several times over
# this budget. Re-derive this number whenever the pattern count moves
# materially; a budget that tracks no measurement is how this one rotted.
echo "500-file scan: ${best}ms (budget 400ms)"
if [ "$best" -ge 400 ]; then
echo "::error::PERF-01 regression: 500 files took ${best}ms, budget is 400ms (issue #29)."
exit 1
fi
Loading