Skip to content

ci: align verify-global-install budget to script default (60s → 120s)#257

Open
theagenticguy wants to merge 1 commit into
mainfrom
ci/raise-gate4-install-budget
Open

ci: align verify-global-install budget to script default (60s → 120s)#257
theagenticguy wants to merge 1 commit into
mainfrom
ci/raise-gate4-install-budget

Conversation

@theagenticguy

Copy link
Copy Markdown
Owner

What

Verify Global Install workflow hardcoded MAX_INSTALL_SECS: "60", but scripts/verify-global-install.sh documents a 120s default and its header explicitly notes the tight 60s "tripped on slow cells despite a clean install." This change deletes the override so the workflow uses the script's documented 120s ceiling.

Why

The macos-x64-node24-nvm leg flaked on the last release push (chore: release main #256, 4c16f5b):

[FAIL] gate 4: install took 61s (> 60s budget)

A 1-second margin miss on a clean install — not a regression. The CI, CodeQL, Semgrep, Self-Scan, and Scorecard suites all passed on the same SHA; only this timing gate tripped.

The gate exists to catch an install that hangs or refetches the native prebuilts (ladybug + duckdb + onnxruntime — though note ADR 0019 has since removed ladybug/duckdb, leaving onnxruntime-node as the only native dep). On a loaded shared macOS runner a cold-cache global install legitimately varies 30–90s, so 60s was below the noise floor. 120s keeps the hang-detection intent while ending the margin flakes.

Blast radius

CI-only. One env value in one workflow step. No source or runtime change.

Fixes the recurring macOS-runner flake class tracked in #163.

🤖 Generated with Claude Code

The macos-x64-node24-nvm leg has been flaking on `gate 4: install took
61s (> 60s budget)` — a 1-second margin miss on a clean install, not a
real regression. scripts/verify-global-install.sh documents a 120s
default and its own header explains the tight 60s "tripped on slow cells
despite a clean install", but the workflow hardcoded MAX_INSTALL_SECS=60,
contradicting that. The gate exists to catch an install that HANGS or
refetches native prebuilts (ladybug + duckdb + onnxruntime), which on a
loaded shared runner legitimately varies 30-90s — it is not a perf
benchmark. Align the workflow to the script's 120s default.

Fixes the recurring macOS-runner flake class tracked in #163.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant