The workflow uses the standard ubuntu-latest GitHub-hosted runner. Standard runner compute is free for public repositories. GitHub Free includes 2,000 Actions minutes per month and 500 MB of artifact storage for private repositories; other repositories owned by the same account share the allowance. These terms were checked on 2026-09-21; consult GitHub's current billing documentation.
This workflow does not use larger paid runners, upload artifacts, enable caches, deploy a site, or publish a package. It uses four Python-version jobs, so private-repository minutes accrue across the matrix. Jobs have a 15-minute timeout and superseded runs are canceled. Those settings limit individual runs, not your account's total spend; account budgets control paid usage.
If the annotation says the job was not started because the account is locked due to a billing issue, GitHub prevented runner execution. That message does not establish a failing test, and changing Python code or deleting/recreating the repository does not resolve the underlying account restriction.
Check the repository owner's Settings → Billing and licensing for the actual account notice, quota, and budget status. If the lock is unexpected—especially for a public repository using standard runners—contact GitHub Support. The workflow cannot inspect or repair account billing. Do not assume that buying a paid plan is necessary based only on this message.
Once the account restriction is resolved, rerun the workflow. Until then, use the local commands in usage. You can disable the workflow from its Actions menu if you prefer local validation only.
Pushes, pull requests, and manual dispatch run the suite for Python 3.10–3.13. fail-fast: false lets each supported version finish when another fails. Each job installs development dependencies, runs make check, and builds wheel/sdist packages. Repository token permissions are read-only, and checkout does not persist credentials.
Local validation on Python 3.10 is recorded separately from hosted CI. An account-blocked run does not validate the other Python versions. Read the first failing step in a job's logs to diagnose an actual installation, test, lint, type, or build error once jobs are allowed to run.