You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Windows Worker deployment convergence
Make Worker health probes use the same standard HTTP_PROXY, HTTPS_PROXY, and NO_PROXY routing model as the relay. The previous direct fetch probe could time out on Windows or managed networks after Wrangler had already completed a valid deployment. Restrict probes to the recorded HTTPS workers.dev origin and reject redirects so persisted state cannot redirect the verifier to an arbitrary network target.
Persist the successful Wrangler deployment URL, content/secret fingerprint, and deployed package version before the secondary health probe. A timeout, proxy failure, TLS failure, network failure, or temporary HTTP 5xx response now stops with corrective guidance and a later start verifies the same Worker instead of uploading it again; only bounded, definitive stale-version/identity evidence triggers automatic same-name redeployment.
Prevent accidental Worker proliferation by rejecting a changed --worker-name for an initialized workspace unless --force-worker explicitly authorizes the replacement. Retain prior names in local inventory so full uninstall can delete intentionally replaced Workers. Add a real local CONNECT-proxy regression, deployment ambiguity/idempotency tests, worker-name replacement tests, expanded CLI/state coverage, and higher critical-module coverage floors.