Baha Aiman

fix(firestore): retry transient connection errors during document reads (#20057) The Firestore client now retries transient connection errors during document reads (GetAll). Previously, if the BatchGetDocuments stream failed mid-stream due to a transient network issue (such as ECONNRESET, ECONNREFUSED, or unexpected EOF), the error was immediately returned to the user. Now, the client tracks outstanding documents and resumes the stream by requesting only the remaining documents. It will attempt to retry up to 5 times, applying backoff between attempts, and respecting the context deadline. This resumption strategy aligns with the behavior of the Java and Node.js Firestore SDKs. Transactional reads are not retried at this level to avoid interfering with transaction retry logic. Queries (`RunQuery`) are excluded from this retry mechanism. Mid-stream query resumption is significantly more complex (requiring tracking cursors, limits, and ordering) and is out of scope for the reported issue which specifically affects document gets (`GetAll`). Fixes #10350

feat(auth): populate http.response.status_code in http transport (#20053) The client-side metric gcp.client.request.duration was missing the http.response.status_code attribute when using the REST/HTTP transport. This change updates the otelAttributeTransport.RoundTrip method in the auth package to extract the HTTP status code from the response and set it on the TransportTelemetryData container in the context. This allows the gax layer to access the status code and record it in the client-side metrics. This change depends on the corresponding changes https://github.com/googleapis/gax-go/pull/513 in the gax-go package, which adds support for the httpStatusCode field in TransportTelemetryData. Fixes b/499375444 --------- Co-authored-by: Wes Tarle <westarle@google.com>