[ismailtrm]

Follow-up to #1035 (comment).

ResponseFormat keeps the escaping parameter only for text/plain; version=1.0.0. For protobuf and OpenMetrics it returns the bare format, so NewDecoder(body, ResponseFormat(h)) falls back to legacy name validation and rejects UTF-8 names that NewEncoder wrote with escaping=allow-utf-8:

Content-Type: application/openmetrics-text; version=1.0.0; charset=utf-8; escaping=allow-utf-8
ResponseFormat: application/openmetrics-text; version=1.0.0; charset=utf-8
Decode: text format parsing error in line 1: invalid metric name "http.server.request.duration"

Prometheus asks for escaping=allow-utf-8 on OpenMetrics 1.0.0 in its default Accept header, so a client that passes that header on and decodes the response this way runs into it.

This reads the parameter the same way for protobuf, text 1.0.0 and OpenMetrics 1.0.0 / 2.0.0, using the value from mime.ParseMediaType so quoted values work too. Without the parameter the returned format is unchanged. For OpenMetrics 2.0 this also means that an explicit scheme other than allow-utf-8 is now honoured instead of dropped. Protobuf is included because NegotiateAccept and the encoder already emit the parameter for it.

Added ResponseFormat table cases and an encode/decode test with a UTF-8 name. The new tests fail without the change, and go test ./... and golangci-lint pass with it.

@roidelapluie

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.