GHSA-jrpm-956j-96jg

## Summary

Ember's chunk decoder parses the size token leniently: it strips leading and trailing whitespace and accepts a leading `+` or `-` sign. RFC9112 §7.1 defines `chunk-size = 1*HEXDIG`. An intermediary that parses the chunk boundary differently (or rejects it) will disagree with Ember on request framing, enabling HTTP request smuggling (TE.TE).

## Impact

### Server

Request smuggling (TE.TE) when ember-server is an origin behind an intermediary that forwards the body verbatim but interprets the malformed chunk sizes differently

- Front-end security bypass: the smuggled request reaches paths the intermediary might have blocked, with attacker-chosen method and headers. - Cache poisoning: the smuggled response is associated with the next request key in a caching proxy. - Request queue hijack:

### Client

ember-client has the same leniencies on the response path, which enables response smuggling when http4s is used as a gateway. This is less severe, as it requires a malicious or compromised upstream rather than an anonymous remote client.

## Preconditions

- Unauthenticated remote attacker (server) - ember-server as origin behind a keep-alive intermediary - Malicious or compromised upstream (client) - Intermediary or upstream forwards chunked bodies without re-encoding and disagrees with Ember on `+` prefix, `-` prefix, or leading and trailing octets rather than rejecting.

## Workarounds

- Intermediary strictly rejects malformed chunk sizes - Intermediary buffers and re-encodes request bodies

## References

- [RFC9112 §7.1: Chunked Transfer Coding](https://www.rfc-editor.org/rfc/rfc9112.html#name-chunked-transfer-coding)