[http] --json collapses duplicate response headers and types body.bytes as a string #277

Activity

1. Antruly

   added

   enhancementNew feature or requestNew feature or request

   on Oct 19, 2026
2. Antruly commented on Oct 19, 2026

   

   Antruly

   on Oct 19, 2026

   OwnerAuthor

   More actions

   Review — checked at bc17f63 (uvp-agents 0.4.0)

   • Duplicate headers: confirmed. A local server sending two Set-Cookie and two X-Dupe lines yields:

     "headers": {
       "set-cookie": "a=1; Path=/, b=2; Path=/",
       "x-dupe": "one, two",
       ...
     }

     No raw list, and header name casing is lower-cased — both as described.

   • body.bytes is a number here, not a string. Same run: "bytes": 11. Against https://www.baidu.com: "bytes": 29506, type number. Forced HTTP/2: "bytes": 129, type number. The schema declares bytes: z.number().int() (src/contracts/webToolContracts.ts:104) and the projection is bytes: exchange.body.byteLength (src/application/WebService.ts:164).

     I could not reproduce the quoted "29506". If it is still visible, please add the exact command and version — it may come from a different path (the HTTP/3 native probe is the one that reads its byte count from a subprocess report) or an older build. My judgment: drop or re-scope the bytes half until it reproduces, and keep the header half, which does.

   The headers_raw suggestion is the right shape: headers stays the convenience view, and the ordered duplicate-preserving list is what a proxy consumer can actually reconstruct from.
3. Antruly commented on Oct 19, 2026

   

   Antruly

   on Oct 19, 2026

   OwnerAuthor

   More actions

   🤖 Antruly-dev-1 · AI Engineer 1 · agent:dev-ai-1

   Claiming this task as Antruly-dev-1. I will follow
   docs/ai-engineer-operations.md and docs/agent-workflow.md.
   Planned scope: add an ordered, duplicate- and case-preserving headers_raw list to uvp_http_request's output (name/value pairs as received), keeping headers as the convenience view, and leave the body.bytes half alone since your review could not reproduce it. The wire data has to come from the HTTP stack's own response object, so the first step is checking what libuvcpp exposes before deciding whether the full raw list is available or a narrower, truthful claim is the only one that can be made.
4. Antruly commented on Oct 19, 2026

   

   Antruly

   on Oct 19, 2026

   OwnerAuthor

   More actions

   🤖 Antruly-dev-1 · AI Engineer 1 · agent:dev-ai-1

   Status: in review — draft PR #285 (fix/277-headers-raw, head f32bc23).

   uvp_http_request now reports headers_raw for HTTP/1.1: wire order, original spelling, duplicates kept; headers stays the convenience view. HTTP/2 and HTTP/3 expose no raw list, so the field is absent there rather than an empty list — asserted against a real h2c session. The fixture has a /duplicate-headers route written as raw header pairs; two mutation probes are killed. The body.bytes half is not changed: your review could not reproduce it, and the schema and projection already type it as a number. Full suite 1674 passed / 14 named skips; the only failures are the two host-environmental real-toolchain files.
5. Antruly

   closed this as completedin #285on Oct 19, 2026