Conversation
|
Firetiger deploy monitoring skipped This PR didn't match the auto-monitor filter configured on your GitHub connection:
Reason: PR title and empty body suggest planning discussion rather than actual code changes to kernel API endpoints or Temporal workflows; please add deploy monitoring manually if this involves kernel API modifications. To monitor this PR anyway, reply with |
plans/cdp-decode.md
Outdated
|
|
||
| **Tests:** | ||
|
|
||
| - `cdp_proto_test.go` (NEW) — golden-fixture tests. For each Layer-1 struct, a captured real Chrome-emitted JSON frame in `testdata/`. Test decodes the frame, then re-marshals and asserts the round-trip preserves all top-level keys (catches dropped fields). This is the mechanical audit. |
There was a problem hiding this comment.
i think you want to compare normalized full JSON after decode/re-marshal and not just top level keys, to get a full structurual round trip check
you could also consider some light codegen around the json schema in json/js_protocol.json in the devtools repo
Note
Low Risk
Documentation-only change that proposes a future refactor; no runtime behavior, APIs, or tests are modified in this PR.
Overview
Adds a new design doc (
plans/cdp-decode.md) proposing a two-layer CDP decode approach: first decode wire events into PDL-faithful structs (retaining all fields, with complex sub-objects kept asjson.RawMessage), then project those typed structs into monitorevents.Eventpayloads.The plan outlines intended file changes (new
cdp_proto.go, handler signature shifts, centralized decode in dispatch) and a golden round-trip fixture test strategy to detect dropped/incorrect fields, but does not implement these code changes.Reviewed by Cursor Bugbot for commit b92a406. Bugbot is set up for automated code reviews on this repo. Configure here.