Conversation
Add the demo/profile harness used to measure large diff rendering through the normal diffs package path. This branch establishes the clean autoresearch baseline without changing library rendering behavior. Experiments: #3 Metric: post_paint_ready_ms baseline 91.3ms
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d865535b25
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| addDuration(leafNodeId, durationUs, 'self'); | ||
|
|
||
| const visitedNodeIds = new Set<number>(); | ||
| let currentNodeId: number | undefined = leafNodeId; | ||
| while (currentNodeId != null && !visitedNodeIds.has(currentNodeId)) { |
There was a problem hiding this comment.
The CPU bottom-up totals are inflated because each sample’s leaf node is counted twice: once via addDuration(..., 'self') and then again when the ancestor walk starts from leafNodeId. Since addDuration always increments totalUs, leaf frames get double totalMs/totalPercent, which skews the reported CPU hotspots in every run with profiling data.
Useful? React with 👍 / 👎.
| continue; | ||
| } | ||
|
|
||
| const [flag, inlineValue] = rawArg.split('=', 2); |
There was a problem hiding this comment.
Using rawArg.split('=', 2) truncates inline values that themselves contain =, e.g. --url=http://127.0.0.1:5173/?a=b becomes http://127.0.0.1:5173/?a. This can silently rewrite profile target URLs (or other inline values), causing the CLI to hit the wrong page and produce misleading failures.
Useful? React with 👍 / 👎.
1 participant
Add the demo/profile harness used to measure large diff rendering through the normal diffs package path. This branch establishes the clean autoresearch baseline without changing library rendering behavior.
Experiments: #3
Metric: post_paint_ready_ms baseline 91.3ms