Repository navigation
Conversation
This comment has been minimized.
This comment has been minimized.
|
Chat: Review posted | View chat Review historydeep-review v0.13.0 | Round 2 | Last posted: Round 2, 1 findings (1 Nit), APPROVE. Review Finding inventoryFindings
Contested and acknowledgedNone. Round logRound 1Full panel (round 1). Netero: no findings, 1 Note. Panel: ging-ts, ging-react, nami, killua, pariston, mafuuu, bisky, gon, leorio, plus wildcard razor (random draw). Size 10 exceeds the focused target of 6-8 because ging-ts, ging-react, nami and killua are triggered specialists, Bisky is required (story changed), and Gon and Leorio have a round 1 floor. The workspace disconnected while several reviewers ran the Storybook suite concurrently; pariston, mafuuu, razor and bisky were resumed on the same agents and wrote their own files. Bisky's draft P3 (story may pass without the Round 2Fix round (one commit, 4fd76a9, renames a predicate parameter; no files added or moved, no signature change). Churn guard PROCEED: CRF-1 addressed. Netero: no findings, verified CRF-1. Panel: gon (raised CRF-1), bisky (test file changed), ging-ts and ging-react ( About deep-reviewCRF = Coder Review Finding (P0-P4, Nit, Note)
|
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
This PR replaces the Radix ScrollArea around markdown code blocks with a native overflow-auto container and updates the LongLineFencedBlock story to find it. 1 Nit. LongLineFencedBlock passes at head and fails when the table min-w-full wrapper is replaced with a bare div, so the story covers the wrapper.
Out of scope (needs a ticket or explicit acceptance by a human):
site/src/pages/AgentsPage/components/ChatElements/tools/ToolCall.tsx:489:ToolCall.Scrollablestill mounts aScrollArea, so each tool row that expands or mounts inserts a document<style>and triggers the same WebKit whole-document restyle.site/src/pages/AgentsPage/components/ChatElements/tools/TerminalOutput.tsx:43:ExecuteTool.tsx:55-62opens execute output by default, so every execute row with output in history mounts aScrollAreaand inserts a<style>; shell-heavy chats keep the WebKit stall this PR removes for code blocks (not benchmarked).site/src/components/ScrollArea/ScrollArea.tsx:40: the shared wrapper inherits Radix's per-instance<style>, so all of its consumers pay the restyle cost on mount.
Notes:
- The PR description lists macOS as untested.
site/src/index.css:737-742states that any non-autoscrollbar-widthopts out of overlay scrollbars on macOS Chromium, so the permanent bar and the 38 to 48 px height change likely apply there too (inferred from that comment, not run on macOS).
🤖 This review was automatically generated with Coder Agents.
|
@codex review
|
|
Codex Review: Didn't find any major issues. Keep it up! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
There was a problem hiding this comment.
Re-review of 4fd76a9, which renames the story predicate parameter to el. The rename is verified and this round has no new findings. LongLineFencedBlock and the rest of Response.stories.tsx pass at head (18 of 18). The three out-of-scope ScrollArea items from the previous review still need a ticket or explicit acceptance by a human.
🤖 This review was automatically generated with Coder Agents.
4fd76a9 to
b65aa98
Compare
b65aa98 to
109d548
Compare
109d548 to
19a0f7e
Compare
19a0f7e to
9848f76
Compare
Radix ScrollArea renders a <style> element per instance, and WebKit restyles the whole document whenever a stylesheet is added or removed, so every streamed code block that mounted or unmounted cost a document-wide style recalculation. Code blocks now scroll in a plain overflow element that uses the global scrollbar color.
…e block scroller by its computed overflow
…oller predicate's element for what it holds
9848f76 to
2d657a1
Compare
Base: #29596 (
perf/chat-interaction). Second of five stacked fixes.Code blocks scrolled in Radix
ScrollArea, which renders a<style>per instance, and WebKit restyles the whole document when a stylesheet is added or removed. They now scroll in a plainoverflow-autoelement. Tool renderers still useScrollArea(ToolCall.tsx:505,TerminalOutput.tsx:43).Visual change: the old block showed a 6 px overlay bar on hover (Radix's default
type="hover").scrollbar-thingives a permanent 10 px bar with classic scrollbars: in Chromium on Linux an overflowing block grows from 38 to 48 px (Windows and macOS not tested).WebKit, #31278 bench (a tool, not for merging), production builds of the pre-rebase versions of
c0dd10843f(base) and2d657a137f(this PR), 2 streaming chats with the same 1,416 stored rows in every run (143k nodes on base, 140k here), 3 alternating runs:<style>inserts, phaseFull history: from
page.gotountil every stored message has a row and the inspector attaches; no run logged a stall or reload. Each recalculation over 1 s took 7.0 to 12.4 s. This PR's one began 1 s into a phase; its cause is not identified. The bench's frame counter misses a frame still running when the phase ends, so the longest-frame row reads frames from the timeline.Chromium, same builds and history, 3 alternating runs: full history loaded in 23.6 to 25.2 s on base and 21.8 to 22.5 s here.