[docs] Fix style prop docs generation

[mj12albert]

In BaseUIComponentProps, style isn't omitted like className even though we redeclared it:

base-ui/packages/react/src/utils/types.ts

Line 39 in 350b6b1

'className' | 'color' | 'defaultValue' | 'defaultChecked'

It should also be excluded from elementProps like className when passing to useRenderElement

Though this doesn't seem to cause any actual type errors but in the docs the description is missing:

• I have followed (at least) the PR section of the contributing guide.

[pkg-pr-new]

• base-ui-tanstack-start

• vite-css-base-ui-example

  pnpm add https://pkg.pr.new/mui/base-ui/@base-ui/react@4406
  

  pnpm add https://pkg.pr.new/mui/base-ui/@base-ui/utils@4406
  

commit: b5f85c0

[mui-bot]

Bundle size report

Bundle	Parsed size	Gzip size
@base-ui/react	🔺+1.69KB(+0.37%)	🔺+490B(+0.34%)

Details of bundle changes

---

Check out the code infra dashboard for more information about this PR.

[netlify]

✅ Deploy Preview for base-ui ready!

Name	Link
🔨 Latest commit	b5f85c0
🔍 Latest deploy log	https://app.netlify.com/projects/base-ui/deploys/69c87804a5fc6500081f0453
😎 Deploy Preview	https://deploy-preview-4406--base-ui.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[atomiks]

Codex Review (GPT-5.4)

1. Bugs / Issues (None)

No concrete correctness bug stood out in the final branch diff.

No blocking issues found in this patch.

2. Pattern Consistency Assessment

Aspect	Consistent?	Notes
Central prop-model change	Yes	Moving the style docs fix into the shared BaseUIComponentProps contract is the right center of gravity for the change
Generated reference output	Yes	The generated API files now consistently describe style instead of leaving it blank
Composite follow-up	Yes	The fix composites commit restores style forwarding in the internal composite path and applies it consistently at the touched call sites

3. Simplification Opportunities

No meaningful simplification change stands out here. The large file count is mostly generated or mechanical churn driven by a small central typing/docs fix.

4. Edge Cases Worth Noting

Edge Case	Status
Function-valued style prop in generated docs	Handled
Composite-backed wrappers after destructuring style	Handled by follow-up
Slider experiment spreading a function-valued style prop	Handled

5. Test Coverage Assessment

Strengths:

• Ran pnpm test:jsdom useRenderElement --no-watch; it passed.
• Ran pnpm test:jsdom RadioGroup --no-watch; it passed.
• The generated JSON and snapshot deltas line up with the intended docs output change.

Gaps:

• One focused regression test in useRenderElement.test.tsx would make this stronger: pass style as a function and verify the callback is resolved to a style object rather than leaking a raw function into the DOM, since that is the shared contract this patch is formalizing.

Merge Confidence Score

Dimension	Score	Notes
Docs-generation correctness	4/5	The central type change and generated output look internally consistent
Regression risk	3/5	The broad mechanical churn and follow-up composite fix raise the value of one targeted regression test
Verification	4/5	Targeted utility and component tests passed

Overall merge confidence is 4/5. This looks safe to merge, and one small composite-style regression test would make the follow-up fix more durable.

[flaviendelangle]

0.37% bundle size increase for that kind of fix is so sad 😢
Could we imagine a way to not omit those props in every component but instead do it in useRenderElement or would that add to much performance overhead?

[mj12albert]

not omit those props in every component but instead do it in useRenderElement

@flaviendelangle Tried this with codex and here's the comparison:

The prototype was validated in a temp worktree with pnpm typescript, pnpm test:jsdom useRenderElement --no-watch, and pnpm size:snapshot.
The bundle result was mixed:

• @base-ui/react barrel entry got smaller: 468,552 -> 467,295 parsed (-0.268%), 146,960 -> 146,666 gzip (-0.200%).
• The full local size snapshot across all measured entrypoints got slightly worse overall: +1,801 parsed (+0.075%) and +440 gzip (+0.054%).
• Distribution matters more than the summed total: 35 entrypoints got bigger, 4 smaller, 40 were unchanged. Many leaf entries moved by about +110 to +130 parsed bytes and +24 to +42 gzip bytes.

Interpretation: the alternative helps the big barrel import a bit, but it pushes extra shared logic into many leaf entrypoints.
Conclusion: If the goal is minimizing leaf entrypoint size and avoiding extra work in a hot shared helper, the current explicit destructuring is the better trade.

[atomiks]

I think these files were intentionally removed by docs-infra upgrades? @mj12albert