fix(server): allow inline styles in CSP so web math renders on non-loopback binds - #1847
Conversation
…opback binds The security CSP fell back to default-src 'self' for styles, which strips the inline style attributes that KaTeX (injected via innerHTML) uses for all glyph positioning — formulas collapsed into overlapping characters on any non-loopback-served web UI. Shiki highlighting and Mermaid diagrams hit the same mechanism. Add style-src 'self' 'unsafe-inline'; scripts remain strictly restricted.
🦋 Changeset detectedLatest commit: f3aa87c The changes in this PR will be included in the next version bump. This PR includes changesets to release 1 package
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
![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: dd79ebbb53
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".
| await hook({} as never, reply as never, 'payload'); | ||
| const csp = headers.get('content-security-policy'); | ||
| expect(csp).toContain("style-src 'self' 'unsafe-inline'"); | ||
| expect(csp).not.toContain("script-src 'self' 'unsafe-inline'"); |
There was a problem hiding this comment.
This negative substring check does not prove the behavior named by the test. If the policy later becomes default-src 'self' 'unsafe-inline', both this assertion and the following default-src 'self' assertion still pass even though inline scripts are then allowed through fallback; explicit variants such as script-src 'unsafe-inline' also evade this exact substring. Parse the directives and assert that the effective script-src (or its default-src fallback) excludes 'unsafe-inline'.
Useful? React with 👍 / 👎.
commit: |
A negative substring check only rejects one exact string: default-src gaining 'unsafe-inline' (inline <script> allowed via fallback) or an explicit script-src 'unsafe-inline' would both slip through. Parse the directives and assert the effective script policy (script-src, falling back to default-src) excludes 'unsafe-inline'/'unsafe-eval'/data:.
|
Codex Review: Didn't find any major issues. Another round soon, please! Reviewed commit: ℹ️ About Codex in GitHubCodex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
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". |
…opback binds (MoonshotAI#1847) * fix(server): allow inline styles in CSP so web math renders on non-loopback binds The security CSP fell back to default-src 'self' for styles, which strips the inline style attributes that KaTeX (injected via innerHTML) uses for all glyph positioning — formulas collapsed into overlapping characters on any non-loopback-served web UI. Shiki highlighting and Mermaid diagrams hit the same mechanism. Add style-src 'self' 'unsafe-inline'; scripts remain strictly restricted. * test(server): assert the effective script-src in the CSP regression test A negative substring check only rejects one exact string: default-src gaining 'unsafe-inline' (inline <script> allowed via fallback) or an explicit script-src 'unsafe-inline' would both slip through. Parse the directives and assert the effective script policy (script-src, falling back to default-src) excludes 'unsafe-inline'/'unsafe-eval'/data:.
…opback binds (MoonshotAI#1847) * fix(server): allow inline styles in CSP so web math renders on non-loopback binds The security CSP fell back to default-src 'self' for styles, which strips the inline style attributes that KaTeX (injected via innerHTML) uses for all glyph positioning — formulas collapsed into overlapping characters on any non-loopback-served web UI. Shiki highlighting and Mermaid diagrams hit the same mechanism. Add style-src 'self' 'unsafe-inline'; scripts remain strictly restricted. * test(server): assert the effective script-src in the CSP regression test A negative substring check only rejects one exact string: default-src gaining 'unsafe-inline' (inline <script> allowed via fallback) or an explicit script-src 'unsafe-inline' would both slip through. Parse the directives and assert the effective script policy (script-src, falling back to default-src) excludes 'unsafe-inline'/'unsafe-eval'/data:.
…opback binds (MoonshotAI#1847) * fix(server): allow inline styles in CSP so web math renders on non-loopback binds The security CSP fell back to default-src 'self' for styles, which strips the inline style attributes that KaTeX (injected via innerHTML) uses for all glyph positioning — formulas collapsed into overlapping characters on any non-loopback-served web UI. Shiki highlighting and Mermaid diagrams hit the same mechanism. Add style-src 'self' 'unsafe-inline'; scripts remain strictly restricted. * test(server): assert the effective script-src in the CSP regression test A negative substring check only rejects one exact string: default-src gaining 'unsafe-inline' (inline <script> allowed via fallback) or an explicit script-src 'unsafe-inline' would both slip through. Parse the directives and assert the effective script policy (script-src, falling back to default-src) excludes 'unsafe-inline'/'unsafe-eval'/data:.
1 participant
Related Issue
N/A — clear, reproducible bug with a focused fix; the problem is explained below.
Problem
When the server binds a non-loopback address (web UI accessed over LAN or a reverse proxy), LaTeX formulas in the chat render as garbled, overlapping characters.
Root cause: the security CSP sent on non-loopback binds is
default-src 'self'; …with nostyle-src, so style handling falls back todefault-src 'self'and the browser strips inlinestyle="…"attributes from injected markup. KaTeX output (rendered by the markdown renderer and injected via innerHTML) carries essentially all glyph positioning — vertical stacking, offsets, delimiter sizing — in inline style attributes (a single fraction formula contains ~68 of them), so formulas collapse into overprinted glyphs. Shiki code highlighting (inline colors) and Mermaid diagrams (inline<style>in the SVG) are degraded by the same mechanism. Loopback binds get no CSP header at all, which is why local development never surfaces this.What changed
style-src 'self' 'unsafe-inline'to the non-loopback CSP. Only styles are relaxed — script execution stays strictly restricted — which is the standard CSP posture for generated markup such as KaTeX/Shiki/Mermaid output. This restores math rendering, code-highlight colors, and Mermaid styling on network-served deployments.Checklist
gen-changesetsskill, or this PR needs no changeset.gen-docsskill, or this PR needs no doc update.