[innocarpe]

Summary

• Stop terminal HTTP links at CJK brackets, punctuation, and Unicode whitespace instead of swallowing adjacent prose.
• Reject non-ASCII characters while scanning the URL authority so missed boundaries cannot be folded into an IDNA hostname.
• Keep non-ASCII document paths and queries valid after the authority boundary.

Fixes #10571.

Why

Terminal URL extraction previously recognized only ASCII boundaries. Text such as a Japanese annotation directly after a GitHub URL could be included in the link, producing a percent-encoded or IDNA-transformed destination instead of the URL the user intended.

Scope and design

• The existing URL extraction and terminal hit-testing flow on main is preserved.
• Boundary predicates are isolated in terminal-http-url-boundary.ts so the scanner remains reviewable and below the module line budget.
• Unicode White_Space, punctuation, and symbol properties cover new scripts without an open-ended glyph list.
• Authority scanning stops at slash, query, or fragment; after that point, Unicode path/query text remains allowed.
• No network, filesystem, or URL-routing behavior changes beyond identifying the correct URL range.

Test plan

• CJK bracket, punctuation, and ideographic-space cases
• Unicode White_Space and property-class terminators
• Authority guard against bare CJK host text
• Non-ASCII path preservation
• Related terminal-link tests: 140 passed
• Typecheck
• Oxlint, React Doctor lint, Oxfmt, and diff checks
• Manual: paste the issue repro lines and verify only the URL is underlined/clickable

Screenshots

No visual chrome changes.

Security audit

The authority guard prevents missed Unicode boundaries from being silently IDNA-folded into a different hostname. The change adds no external requests, persistence, or new routing destination.

Platform review

The scanner uses standard JavaScript Unicode property escapes and URL parsing already supported by the desktop terminal path. The behavior is platform-independent across macOS, Linux, and Windows.

ELI5

If a terminal prints a URL followed immediately by Japanese or other Unicode notes, Orca now links only the real URL. Unicode text inside the URL path still works.

[coderabbitai]

📝 Walkthrough

Walkthrough

Adds shared URL boundary predicates for ASCII whitespace, delimiters, trailing punctuation, and CJK/fullwidth punctuation. Terminal URL hit testing now imports these predicates instead of local implementations. Tests cover URL termination around fullwidth parentheses, CJK corner brackets, ideographic spaces, and half-width separators, including extracted URL spans.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	The changes directly implement #10571 by adding CJK whitespace, bracket, and punctuation boundaries plus regression tests.
Out of Scope Changes check	✅ Passed	The diff stays focused on terminal URL boundary logic and tests, with no unrelated feature or refactor changes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Title check	✅ Passed	The title clearly and concisely describes the main change: terminating terminal HTTP links at CJK brackets and whitespace.
Description check	✅ Passed	The description covers the user-visible change, scope, tests, screenshots, security, and platform behavior, but omits separate AI Review Report and Notes sections.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

src/renderer/src/components/terminal-pane/terminal-url-link-hit-testing.test.ts (1)

63-81: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Add coverage for trailing CJK punctuation.

These tests cover bracket and whitespace boundaries, but not isCjkUrlTrailingPunctuation for characters such as 。, 、, ！, or ？. Add a case asserting both the extracted URL and endIndex exclude trailing punctuation.

Based on the PR objective to cover trailing CJK punctuation, this path needs a direct regression case.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 7e9cbe17-e47c-4d28-b873-7247211b83d2

📥 Commits

Reviewing files that changed from the base of the PR and between 505967e and 2a4bb28a2a514c571dc2350ebb856a09ffc9bfcd.

📒 Files selected for processing (3)

• src/renderer/src/components/terminal-pane/terminal-http-url-boundary.ts
• src/renderer/src/components/terminal-pane/terminal-url-link-hit-testing.test.ts
• src/renderer/src/components/terminal-pane/terminal-url-link-hit-testing.ts

[AmethystLiang]

Thanks for turning this around so fast — the diagnosis in #10571 is spot on, the ASCII-only predicates are exactly the bug.

I dug in before approving and think the fix needs a different shape. Running this PR's predicates against a wider case set, the punctuation-only half of isCjkUrlTrailingPunctuation — 、。・…，．！？：；, everything it adds beyond isCjkUrlBracketOrQuote — never actually fires:

line	this PR resolves to
https://example.com、それから	https://example.xn--com-6y3b8ordxnwa/
https://example.com。次の行	https://example.com.xn--u9j508ocks/
https://example.com・次	https://example.xn--com-rs4b4090a/

Trailing-trim only strips from the end of an already-terminated candidate. 、 isn't in the body-terminator set and CJK letters don't terminate either, so findHttpUrlCandidateEnd runs to end of line; the last character is ら, which isn't trailing punctuation, so the trim loop strips nothing. The three lines in the issue pass because （ and 「 are real body terminators — the ・ in （作成済み・マージ待ち） never gets exercised.

Worth flagging what that output actually is: with no path, the WHATWG parser folds the CJK into the authority and IDNA-encodes it, so the link silently retargets to a different, registrable domain rather than 404ing. I confirmed it against the live module in a running renderer. That makes this a link-spoofing surface, which is why I want to be careful here.

CodeRabbit's one nitpick on this PR was "add coverage for trailing CJK punctuation (。, 、, ！, ？)" — that test is exactly what would have surfaced it.

Two changes I'd make:

1. Move CJK punctuation into the body-terminator set, and go property-based — \p{White_Space}, \p{P} and \p{Sm}, scoped to code points above U+007F — instead of enumerating. This is the third boundary bug in this scanner after PR auto-link fails for Claude Code PRs: ANSI reset code glued to URL breaks terminal scanner #8126 and [Bug]: Cmd-click on a URL opens it with the next line's text glued on #8832, and each enumeration only covers what someone thought of: 〖〗 and ｟｠ are missing from the bracket list here, and 〜 U+301C is Pd while ～ U+FF5E is Sm, so even "all the CJK punctuation" wouldn't have caught them. I got that wrong myself on the first pass and had to check the categories rather than assume them.
2. Add a hostname guard so the authority case can't happen at all, independent of which terminators are in the set.

There's also a second detector in play: pane-dom-creation.ts constructs WebLinksAddon with no urlRegex, so the hover underline and tooltip come from xterm's default regex while the click re-resolves through extractTerminalHttpLinks. The two disagree in both directions today, which is why the underline stays wrong even with this patch applied.

That adds up to a good deal more than this PR signed up for, so I'm going to pick it up in a follow-up with you as co-author rather than ask you to build the whole thing. Hope that's alright — the report and repro were genuinely useful and made this much faster to work through.

[innocarpe]

Sync update (7d9e85301)

Address Amethyst review: property-based non-ASCII terminators + hostname guard against IDNA spoof; broader Unicode whitespace.

• Commit: 7d9e85301e0160d4bcef69a5fa58f1c1dbb0d16e
• Upstream: fix(terminal): terminate HTTP links on CJK brackets and whitespace #10573
• Fork portfolio: [upstream #10573] fix(terminal): terminate HTTP links on CJK brackets and whitespace innocarpe/orca#73

[innocarpe]

Closing in favor of merged #15240, which already closes #10571.

Current main now terminates HTTP URLs at CJK punctuation / ideographic space in terminal-http-url-extraction.ts (NON_ASCII_URL_TERMINATOR). That is the same user-facing gap this PR targeted. Thank you @nwparker.