[autofix] xiaohongshu/search: EMPTY_RESULT — page.evaluate envelope not unwrapped

[smallfoolh]

Summary

OpenCLI autofix repaired this adapter locally, and the retry passed.

Adapter

• Site: xiaohongshu
• Command: search
• OpenCLI version: 1.7.18

Original failure

• Error code: EMPTY_RESULT (silent — exit 0 with [], no error envelope)

The adapter at clis/xiaohongshu/search.js returns [] for every query, including queries that visibly have results in a logged-in Chrome tab. opencli xiaohongshu feed (intercept-mode) works fine in the same browser, ruling out auth/session issues.

Root cause

page.evaluate() returns the daemon exec envelope { session, data }, not the raw evaluated value. The adapter compares/iterates it directly:

• if (waitResult === 'login_wall') — never true because waitResult is {session, data: 'login_wall'}, so an actual login wall would be silently mis-classified.
• Array.isArray(payload) — always false because payload is {session, data: [...items...]}, so the 10+ extracted items fall through to the [] fallback.

Secondary issue: WAIT_FOR_CONTENT_JS resolves on the first section.note-item element, but xiaohongshu's virtual list renders section shells before their inner titles populate, racing autoScroll/extract against lazy fill-in. The trace-time state snapshot looks healthy because it's captured after the extract returned empty.

Local fix summary

1. Added an unwrap(v) helper that returns v.data when v looks like the {session, data} envelope, else v (tolerant of any pre-envelope return shape).
2. Applied it to both page.evaluate(WAIT_FOR_CONTENT_JS) and page.evaluate(buildSearchExtractJs(...)).
3. Tightened WAIT_FOR_CONTENT_JS to require ≥3 section.note-item elements with non-empty title text, with timeout raised 5s → 8s.

Verified locally: opencli xiaohongshu search "龙虾电脑" --limit 10 now returns 10 results.

Issue filed by OpenCLI autofix after a verified local repair.

[jackwener]

已由 #1518 修复并合入 main (2026-05-13)。

根因和你贴的一样：extension 的 pageScopedResult() 在 #1461 重构后给所有 exec 结果加了 {session, data} 信封，导致 Array.isArray(payload) 永远 false，提取出来的 10+ items 落回 []。

PR #1518 直接在 extension 端把 session 注入去掉（不再走 client-side unwrap workaround），extension bump 到 1.0.14。同时还做了两件相关的事：

• xiaohongshu/search 在 scroll 之前先 extract 一次再合并，避免虚拟化 masonry 在 scroll 后把初始 note cards 从 DOM 里 evict 掉（详见 clis/xiaohongshu/search.js 当前实现，PR fix(xiaohongshu+rednote): scroll until enough rows for --limit > 13 (#1471) #1487 也加了 buildScrollUntilJs 用来打破固定 autoScroll({times:2}) 的 13 条上限）。
• section.note-item class 偶尔被换掉时，fallback 到 href-based 选择器（PR [BUG] xiaohongshu search returns empty array due to outdated DOM selectors #1506 / fix(xiaohongshu+rednote/search): fall back to href-based note cards when section.note-item class is dropped #1507）。

升级到 OpenCLI 1.7.19 + Browser Bridge 1.0.14 即可。如果仍能复现请重开。