diff --git a/vendor/dsh-remote-ssh/tests/remote-search.test.mjs b/vendor/dsh-remote-ssh/tests/remote-search.test.mjs new file mode 100644 index 0000000..f6cdeca --- /dev/null +++ b/vendor/dsh-remote-ssh/tests/remote-search.test.mjs @@ -0,0 +1,193 @@ +// 远程文件名搜索的两趟策略 / 预算 / 缓存回归测试(issue #13 实测跟进)。 +// +// 背景:侧栏「按文件名搜索」在真实的大型远程项目工作区一直「加载中…」——旧命令 +// `find -name '*q*' -printf '%p\n' | sort | head -n 500` +// 有两个致命点:① `sort` 缓冲全部 find 输出,`head` 的提前短路失效;② 无深度/时间约束, +// 而 find 是深度优先,巨型子目录会吃光预算,连顶层文件都轮不到(实测 5 分钟无输出)。 +// 实测该工作区:顶层 12 个条目、命中集中在 depth 1–3(0.02–0.42s),depth 4 起跳到 3s+。 +// +// 现在:浅层趟(maxdepth 3 / 3s 预算)→ 未找满再深挖趟(maxdepth 8 / 5s)→ 合并去重排序, +// 全程无 sort、剪噪声目录、timeout 到点返回部分结果并标记 truncated,另加同 query 缓存与并发合并。 +import { readFileSync } from 'node:fs' + +const root = new URL('..', import.meta.url).pathname.replace(/^\/([A-Za-z]:)/, '$1') +const src = readFileSync(root + '/lib/index.js', 'utf8') + +function grab(name) { + const start = src.indexOf(name) + if (start < 0) throw new Error(name + ' not found') + const declStart = src.lastIndexOf('\n', start) + let i = src.indexOf('{', start) + let depth = 0 + for (; i < src.length; i++) { + if (src[i] === '{') depth++ + else if (src[i] === '}') { depth--; if (depth === 0) break } + } + return src.slice(declStart + 1, i + 1) +} +const pick = (re) => { const m = src.match(re); if (!m) throw new Error('missing ' + re); return m[0] } + +const code = [ + pick(/const MAX_BYTES = [^;]+;/), + pick(/const REF_EXCLUDED_DIRS = new Set\(\[[\s\S]*?\]\);/), + pick(/const REMOTE_SEARCH_SKIP_DIRS = new Set\(\[[\s\S]*?\]\);/), + pick(/const REF_SEARCH_MAX_MATCHES = \d+;/), + pick(/const REF_SEARCH_SHALLOW_DEPTH = \d+;/), + pick(/const REF_SEARCH_SHALLOW_BUDGET_SEC = \d+;/), + pick(/const REF_SEARCH_DEEP_DEPTH = \d+;/), + pick(/const REF_SEARCH_DEEP_BUDGET_SEC = \d+;/), + pick(/const REF_SEARCH_TIMEOUT_MS = \d+;/), + pick(/const REF_SEARCH_TTL_MS = \d+;/), + pick(/const remoteSearchCache = new Map\(\);/), + pick(/const remoteSearchInflight = new Map\(\);/), + grab('function shellQuote'), + grab('function shellQuotePath'), + grab('async function remoteGlob'), + grab('function profileKey'), + grab('function cacheEpoch'), + grab('function bumpCacheEpoch'), + pick(/const cacheEpochs = new Map\(\);/), + grab('async function remoteSearch'), +].join('\n') + +const api = new Function(code + ` + return { remoteGlob, remoteSearch, remoteSearchCache, remoteSearchInflight, bumpCacheEpoch, profileKey, + REF_SEARCH_MAX_MATCHES, REF_SEARCH_SHALLOW_DEPTH, REF_SEARCH_DEEP_DEPTH, REMOTE_SEARCH_SKIP_DIRS };`)() + +let pass = 0, fail = 0 +const check = (cond, label) => { if (cond) { pass++; console.log(' ✓ ' + label) } else { fail++; console.log(' ✗ FAIL: ' + label) } } + +const ROOT = '~/proj' +const PROFILE = { id: 'p1', host: 'h', user: 'u' } + +/** 可编排的桩 runner:按调用序返回预设结果,并记录命令。 */ +function stubRunner(results) { + const cmds = [] + let i = 0 + return { + cmds, + runner: async (p, cmd) => { cmds.push(cmd); const r = results[Math.min(i, results.length - 1)]; i++; return r }, + } +} +const info = (cmd) => ({ + maxdepth: (/-maxdepth (\d+)/.exec(cmd) || [])[1], + budget: (/"\$T" (\d+) /.exec(cmd) || [])[1], + sort: cmd.includes('| sort'), + fmt: cmd.includes("'%P") ? '%P' : (cmd.includes("'%p") ? '%p' : '?'), + head: (/head -n (\d+)/.exec(cmd) || [])[1], + prune: cmd.includes('-prune'), + guard: cmd.includes('command -v timeout'), +}) + +console.log('A · 命令生成(两趟)') +{ + const s = stubRunner([{ ok: true, stdout: '' }]) + await api.remoteSearch(s.runner, PROFILE, ROOT, 'solver') + check(s.cmds.length === 2, '未找满时跑两趟(浅层 + 深挖)') + const a = info(s.cmds[0]), b = info(s.cmds[1]) + check(a.maxdepth === String(api.REF_SEARCH_SHALLOW_DEPTH), `浅层趟 -maxdepth ${a.maxdepth}`) + check(b.maxdepth === String(api.REF_SEARCH_DEEP_DEPTH), `深挖趟 -maxdepth ${b.maxdepth}`) + check(a.budget === '3' && b.budget === '5', `两趟墙钟预算 ${a.budget}s / ${b.budget}s`) + check(!a.sort && !b.sort, '两趟都不用 sort(head 才能提前短路)') + check(a.fmt === '%P' && b.fmt === '%P', '两趟都用 %P(cwd 相对路径,matches 契约)') + check(a.head === String(api.REF_SEARCH_MAX_MATCHES), `head -n ${a.head}(与上游 200 一致)`) + check(a.prune && b.prune, '两趟都剪噪声目录(-prune)') + check(a.guard && b.guard, '两趟都有 timeout/gtimeout 探测与回退分支') + check(s.cmds[0].includes("-name '.git'") && s.cmds[0].includes("-name 'node_modules'"), '剪枝表含 .git / node_modules') + check(s.cmds[0].includes("-name '.pnpm-store'") && s.cmds[0].includes("-name '.umi-production'"), '剪枝表含上游 SEARCH_SKIP_DIRS 的条目') + check(s.cmds[0].includes('|| command -v gtimeout'), 'macOS 回退到 gtimeout') +} + +console.log('B · 两趟合并与降级(延迟策略:有结果即返回,深挖后台预热)') +{ + api.remoteSearchCache.clear() + const s = stubRunner([ + { ok: true, stdout: 'a.jl\nb/c.jl\n' }, + { ok: true, stdout: 'b/c.jl\nd/e/f.jl\n' }, + ]) + const r = await api.remoteSearch(s.runner, PROFILE, ROOT, 'q1') + check(r.ok && r.files.length === 2, `浅层有结果 → 立即返回浅层 ${r.files.length} 条(不等深挖)`) + check(JSON.stringify(r.files) === JSON.stringify(['a.jl', 'b/c.jl']), '返回内容即浅层结果(本地排序)') + check(r.truncated === false, '未触预算 → truncated false') + await new Promise((res) => setTimeout(res, 5)) // 等后台预热落地 + const warmed = await api.remoteSearch(s.runner, PROFILE, ROOT, 'q1') + check(warmed.files.length === 3, `后台预热写回缓存:同一 query 再查得到合并后的 ${warmed.files.length} 条`) + check(JSON.stringify(warmed.files) === JSON.stringify(['a.jl', 'b/c.jl', 'd/e/f.jl']), '合并结果去重且排序') + + api.remoteSearchCache.clear() + const empty = stubRunner([ + { ok: true, stdout: '' }, + { ok: true, stdout: 'deep/only.jl\n' }, + ]) + const re = await api.remoteSearch(empty.runner, PROFILE, ROOT, 'q-empty') + check(re.files.length === 1 && re.files[0] === 'deep/only.jl', '浅层零命中 → 同步等深挖(用户此时没别的可看)') + check(empty.cmds.length === 2, '零命中时两趟都在响应前完成') + + api.remoteSearchCache.clear() + const full = stubRunner([{ ok: true, stdout: Array.from({ length: api.REF_SEARCH_MAX_MATCHES }, (_, i) => 'f' + i).join('\n') }]) + const rf = await api.remoteSearch(full.runner, PROFILE, ROOT, 'q2') + check(full.cmds.length === 1, `浅层已满 ${api.REF_SEARCH_MAX_MATCHES} 条 → 不深挖也不预热(无法再展示更多)`) + check(rf.truncated === true, '满额 → truncated true') + + api.remoteSearchCache.clear() + const budget = stubRunner([ + { ok: false, exitCode: 124, stdout: 'p.jl\n' }, // 浅层被预算杀掉,但有部分输出 + { ok: false, exitCode: 124, stdout: 'q/r.jl\n' }, // 后台预热同样 + ]) + const rb = await api.remoteSearch(budget.runner, PROFILE, ROOT, 'q3') + check(rb.ok && rb.files.length === 1, '预算到点:立即返回已收集的部分结果(不再无限加载)') + check(rb.truncated === true, '部分结果标记 truncated(客户端可提示不完整)') + + api.remoteSearchCache.clear() + const dead = stubRunner([{ ok: false, exitCode: 1, stdout: '', error: 'ssh down' }]) + const rd = await api.remoteSearch(dead.runner, PROFILE, ROOT, 'q4') + check(rd.ok === false && rd.files.length === 0, '硬失败且无输出 → ok:false(不缓存)') + const deadKey = api.profileKey(PROFILE) + '|' + ROOT + '|q4' + check(!api.remoteSearchCache.has(deadKey), '失败结果不进缓存(按 key 判定,避免后台预热干扰)') + + api.remoteSearchCache.clear() + const half = stubRunner([ + { ok: false, exitCode: 1, stdout: '', error: 'boom' }, + { ok: true, stdout: 'deep/only.jl\n' }, + ]) + const rh = await api.remoteSearch(half.runner, PROFILE, ROOT, 'q5') + check(rh.ok === false && rh.files.length === 0, '浅层硬失败(无任何输出)→ ok:false') + check(half.cmds.length === 1, '硬失败不再深挖(连接已坏,省一趟往返;预算到点带部分输出时才继续)') +} + +console.log('C · 缓存与并发合并') +{ + api.remoteSearchCache.clear() + const s = stubRunner([{ ok: true, stdout: 'x.jl\n' }, { ok: true, stdout: '' }]) + const r1 = await api.remoteSearch(s.runner, PROFILE, ROOT, 'same') + const callsAfterFirst = s.cmds.length + const r2 = await api.remoteSearch(s.runner, PROFILE, ROOT, 'same') + check(s.cmds.length === callsAfterFirst, '相同 query 30s 内直接命中缓存(不再发起 SSH)') + check(JSON.stringify(r1.files) === JSON.stringify(r2.files), '缓存结果一致') + + api.bumpCacheEpoch(PROFILE) + await api.remoteSearch(s.runner, PROFILE, ROOT, 'same') + check(s.cmds.length > callsAfterFirst, 'epoch 变更(写/exec 后)→ 缓存失效并重新遍历') + + api.remoteSearchCache.clear() + const conc = stubRunner([{ ok: true, stdout: 'a\n' }, { ok: true, stdout: '' }]) + const [c1, c2, c3] = await Promise.all([ + api.remoteSearch(conc.runner, PROFILE, ROOT, 'race'), + api.remoteSearch(conc.runner, PROFILE, ROOT, 'race'), + api.remoteSearch(conc.runner, PROFILE, ROOT, 'race'), + ]) + check(conc.cmds.length === 2, `三个并发同 query 只跑一趟遍历(实际 ${conc.cmds.length} 次命令 = 浅层+深挖)`) + check(c1.files.length === 1 && c2.files.length === 1 && c3.files.length === 1, '并发调用都拿到同一结果') + check(api.remoteSearchInflight.size === 0, 'in-flight 表已清理(无泄漏)') +} + +console.log('D · 接线') +{ + const branch = src.slice(src.indexOf('} else if (method === "fs.search") {'), src.indexOf('} else if (method === "fs.rename") {')) + check(/await remoteSearch\(runPooled, profile, rp, payload\.query \|\| ""\)/.test(branch), 'fs.search 远程分支改走 remoteSearch(两趟 + 缓存)') + check(!/remoteGlob\(runPooled, profile, "\*"/.test(branch), '不再直接调用 remoteGlob(避免无预算的单趟遍历)') + check(/fsSearchResult\(matches/.test(branch), '仍经 fsSearchResult 产出 matches 契约(issue #13 主修复不回退)') +} + +console.log(`\n结果: ${pass} 通过, ${fail} 失败`) +process.exit(fail ? 1 : 0)