Retire app-boot helpers used only by fixtures
删除仅供 fixture 使用的 app-boot 辅助函数
[App boot](../../../../packages/boot/app-boot/src/index.ts) exports `resolveConfigPath`, including a basename-specific replay substitution, and `loadEnv`, which reads one optional `.env`. The supported [CLI launcher](../../../../apps/cli/src/bin.ts) uses `loadLayeredEnv`; [profile boot](../../../../apps/cli/src/profile-boot.ts) supplies an absolute path directly. Neither production path calls the older helpers. An ex
English
Problem
App boot exports resolveConfigPath, including a basename-specific replay substitution, and loadEnv, which reads one optional .env. The supported CLI launcher uses loadLayeredEnv; profile boot supplies an absolute path directly. Neither production path calls the older helpers.
An exact-symbol search across packages, apps, and scripts finds ten resolveConfigPath calls in eight fixture drivers, all passing undefined for replay mode. The only non-unit loadEnv caller is the Loader smoke fixture. The snapshot launcher separately owns replay patch selection. The old helpers consequently retain two public APIs, unused production replay policy, and dedicated tests for fixture-only needs.
The shared app-bin glue record explains the original shared-bin use. The active single-launcher decision replaces those supported entry paths. This is partial supersession: boot and fail-loud handling still have owners, and the archived record remains frozen.
Proposal
Delete resolveConfigPath and loadEnv from the production export. Use Node's path.resolve in the eight fixture drivers. Keep optional one-directory .env handling local to the sole fixture that needs it, including its absent-file and warning behavior; do not introduce a replacement production package.
Remove helper-specific tests and update app-boot's package description, README pair, and boot JSDoc. Keep boot, loadLayeredEnv, shipped profile composition, and snapshot replay selection. The identified removal is 39 production source/JSDoc lines and about 73 dedicated test lines, less the small fixture-local environment handling and builtin imports; the estimate is not an implemented diff.
Alternatives considered
Keep public conveniences for low-level embedders. They can be useful, but current fixtures request only path resolution and one optional file read. Maintaining a second production replay selector and obsolete app-bin guidance is disproportionate to those uses.
Replace every use with loadLayeredEnv. Rejected because layered discovery, filtering, and precedence differ from reading one fixture-local .env. Sharing a name does not establish equivalent behavior.
Acceptance criteria
- Exact-name searches find no live helper imports or calls. All ten path replacements remain absolute and select the same files.
- The sole environment-loading fixture preserves missing-file, present-file, and warning behavior; helper-only tests disappear without dropping retained boot coverage.
- Run focused app-boot and affected fixture-entry tests, including built entry checks where required, a shipped profile smoke, and a keyless replay scenario.
- Update README pairs and package/JSDoc descriptions; run build, relevant hygiene, typecheck, doc-sync, and lint. Add no compatibility aliases or second replay selector.
Risks
Out-of-tree callers importing these pre-stable helpers must use the retained profile APIs or own their low-level path/environment policy. Accidentally switching the fixture to layered environment discovery would change its test inputs; preserve that distinction explicitly.
中文
问题
App boot 导出 resolveConfigPath，其中包含按文件名替换重放配置的逻辑，以及读取单个可选 .env 的 loadEnv。受支持的 CLI 启动器 使用 loadLayeredEnv；profile boot 直接提供绝对路径。这两条生产路径都不调用旧辅助函数。
在 packages、apps 和 scripts 中搜索精确符号，可找到八个 fixture（测试前置数据）驱动器中的十次 resolveConfigPath 调用，重放模式全部传入 undefined。唯一非单元测试的 loadEnv 调用方是 Loader 冒烟 fixture。快照启动器 单独拥有重放补丁选择。因此，旧辅助函数仅为 fixture 的需要保留了两个公共 API、生产中未使用的重放策略及专用测试。
共享 app-bin 胶水代码记录 解释了最初的共享可执行入口用途。当前单一启动器决策 已替代这些受支持入口。这属于部分取代：boot 和明确失败处理仍有归属，归档记录保持冻结。
提案
从生产导出中删除 resolveConfigPath 和 loadEnv。八个 fixture 驱动器改用 Node 的 path.resolve。仅在需要它的 fixture 内保留读取单目录可选 .env 的处理，包括文件缺失与警告行为；不引入替代生产包。
删除辅助函数专用测试，并更新 app-boot 的包描述、README 配对及 boot JSDoc。保留 boot、loadLayeredEnv、随产品发布的 profile 组合和快照重放选择。已定位可删除 39 行生产源码/JSDoc 和约 73 行专用测试，需扣除少量 fixture 本地环境处理及内置模块导入；这是估算，尚非实现 diff。
考虑过的替代方案
为底层嵌入方保留公共便利函数。 它们可能有用，但当前 fixture 仅需要路径解析和单个可选文件读取。为这些用途维护第二套生产重放选择器及过时的 app-bin 指南，成本过高。
将所有调用替换为 loadLayeredEnv。 否决，因为分层发现、过滤和优先级不同于读取单个 fixture 本地 .env。名称相近不能证明行为等价。
验收标准
- 精确名称搜索不再发现有效的辅助函数导入或调用。十处路径替换仍为绝对路径并选择相同文件。
- 唯一加载环境的 fixture 保持文件缺失、文件存在和警告行为；删除辅助函数专用测试时不丢失保留的 boot 覆盖。
- 运行聚焦的 app-boot 和受影响 fixture 入口测试，包括所需的构建产物入口检查、随产品发布的 profile 冒烟，以及无密钥重放场景。
- 更新 README 配对及包/JSDoc 描述；运行 build、相关 hygiene、typecheck、doc-sync（文档同步门禁）和 lint。不添加兼容别名或第二个重放选择器。
风险
导入这些稳定前辅助函数的仓库外调用方，需要改用保留的 profile API，或自行拥有底层路径与环境策略。意外将 fixture 改为分层环境发现会改变其测试输入；必须明确保留这一差别。