start:dev now aborts on local R2 seed failure and status:dev probes the newest catalog tool page so home-200 plus new-tool-404 cannot look healthy.

Date: 2026-09-01 13:24 Summary: start:dev now aborts on local R2 seed failure and status:dev probes the newest catalog tool page so home-200 plus new-tool-404 cannot look healthy.

[question] 启动测试本地流程中修正这个bug

[try to solve] The local bug was: build:site wrote new _pages HTML, then upload:r2:local crashed, but start-dev only warned and still started wrangler. Home and old tools stayed 200 from Assets / the old mock bucket; new slugs 404’d. --no-build also skipped seed entirely.

Changes in the start/test flow:

runSeedLocalR2 throws; wrangler is not started if seed fails (--no-seed-r2 still skips on purpose).
Seed also runs with --no-build and when wrangler is already up.
After Ready, GET the newest catalog updatedAt slug (and the newest mtime file) with Accept: text/html. Disk mtime alone is wrong after a full gzip (it picked yaml-json while open-graph-image-size was the new page).
status:dev FAIL if those tool URLs are not 200.
Docs: ops/README.md (“新工具 404、旧工具正常”), ops/worker-r2-ops.md, cloudflare-workflow.mdc; PowerShell -NoSeedR2.

Verified: npm run status:dev reports zh/open-graph-image-size and zh/yaml-json both HTTP 200.

[actions]

ops/dev/start-dev.mjs: fail-hard seed; seed on --no-build; tool-page smoke
ops/lib/dev-process.mjs: catalog + mtime probes
ops/dev/status-dev.mjs: tool-page health in status
ops/dev/start-dev.ps1 / start-dev.sh
ops/README.md, ops/worker-r2-ops.md, .cursor/rules/cloudflare-workflow.mdc

start:dev now aborts on local R2 seed failure and status:dev probes the newest catalog tool page so home-200 plus new-tool-404 cannot look healthy. | Dev Logs