朱子纯 / erp-framework-wiki

docs: slice 01 — close the live-save verification gap

Drove the BACK SPA's universal-CRUD save end-to-end via real
keystrokes + clicks. Captured:

- Network: POST /xlyEntry/business/addUpdateDelBusinessData
  ?sModelsId=13 → 200 OK (sModelsId only as query, no path
  variable — matches the SPA bundle's saveData call site:
  business/addUpdateDelBusinessData?sModelsId=" + sModelsId).
- DB side-effect: directly after save, gdsformconst.sChinese for
  sId=20260123… returned the modified value with the typed audit-
  tag suffix (since reverted). Empirical proof that the body's
  updateData[].column.sChinese landed in the named column on the
  named sTable — the universal-CRUD path verified end-to-end.
- Body shape: still rests on Javadoc + the SPA bundle's value
  parameter. The exact byte sequence the SPA serializes wasn't
  captured because the SPA's whatwg-fetch polyfill caches the
  XMLHttpRequest constructor at startup and bypasses any post-
  startup XHR monkey-patch (constructor wrapping, prototype
  override, fetch hook all silently bypassed). Documented this
  technical reason inline.
- Pre-save addSysLocking POST also confirmed in the same network
  capture (optimistic-lock flow already documented in internal.md).

Item 3 (SQL-layer capture from syslog4j / MyBatis debug log)
demoted to "Future-work backlog" — the dev DB's logging level
doesn't include MyBatis statements; closing it requires a
deployment with MyBatis debug logging enabled.