[bpamiri]

Summary

Comments only. Three docblocks now match the behaviour and the docs:

• WheelsTest.recordQueries() and Base.$recordQuery() said values are "never recorded". Bound values become ?, but a value written into the SQL itself (parameterize=false) appears as written. This matches the testing guide and .ai/testing.md since feat(test): assert how many SQL statements a spec's call sends #4357.

• appCacheFetch() and appCacheRead() now say:

  • keys are shared by every user, session, tenant and host, so per-user, per-role or per-tenant data needs that identity in the key ("digest-" & userId);
  • when the cache is full, appCacheFetch() returns the computed value without caching it.

  This matches the caching guide and .ai/caching.md since feat: add appCacheFetch() and friends for caching your own data #4388.

Tests

No code changed (the diff touches only * comment lines). Compile check on Lucee 7 + SQLite:

Specs	Pass	Fail	Error
caching	50	0	0
wheelstest	277	0	0
global	338	0	0

🤖 Generated with Claude Code

[bpamiri]

Review: head f9159e9. APPROVE (comment-only).

• All 16 changed lines in WheelsTest.cfc, databaseAdapters/Base.cfc and global/cache.cfm are docblock lines (*-prefixed), so no executable line changed.
• recordQueries() / $recordQuery(): "bound values become ?; a value written into the SQL with parameterize=false appears as written" is accurate, and says more precisely what "values are never recorded" meant.
• appCacheFetch() / appCacheRead(): the full-cache behaviour and the key-scope sentence match the shipped feat: add appCacheFetch() and friends for caching your own data #4388 guide and .ai text.

The gate is PR CI (7 pass, 14 pending at review time). The final follows once it's green.

[bpamiri]

FINAL: APPROVE. Head f9159e9.

• Comment-only: all 16 changed lines are docblock lines in WheelsTest.cfc, databaseAdapters/Base.cfc and global/cache.cfm, so no executable code changed and no compat matrix is needed.
• PR CI: 21 pass, 0 fail, 0 pending, 3 skipping.