chore: ignore the runtime artifacts a local test run writes into the tree

[bpamiri]

Removes seven generated TestBox stub files that should never have been committed, and stops the whole class of mistake at the source.

What happened

public/testbox/system/stubs/*.cfm, rewrite.config and a root box.json are all written into the working tree by a local test run or dev server. None are source. None were ignored either — so any git add -A after running the suite sweeps them into the diff.

That happened twice:

• Caught in self-review across the SQL Generation Bug: Variable state leakage in findAll(include="...") when mixed nested and shallow joins are used #3334 / validatesUniquenessOf(scope=...) dereferences scope properties unguarded — throws 'has no accessible Member' instead of validating #3350 / PostgreSQL adapter: $getColumnInfo() pulls cross-schema columns, crashing any model whose table name collides with an information_schema view #3349 / super<name> originals are only registered for model overrides — controller/view helper overrides get no superLinkTo() #3325 / Job.cfc persists jobClass from GetMetadata(this).name, whose casing is not guaranteed stable across reloads/engines #3351 batch, and all five branches were cleaned before merging.
• Missed on fix(security): validate the CSRF decrypt result instead of trusting that it did not throw #3362, which merged seven stubs onto develop. Same mistake, same cause, one PR later.

Cleaning the branches was treating the symptom. This is the fix.

Verified rather than assumed

Removing tracked files is only safe if they are genuinely regenerated. Deleting every file in public/testbox/system/stubs/ and running the full core suite:

✓ 4757 passed (0 fail, 0 error)
regenerated: 7 files

And with the ignore in place, a complete suite run now leaves git status entirely clean — which is the property that was missing and the reason this kept happening.

One file deliberately left alone

F952D54F… was already tracked before any of this. Ignoring a directory does not untrack what is already in the index, so it stays exactly as it is — it is not mine to remove in a cleanup commit. Worth knowing it is there, because its presence is what makes a stray stub look unremarkable at a glance.

🤖 Generated with Claude Code

[wheels-bot]

Wheels Bot — Reviewer

TL;DR — This chore removes seven generated TestBox mock stubs that were accidentally committed onto develop (via #3362) and adds three .gitignore entries so the whole class of git add -A-after-a-test-run mistake stops at the source. The change is correctly scoped, the deletions are verified-regenerable, and there are no code paths to break. Verdict: approve.

Conventions

• .gitignore — The root-anchored form is the right call and I verified it does not over-reach. /box.json and /rewrite.config only match the repo root, so the 15+ legitimately-tracked box.json files (tools/build/core/box.json, examples/tweet/box.json, vendor/wheels/tests/_assets/plugins/**/box.json, …) and cli/lucli/templates/app/rewrite.config are all untouched. No root-level box.json/rewrite.config is tracked, so nothing source is being shadowed.
• The added block also fixes the pre-existing missing trailing newline on the file — small but nice.

Correctness

• The seven deleted files are TestBox-generated _mock*/tmp_* statemachine stubs (e.g. tmp_closeSSEStream_…, tmp_$getChannelEngine_…), which are runtime mock artifacts, not source. The PR body documents a full-suite run (4757 pass / 0 fail / 0 error) that regenerates them, which matches their content. Safe to remove.

Non-blocking note

• public/testbox/system/stubs/ ignores the whole directory, which leaves the already-tracked F952D54F1096E25C030C8E3149ABD8C4.cfm in a tracked-but-ignored state (a gitignore entry does not untrack an indexed file). This is intentional and clearly documented in both the commit message and the .gitignore comment — flagging only so a future reader is not surprised that one stub is still versioned while its siblings are ignored. No action needed.

Clean chore with an unusually thorough justification. Nice work closing the loop at the source rather than re-cleaning branches.