[bpamiri]

callbacks = false also skips the commit callbacks

save(callbacks = false) and delete(callbacks = false) skipped every callback except afterCommit and afterRollback, which they still queued. They now skip those too.

$enqueueTransactionCallbacks() takes a callbacks argument. With callbacks = false it queues only the saved-change restore, the same path a model without commit callbacks already takes. A callbacks = false save that rolls back therefore still puts back what savedChanges() reported before it.

A save that changes nothing still queues afterCommit, as before.

Specs

vendor/wheels/tests/specs/model/callbacksFalseAfterCommitSpec.cfc, on the tag model with afterCommit and afterRollback registered:

• create, update and delete with callbacks = false: no commit callback runs;
• a rolled-back save with callbacks = false: no afterRollback;
• a rolled-back callbacks = false save: savedChanges() equals its value from before the save;
• with callbacks = true: they run, as before.

RED on develop: the four callbacks = false cases fail. The savedChanges() case was checked against a version that skipped the queue entirely, and it fails there.

GREEN, Lucee 7 + SQLite: full suite 6852/6852 after merging develop; complexity gate passes.

🤖 Generated with Claude Code

[bpamiri]

Review @ 7d225d3: source looks good. Final verdict after matrix 37344694616 and PR CI

• The change. $enqueueTransactionCallbacks() gains callbacks = true, and !arguments.callbacks now takes the same branch as "no transaction callbacks": it queues only $enqueueSavedChangesRestore(). All three call sites pass callbacks = arguments.callbacks: the create and update branches of $save() and delete(). There are no other call sites. With callbacks = true the path is untouched.
• savedChanges() stays correct after a rolled-back callbacks = false save. The restore entry (feat(model): report what the last save wrote with savedChanges() #4349's callbacks = false queue entry) is still queued inside a real transaction, so $restoreSavedChangesBeforeQueue() puts back the pre-save state on rollback. "still puts back savedChanges() when a callbacks = false save rolls back" asserts t.savedChanges() equals the earlier snapshot. A version that skipped the enqueue entirely would leave the rolled-back save's changes visible and fail it, which is the guard orch1 asked for.
• No change with callbacks = true. "still runs them with callbacks = true" checks the exact log (commit:…, commit:…, rollback:…), and "still queues afterCommit for a save that changes nothing" pins the no-op case.
• dependent= deletes ignore callbacks=false; invalid dependent= fails late #4373 (dependent= deletes ignoring callbacks = false) is not folded in. There's no change to the association dependent path, and dependent= deletes ignore callbacks=false; invalid dependent= fails late #4373 is still open, so it's separate.
• Branch freshness. The branch is behind develop, but develop hasn't touched callbacks.cfm, create.cfm or delete.cfm since its base, and git merge-tree reports a clean merge.

INFO (wording): the fragment's last sentence, "A save that changes nothing still queues afterCommit", could be read as applying with callbacks = false too. What the spec pins is the callbacks = true case. "(with callbacks on)" would remove the ambiguity.

No findings.

[github-actions]

Wheels Test Results

224 212 tests   219 091 ✅  1h 26m 19s ⏱️
 17 504 suites    5 121 💤
     32 files          0 ❌

Results for commit 7d225d3.

[bpamiri]

FINAL: APPROVE. Head c1c50b8.

• Compat matrix 37344694616 ran on 7d225d3: 32/32 legs green, 0 fail, 0 error. callbacksFalseAfterCommitSpec is 7/7 with 0 skipped on every leg.
• The only commit after the matrix (c1c50b8) changes one line of changelog.d/callbacks-false-commit-callbacks.fixed.md. It now says "With callbacks on, a save that changes nothing still queues afterCommit", which addresses my INFO about the fragment wording. Nothing executable changed, so the matrix covers this head.
• PR CI on c1c50b8: 18 pass, 0 fail, 0 pending, 2 skipping.