[yeroo]

A draft stays addressable by its local: id after Store::reconcile_id renames it to its Graph id. A compose window left open across a SaveDraft keeps working: later edits, attachments, a second SaveDraft and SendDraft reach the draft. A queued SendDraft { local: } retried after a failed send finds its draft again.

Closes #1884
Closes #1881

What changed

• Alias on the row. messages.local_id: reconcile_id keeps the local: id there, only for a local: id. The issue proposed an alias table; a column is used instead. It goes with the message on every delete path (delete_message, a delta @removed, an FK cascade), so no trigger or cleanup is needed and no orphan alias is possible. Migration: the swallowed ALTER TABLE like the other columns, then a UNIQUE index created in Store::init.
• Resolve inside the statement. draft_row_sql! (on messages) and current_id_sql! (for bodies/outbound_attachments) are spliced into the one SQL statement that acts. That keeps it atomic against a reconcile_id the sync thread commits on its own connection. Graph ids are never looked up (substr(?1,1,6) = 'local:' guard).
• Draft calls that resolve: draft (now one JOIN bodies statement; the returned row's id is the current id), update_draft_fields, mark_sent, move_message, delete_message, add/remove/clear_outbound_attachment(s), outbound_attachments.
• New Store calls:
  • current_id(id);
  • delete_message_resolved(id) -> Option<String>: DELETE … RETURNING id, one statement.
• Outbox. ensure_draft_on_graph decides create vs update from the stored row's current id, so an already-reconciled local: id is patched, never created twice. A queued Delete resolves its id; one still local: makes no Graph call (it never reached Graph, short of a delete racing the create).
• Deletes name the row they removed. The engine's SyncCommand::Delete, lookxy's UI deletes (App::delete_and_send: the DeleteThread confirm and both delete_selected paths) and the mail.delete automation verb delete through delete_message_resolved. They send SyncCommand::Delete with the id of the row actually deleted. The list can still hold a local: id after the sync thread reconciled it.
• Payloads unchanged. SaveDraft/SendDraft payloads keep the id they were addressed with, and so does SyncEvent::Sent { id }, so a UI tracking pending_ops by its id still finds them.
• Test server. FakeServer::start_limited (routes with a use limit; start is unchanged). Also the 202/503 reason phrases.
• Docs. CLAUDE.md: "Mail drafts keep their local: id (mailcore: a draft stays addressable by its local: id after reconcile_id #1884, mailcore: a queued draft op keeps its local: id after reconcile_id, so it can never be retried #1881)".

The four assertions the issue named as intentionally changing (draft(local) was None after reconcile) now assert the row with the Graph id. A fifth, that outbound_attachments(local) was empty after reconcile, now asserts the stored key is the Graph id and the local: id lists the same row.

Testing

cargo fmt --check; cargo clippy -p mailcore -p lookxy --all-targets -- -D warnings; cargo test -p mailcore -p lookxy: mailcore 284 passed, lookxy 332 passed (24 new).

Acceptance cases from #1884 (test server)

• Save, edit, Send by the local: id, the edit through a second Store connection: 1 POST, 1 PATCH carrying the edit, 1 send, Sent { id: local }.
• Save, edit, Save: 1 POST, then 1 PATCH carrying the edits; nothing left queued or quarantined.
• Offline, two queued saves: 1 POST, then 1 PATCH.
• mailcore: a queued draft op keeps its local: id after reconcile_id, so it can never be retried #1881: the first send returns 503 (Throttled). A command barrier confirms the op is still queued; then Refresh: 1 POST, 1 PATCH, 2 sends, one Sent { id: local }.
• A queued save then send of the same local: id: the message is sent.

Other tests

• Store: resolution of every draft call, the alias dropped on delete and on a folder cascade, kept across a delta upsert, the migration of a pre-local_id file, and delete_message_resolved.
• Outbox: patch-not-create, delete by a reconciled id, no Graph call for an unresolved local: delete.
• lookxy: deleting a list row and a thread that still hold a reconciled local: id sends the Graph id; mail.delete likewise.

Review record

revmux, 2 rounds (Claude-only profile).

r1:

• Major, fixed: lookxy's UI delete paths with a stale local: id deleted the draft locally but never on Graph. Fixed with delete_message_resolved everywhere.
• Minor, fixed: mail.delete resolved and deleted in two statements.
• Minor, fixed: a vacuous wait in the mailcore: a queued draft op keeps its local: id after reconcile_id, so it can never be retried #1881 test. A 503 is GraphError::Throttled, which emits no state change, so the test now waits on a command barrier.
• Minor, fixed: an over-strong "never reached Graph" comment.

r2:

• Minor, fixed: a stale SyncCommand::SaveDraft doc.
• Pre-existing, folded: Compose::draft_id's doc claimed it becomes the Graph id.

Nothing disputed. review stopped: round 2 had no Major; all findings fixed

Out of scope, filed as follow-ups in #1892:

• Move/MarkRead/SetFlag/SetCategories ops with a local: id;
• calendar events' local: ids;
• attachment re-upload on a retried send;
• a queued save of a deleted never-synced draft quarantining;
• busy_timeout on the store's connections.

🤖 Generated with Claude Code

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 99.22879% with 6 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
mailcore/src/store/mod.rs	98.34%	4 Missing ⚠️
lookxy/src/app.rs	91.66%	1 Missing ⚠️
lookxy/src/ui/mod.rs	98.57%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[yeroo]

[linux] Merged automatically (auto-merge is on for this checkout): review stopped: round 2 had no Major; all findings fixed; merge rounds: none; whole suite green on 4933814 (cargo test --workspace --locked, 9025 tests; the html-export feature needs the wasm target this machine lacks, and CI ran it green); CI 14/14 passed; merge-check ok.

Follow-ups: the leftovers issue #1892, with the plan's five out-of-scope items:

• Move/MarkRead/SetFlag/SetCategories ops push a draft's local: id to Graph
• check calendar events for the same stale local: id after reconcile_event_id
• a retried SendDraft re-uploads its attachments
• a queued SaveDraft of a deleted never-synced draft quarantines instead of dropping
• busy_timeout on the store's connections

No finding was disputed or deferred; no duplicate reports.