DanS 828018de2b feat(settings): sweep a private key's funds to your own wallet ...

Add a "Sweep to my wallet (don't keep the key)" option to the Import Private
Key dialog (spending keys only). Rather than keeping the key, it imports the
key, then sends ALL its funds to a destination you control — a freshly
generated shielded address by default, or an existing wallet address you pick.

Flow (App::sweepPrivateKey): import the key (rescan; the stock node has no
address index, so its UTXOs/notes can't be enumerated otherwise) → confirm it
holds spendable funds → resolve the destination → z_sendmany balance-minus-fee
from the key's address to the destination, via the existing tested submitZSendMany
wrapper + async-operation tracker. The imported key is left in the wallet with an
empty balance (there is no remove-key RPC) — stated plainly in the UI caveat.

Design notes forced by this daemon (verified against external/dragonx):
- z_importkey returns null here, so the shielded source address is found by
  diffing z_listaddresses before/after import (aborts if the pre-snapshot fails,
  so a wrong address can never be picked). Transparent uses importprivkey's return.
- z_sendmany, NOT z_mergetoaddress: moves a single-UTXO transparent source (the
  common paper-wallet case) and fails loudly instead of silently leaving a
  remainder. Amount computed in integer satoshis for an exact fixed-decimal string.
- Clear errors for no-funds / unconfirmed / dust-below-fee / already-imported.
- The mode + destination are locked during a running sweep; a sweep that finishes
  while the dialog is closed still shows its Done/Error+txid on reopen; sweep UI
  can't bleed into the watch-only viewing dialog.

i18n: 8 new sweep_* keys (en + 8 langs); dynamic status strings are English,
matching the beginSweepToSeedWallet precedent.

Reviewed across three adversarial multi-agent rounds (fund-safety focus): round 1
caught the feature was broken for shielded keys (z_importkey null) and rejected
single-UTXO transparent sources; round 2 caught reopen/double-submit/precision
issues; round 3 verified the fixes. NOT yet exercised against a live daemon —
needs a small-amount mainnet test before it's trusted, like migrate-to-seed.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-07-13 17:29:10 -05:00

611 KiB

Raw History

View Raw