@toon-protocol/views@0.10.8

Patch Changes

• 0f6fc74: Fix three wallet-overview bugs that made the wallet card look broken in the host iframe:

  • Copy button was a silent no-op. The TOON app runs in the host iframe, which
    isn't granted the clipboard-write permission policy, so
    navigator.clipboard.writeText rejects there — and the click had no rejection
    handler. CopyButton now falls back to the legacy document.execCommand('copy')
    over a hidden textarea (works in a sandboxed frame), same iframe-limitation
    class as the window.confirm → consent fix.
  • Fund button gave no feedback. Tapping "Fund" fired the faucet but the card
    never changed, so it read as broken. The button now shows a Funding… →
    Requested (or Retry fund on failure) state and re-reads balances after a
    successful drip.
  • Balances rendered blank on a flaky read. The toon_balances control plane
    can transiently refuse on :8787 while the websocket transport is healthy
    (toon-client#186) — it succeeds on retry. readBalances now retries before
    giving up and throws on persistent failure, and wallet-overview shows a
    "Balances temporarily unavailable — Retry" state instead of a blank card that's
    indistinguishable from a real zero balance.

  Also fixes the daemon-side root cause of that flakiness (toon-client#186), not
  just the symptom:

  • Stale keep-alive socket race. The long-lived MCP server calls the localhost
    control plane infrequently, so the daemon (Node's default 5s keep-alive) reaped
    idle sockets the undici client pool still held — the next request reused a dead
    socket and failed with ECONNRESET, mislabeled as "daemon not reachable". The
    daemon now keeps idle sockets alive past the client's pool window
    (keepAliveTimeout: 650s), and the ControlClient transparently retries
    idempotent (GET/DELETE) requests on a transient connection failure. Mutating
    POSTs are not retried (no double publish/fund/deposit).
  • Timeouts no longer masquerade as "daemon not reachable". A request the
    client aborts on its own timeout (e.g. a hung on-chain balance read) is now a
    retryable 504, so the surfaced message says "retry" instead of "the daemon
    failed to start — check the log".