test(sdk): add deterministic SubRosaClient unit tests and include in …

[Ce-Royals]

closes #2

PR description
Title

Add deterministic, network-free unit tests for SubRosaClient and update SDK test script
Summary

Adds a deterministic, network-free unit test suite for SubRosaClient covering state-changing methods, read-only views, conversions, and both direct RPC and injected TransactionSubmitter submission paths. Updates the @sub-rosa/sdk test script so pnpm sdk:test runs the new tests.
Files changed

client.test.ts — new tests for SubRosaClient (createRound, commit, openReveal/reveal, clear/settle/void flows via mocks; read-only views; error wrapping; injected submitter behaviors).
package.json — include src/client.test.ts in the test script so pnpm sdk:test runs the suite.

[drips-wave]

@Ce-Royals Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits

[karagozemin]

Thanks for the PR. I tested the latest head (4cdea8b) in a clean merge simulation against current main, and I cannot merge it yet because pnpm sdk:test hangs/fails.

Command run:

pnpm sdk:test

Observed result: src/client.test.ts stayed pending for over 75s and had to be interrupted. Node then reported:

✖ src/client.test.ts
'Promise resolution is still pending but the event loop has already resolved'

The likely culprit is the timeout test: it sets Date.now to a constant large value while _sleep is a no-op. Since the SDK computes deadline = Date.now() + confirmTimeout, a constant Date.now() never advances past the deadline, so the poll loop never exits. Please make the fake clock advance deterministically or configure a timeout path that actually terminates without real sleeps.

Also, #2 asks for isolated coverage for every SubRosaClient call path. The current suite covers some important pieces, but it still needs more explicit coverage for:

• openReveal, reveal, clear, settle, and void method wiring
• default bidder behavior in commit, not only missing-source behavior
• direct successful signAndSend, not only wrapped failure
• injected submitter timeout without hanging

After pnpm sdk:test and pnpm sdk:typecheck pass deterministically, I will review again.

[karagozemin]

Thanks for the update. I re-tested the current head (ad7d80c) against current main.

I cannot merge this yet because the PR no longer merges cleanly after the SDK configuration coverage work that landed on main.

Merge simulation result:

• packages/sdk/package.json has a content conflict.
• packages/sdk/src/client.test.ts has an add/add conflict.

Please rebase this branch on current main and keep the SDK config validation tests that are already on main. After resolving the overlap, please re-run:

• pnpm sdk:test
• pnpm sdk:typecheck

I will review again once the branch merges cleanly and the isolated SubRosaClient call-path tests still pass deterministically.

[karagozemin]

Thanks for rebasing and for the deterministic timeout update. I rechecked the current head against #2. The branch is now mergeable, but I still cannot merge because the required SDK checks fail.\n\nCommands run:\n\n- pnpm --filter @sub-rosa/sdk test\n- pnpm --filter @sub-rosa/sdk typecheck\n- git diff --check origin/main...HEAD\n\ngit diff --check passes, but the SDK test command fails in src/client.test.ts:\n\n- commit defaults to the configured bidder when one is not supplied fails with SubRosaClientConfigError: auditor blob must not be empty.\n- direct RPC signAndSend errors are wrapped in SubRosaSubmitError also hits auditor blob must not be empty instead of reaching the mocked submit failure.\n\nThe SDK typecheck also fails with:\n\n- src/client.test.ts(208,27): error TS2339: Property 'operator' does not exist on type 'never'.\n- same never issue for reveal_round, commit_deadline, reveal_deadline, item_ref, and auditor_pubkey.\n\nPlease update the tests to use a non-empty auditorBlob where the test needs to reach commit submission, and adjust the typed capture for seenArgs so the test file typechecks. After those two SDK commands pass, I can re-review again.