test: failing repros for confirmed v2-dev bugs (fixes to follow)

[MuncleUscles]

What

Adds tests/bug-hunt-v2dev.test.ts — 12 failing tests, each reproducing a distinct, confirmed defect on v2-dev (the v0.6 integration branch). This PR intentionally contains no fixes; the tests are the deliverable so a follow-up change can be validated against them. Every test documents the file:line and root cause inline.

All 114 pre-existing tests still pass; only these 12 new tests fail.

Wrong results

#	Area	File:line	Bug
1	calldata toString()	src/abi/calldata/string.ts:8-16	Map entries emitted with no separator → {"a":1"b":2} (array branch adds ,; the map branch doesn't). Corrupts the readable field of decoded receipts.
2	calldata toString()	src/abi/calldata/string.ts:48-52, 62-66	Per-byte hex uses b.toString(16) with no padStart(2,"0") → [0x01,0x02] and [0x12] both render b#12; addresses render with <40 hex chars.
3	toJsonSafeDeep()	src/utils/jsonifier.ts:~90-94	Cycle-guard seen.add(value) is never released, so a DAG (same object referenced twice) has its 2nd occurrence replaced with null. Reachable from readContract output.
4	createClient({endpoint})	src/client/client.ts:~132-135	chainConfig.rpcUrls.default.http = [endpoint] mutates the shared imported chain singleton in place; one client's endpoint override leaks into every other client (and later createClient() calls).
5	schema strictTypes	src/contracts/schema.ts:101-113	Struct validation only runs if (isPlainObject(value)); a non-object scalar falls through to return true and is accepted as a struct.
6	schema strictTypes	src/contracts/schema.ts:55	'int' accepts any typeof === "number", so 1.5/NaN pass validation and then crash later inside calldata.encode.
7	staking getValidatorInfo()	src/staking/actions.ts:473	toHex(identityRaw.extraCid) double-encodes — viem already returns bytes as hex, so 0xdeadbeef becomes 0x30786465616462656566 (hex of the ASCII "0xdeadbeef"). Breaks the SDK's own setIdentity→getValidatorInfo round-trip.
8	vesting encodeExtraCid()	src/vesting/actions.ts:95-99	Blind startsWith("0x") cast with no isHex/even-length check; an odd-length CID is silently right-padded by viem (0x123→0x1230) and written on-chain with no error.
9	isSuccessful()	src/transactions/actions.ts:68-86	Requires executionResultName === FINISHED_WITH_RETURN, derived only from txExecutionResult(Name) — fields the studio/localnet getTransaction path never populates. A finalized, successful studio tx is reported as unsuccessful.

Instability

#	Area	File:line	Bug
10	calldata decode()	src/abi/calldata/decoder.ts:52-73	TYPE_ARR/TYPE_MAP element counts are taken straight off the wire and drive a while/slice loop with no bound against remaining bytes. A tiny malformed payload forces multi-second allocation before erroring (a fixed impl should reject in O(1)). Relevant because decode() runs on consensus/validator-supplied bytes.
11	vesting exit	src/vesting/actions.ts:310, 362	BigInt(options.shares) turns "" into 0n, silently signing & broadcasting a zero-share exit instead of rejecting the input.
12	staking API drift	src/staking/actions.ts:550	Commit a338ad5 advertised the helper as isValidatorBelowMinStake, but it's implemented as isValidatorBelowMin, so the documented name is missing at runtime (caller gets is not a function). It is also absent from the StakingActions type.

Why

These were surfaced by a focused bug hunt across the recent v0.6 work (calldata, fees/transactions, staking, vesting, accounts/contracts, utils/client). Each one can produce silently wrong results or client instability. Landing the repros first lets the fix PR be verified objectively.

Testing done

• vitest run → 114 pre-existing tests pass, the 12 new tests fail on their intended assertions (no harness errors).
• vitest --typecheck → the new file typechecks clean (all reds are assertions).
• Verified test isolation: the createClient singleton-mutation test restores localnet in a finally, so it does not pollute the rest of the run.

Decisions made

• Tests only, no fixes — per the task, a separate change will implement the fixes and turn these green.
• Where encodeExtraCid / the double-encode are internal, tests exercise them black-box through the public actions and decode the resulting calldata.
• Not turned into a failing test (flagged for the fixer): accountActions.transfer (src/accounts/actions.ts:95) unconditionally calls waitForTransactionReceipt, but every other send lane returns early on chain.isStudio (see src/contracts/actions.ts:2363-2367). On a Studio localnet there is no EVM receipt, so a transfer would block ~180s then throw. No test is included because the correct fix is genuinely ambiguous (reject Studio early vs. synthesize a receipt — the method's return type is TransactionReceipt) and could not be reproduced against a live Studio backend.
• Investigated and cleared (not bugs): the baseline Position 42 is out of bounds RLP stderr in transactions.test.ts (benign — decodeInputData catches it and returns null); fee/budget math vs. the studio reference; call-key derivation; v0.6 enum remaps; vesting ABI signatures vs. the on-chain contracts.

Checks

• I have tested this code
• I have reviewed my own PR
• I have created an issue for this PR
• I have set a descriptive PR title compliant with conventional commits

Reviewing tips

Read the inline comments in tests/bug-hunt-v2dev.test.ts — each describe block names the file:line and root cause, and each assertion notes the actual (buggy) value it observes today.

User facing release notes

None yet — this PR only adds failing regression tests. User-facing notes will accompany the fixes.

🤖 Generated with Claude Code

---

Generated by Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: bf7db6c7-da57-47a0-8a36-1a4d190cdc79

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/new-dev-bug-hunting-eo2cic

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}