fix: ERC20 approve state propagation for Uniswap workflow simulation

[chrisli30]

• fix: inject ERC20 allowance override after approve simulation for state propagation

Summary

Tenderly HTTP /simulate returns raw_state_diff: null for approve() calls, so MergeRawStateDiff was always a no-op for the approve step. Downstream nodes (e.g. Uniswap exactInputSingle) never saw the new allowance and failed with "ERC20: transfer amount exceeds allowance."

After a successful approve() simulation, we now explicitly compute and set the allowance[owner][spender] storage slot in SimulationStateMap across common ERC20 slot indices (1, 2, 3, 10 — covering OpenZeppelin, legacy, and USDC implementations).

Changes

• simulation_state.go: add erc20AllowanceSlot() (double-keccak for nested mapping) and commonAllowanceSlots
• vm_runner_contract_write.go: add isMethodWithParams() predicate, extract simSuccess guard, inject allowance slots post-approve
• simulate_uniswap_workflow_test.go: comprehensive integration test replicating the full Studio stop-loss-on-uniswap template (8 nodes: trigger, balance, code, branch, approve, contractRead/quote, code/slippage, swap) on Sepolia with zero on-chain allowance
• docs/changes/0002: changelog documenting root cause (Tenderly null raw_state_diff) and fix

How it was diagnosed

1. Wrote integration test using salt:0 wallet (6.78 USDC, zero allowance to SwapRouter02)
2. Confirmed approve succeeded but swap failed with allowance error
3. Raw HTTP diagnostic confirmed Tenderly returns raw_state_diff: null (field present, value null)
4. Extraction path (transaction.transaction_info.raw_state_diff) was correct — no data to extract
5. Fix: explicit allowance injection after approve, independent of raw_state_diff

Test plan

• go vet clean
• TestSimulateTask_StopLossWorkflow_Sepolia — all 8 steps pass (trigger, balance, JS code, branch, approve, quote, slippage, swap)
• Wallet has zero on-chain allowance — swap success proves propagation works
• Post-deploy: verify stop-loss template simulation in Studio UI

[Copilot]

Pull request overview

Fixes missing ERC20 allowance state propagation between sequential Tenderly simulations (approve → Uniswap swap) by explicitly injecting the allowance storage slot into the accumulated SimulationStateMap when Tenderly returns raw_state_diff: null for approve().

Changes:

• Add ERC20 allowance storage slot computation and a set of common allowance slot indices.
• Inject allowance overrides into SimulationStateMap after successful approve() simulations.
• Add a Sepolia Uniswap stop-loss workflow integration test and a changelog entry documenting the root cause/fix.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
docs/changes/0002-uniswap-simulation-state-propagation.md	Documents the Tenderly raw_state_diff: null behavior and the explicit allowance injection workaround.
core/taskengine/vm_runner_contract_write.go	Adds simSuccess guard + approve()-specific post-simulation allowance injection into SimulationStateMap.
core/taskengine/simulation_state.go	Adds erc20AllowanceSlot() and commonAllowanceSlots to support allowance storage override injection.
core/taskengine/simulate_uniswap_workflow_test.go	Adds an end-to-end Sepolia workflow simulation test that exercises approve → quote → swap state propagation.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

amount.SetString return value is ignored, so parse failures (e.g. when the approve amount is provided as hex like 0x...) will silently leave amount at 0 and inject a zero allowance. Also, amount.Sign() < 0 doesn’t detect parse failures, so the "try hex" fallback won’t run when base-10 parsing fails.

Consider parsing once with base 0 (auto-detect 0x), checking the ok boolean, and skipping injection (or logging a warning) when parsing fails; also ensure the parsed value is non-negative.

Suggested change

	amount := new(big.Int)
	amount.SetString(resolvedMethodParams[1], 10)
	if amount.Sign() < 0 {
	amount.SetString(resolvedMethodParams[1], 0) // try hex
	}
	valueHex := fmt.Sprintf("0x%064x", amount)
	for _, candidateSlot := range commonAllowanceSlots {
	slotHash := erc20AllowanceSlot(senderAddress, spender, candidateSlot)
	r.vm.simulationState.SetStorageSlot(contractAddress.Hex(), slotHash.Hex(), valueHex)
	}
	r.vm.logger.Debug("Injected allowance override after approve simulation",
	"token", contractAddress.Hex(),
	"owner", senderAddress.Hex(),
	"spender", spender.Hex(),
	"amount", amount.String(),
	"slots_set", len(commonAllowanceSlots))
	amount, ok := new(big.Int).SetString(resolvedMethodParams[1], 0)
	if !ok {
	r.vm.logger.Debug("Skipping allowance override after approve simulation due to invalid amount",
	"token", contractAddress.Hex(),
	"owner", senderAddress.Hex(),
	"spender", spender.Hex(),
	"amount_raw", resolvedMethodParams[1])
	} else if amount.Sign() < 0 {
	r.vm.logger.Debug("Skipping allowance override after approve simulation due to negative amount",
	"token", contractAddress.Hex(),
	"owner", senderAddress.Hex(),
	"spender", spender.Hex(),
	"amount_raw", resolvedMethodParams[1])
	} else {
	valueHex := fmt.Sprintf("0x%064x", amount)
	for _, candidateSlot := range commonAllowanceSlots {
	slotHash := erc20AllowanceSlot(senderAddress, spender, candidateSlot)
	r.vm.simulationState.SetStorageSlot(contractAddress.Hex(), slotHash.Hex(), valueHex)
	}
	r.vm.logger.Debug("Injected allowance override after approve simulation",
	"token", contractAddress.Hex(),
	"owner", senderAddress.Hex(),
	"spender", spender.Hex(),
	"amount", amount.String(),
	"slots_set", len(commonAllowanceSlots))
	}

[Copilot]

The doc references an integration test named TestSimulateTask_UniswapApproveSwap_Sepolia, but the test added in this PR is TestSimulateTask_StopLossWorkflow_Sepolia (and there’s no TestSimulateTask_UniswapApproveSwap_Sepolia in core/taskengine). Please update this reference to match the actual test name so the diagnosis steps remain reproducible.

Suggested change

	1. Wrote an integration test (`TestSimulateTask_UniswapApproveSwap_Sepolia`) using a salt:0 wallet with real Sepolia USDC balance but **zero** on-chain allowance to SwapRouter02. This made the test conclusive: swap can only succeed if the approve's allowance propagates.
	1. Wrote an integration test (`TestSimulateTask_StopLossWorkflow_Sepolia`) using a salt:0 wallet with real Sepolia USDC balance but **zero** on-chain allowance to SwapRouter02. This made the test conclusive: swap can only succeed if the approve's allowance propagates.