Oev 671 txm improvements #294

dimriou · 2025-11-07T18:38:27Z

Each commit corresponds to the following features:

OEV-676 Sync broadcasting and backfilling
OEV-680 Trigger instant retransmission of purgeable transactions
OEV-675 Reduce empty transaction gas limit
OEV-677 Add Meta Error Handler
OEV-668 Add agnostic gas bumping
OEV-679 Add separate pricing for empty transactions

augustbleeds · 2025-11-11T01:58:56Z

pkg/txm/txm.go

-	go t.broadcastLoop(address, triggerCh)
-	go t.backfillLoop(address)
+	t.wg.Add(1)
+	go t.loop(address, triggerCh)


can you remind me what the advantage was of combining these two loops?

It enables easier nonce management via error handling when broadcasting a transaction.

augustbleeds · 2025-11-11T02:03:35Z

pkg/txm/txm.go

+		return err
 	}

 	if tx == nil || *tx.Nonce != latestNonce {


when does this scenario happen? hard for me to think of why the local nonce would be behind the chain nonce if it has to originate from the local txm.

Not behind but in front. Here's an example: you broadcast tx A to the mempool. You restart the TXM, fetch the pending nonce, which is A+1, and start sending more transactions to the mempool. Since you restarted the TXM, transaction A is not being tracked. If it gets dropped by the RPC or the builders, TXM won't be able to recover unless it fills the nonce. That's what this case does.

I noticed TXM uses pending for the block parameter to capture pending transactions.
Does the Flashbots client behave accordingly (i.e., takes into account inflight transactions)?

augustbleeds · 2025-11-11T02:29:17Z

pkg/txm/attempt_builder.go

+			// TODO: add better handling
+			bumpedAttempt, err := a.NewBumpAttempt(ctx, lggr, tx, *attempt)
+			if err != nil {
+				return attempt, nil


can we select for that specific error so we reduce possibility of halting on a different error?

pkg/txm/clientwrappers/dualbroadcast/meta_client.go

augustbleeds

Changes look good to me, just want to clarify the error handling behavior for empty tx gas bumping.

cll-gg · 2025-11-17T12:28:31Z

pkg/txm/clientwrappers/dualbroadcast/meta_error_handler.go

+}
+
+func (e *errorHandler) HandleError(ctx context.Context, tx *types.Transaction, txErr error, txStore txm.TxStore, setNonce func(common.Address, uint64), isFromBroadcastMethod bool) error {
+	// If this isn't the first broadcast, don't mark the tx as fatal as other txs might be included on-chain.


Suggested change

// If this isn't the first broadcast, don't mark the tx as fatal as other txs might be included on-chain.

// Only if this is the first broadcast, mark the tx as fatal. In other cases, other txs might be included on-chain

cll-gg · 2025-11-17T12:31:01Z

pkg/txm/attempt_builder.go

+
+	// bump purge attempts
+	if tx.IsPurgeable {
+		// TODO: add better handling


Would be good to make a separate ticket with an explanation of what would be better here

lhmerino · 2025-11-18T19:26:19Z

pkg/txm/attempt_builder.go

+		for {
+			bumpedAttempt, err := a.NewBumpAttempt(ctx, lggr, tx, *attempt)
+			if err != nil {
+				if errors.Is(err, fees.ErrConnectivity) {


I'm looking at the error choice. I believe this is due to the code located here

if bumpedMaxPriorityFeePerGas.Cmp(priorityFeeThreshold) > 0 { return ..., ErrConnectivity }

It feels slightly counterintuitive to return ErrConnectivity for a value check, where ErrBumpFeeExceedsLimit might seem more natural.

Is the intention to treat priorityFeeThreshold as a sanity guardrail (implying the network/estimator is broken if fees get this high) rather than a budget limit (implying the tx just needs to pay more)? If so, adding a small comment explaining why this triggers a 'connectivity' error would be helpful for readability.

lhmerino · 2025-11-18T22:09:20Z

pkg/txm/txm.go

 	defer cancel()
 	broadcastWithBackoff := newBackoff(1 * time.Second)
 	var broadcastCh <-chan time.Time
+	backfillCh := time.After(utils.WithJitter(t.config.BlockTime))


utils.WithJitter is a deprecated function.
It says to use timeutils package instead.

lhmerino · 2025-11-18T22:39:30Z

pkg/txm/clientwrappers/dualbroadcast/meta_client.go

 	metrics, err := NewMetaMetrics(chainID.String())
 	if err != nil {
 		return nil, fmt.Errorf("failed to create Meta metrics: %w", err)
 	}


Is the MetaClient currently active -- it looks A-specific?
I noticed a potential dependency in the metrics setup. If beholder is unavailable, will this cause the client to error out as well?

Ideally, we want to ensure the client is decoupled so that a beholder outage doesn't block the main flow. For example, on 10/10 I saw many alerts about beholder potentially having issues.

lhmerino · 2025-11-18T22:45:32Z

pkg/txm/clientwrappers/dualbroadcast/selector.go

-		return NewFlashbotsClient(client, keyStore, url), nil
+		return NewFlashbotsClient(client, keyStore, url), nil, nil
 	default:
-		return NewMetaClient(lggr, client, keyStore, url, chainID)


This code could then be made cleaner if NewMetaClient does not fail.

lhmerino · 2025-11-18T23:10:07Z

pkg/txm/storage/inmemory_store.go

+
+	// TODO: for now do the simple thing and drop the transaction instead of adding it to the fatal queue.
+	delete(m.UnconfirmedTransactions, *txToMark.Nonce)
+	delete(m.Transactions, txToMark.ID)


I have a few safety concerns about deleting these entries:

Nonce Gaps: Could deleting this nonce create a gap in our internal inflight list? I'm concerned that higher-nonce transactions currently inflight might stall indefinitely waiting for the tx associated with this nonce to fill.

Panic Risk: Dereferencing *txToMark.Nonce without checking if txToMark.Nonce is not nil can cause a panic.

Less important but still relevant: could this complicate debugging by deleting this information?

lhmerino · 2025-11-18T23:26:46Z

pkg/txm/storage/inmemory_store.go

 		FromAddress:       m.address,
 		ToAddress:         common.Address{},
 		Value:             big.NewInt(0),
 		SpecifiedGasLimit: gasLimit,


Suggestion: We could simplify this by hardcoding the gas limit to 21_000.

Since a standard ETH transfer to an EOA (0x0) has a fixed cost (21_000), passing in gasLimit and maintaining a separate EmptyTxLimitDefault config seems like unnecessary overhead. In addition, it would also remove the need to add gasEstimatorConfig parameter to func NewTxmV2.

If we were to instead send ETH to/from a smart contract, only then would the gas amount differ.

dimriou added 5 commits November 5, 2025 17:53

OEV-676 Sync broadcasting and backfilling

976ba44

OEV-680 Trigger instant retransmission of purgeable transaction

e533378

OEV-675 Reduce empty transaction gas limit

a822f17

OEV-677 Add Meta Error Handler

91c5a2d

OEV-668 Add gas bumping

81ef8e6

dimriou added the do-not-merge label Nov 7, 2025

OEV-679 Add separate pricing for empty txs

64f4d2f

dimriou force-pushed the oev-671_txm_improvements branch from 8276ab6 to 64f4d2f Compare November 10, 2025 10:51

Add tests

7ef98a2

augustbleeds reviewed Nov 11, 2025

View reviewed changes

pkg/txm/clientwrappers/dualbroadcast/meta_client.go Show resolved Hide resolved

Add more tests

c3d4502

augustbleeds reviewed Nov 11, 2025

View reviewed changes

Add txm test cases for purgeable retransmission and error handler

115e871

dimriou force-pushed the oev-671_txm_improvements branch from 29a1813 to 115e871 Compare November 11, 2025 15:51

cll-gg reviewed Nov 17, 2025

View reviewed changes

lhmerino reviewed Nov 18, 2025

View reviewed changes

	// If this isn't the first broadcast, don't mark the tx as fatal as other txs might be included on-chain.
	// Only if this is the first broadcast, mark the tx as fatal. In other cases, other txs might be included on-chain

Oev 671 txm improvements #294

Are you sure you want to change the base?

Oev 671 txm improvements #294

Uh oh!

Conversation

dimriou commented Nov 7, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhmerino Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

augustbleeds left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhmerino Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhmerino Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhmerino Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhmerino Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

lhmerino Nov 18, 2025 •

edited

Loading

lhmerino Nov 18, 2025 •

edited

Loading

lhmerino Nov 18, 2025 •

edited

Loading

lhmerino Nov 18, 2025 •

edited

Loading

lhmerino Nov 18, 2025 •

edited

Loading