FeeQuoter support for token transfers by vicentevieytes · Pull Request #872 · smartcontractkit/chainlink-ton

vicentevieytes · 2026-09-04T14:30:35Z

No description provided.

… vv/fee-quoter-token-transfers

duck-types

The fee calculation logic looks correct compared to EVM's, but I've left some small suggestions

duck-types · 2026-09-09T17:26:21Z

+// SnakedCell<T> (= cell) is passed directly as this content cell - no extra ref indirection, since
+// get-method args aren't loaded via a struct's loadRef()-based field deserialization.
+function buildSnakedCellOf<T>(v: SnakedCell<T>, storeFn_T: StoreCallback<T>): c.Cell {
    if (v.length === 0) {


Nice catch. nit: remove the LLM reasoning comment

duck-types · 2026-09-09T20:39:00Z

+    var tokenCount = 0;
+
+    var tokenAmountsIter = msg.tokenAmounts.iter();
+    if (tokenAmountsIter.empty()) {


Invert this if-else clauses for code to be more similar to EVM's

duck-types · 2026-09-09T20:53:29Z

+    if (overrideEntry.isFound) {
+        val config = overrideEntry.loadValue();
+        if (config.isEnabled) {


There are too many anidated ifs and it gets difficult to follow the logic. We could have a helper function to fetch the config and check if it is enabled, and return TokenTransferFeeConfig?, then we do

Suggested change

if (overrideEntry.isFound) {

val config = overrideEntry.loadValue();

if (config.isEnabled) {

if (config == null) {

// return defaults

}

// rest of the calculations

duck-types · 2026-09-09T21:02:05Z

+            if (premiumFeeUsdWei < minFeeUsdWei) {
+                premiumFeeUsdWei = minFeeUsdWei;
+            } else if (premiumFeeUsdWei > maxFeeUsdWei) {
+                premiumFeeUsdWei = maxFeeUsdWei;
+            }


I know this is what EVM does, but it could use a min and max functions, right? We have thoes in the std lib in Tolk

duck-types · 2026-09-09T21:04:46Z

-        // + tokenCount as uint256 * TON_2_EVM_MESSAGE_FIXED_BYTES_PER_TOKEN
-        // + tokenTransferBytesOverhead
+        + msgDataLen
+        + mustProd(tokenCount, TON_2_EVM_MESSAGE_FIXED_BYTES_PER_TOKEN, FeeQuoter_Error.DataAvailabilityCostOverflow)


We should probably use mustAdd() here, right?

krebernisak · 2026-09-11T12:20:07Z

@@ -368,7 +427,6 @@
 fun validateMessageAndResolveGasLimitForDestination(extraArgs: Cell<ExtraArgs>, config: FeeQuoterDestChainConfig, message: Router_CCIPSend, msgDataLen: uint256): int {


F2 (correctness gap): SVM/SUI maxDataBytes check ignores tokens

contract.tolk:427–497 (validateMessageAndResolveGasLimitForDestination): EVM v1.6.3 expands the payload before the maxDataBytes assert — numberOfTokens × SVM_TOKEN_TRANSFER_DATA_OVERHEAD (v1.6.3:1145–1155) and × SUI_TOKEN_TRANSFER_DATA_OVERHEAD (:1088) plus a per-token destBytesOverhead loop (:1150–1160). TON's validate function has neither, even though SVM_TOKEN_TRANSFER_DATA_OVERHEAD and SUI_TOKEN_TRANSFER_DATA_OVERHEAD already exist in types.tolk:108,126 unused.

Pre-PR this was theoretical (tokens rejected with TokenTransfersNotSupported). This PR allows tokens, so a SVM/SUI-destined message with N tokens now passes maxDataBytes while actually being N × (36 + destBytesOverhead) bytes larger — the exact overrun the EVM check exists to prevent.

krebernisak · 2026-09-11T12:22:44Z

+        while (!tokenAmountsIter.empty()) {
+            val tokenAmount = tokenAmountsIter.next();
+            tokenCount += 1;
+            assert (tokenCount <= destChainConfig.config.maxNumberOfTokensPerMsg) throw FeeQuoter_Error.UnsupportedNumberOfTokens;


F3 (error-precedence deviation): maxNumberOfTokensPerMsg checked inside the fee loop

contract.tolk:293–294 bounds tokenCount inside the tokenAmounts iteration. EVM checks it once in the validate step (v1.6.3:1032, "too many tokens" fires before extraArgs/address validation and before any fee math). Consequences:

A message that is both oversized and, say, has an invalid SVM receiver now reverts with UnsupportedNumberOfTokens on TON but InvalidSVMReceiverAddress-class errors on EVM — fine for users, but it silently skips the maxDataBytes expansion (F2) in cases EVM would still measure.

It's O(n) checks where one would do, and it forces the loop to run per-token fee math for tokens 1..max before rejecting token max+1 — EVM rejects before touching any fee state.

krebernisak · 2026-09-11T12:25:34Z

    }
 }

 fun updateTokenTransferFeeConfigs(mutate st: Storage, msg: FeeQuoter_UpdateTokenTransferFeeConfigs) {


F4 (missing admin validation vs EVM): updateTokenTransferFeeConfigs accepts invalid configs

contract.tolk:127–154. EVM v1.6.3/2.0 _updateTokenTransferInitialConfigs/ applyUpdateConfig revert on:

TokenTransferConfigMustBeEnabled — you can't store a config with isEnabled == false (v1.6.3:~1310); disabling is done by removing;

InvalidDestBytesOverhead — destBytesOverhead < CCIP_LOCK_OR_BURN_V1_RET_BYTES (32) rejected;

dest chain with config.selector == 0 rejected (InvalidDestChainConfig).

TON accepts isEnabled=false entries with arbitrary values (they're then silently treated as "no config" at billing time — a stored-but-inert state EVM forbids), stores destBytesOverhead < 32 (which under-counts DA/calldata bytes forever), and noisily no-ops for unknown destChainSelector (replaceIfExists at :151 swallows the whole update).

Owner-only, so severity is low, but these are exactly the foot-guns EVM added reverts for, and "2.0 parity" is a stated goal. Ask: add the three asserts (FeeQuoter_Error extensions or reuse InvalidDestChainConfig-style codes), and change the unknown-selector path to throw instead of silently dropping — an operator who "successfully" updates a nonexistent lane will otherwise pay for gas and get nothing, and updateDestChainConfigs in the same file sets the precedent for how you want this to behave.

krebernisak · 2026-09-11T12:31:31Z

    tokenCount: uint256,
    tokenTransferBytesOverhead: uint256,
 ): uint256 {
-    assert (tokenCount == 0) throw FeeQuoter_Error.TokenTransfersNotSupported;


F7 (hygiene): stale error comments and a now-dead error code

errors.tolk:24–29: TokenTransfersNotSupported has no throw site anymore (grep: only the enum decl), and PremiumFeeOverflow / DataAvailabilityCostOverflow / FeeCalculationOverflow carry // Unreachable comments your own tests disprove (the 2^119 × 2^223 case exercises PremiumFeeOverflow). Ask: delete the // Unreachable annotations, and remove TokenTransfersNotSupported

… vv/fee-quoter-token-transfers

vicentevieytes added 2 commits September 4, 2026 11:29

implement feequoter token transfer fee math

80a10b5

fix unwrapSnakedCell.ts and wrapper for feequoter getter

4a78066

vicentevieytes marked this pull request as ready for review September 8, 2026 13:21

vicentevieytes requested a review from a team as a code owner September 8, 2026 13:21

Merge branch 'main' of github.com:smartcontractkit/chainlink-ton into…

6bf29e3

… vv/fee-quoter-token-transfers

duck-types requested changes Sep 9, 2026

View reviewed changes

duck-types added the Flag label Sep 10, 2026

krebernisak reviewed Sep 11, 2026

View reviewed changes

vicentevieytes added 3 commits September 15, 2026 16:53

remove data availability cost and match solana/evm impl

3354fc5

address PR comments

e7e192a

Merge branch 'main' of github.com:smartcontractkit/chainlink-ton into…

b6e3048

… vv/fee-quoter-token-transfers

vicentevieytes requested review from duck-types and krebernisak September 18, 2026 13:47

Navigation Menu

FeeQuoter support for token transfers - #872

FeeQuoter support for token transfers#872
vicentevieytes wants to merge 6 commits into
mainfrom
vv/fee-quoter-token-transfers

vicentevieytes commented Sep 4, 2026

Uh oh!

duck-types left a comment

Uh oh!

duck-types Sep 9, 2026

Uh oh!

duck-types Sep 9, 2026

Uh oh!

duck-types Sep 9, 2026

Uh oh!

duck-types Sep 9, 2026

Uh oh!

duck-types Sep 9, 2026

Uh oh!

krebernisak Sep 11, 2026

Uh oh!

krebernisak Sep 11, 2026

Uh oh!

krebernisak Sep 11, 2026

Uh oh!

vicentevieytes Sep 17, 2026

Uh oh!

krebernisak Sep 11, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		@@ -368,7 +427,6 @@
		fun validateMessageAndResolveGasLimitForDestination(extraArgs: Cell<ExtraArgs>, config: FeeQuoterDestChainConfig, message: Router_CCIPSend, msgDataLen: uint256): int {

Conversation

vicentevieytes commented Sep 4, 2026

Uh oh!

duck-types 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!

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!

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!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants