feat: 7702 in RLP_TXN by letypequividelespoubelles · Pull Request #362 · Consensys/linea-specification

letypequividelespoubelles · 2026-01-20T09:03:22Z

Note

Medium Risk
Medium risk because it extends core rlp_txn constraint/spec logic with a new transaction type, new phase transitions, and new counters/columns; some of the new authorization-list constraints still contain TODO/incomplete pieces that could leave behavior underspecified.

Overview
Adds support for 7702 transactions by enabling typeFourTx and introducing a new \phaseAuthorizationList phase (plus NUMBER_OF_AUTHORIZATIONS) across the rlp_txn column definitions and general constraints.

Updates phase flag sums, binarity/constancy requirements, LT/LX participation, and transaction-type-specific phase-transition/RLP-component constraints to account for the new phase, and documents that 7702 txs cannot be deployment transactions.

Introduces a new phase_constraints/authorization_list/* section defining shorthands and RLP-ization/tuple-index/countdown constraints for authorization lists, adjusts auxiliary data column counts accordingly, and removes obsolete access_list/constraints.tex plus a stale version.md.

^{Written by Cursor Bugbot for commit a9d5cbb. This will update automatically on new commits. Configure here.}

Signed-off-by: F Bojarski <[email protected]>

rlp_txn/phase_constraints/authorization_list/shorthands.tex

rlp_txn/phase_constraints/authorization_list/constraints.tex

cursor · 2026-01-29T00:31:30Z

rlp_txn/phase_constraints/authorization_list/constraints.tex

+\begin{enumerate}
+	\item \If $\locAuthorizationPrefix _{i} = 1$ \Then
+		\ob{We require a constraint that connects \rlpTxnCommonColumnNumberOfAuthorization{}
+		to an eponymous column in the $\isTxn{}$-perspective.}


Required constraint explicitly marked as missing TODO

High Severity

Constraint item 1 contains a red TODO marker (\ob{}) stating "We require a constraint that connects \rlpTxnCommonColumnNumberOfAuthorization to an eponymous column in the \isTxn{}-perspective." This explicitly documents a missing constraint that is required but not specified. Without this linkage, the number of authorizations from transaction data won't be properly available during authorization list processing.

@OlivierBBB we can discard thos no ? \rlpTxnCommonColumnNumberOfAuthorization is a common colun, so no need to duplicate it in the txn perpective ?

We can indeed.

cursor · 2026-01-29T00:40:24Z

rlp_txn/phase_constraints/authorization_list/constraints.tex

+			\left\{ \begin{array}{lclr}
+				\locAuthorizationIndex  _{i -                             1} & = & 0               & (\sanityCheck) \\
+				\locAuthorizationIndex  _{i}                                 & = & 0              \\
+				\locAuthorizationIndex  _{i +                             1} & = & 1 \vspace{2mm} \\


Empty authorization list causes unsatisfiable constraints

High Severity

The specification cannot handle type-4 transactions with empty authorization lists. Line 45 unconditionally sets \locAuthorizationIndex _{i + 1} = 1 when entering the phase, but line 97 requires \locAuthorizationIndex = \rlpTxnCommonColumnNumberOfAuthorizations to exit. Combined with the monotonicity constraint (line 62) that only allows the index to increase, if NumberOfAuthorizations is 0, the index becomes 1 and can never return to 0, making the exit condition unsatisfiable.

Additional Locations (1)

rlp_txn/phase_constraints/authorization_list/constraints.tex#L96-L97

that's the aim, yes.

OlivierBBB

Several comments, also made some changes to the counter of authorization tuples.

OlivierBBB · 2026-01-28T15:54:40Z

pkg/rlp_txn_v2.sty

+
 \newcommand{\rlpTxnCommonColumnReplayProtection}          {\col{REPLAY\_PROTECTION}}
 \newcommand{\rlpTxnCommonColumnYparity}                   {\col{Y\_PARITY}}
+\newcommand{\rlpTxnCommonColumnNumberOfAuthorization}	  {\col{NUMBER\_OF\_AUTHORIZATION}}


Is that the number of an authorization tuple in the authorization list ? If so the name we use in the HUB and the RLP_TXN module is TUPLE_INDEX.

Edit. Since this column is constant the above doesn't apply. The only comment I would make is spelling, it has to be pural

\newcommand{\rlpTxnCommonColumnNumberOfAuthorizations} {\col{NUMBER\_OF\_AUTHORIZATIONS}}

rlp_txn/_local.tex

OlivierBBB · 2026-01-28T17:16:13Z

rlp_txn/columns/shared.tex

+	\markAsJustifiedHere{}
+	\rlpTxnCommonColumnNumberOfAuthorization{}:
+	the number of authorization in the authorization list;
+\end{enumerate}


BTW we also (currently) count the number of auth tuples in the HUB: as it stands the HUB counts, for every type 4 transaction

the number of authorization tuples

the number of valid authorization tuples with sender == authority

Counting the valid ones with address collision in the HUB makes sense. I will remove the generic counting from the HUB as counting it in RLP_TXN makes more sense.

Is this column constant for the transaction ? Is it a countdown ? I ask this question in light of my

\rlpTxntransactionColumnLengthOfAuthorizationList

comment

yes, \rlpTxnCommonColumnNumberOfAuthorization{}: is transaction constant (and must be 0 for non-7702 txs)

rlp_txn/columns/transaction.tex

OlivierBBB · 2026-01-28T17:23:34Z

rlp_txn/generalities/constancies.tex

 		\item \CFI
 		\item \rlpTxnCommonColumnReplayProtection
 		\item \rlpTxnCommonColumnYparity
+		\item \rlpTxnCommonColumnNumberOfAuthorization


So \rlpTxnCommonColumnNumberOfAuthorization isn't a countdown given it's tx-constant.

rlp_txn/generalities/transaction_decoding.tex

OlivierBBB · 2026-01-28T17:43:18Z

rlp_txn/phase_constraints/authorization_list/constraints.tex

+	}
+\end{center}
+
+We constrain \locAuthorizationIndex to increase by at most one:


Missing braces to separate the macro from "to"

... \locAuthorizationIndex{} to ...

OlivierBBB · 2026-01-28T18:11:36Z

rlp_txn/phase_constraints/authorization_list/shorthands.tex

+	\end{array} \right.
+\]
+
+\saNote{} We don't require any counter constancy for those columns.


Do we have a constraint imposing that authorization lists are nonempty ? Likely something à la

If TYPE_4[i] = true is_first_row_of_phase_y[i] = true Then PHASE_AUTHORIZATION[i - 1] = true

Should do.

yes, the "listCantBeEmpty" bit of the rlpization of the prefix of the authorization list

OlivierBBB · 2026-01-28T18:37:37Z

rlp_txn/generalities/phase_end.tex

Cursor comment is right (Spurious text appended to section description)

OlivierBBB · 2026-01-28T18:41:43Z

rlp_txn/phase_constraints/authorization_list/constraints.tex

Is there no constraint imposing that \locAuthorizationIndex _{i} should start at the value 1 for the first authorization tuple ?

Should we not impose e.g.

If AUTHORIZATION_PHASE[i - 1] = faux and AUTHORIZATION_PHASE[i] = true Then auth_index[i] = 1 If AUTHORIZATION_PHASE[i - 1] = true and AUTHORIZATION_PHASE[i] = true Then auth_index[i] is one of { auth_index[i - 1], 1 + auth_index[i - 1] }

Or do we know for sure that auth_index[i - 1] = 0 if i is the row where you compute the RLP prefix for the entire auth list ? If that is the case we should state it. Currently I'm not sure we couldn't start at the value auth_index = 13 for instance.

cursor · 2026-01-29T00:47:00Z

rlp_txn/phase_constraints/authorization_list/constraints.tex

+				\phaseAuthorizationList _{i +                             1} & = & 1              \\
+				\phaseAuthorizationList _{i +                            10} & = & 1              \\
+				\phaseAuthorizationList _{i + \locAuthorizationNumberOfRows} & = & 1              \\
+			\end{array} \right.


Authorization list constraints unsatisfiable for empty lists

High Severity

The authorization list constraints are unsatisfiable when rlpTxnCommonColumnNumberOfAuthorizations is 0 (empty list). Constraint 2 unconditionally sets locAuthorizationIndex[i+1] = 1 and forces the phase to last at least locAuthorizationNumberOfRows (19) rows. However, constraint 5 requires locAuthorizationIndex = numberOfAuthorizations when exiting the phase. For an empty list (count=0), these constraints contradict each other - the index is forced to 1 but must equal 0 to exit. No constraint forbids empty authorization lists for type-4 transactions.

Additional Locations (1)

rlp_txn/phase_constraints/authorization_list/constraints.tex#L96-L97

cursor · 2026-01-29T15:06:36Z

rlp_txn/phase_constraints/authorization_list/generalities.tex

+
+		\ob{TODO: We require a constraint that connects \rlpTxnCommonColumnNumberOfAuthorizations{}
+		to an eponymous column in the $\isTxn{}$-perspective.}
+\end{enumerate}


Missing constraint connecting authorization count to transaction data

High Severity

The code contains an explicit TODO (lines 11-12) marking a required but missing constraint: "We require a constraint that connects \rlpTxnCommonColumnNumberOfAuthorizations to an eponymous column in the \isTxn-perspective." Without this constraint, the authorization count column is never actually set from transaction data for type-4 transactions, yet this value is used to control when authorization tuple processing terminates.

cursor

Cursor Bugbot has reviewed your changes and found 1 potential issue.

cursor · 2026-01-29T15:34:45Z

rlp_txn/phase_constraints/authorization_list/generalities.tex

+
+		\ob{TODO: We require a constraint that connects \rlpTxnCommonColumnNumberOfAuthorizations{}
+		to an eponymous column in the $\isTxn{}$-perspective.}
+\end{enumerate}


Missing constraint leaves NumberOfAuthorizations unvalidated for type-4 transactions

High Severity

The constraint at lines 8-13 has an empty body after \If $\locAuthorizationPrefix _{i} = 1$ \Then - only a TODO comment exists where the actual constraint should be. This means \rlpTxnCommonColumnNumberOfAuthorizations is never connected to actual transaction data for type-4 transactions. While non-type-4 transactions force this to 0, type-4 transactions allow any prover-supplied positive value. This is a soundness issue where the count of authorization tuples processed may not match the actual transaction.

letypequividelespoubelles · 2026-01-30T04:54:59Z

rlp_txn/phase_constraints/authorization_list/index.tex

+		\If $\locAuthorizationPrefix _{i} = 1$ \Then
+		\[
+			\left\{ \begin{array}{lclr}
+				\locAuthorizationIndex  _{i -                             1} & = & 0              \\


? the previous row is a TXN row tuple_index is a CMP row. This has to be deleted imho.

👍 Good point

letypequividelespoubelles · 2026-01-30T04:55:41Z

rlp_txn/phase_constraints/authorization_list/index.tex

+				\locAuthorizationIndex  _{i +                             1} & = & 1 \vspace{2mm} \\
+				\isCmp                  _{i +                             1} & = & 1 \vspace{2mm} \\
+				\phaseAuthorizationList _{i +                             1} & = & 1              \\
+				\phaseAuthorizationList _{i +                            10} & = & 1              \\


what is this +10 about ?

The above is a precaution to have checkpoints that are 100% reachable in any trace from the current vantage point to prevent early exits from the current phase. It's related to how constraints with large offsets get ignored if the trace runs out of rows such as how certain CALL constraints get clipped due to the presence of large MODEXP related shifts.

There are 1 + 3 rows per follow-up phase where integer rlp-utils calls are done on y, r, s. Three such phases are expected and must follow the current phase, so we are guaranteed that 12 rows must follow after the first CMP row of the AUTHORIZATION_LIST phase. The constraint with offset 10 falls squarely in that range and forces there to be at least 10 + 12 rows past the current vantage point.

The 3rd checkpoint isn't strictly speaking necessary due to this and will already be covered by the constraint saying that if tuple_index ≠ NUMBER_OF_AUTHORIZATIONS then IS_AUTHORIZATION[i + 19] = 1.

E.g. implemented as follows

(defconst AUTHORIZATION_PHASE___1ST_CHECKPOINT_OFFSET 1 AUTHORIZATION_PHASE___2ND_CHECKPOINT_OFFSET 10 AUTHORIZATION_PHASE___3RD_CHECKPOINT_OFFSET 19 ) (defconstraint authority-list-phase---ensuring-minimum-phase-length (:guard ...) (begin (eq! (shift IS_AUTHORIZATION_LIST AUTHORIZATION_PHASE___1ST_CHECKPOINT_OFFSET ) 1) (eq! (shift IS_AUTHORIZATION_LIST AUTHORIZATION_PHASE___2ND_CHECKPOINT_OFFSET ) 1) (eq! (shift IS_AUTHORIZATION_LIST AUTHORIZATION_PHASE___3RD_CHECKPOINT_OFFSET ) 1) ))

letypequividelespoubelles · 2026-01-30T04:56:16Z

rlp_txn/phase_constraints/authorization_list/index.tex

+				\locAuthorizationIndex  _{i -                             1} & = & 0              \\
+				\locAuthorizationIndex  _{i}                                 & = & 0              \\
+				\locAuthorizationIndex  _{i +                             1} & = & 1 \vspace{2mm} \\
+				\isCmp                  _{i +                             1} & = & 1 \vspace{2mm} \\


We can keep only of the two folowing: we have general constrains that a phase change <=> TXN row.

letypequividelespoubelles · 2026-01-30T04:58:06Z

rlp_txn/phase_constraints/authorization_list/index.tex

+		where $\ell$ is the length of the \texttt{authorization\_list} in the
+		current ``set code transaction'' i.e. $\ell \equiv \rlpTxnCommonColumnNumberOfAuthorizations$.
+	\item\label{rlp txn: phase constraints: authorization list: constraints: monotonicity of tuple index}
+		\If $\phaseAuthorizationList _{i - 1} = \true$ \Then $\locAuthorizationIndex _{i} \in \{ \locAuthorizationIndex _{i - 1}, 1 + \locAuthorizationIndex _{i - 1} \}$


why not using shorthand again_CMP_row_of_auth_phase ? We have already the constraint for the two first rows: 0 and then 1.

letypequividelespoubelles · 2026-01-30T05:00:35Z

rlp_txn/phase_constraints/authorization_list/index.tex

+		\Then
+		\[
+			\left\{ \begin{array}{lcl}
+				\phaseAuthorizationList _{i + 1}                             & = & 1                               \\


one of the two is enough

I mean one of the CMP or IS_AUTH_LIST, but the two offset are ok.

rlp_txn/phase_constraints/authorization_list/generalities.tex

letypequividelespoubelles · 2026-01-30T05:07:38Z

rlp_txn/phase_constraints/authorization_list/shorthands.tex

+		\locAuthorizationLengthCountdown _{i} & \define & \rlpTxnComputationColumnAuxiliaryData                {1} _{i} \\
+		\locAuthorizationItemCountdown   _{i} & \define & \rlpTxnComputationColumnAuxiliaryData                {2} _{i} \\
+		\locAuthorizationIndex           _{i} & \define & \rlpTxnComputationColumnCounterConstantAuxiliaryData {1} _{i} \\
+		\locAuthorizationChainId         _{i} & \define & \rlpTxnComputationColumnCounterConstantAuxiliaryData {2} _{i} \\


I guess we want a chainId hi as well ?

letypequividelespoubelles · 2026-01-30T05:09:11Z

rlp_txn/phase_constraints/authorization_list/generalities.tex

+We enforce that authorization lists must be nonempty
+\begin{enumerate}[resume]
+	\item\label{rlp txn: phase constraints: authorization list: constraints: authorization lists must be nonempty}
+		$\rlpTxnCommonColumnNumberOfAuthorizations _{i} \neq 0$


it's a sanity check, we already ask thge list to be non empty:

and

letypequividelespoubelles · 2026-01-30T05:11:37Z

rlp_txn/phase_constraints/authorization_list/rlp.tex

+					\rlpProcessInteger {
+						anchorRow  = i                        ,
+						relOffset  = \yellowm{1}              ,
+						integerHi  = 0                        ,


chainId Hi as well ?

letypequividelespoubelles · 2026-01-30T05:13:10Z

rlp_txn/phase_constraints/authorization_list/rlp.tex

+	\end{enumerate}
+	And a finalization constraint:
+	\begin{enumerate}[resume]
+		\item \If $\phaseEnd _{i} = 1$ \Then $\locAuthorizationLengthCountdown _{i} = 0$


I would put it as the same place as

(or at least a reminder ...)

feat: 7702 in RLP_TXN

d0671b8

Signed-off-by: F Bojarski <[email protected]>

letypequividelespoubelles requested a review from OlivierBBB January 20, 2026 09:03

letypequividelespoubelles self-assigned this Jan 20, 2026

letypequividelespoubelles linked an issue Jan 20, 2026 that may be closed by this pull request

7702 in RLP-TXN #361

Open

This comment was marked as outdated.

Sign in to view

ras

2f4c240

Signed-off-by: F Bojarski <[email protected]>

This comment was marked as outdated.

Sign in to view

reorder for better typing

54c1630

Signed-off-by: F Bojarski <[email protected]>

This comment was marked as outdated.

Sign in to view

fix: ensure 0 authorizations when not typoe 4 tx

d1bb338

Signed-off-by: F Bojarski <[email protected]>

cursor bot reviewed Jan 23, 2026

View reviewed changes

rlp_txn/phase_constraints/authorization_list/shorthands.tex Outdated Show resolved Hide resolved

rlp_txn/phase_constraints/authorization_list/constraints.tex Outdated Show resolved Hide resolved

rlp_txn/phase_constraints/authorization_list/constraints.tex Outdated Show resolved Hide resolved

OlivierBBB added 3 commits January 28, 2026 18:53

ras: formatting

0be34f6

feat: some changes (mostly wrt tuple_index)

0ace381

todo: record the number of auth_tuples in the TXN perspective

f372402

cursor bot reviewed Jan 29, 2026

View reviewed changes

ras: renaming one column

1ac60d5

cursor bot reviewed Jan 29, 2026

View reviewed changes

ras

403f2d4

OlivierBBB previously approved these changes Jan 29, 2026

View reviewed changes

cursor bot reviewed Jan 29, 2026

View reviewed changes

ras: restructuring

2c53ebb

OlivierBBB dismissed their stale review via 2c53ebb January 29, 2026 15:02

cursor bot reviewed Jan 29, 2026

View reviewed changes

ras

a9d5cbb

OlivierBBB force-pushed the 361-7702-in-rlp-txn branch from a2bf987 to a9d5cbb Compare January 29, 2026 15:25

cursor bot reviewed Jan 29, 2026

View reviewed changes

letypequividelespoubelles commented Jan 30, 2026

View reviewed changes

Conversation

letypequividelespoubelles commented Jan 20, 2026 • edited by cursor bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as outdated.

Uh oh!

This comment was marked as outdated.

Uh oh!

This comment was marked as outdated.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cursor bot Jan 29, 2026

Choose a reason for hiding this comment

Required constraint explicitly marked as missing TODO

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cursor bot Jan 29, 2026

Choose a reason for hiding this comment

Empty authorization list causes unsatisfiable constraints

Uh oh!

Choose a reason for hiding this comment

Uh oh!

OlivierBBB 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!

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!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

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!

cursor bot Jan 29, 2026

Choose a reason for hiding this comment

Authorization list constraints unsatisfiable for empty lists

Uh oh!

cursor bot Jan 29, 2026

Choose a reason for hiding this comment

Missing constraint connecting authorization count to transaction data

Uh oh!

cursor bot left a comment

Choose a reason for hiding this comment

Uh oh!

cursor bot Jan 29, 2026

Choose a reason for hiding this comment

Missing constraint leaves NumberOfAuthorizations unvalidated for type-4 transactions

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!

OlivierBBB Jan 30, 2026 • 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!

letypequividelespoubelles commented Jan 20, 2026 •

edited by cursor bot

Loading

OlivierBBB Jan 30, 2026 •

edited

Loading