Implement glamsterdam-devnet-8

[flcl42]

Tracking issue for glamsterdam-devnet-8 (target: early August 2026).

Spec updates (vs devnet-7)

• EIP-2780: fold transfer log cost into TX_VALUE_COST (create-with-value cost changes) — EIPs#11997, execution-specs#3214 @hudem1
  feat(eip-2780): fold EIP-7708 transfer log cost into TX_VALUE_COST #12713
• EIP-7928: exclude the recipient when a runtime halt precedes — EIPs#11902 @hudem1
  It seems this was already done on our side in Devnet-7 tests and repricing #12476. It seems the EIP spec clarification came afterward to codify what implementations were already doing.
• EIP-7997: activation-block deployment + BAL recording — EIPs#11991, execution-specs#3200 @hudem1
  feat(evm): install EIP-7997 deterministic factory on fork activation #12714
• 👌 EIP-8037: rename "regular gas" to "execution gas" (terminology only) — EIPs#11998 @flcl42
  Rename EIP-8037 regular gas dimension to execution gas #12600
• - Benchmark-driven repricing updates to EIP-8037 / EIP-8038 — EIPs#12083, execution-specs#3293 @flcl42
  - Revised EIP-8038 values: COLD_STORAGE_ACCESS 2,100, ACCOUNT_WRITE 9,000, CALL_VALUE 11,300, CREATE_ACCESS 12,000, storage-clear refund 11,616, and access-list costs 2,900 / 2,000.
  - Derive access-list address and storage-key costs from cold access minus warm access — EIPs#12056, execution-specs#3271.
  Apply final EIP-8038 state-access repricing #12633

JSON-RPC (execution-apis)

• 👌 execution-apis#851: eth/debug block access list getter semantics @hudem1
  checked and everything seems already fine, nothing to do here
• 👌 execution-apis#852: callTracer output + EIP-8037 two-dimensional gas in tracing @hudem1
  feat: EIP-8037 two-dimensional gas tracing (stateGasTracer + callTracer) #12628
• 👌 execution-apis#854: eth_createAccessList gas-fee affordability clarification @hudem1
  test: eth_createAccessList affordability with omitted fee fields #12629

Actually those 2 items below are already existing bugs regarding glamsterdam :

• contract sender divergence @hudem1
  fix(simulate): skip EIP-3607 on the EIP-7928 BAL path in eth_simulateV1 #12691
• eth_simulateV1: EIP-7928 BAL execution pool bypasses the simulate tx-processor adapter, see eth_simulateV1: BAL path (EIP-7928) bypasses simulate adapter behaviours (gas default, GasCap, gas accounting) #12692 @hudem1
  fix(simulate): route the EIP-7928 BAL path through the simulate tx adapter (#12692) #12721
  fix(simulate): default a no-gas eth_simulateV1 call to the block gas budget (#12692) #12722

Networking

• 👌 discv5-only EL discovery (--Discovery.DiscoveryVersion V5): publish and refresh the eth forkid ENR entry at the fork — geth's dial-candidate filter drops nodes without it

Testing

• Update to tests-glamsterdam-devnet@v8.1.0 (release tracker) @flcl42
  Update Glamsterdam devnet tests to v8.1.0 #12634
• 👌 Txpool revalidation across the fork (pools full of pre-fork-priced txs; EIP-2780 intrinsic gas / create-with-value changes) @flcl42
  Revalidate the txpool across fork boundaries #12755
• Fix testing failing tests (like nonce related ones)@flcl42

Networking

• 👌 EIP-8070: eth/72 sparse blobpool (required) @flcl42
  Add sparse blob pool #11094
• 👌 EIP-8189: snap/2 BAL-Based State Healing (nice to have) @batrr
  feat(snap): add snap/2 protocol and bal healing #12578

Spec

• Support --network plataberget

---

👌 marks items that do not break compatibility with devnet-7, so you can create prs right into master