feat(tooling): real ethrex block fixture

.github/workflows/pr_main.yaml

@@ -101,15 +101,33 @@ jobs:
         run: |
           cargo test --release -p executor test_ckzg -- --ignored
 
+      # The real-block fixture (~1 MB) is gitignored and generated on demand;
+      # test_ethrex_real_block_native below needs it present. This also
+      # downloads the ethrex-replay cache the converter's own tests read.
+      - name: Generate ethrex real-block fixture
+        run: make ethrex-real-block-fixture
+
+      # Detached workspace. Executes the block through `LambdaVmEcsmCrypto`, the
+      # guest's own `Crypto` impl, and pins the fixture's sha256. Note it does NOT
+      # screen KZG (its graph links c-kzg via ethrex-config — see that crate's
+      # Cargo.toml); the ethrex-tests step below is what covers 0x0a.
+      - name: Run ethrex real-block converter tests (detached workspace)
+        run: |
+          cd tooling/ethrex-real-block && cargo test --release
+
       # ethrex host-reference tests live in the detached `tooling/ethrex-tests`
       # workspace (ethrex pins rkyv's `unaligned` feature, which must not
       # feature-unify with the main workspace's aligned proof format), so run
       # them from that directory to use its isolated Cargo.lock. The guest ELF
       # and committed fixtures are already present from the steps above.
       # --include-ignored also runs the heavier synthetic-block test.
+      # `test_ethrex_real_block_vm` is skipped: it drives a real block through
+      # the VM and its runtime is not yet measured, so it stays opt-in rather
+      # than sitting in the PR gate. Run it manually on a build server.
       - name: Run ethrex host-reference tests (detached workspace)
         run: |
-          cd tooling/ethrex-tests && cargo test --release -- --include-ignored
+          cd tooling/ethrex-tests && \
+            cargo test --release -- --include-ignored --skip test_ethrex_real_block_vm
 
   test-cli:
     name: CLI tests

Makefile

@@ -1,12 +1,12 @@
 .PHONY: deps deps-linux deps-macos compile-programs-asm compile-programs-rust compile-bench \
 compile-programs compile-recursion-elfs clean-asm clean-rust clean-bench clean-shared \
 clean-recursion-elfs clean test test-asm \
-test-rust test-ethrex test-executor test-syscalls test-flamegraph flamegraph-prover test-profile-recursion test-profile-recursion-single test-profile-recursion-multi \
+test-rust test-ethrex test-ethrex-offline test-executor test-syscalls test-flamegraph flamegraph-prover test-profile-recursion test-profile-recursion-single test-profile-recursion-multi \
 test-profile-recursion-block recursion-profile-block-input \
 test-fast test-prover test-prover-all test-prover-debug test-disk-spill test-math-cuda test-cuda-integration test-cuda-fallback \
 test-prover-cuda test-prover-comprehensive-cuda \
 bench-math-cuda bench-prover bench-prover-cuda build check clippy fmt lint regen-ethrex-fixtures \
-update-ethrex-fixture-checksums check-ethrex-fixture-checksums
+update-ethrex-fixture-checksums check-ethrex-fixture-checksums ethrex-real-block-fixture
 
 UNAME := $(shell uname)
 
@@ -271,10 +271,45 @@ test-asm: compile-programs-asm
 test-rust: compile-programs-rust
 	cargo test -p executor --test rust
 
+# Real-block fixture: a genuine Hoodi block (4.4M gas, 11 txs, ~124 KB of
+# contract bytecode, 1705 state-trie nodes), as opposed to the synthetic
+# N-plain-transfer blocks from tooling/ethrex-fixtures. ~1 MB, so it is
+# generated on demand and gitignored rather than committed.
+#
+# The source is ethrex-replay's cache JSON, not that tool's own rkyv output:
+# ethrex-replay tracks ethrex `main`, where `ProgramInput` has diverged from the
+# rev our guest pins, so only the JSON is safe to read across the two.
+ETHREX_REAL_BLOCK := 1265656
+# Pinned by immutable `rev`, as the guest pins ethrex itself: a branch ref would
+# let the benchmark's input change under a fixed fixture name, silently breaking
+# comparability across runs. Re-pin deliberately when adopting a new block.
+ETHREX_REPLAY_REV := 2693e0182a8734117151d8ea2891eda5afc60383
+ETHREX_REAL_BLOCK_CACHE := tooling/ethrex-real-block/caches/cache_hoodi_$(ETHREX_REAL_BLOCK).json
+ETHREX_REAL_BLOCK_FIXTURE := executor/tests/ethrex_hoodi_$(ETHREX_REAL_BLOCK).bin
+
+ethrex-real-block-fixture: $(ETHREX_REAL_BLOCK_FIXTURE)
+
+$(ETHREX_REAL_BLOCK_CACHE):
+	mkdir -p $(dir $@)
+	curl -fsSL --retry 3 --retry-delay 2 --retry-all-errors -o $@.tmp \
+		https://raw.githubusercontent.com/lambdaclass/ethrex-replay/$(ETHREX_REPLAY_REV)/caches/cache_hoodi_$(ETHREX_REAL_BLOCK).json
+	mv $@.tmp $@
+
+$(ETHREX_REAL_BLOCK_FIXTURE): $(ETHREX_REAL_BLOCK_CACHE) tooling/ethrex-real-block/src/main.rs tooling/ethrex-real-block/Cargo.toml tooling/ethrex-real-block/Cargo.lock
+	cd tooling/ethrex-real-block && \
+		cargo run --release -- ../../$(ETHREX_REAL_BLOCK_CACHE) ../../$@
+
 # ethrex host-reference tests live in the detached `tooling/ethrex-tests`
 # workspace (ethrex pins rkyv's `unaligned` feature; isolated Cargo.lock).
-test-ethrex: compile-programs-rust
-	cd tooling/ethrex-tests && cargo test --release -- --include-ignored
+test-ethrex: compile-programs-rust $(ETHREX_REAL_BLOCK_FIXTURE)
+	cd tooling/ethrex-tests && cargo test --release -- --include-ignored --skip test_ethrex_real_block_vm
+
+# Offline variant: no network. `--skip test_ethrex_real_block` is a substring
+# match, so it drops both real-block tests — the `_vm` one and the `_native` one,
+# which reads the downloaded fixture and would otherwise fail on a clean checkout.
+# The committed synthetic fixtures and `no_kzg_backend_linked` still run.
+test-ethrex-offline: compile-programs-rust
+	cd tooling/ethrex-tests && cargo test --release -- --include-ignored --skip test_ethrex_real_block
 
 test-flamegraph:
 	cargo test -p executor --test flamegraph

executor/.gitignore

@@ -1,6 +1,8 @@
 /target
 /program_artifacts/rust
-/tests/ethrex_hoodi.bin
+# Real-block fixtures (~1 MB): generated by `make ethrex-real-block-fixture`
+# from an ethrex-replay cache, never committed. See tooling/ethrex-real-block.
+/tests/ethrex_hoodi*.bin
 /tests/ethrex_bench_*.bin
 # _4 is committed (~17 KB): used by `make recursion-profile-block-input` and
 # scripts/bench_recursion_scaling.sh. Other sizes stay ignored — scaling.sh

executor/tests/README.md

@@ -52,3 +52,13 @@ ethrex_10_transfers.bin
   sha256: 38901ee4d40b99cf0aa7f642a92f0fc8db76d974bf43033a1673839020c3c28e
   contents: stateless ethrex block with ten plain ETH transfer transactions
 ```
+
+## Real-block fixtures
+
+The blocks above are synthetic (N plain ETH transfers over a small genesis).
+For a representative workload — real contract execution, real trie depth, real
+bytecode — `make ethrex-real-block-fixture` generates
+`ethrex_hoodi_1265656.bin` from an ethrex-replay cache. It is ~1 MB, so it is
+gitignored and generated on demand rather than committed, and its checksum is
+pinned in `tooling/ethrex-real-block/README.md` rather than above (the
+checksum script only covers committed fixtures).

tooling/ethrex-real-block/.gitignore

@@ -0,0 +1,4 @@
+/target
+# Downloaded ethrex-replay cache JSONs (~1.5 MB each), fetched on demand by
+# `make ethrex-real-block-fixture`.
+/caches/