ipsilon/evmone - CodSpeed

evmone

Blog Docs Changelog

Performance History

Latest Results

test: Collect the test files in one place Both drivers walked the directory themselves, with the same recursive search, the same index.json skip, the same sort and the same suite naming written out twice. Only the registration at the end differed. collect_test_files() in testutils does the walk once and returns each file with the suite it belongs to. The listed tests are unchanged, name for name: 20245 for the state tool and 21333 for the blockchain one, in the same order.

test/collect-test-files

4 hours ago

ci: Build without testing in the min-compiler jobs (#1679) `EVMONE_TESTING` defaults to `OFF`, and that is the only configuration where GTest is never fetched — but every job funnels through a `build` command that passes `-DEVMONE_TESTING=ON`, so the default has never been built in CI. Nothing would notice if the code built under `EVMONE_TOOLS` grew a dependency on GTest, which matters now that `evmone::testutils` carries the fixture runners and #1676 relies on them staying GTest-free. `gcc-min` and `clang-min` now configure the default first, build it, run `evmone run` once to check the tool works at all, and then reconfigure with testing on in the same directory. Measured locally: the first pass compiles 54 objects, the second 111, and **none of the 54 is recompiled** — 165 distinct in total, the same as a single testing-on build. So this costs an extra CMake configure and a smoke test, not a second build.

master

4 hours ago

ci: Build without testing in the min-compiler jobs EVMONE_TESTING defaults to OFF and is the only configuration where GTest is never fetched, but every job passed it ON, so nothing noticed if the code built under EVMONE_TOOLS grew a dependency on GTest. That matters now that evmone::testutils carries the fixture runners. gcc-min and clang-min build the default configuration first, run the tool once to check it works at all, then reconfigure with testing on. The second pass reuses every object of the first, so this costs a configure, not a build.

ci/tools-only-build

6 hours ago

ci: Build without testing in the min-compiler jobs EVMONE_TESTING defaults to OFF and is the only configuration where GTest is never fetched, but every job passed it ON, so nothing noticed if the code built under EVMONE_TOOLS grew a dependency on GTest. That matters now that evmone::testutils carries the fixture runners. gcc-min and clang-min build the default configuration first, run the tool once to check it works at all, then reconfigure with testing on. The second pass reuses every object of the first, so this costs a configure, not a build.

ci/tools-only-build

9 hours ago

ci: Build without testing in the min-compiler jobs EVMONE_TESTING defaults to OFF and is the only configuration where GTest is never fetched, but every job passed it ON, so nothing noticed if the code built under EVMONE_TOOLS grew a dependency on GTest. That matters now that evmone::testutils carries the fixture runners. gcc-min and clang-min build the default configuration first, run the tool once to check it works at all, then reconfigure with testing on. The second pass reuses every object of the first, so this costs a configure, not a build.

ci/tools-only-build

10 hours ago

test: Move the fixture runners into evmone::testutils (#1676) Now that the runners no longer use gtest, they can live in `evmone::testutils` instead of in the test binaries. `statetest_runner.cpp` moves unchanged; `blockchaintest_runner.cpp` only swaps its own header for `blockchaintest.hpp`, where `run_blockchain_tests()` now sits next to `load_blockchain_tests()` — that retires `blockchaintest_runner.hpp`. `run_state_test()` ends up defined in the same target as the `statetest.hpp` declaration it implements. The point is the build configuration: `evmone::testutils` is built by `EVMONE_TOOLS` alone, where GTest is never fetched, so this is what lets the `evmone` CLI reach the runners later. Configuring with `EVMONE_TOOLS=ON, EVMONE_TESTING=OFF` builds both runner objects into testutils with no mention of GTest; before this commit that configuration had no runners at all.

master

12 hours ago

test: Fix test name Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

test/runners-in-testutils

14 hours ago

test: Cover the fixture runners' failure paths The suites the coverage job runs, unit and integration, reach little of the runners: blockchaintest_runner.cpp sat at 61% lines, with validate_block()'s rejections and the state dumps never executed. EEST does reject blocks on a green run, but never reports a disagreement, so the verdicts below a rejection are unreachable from it. Both runners now live in evmone::testutils, which evmone-unittests links, so a test can drive them directly with a fixture broken on purpose. This lifts them to 75% and 96%. tx_invalid_nonce moves out of the integration tests, which keep the remaining cases as end-to-end smoke.

test/runners-in-testutils

15 hours ago

Latest Branches

0%

test: Collect the test files in one place#1680

4 hours ago

7ac5e0b

test/collect-test-files

0%

ci: Build without testing in the min-compiler jobs#1679

10 hours ago

ea72099

ci/tools-only-build

0%

test: Move the fixture runners into evmone::testutils#1676

14 hours ago

649799b

test/runners-in-testutils

© 2026 CodSpeed Technology

Home Terms Privacy Docs