Closed alxiord closed 4 years ago
What's the plan going forward?
- Merge just the bench tests in some crate.
linux-loader
orvm-memory
are good candidates. This is needed for the very first run of the test, when it tries tocargo bench
the upstream tip.- Merge this PR to have a starting point, and iterate on it.
- Figure out an acceptable tolerance for the bench numbers, or if we trust
critcmp
enough to give it authority to fail a PR. Use the printed output to get a sense of how the test behaves in the wild.- Update the test with PASS/FAIL conditions in a new PR.
Does it make sense?
Please also update the README.md to specify how to run this test and how it works.
Done, PTAL
24
This PR introduces a new integration test that can be included in per-crate Buildkite pipelines to run the crate's benchmark tests and assess the impact of a PR. The test runs
cargo bench
on the tip of the upstream crate and on theHEAD
of the PR and compares the results withcritcmp
. For the moment, the test only prints the results.