kevinjqliu opened a new pull request, #3520: URL: https://github.com/apache/iceberg-rust/pull/3520
## Which issue does this PR close? <!-- We generally require a GitHub issue to be filed for all bug fixes and enhancements and this helps us generate change logs for our releases. You can link an issue to this PR using the GitHub syntax. For example `Closes #123` indicates that this PR will close issue #123. --> - N/A (CI maintenance). ## What changes are included in this PR? <!-- Provide a summary of the modifications in this PR. List the main changes such as new features, bug fixes, refactoring, or any other updates. --> `rust-cache` hashes every installed toolchain into its cache key, including the runner's preinstalled stable. We build with the nightly in `rust-toolchain.toml` and never use stable. But runner images roll out with different stable versions (1.99.0 vs 1.99.1 this week, on Linux, macOS and Windows), so each job ends up with 2 cache keys that can't restore from each other. What that looks like over ~45 recent runs per workflow: - The less common image misses a lot more. macOS `build` missed 4/4, Windows bindings 7/10, `check-public-api` 64% vs 37%. - A miss rebuilds every dependency. For `check-public-api` that's ~6 min vs ~1 min. - The duplicate entries take ~2.8GB. We're at 10.7GB of the 10GB cache limit, so they push out caches we still need. Remove stable in `setup-builder` when it uses `rust-toolchain.toml`. Jobs that pass `rust-version` (audit, MSRV, publish, Python release) are unchanged. `rust-toolchain.toml` is still part of the key, so a nightly bump still invalidates the cache. [Apache Iggy does the same](https://redirect.github.com/apache/iggy/pull/4624) in its setup action. ## Are these changes tested? <!-- Specify what test covers (unit test, integration test, etc.). If tests are not included in your PR, please explain why (for example, are they covered by existing tests)? --> CI only. `actionlint` and `zizmor --offline` show no new findings. Checked locally that `rustup toolchain uninstall stable` exits 0 when stable is the default toolchain and when it's not installed. The hit rate only shows after this lands on `main` and the caches are re-saved. ## AI Disclosure <!-- https://iceberg.apache.org/contribute/#guidelines-for-ai-assisted-contributions --> Prepared with Copilot, including pulling the cache hit/miss numbers from run logs. Worth a look: that nothing in these jobs calls `cargo +stable` or relies on stable being installed. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
