james-willis opened a new pull request, #841: URL: https://github.com/apache/sedona-db/pull/841
**This is an experimental PR to validate a hypothesis. Do not review or merge.** #838 fixed the macos-latest flake from #837 by adding a \`rustup toolchain install nightly --component rust-src\` step. The mechanism, as far as I can tell from CI logs, is unrelated to nightly itself: - The error in #837 (\`unexpected argument 'build' found\` from \`rustup-init\`) comes from the regular stable \`cargo build\` line in \`r/sedonadb/src/Makevars.in\`, not the wasm/nightly branch. Make echoes the entire \`if/then/else\` recipe as one line, which gave the misleading impression nightly was firing. - The configure script's \`cargo --version\` call reports \`rustup-init 1.29.0\` on failing runs, meaning the \`cargo\` on PATH is acting as the rustup installer rather than the toolchain proxy. - Failing runs all happened on cache-hit of \`Swatinem/rust-cache@v2\` with \`cache-bin: true\` (the default). My read: the cached \`~/.cargo/bin\` was saved on an older macos runner image generation and conflicts with the current one on restore. - After #838, the env hash that rust-cache uses changed (\`a124ad58\` → \`06097aa2\`) because the list of installed toolchains grew, evicting the bad cache. Post-#838 caches are healthy. This branch removes the nightly install and instead sets \`cache-bin: false\`, which prevents \`~/.cargo/bin\` from being cached at all. **Predicted outcome:** macos-latest passes consistently, no nightly toolchain required. If that holds, the follow-up PR is just this diff with a real commit message and #838 reverted. If it doesn't, my theory is wrong and #838 stays. Refs #837, #838. -- 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]
