ci: free runner disk space before heavy cargo builds

[jensholdgaard]

Problem

cargo test --all-features on main started failing with disk exhaustion mid-link:

error: failed to write to `.../full.rmeta`: No space left on device (os error 28)
rustc-LLVM ERROR: IO failure on output stream: No space left on device
collect2: fatal error: ld terminated with signal 7 [Bus error], core dumped

The DataFusion-heavy workspace plus a growing set of integration-test binaries — all linked with full debug info — outgrew ubuntu-latest's ~14 GiB root disk. #243 added one more test binary (rfc0014_ingest_write_path) and tipped target/ over the edge. Two consecutive post-merge re-runs hit the identical wall, so this is a capacity problem, not the transient rust-lld bus-error flake (the bus error here is a symptom of the full disk).

The job never reached a single test — it failed while compiling/linking. Locally the suite is green (cargo test -p ourios-ingester --all-features passes; the merged code is correct).

Fix

Reclaim the large preinstalled SDKs this Rust workspace never uses (Android ~9 GiB, dotnet, GHC, CodeQL, boost) via a shared .github/scripts/free-disk-space.sh, run before the build on the two heavy whole-workspace --all-features jobs (test, coverage). This buys ~12 GiB of headroom — also necessary ahead of RFC 0014 part 2, which adds further test binaries.

No third-party action is introduced (just sudo rm + df), so the SHA-pinned-actions / OpenSSF Scorecard posture (#211–#213) is unchanged.

Verification

• ci.yml validates as YAML; free-disk-space.sh passes bash -n and shellcheck.
• The test job on this PR exercises the fix end-to-end (this is the gate that was failing).

No invariant (§3) or hazard (§4) is touched — CI infrastructure only.

🤖 Generated with Claude Code

Summary by CodeRabbit

• Chores
  • Enhanced CI/build reliability by optimizing disk space usage during automated testing and coverage reporting.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 23d5a93b-1d40-427b-9043-e828c63cc9d5

📥 Commits

Reviewing files that changed from the base of the PR and between 5661984 and 9ff4571.

📒 Files selected for processing (2)

• .github/scripts/free-disk-space.sh
• .github/workflows/ci.yml

---

📝 Walkthrough

Walkthrough

A new bash script .github/scripts/free-disk-space.sh is added that removes several large preinstalled SDK and toolchain directories on Ubuntu runners using best-effort sudo rm -rf. The script is wired as an early step in both the test and coverage CI workflow jobs.

Changes

CI Disk Space Reclamation

Layer / File(s)	Summary
Disk space cleanup script .github/scripts/free-disk-space.sh	New script with set -euo pipefail that reports disk usage, then best-effort removes Android SDK, dotnet, GHC, ghcup, CodeQL, and boost directories via sudo rm -rf ... || true, then re-reports usage.
CI job integration .github/workflows/ci.yml	Adds a Free runner disk space step after checkout and before Rust toolchain setup in both the test job (line 70) and the coverage job (line 223).

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

Poem

🐇 Hop hop, clear the way!
Android and dotnet, swept away,
GHC and ghcup — gone today,
Cargo needs room to compile and play.
More gigabytes for Rust, hooray! 🦀

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main change: adding a disk-freeing step before Cargo builds in CI.
Description check	✅ Passed	The description provides context (problem statement with error logs), explains the root cause, describes the solution, and mentions verification steps, covering all essential information.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch ci/free-disk-space-heavy-jobs

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[Copilot]

Pull request overview

This PR mitigates GitHub-hosted runner disk exhaustion during heavy Rust builds by reclaiming unused preinstalled SDK space before running workspace-wide --all-features compilation/link steps.

Changes:

• Add a shared .github/scripts/free-disk-space.sh that removes large, unused SDK/tool directories and prints df before/after.
• Invoke the script at the start of the test and coverage jobs in CI to increase available disk headroom before cargo test --all-features / cargo llvm-cov.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
.github/workflows/ci.yml	Runs the new disk-reclamation script early in the two heaviest jobs (test, coverage) before toolchain install/cache/build.
.github/scripts/free-disk-space.sh	Best-effort sudo rm -rf of large unused directories (Android, dotnet, GHC, CodeQL, boost) to recover runner disk space and avoid link-time ENOSPC failures.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.