fix(ci): free runner disk before Android release build

[thomasluizon]

Problem

Android Release run #29030329772 failed during Build signed AAB with Gradle with:

System.IO.IOException: No space left on device : '/home/runner/actions-runner/cached/2.335.1/_diag/Worker_...log'

The runner exhausted its disk mid-bundleRelease. It ran out of space so hard it couldn't even flush its own diagnostic log, so the step log blob was never uploaded (gh run view --log returns BlobNotFound). This is a runner-capacity problem, not a code defect: the isolated mobile node_modules copy + expo prebuild + the full signed AAB (plus Sentry sourcemaps) overrun the ~14 GB free on a stock ubuntu-latest runner.

Fix

Add a first step that reclaims the unused preinstalled toolchains (.NET, Haskell, large apt packages, Docker images, swap) via jlumbroso/free-disk-space, pinned by commit SHA (matching this repo's SonarSource pinning convention).

Critically, android: false — the action removes /usr/local/lib/android/sdk by default, which this Gradle build depends on. All other reclamation targets keep their safe defaults, freeing ~15+ GB of headroom.

Verification

• Workflow YAML parses; new step is first, followed by the unchanged checkout.
• The only reachable check is re-dispatching the Android Release workflow (it's workflow_dispatch-only), which the maintainer runs to cut a build.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
orbit-ui-mobile-web	Ignored		Jul 9, 2026 4:08pm

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[claude]

Code Review: PR #435 — fix(ci): free runner disk before Android release build

Scope: PR #435 in thomasluizon/orbit-ui-mobile (.github/workflows/android-release.yml, +11/-0, single file)
Recommendation: APPROVE

Summary

The diff adds one step — jlumbroso/free-disk-space@54081f138730dfa15788a46383842cd2f914a1be (pinned by full commit SHA, matching the repo's existing SonarSource pinning convention) — as the first step in the Android Release workflow, before checkout, with android: false explicitly set so the Gradle build's dependency on the preinstalled Android SDK survives. The change is narrowly scoped to CI infrastructure, correctly placed (runs before any disk-hungry step), carries a WHY-with-URL comment explaining the runner-capacity root cause, and does not touch any application code, contract, or user-facing surface.

Findings

Critical

None

High

None

Medium

None

Low / Info

[Info] Swap-storage reclamation trades disk headroom for memory headroom
· dimension: 1 (Correctness) / 12 (Security — configuration)
· location: .github/workflows/android-release.yml:60
· issue: `swap-storage` is left at the action's default (removed). `./gradlew --no-daemon bundleRelease` (line ~230) is memory-hungry; removing swap means a JVM heap spike that used to page to swap now risks an OOM kill instead.
· risk: Low — ubuntu-latest runners have enough RAM that this is unlikely to bite, and the PR's own stated failure was disk, not memory. Flagging only as a thing to watch if a future run fails with an OOM instead of a disk error.
· fix: No action needed now; if a future release run OOMs, the first thing to try is `swap-storage: false` while keeping the disk reclamation.

Per the rubric's signal gate, this Info item is not a blocking finding — included for completeness only, not counted toward the recommendation.

Subagents

Agent	Verdict
parity-checker	N/A — diff touches only .github/workflows/android-release.yml, no apps/web/** or apps/mobile/** source file
i18n-syncer	N/A — no user-facing strings changed
contract-aligner	N/A — no packages/shared/src/types/* or endpoints.ts change, orbit-api not touched
security-reviewer	N/A — orbit-api not touched
design-reviewer	N/A — no apps/* UI files changed

None of the five subagents gate in for this diff — it is CI-workflow-only.

Validation

Phase 6 (/validate) skipped per this PR's CI adaptation — Build / Unit Tests / SonarCloud run as separate required checks.

Deferred — N/A dimensions & files not verdicted

• Dimensions 3 (SOLID/clean-arch), 6 (type safety), 7 (console.log), 8 (DESIGN.md/AI-slop), 9 (parity), 10 (i18n), 11 (contract drift), 13 (backend hard rules), 14 (FEATURES.md parity): all N/A — the diff is a single GitHub Actions YAML file with no application, contract, UI, or feature-surface code.
• Dimension 12 (security): API-side and frontend-XSS/auth-leakage categories N/A (no app or API code); the CI-supply-chain angle (pinning, android: false) was reviewed inline above.
• Not independently verifiable in this environment: whether commit SHA 54081f138730dfa15788a46383842cd2f914a1be actually corresponds to the jlumbroso/free-disk-space v1.3.1 tag claimed in the trailing comment — no outbound network access to GitHub's API for the third-party repo in this job. The pinning style (full SHA + version comment) matches the existing SonarSource precedent in sonarcloud.yml, so the practice is sound; only the specific SHA↔tag correspondence is unverified. Recommend a maintainer spot-check this once, not blocking.
• orbit-api not checked out in this job — no cross-repo dimension applies to this diff anyway (CI-workflow-only change).
• No other changed files — this is the full diff.

What's good

• Root-cause diagnosis is precise and cited (no space left on device in a diagnostic log write, not a code defect) — matches CLAUDE.md rule 1 (root cause over workarounds).
• WHY comment carries a real URL to the failed run, exactly the shape the comment-policy dimension asks for.
• android: false shows the author read the action's docs rather than accepting all defaults blind — the Android SDK removal would have broken the very build this fixes.
• SHA-pinning matches the repo's own established convention rather than a mutable tag/branch ref.
• Correct step placement (first, before checkout) so reclamation happens before anything else consumes disk.

Recommendation

Approve as-is. Nothing here needs a code change. Optional, non-blocking follow-up: after the next real release run, spot-check that the SHA pin still matches v1.3.1 and that swap-storage removal hasn't introduced a memory-pressure failure mode in place of the disk one.

Severity	Count
Critical (incl. ⚠️ old-client breaks)	0
High	0
Medium	0
Low / Info	1