fix: wait for image conversion before reporting build ready (KERNEL-863)

[hiroTamada]

Summary

• Fixes race condition where build reports "ready" before image conversion completes
• Adds waitForImageReady() to poll image status before marking build complete
• If image conversion fails or times out, build is marked as failed

Problem

The registry's triggerConversion() runs asynchronously after returning 201 to the builder:

if wrapper.statusCode == http.StatusCreated {
    go r.triggerConversion(...)  // ASYNC!
}

This caused a race where:

1. Builder pushes image → Registry returns 201
2. Builder reports success → Build status = "ready"
3. User creates instance → FAILS (image still converting)

Solution

Build manager now waits for image to be ready before reporting build complete:

// Wait for image to be ready before reporting build as complete
if err := m.waitForImageReady(ctx, id); err != nil {
    // Mark build as failed
}
m.updateBuildComplete(id, StatusReady, ...)

Test plan

• TestWaitForImageReady_Success - image already ready
• TestWaitForImageReady_WaitsForConversion - polls until ready
• TestWaitForImageReady_ContextCancelled - respects context timeout
• TestWaitForImageReady_Failed - handles failed conversion
• All existing builds tests pass

Fixes KERNEL-863

---

Note

Addresses a race where builds reported ready before image conversion finished.

• Integrates images.Manager into builds manager and wiring (ProvideBuildManager, wire_gen.go)
• Adds waitForImageReady() polling and invokes it in runBuild before setting status to ready; uses build context for accurate metrics/timeout
• Updates metrics recording to use buildCtx
• Adds tests: race demonstration and waitForImageReady success, conversion wait, context-cancel timeout, and failed conversion

^{Written by Cursor Bugbot for commit 450d496. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

[cursor]

Test has data race when modifying image status

Low Severity

TestWaitForImageReady_WaitsForConversion has a data race: the goroutine at lines 105-110 directly writes to imageMgr.images[imageRef].Status while waitForImageReady concurrently reads the same Status field through GetImage. The mock's GetImage returns a pointer to the same Image struct without copying, so both goroutines access the same memory location without synchronization. This will cause failures when running tests with -race.

 

[hiroTamada]

this is fine for now. We are not running test with race.