test(build): fix data race in TestBuildLayer under -race

[aftersnow]

Summary

Fix a data race in TestBuilderSuite/TestBuildLayer reported by go test -race ./pkg/backend/build/.... Test-only change; production code is untouched.

Root cause

BuildLayer hands OutputLayer a live *io.PipeReader whose backing tar-writer goroutine (archiver.Tar) is mid-write. The subtest uses the generated testify mock for OutputStrategy, and testify's mock.Arguments.Diff formats every actual argument with fmt (fmt.Sprintf("(%[1]T=%[1]v)", actual)) to match the call. Formatting the *io.PipeReader reflects into the pipe's internal sync.Mutex state concurrently with the writer goroutine's Lock() — the reported race.

Write by goroutine (tar): io.(*pipe).write -> sync.(*Mutex).Lock
Read  by goroutine (test): fmt.Sprintf -> reflect.Value.Interface  <- testify Arguments.Diff

No mock matcher avoids it: Diff formats the actual argument before any matcher logic runs, so mock.Anything / AnythingOfType / MatchedBy make no difference.

Fix

Replace the testify mock in the "successful build layer" subtest with a small hand-written OutputStrategy that reads the reader (exactly as the real localOutput/remoteOutput strategies do). This avoids reflecting over the live pipe and unblocks the writer goroutine, while still asserting the call arguments and returned descriptor.

Notes

• Production is unchanged. remoteOutput.OutputLayer relies on the reader's concrete *io.PipeReader type (to drain it when a blob already exists), so wrapping the reader was deliberately avoided.
• Pre-existing: this race is present on main today and is not currently caught by CI (the test targets don't pass -race).

Test plan

• go test -race -count=20 -run TestBuilderSuite/TestBuildLayer ./pkg/backend/build/... — clean (was failing).
• go test -race ./... — all packages clean.
• go vet ./... / gofmt clean.

[gemini-code-assist]

Code Review

This pull request replaces a testify mock with a hand-written recordingStrategy in builder_test.go to resolve a data race caused by concurrent formatting of a live *io.PipeReader. The feedback recommends restoring the original s.builder.strategy at the end of the subtest to prevent test pollution, and initializing the embedded OutputStrategy interface to avoid potential nil pointer dereference panics.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

Overwriting the shared suite-level s.builder.strategy field inside a subtest without restoring it can lead to test pollution and unexpected side effects in subsequent subtests or other tests.

Additionally, since recordingStrategy embeds the OutputStrategy interface anonymously, leaving it uninitialized (nil) means any unexpected calls to other methods (like OutputConfig or OutputManifest) will cause a nil pointer dereference panic.

To make the test more robust and isolated, initialize the embedded OutputStrategy with s.mockOutputStrategy and use defer to restore the original strategy after the subtest completes.

Suggested change

	strategy := &recordingStrategy{desc: expectedDesc}
	s.builder.strategy = strategy
	strategy := &recordingStrategy{
	OutputStrategy: s.mockOutputStrategy,
	desc: expectedDesc,
	}
	s.builder.strategy = strategy
	defer func() {
	s.builder.strategy = s.mockOutputStrategy
	}()