Improve integration test reliability and diagnostics

[David Negstad (danegsta)]

Recent CI failures exposed several reliability gaps: standard integration environments could still be shutting down after a test completed, the physical-volume deletion test could observe an intermediate controller state, and intermittent PTY failures did not report enough process information to identify the failing layer.

This change:

• makes standard test environments own cancellation and synchronously wait for API server disposal during test cleanup, with an explicit post-shutdown callback phase for teardown assertions
• waits for the complete expected terminating/creating volume state instead of asserting against the first observed deletion timestamp
• adds failure-only PTY diagnostics covering read attempts, bytes received, elapsed time, child PID, exit status, exit error, and captured output

The historical PTY empty-EOF failure was not reproduced locally. Sustained concurrent stress did expose a separate /dev/ptmx allocation-pressure failure, so the new diagnostics are intended to distinguish child execution, PTY delivery, and process-exit behavior the next time CI encounters the original failure.

Validation:

• make lint
• make test
• volume deletion test passed in ten separate test processes

[Copilot]

Copilot review overview

🟢 Approval recommended

The behavioral changes are sound; the identified cleanup duplication and error-message casing are non-blocking refinements.

Review effort: Balanced
Findings: 1 · 1

Open (2)

• Remove legacy deferred teardown waits · New
• Lowercase the PTY error prefix · New

What changed in this PR

Improves integration-test teardown reliability, volume-state synchronization, and PTY failure diagnostics.

Changes:

• Centralizes synchronous test-environment shutdown and post-shutdown assertions.
• Waits for the complete expected volume deletion state.
• Adds detailed PTY read and child-process diagnostics.

File	Description
test/​integration/​standard_test_env.go	Adds cleanup-owned environment shutdown and callbacks.
test/​integration/​controllers_common_test.go	Preserves manual lifecycle for TestMain.
test/​integration/​container_network_tunnel_proxy_test.go	Migrates teardown and post-shutdown assertions.
test/​integration/​container_controller_test.go	Uses cleanup-aware environment startup.
test/​integration/​executable_controller_test.go	Uses cleanup-aware environment startup.
test/​integration/​network_controller_test.go	Uses cleanup-aware environment startup.
test/​integration/​volume_controller_test.go	Uses cleanup-aware environment startup.
test/​integration/​v2_physical_container_controller_test.go	Migrates dedicated environment cleanup.
test/​integration/​v2_physical_container_image_controller_test.go	Migrates dedicated environment cleanup.
test/​integration/​v2_physical_container_network_controller_test.go	Uses cleanup-aware environment startup.
test/​integration/​v2_physical_container_volume_controller_test.go	Strengthens deletion-state synchronization.
internal/​termpty/​pty_test.go	Adds reusable PTY failure diagnostics.
internal/​termpty/​pty_unix_test.go	Applies diagnostics to Unix output tests.
internal/​termpty/​conpty_windows_test.go	Applies diagnostics to Windows output tests.

---

💡 Configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🔵 Needs a closer look

Context-timeout failures still bypass collection of the child process exit status and error.

Review effort: Balanced
Findings: None

Resolved since last review (2)

• Remove legacy deferred teardown waits
• Lowercase the PTY error prefix

Previously missed (1)

In code that hasn't changed since last review

Wait for process exit after context expiration read failure

internal/​termpty/​pty_test.go:194

When readUntil fails because ctx expired, this same ctx.Done() case is already ready, so the diagnostic path returns immediately without waiting for the context-driven process termination. That failure mode therefore omits the child exit status/error the helper is intended to capture. Keep the bounded exitTimer, but wait for either Exited() or that timer after a read failure.

[Copilot]

Copilot review overview

🟢 Approval recommended

The cleanup migration is consistent across callers, and no unresolved correctness issues were identified.

Review effort: Balanced
Findings: None