[AI Improvement] [Task] Replace arbitrary 5s sleep delays in storage-emulator-integration with active port polling

[joehan]

Resolves Buganizer b/563410680

Proposed Improvement

Replaces 4 consecutive sleep 5 calls in scripts/storage-emulator-integration/run.sh (20 seconds total of idle wait time) with an active port polling helper (wait_for_emulators_shutdown). The helper checks whether the storage emulator port (9199), hub ports (4400/4000), and auth emulator port (9099) have been released using nc -z, bash /dev/tcp, or inline Node socket checks with a 100ms poll interval up to a 5-second timeout.

This eliminates up to ~19 seconds of unnecessary idle delays during normal test execution while preventing EADDRINUSE port collision failures on slower machines.

Verification

• npm run build: Passed cleanly.
• npm run lint:quiet: Passed cleanly.
• npm run test:storage-emulator-integration: Executed full storage integration test suite (123 tests passing across all 5 sub-suites: internal, rules, import, multiple-targets, and conformance).

[gemini-code-assist]

Code Review

This pull request replaces static sleep delays in the storage emulator integration script with a dynamic port-polling function, wait_for_emulators_shutdown, to wait for emulators to release their ports. Feedback suggests optimizing this polling loop by using the Bash built-in $SECONDS variable to avoid spawning external date processes, and using : > /dev/tcp/... instead of echo > /dev/tcp/... to prevent sending unnecessary payload data to active emulators.

[hchzh]

Wondering if this is expected that each port will have the same timeout? For 5s timeout, 4 ports means total 4*5 seconds.

[joehan]

Great catch! Moved local end=$((SECONDS + timeout)) outside the port loop so that all ports share the single overall timeout budget, ensuring total wait time across all ports is strictly capped at timeout (5s) rather than multiplying per port.

[hchzh]

This check is on each loop, could we move it globally, so only need to check once?

[joehan]

Good suggestion! Moved the capability check outside the function to define is_port_in_use once globally, avoiding repeated command -v lookups on every loop iteration.