Fix deadlock when UDP port is already in use

[greynewell]

Summary

• listenUDP now signals startup success/failure via a buffered channel instead of logging and silently returning
• If the port is already bound and FSWatch is disabled, Run() returns immediately with a clear error: UDP port 7734 already in use — is supermodel watch already running?
• If FSWatch is active and UDP fails, it logs a warning and continues (fs-watch is the primary trigger)

Root cause

When the previous watch process left port 7734 bound, listenUDP returned early from its goroutine. The Run() select loop had ctx.Done() as its only exit, but in the test scenario the process was killed — triggering Go's deadlock detector.

Test plan

• Kill a running supermodel watch, immediately rerun — should get a clear error instead of hanging
• supermodel watch with no prior instance — starts normally
• go test ./internal/files/... passes

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes
  • UDP listener startup now blocks until the listener reports success or failure, providing clearer error handling when the port is already in use.
  • Improved graceful degradation on UDP bind failure: startup returns an explicit error when file-watching is disabled, and logs a warning and continues when enabled.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 162d26f4-3f74-48a0-a217-73206594823f

📥 Commits

Reviewing files that changed from the base of the PR and between 81bbfb8 and d6d4e36.

📒 Files selected for processing (1)

• internal/files/daemon.go

---

Walkthrough

Run now synchronizes UDP listener startup using a buffered error channel. It waits for the listener to report bind success or an error; on bind failure it returns an error when FSWatch is disabled, or logs a warning and continues when FSWatch is enabled.

Changes

Cohort / File(s)	Summary
UDP listener startup internal/files/daemon.go	Reworked startup to use a buffered ready error channel. Run blocks until listenUDP sends nil (success) or an error. listenUDP now sends the bind error (or nil) instead of logging directly. Bind failure: return error when FSWatch is off; log warning and continue when FSWatch is on.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Poem

📡 A tiny channel hums, "I'm here" or "No,"
The daemon waits—no race, no shadowed woe.
If files watch sleeps, a crash will show the cue;
If it watches on, a warning sees us through. 🎶

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately describes the main issue being fixed: a deadlock that occurred when the UDP port was already in use.
Description check	✅ Passed	The description covers the key changes, root cause, and test plan, matching the template structure with detailed context about the fix.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/udp-port-deadlock

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@internal/files/daemon.go`:
- Around line 87-94: The current UDP bind error handling assumes every failure
is "address in use"; update the handler where d.listenUDP sends errors to
udpReady (the go routine started with go d.listenUDP(ctx, udpReady)) to inspect
the returned error before emitting the "already in use" message: use
errors.Is(err, syscall.EADDRINUSE) (or equivalent platform-aware check) and only
return the "port already in use — is `supermodel watch` already running?"
message when that specific error matches; for any other error, return or log a
generic "failed to start UDP listener" message including the real error and
include d.cfg.NotifyPort in context and preserve the existing FSWatch fallback
path that logs a warning when FSWatch is active.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: d061b9cd-a3f6-4db4-b48d-9e15d46ba2b9

📥 Commits

Reviewing files that changed from the base of the PR and between 1ca605e and 81bbfb8.

📒 Files selected for processing (1)

• internal/files/daemon.go