fix: remove vestigial state.json files from agent directories
Agent directories (witness/, refinery/, mayor/) contained state.json files with last_active timestamps that were never updated, making them stale and misleading. This change removes: - initAgentStates function that created vestigial state.json files - AgentState type and related Load/Save functions from config package - MayorStateValidCheck from doctor checks - requesting_* lifecycle verification (dead code - flags were never set) - FileStateJSON constant and MayorStatePath function Kept intact: - daemon/state.json (actively used for daemon runtime state) - crew/<name>/state.json (operational CrewWorker metadata) - Agent state tracking via beads (the ZFC-compliant approach) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
splendid
@@ -3,23 +3,15 @@ package doctor
|
||||
import (
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"os"
|
||||
"os/exec"
|
||||
"path/filepath"
|
||||
"strings"
|
||||
|
||||
"github.com/steveyegge/gastown/internal/session"
|
||||
)
|
||||
|
||||
// LifecycleHygieneCheck detects and cleans up stale lifecycle state.
|
||||
// This can happen when:
|
||||
// - Lifecycle messages weren't properly deleted after processing
|
||||
// - Agent state.json has stuck requesting_* flags
|
||||
// - Session was manually killed without clearing state
|
||||
// This can happen when lifecycle messages weren't properly deleted after processing.
|
||||
type LifecycleHygieneCheck struct {
|
||||
FixableCheck
|
||||
staleMessages []staleMessage
|
||||
stuckStateFiles []stuckState
|
||||
staleMessages []staleMessage
|
||||
}
|
||||
|
||||
type staleMessage struct {
|
||||
@@ -28,19 +20,13 @@ type staleMessage struct {
|
||||
From string
|
||||
}
|
||||
|
||||
type stuckState struct {
|
||||
stateFile string
|
||||
identity string
|
||||
flag string
|
||||
}
|
||||
|
||||
// NewLifecycleHygieneCheck creates a new lifecycle hygiene check.
|
||||
func NewLifecycleHygieneCheck() *LifecycleHygieneCheck {
|
||||
return &LifecycleHygieneCheck{
|
||||
FixableCheck: FixableCheck{
|
||||
BaseCheck: BaseCheck{
|
||||
CheckName: "lifecycle-hygiene",
|
||||
CheckDescription: "Check for stale lifecycle messages and stuck state flags",
|
||||
CheckDescription: "Check for stale lifecycle messages",
|
||||
},
|
||||
},
|
||||
}
|
||||
@@ -49,36 +35,21 @@ func NewLifecycleHygieneCheck() *LifecycleHygieneCheck {
|
||||
// Run checks for stale lifecycle state.
|
||||
func (c *LifecycleHygieneCheck) Run(ctx *CheckContext) *CheckResult {
|
||||
c.staleMessages = nil
|
||||
c.stuckStateFiles = nil
|
||||
|
||||
var details []string
|
||||
|
||||
// Check for stale lifecycle messages in deacon inbox
|
||||
staleCount := c.checkDeaconInbox(ctx)
|
||||
if staleCount > 0 {
|
||||
details = append(details, fmt.Sprintf("%d stale lifecycle message(s) in deacon inbox", staleCount))
|
||||
}
|
||||
|
||||
// Check for stuck requesting_* flags in state files
|
||||
stuckCount := c.checkStateFiles(ctx)
|
||||
if stuckCount > 0 {
|
||||
details = append(details, fmt.Sprintf("%d agent(s) with stuck requesting_* flags", stuckCount))
|
||||
}
|
||||
|
||||
total := staleCount + stuckCount
|
||||
if total == 0 {
|
||||
if staleCount == 0 {
|
||||
return &CheckResult{
|
||||
Name: c.Name(),
|
||||
Status: StatusOK,
|
||||
Message: "No stale lifecycle state found",
|
||||
Message: "No stale lifecycle messages found",
|
||||
}
|
||||
}
|
||||
|
||||
return &CheckResult{
|
||||
Name: c.Name(),
|
||||
Status: StatusWarning,
|
||||
Message: fmt.Sprintf("Found %d lifecycle hygiene issue(s)", total),
|
||||
Details: details,
|
||||
Message: fmt.Sprintf("Found %d stale lifecycle message(s) in deacon inbox", staleCount),
|
||||
FixHint: "Run 'gt doctor --fix' to clean up",
|
||||
}
|
||||
}
|
||||
@@ -121,139 +92,7 @@ func (c *LifecycleHygieneCheck) checkDeaconInbox(ctx *CheckContext) int {
|
||||
return len(c.staleMessages)
|
||||
}
|
||||
|
||||
// checkStateFiles looks for stuck requesting_* flags in state.json files.
|
||||
func (c *LifecycleHygieneCheck) checkStateFiles(ctx *CheckContext) int {
|
||||
stateFiles := c.findStateFiles(ctx.TownRoot)
|
||||
|
||||
for _, sf := range stateFiles {
|
||||
data, err := os.ReadFile(sf.path)
|
||||
if err != nil {
|
||||
continue
|
||||
}
|
||||
|
||||
var state map[string]interface{}
|
||||
if err := json.Unmarshal(data, &state); err != nil {
|
||||
continue
|
||||
}
|
||||
|
||||
// Check for any requesting_* flags
|
||||
for key, val := range state {
|
||||
if strings.HasPrefix(key, "requesting_") {
|
||||
if boolVal, ok := val.(bool); ok && boolVal {
|
||||
// Found a stuck flag - verify session is actually healthy
|
||||
if c.isSessionHealthy(sf.identity, ctx.TownRoot) {
|
||||
c.stuckStateFiles = append(c.stuckStateFiles, stuckState{
|
||||
stateFile: sf.path,
|
||||
identity: sf.identity,
|
||||
flag: key,
|
||||
})
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return len(c.stuckStateFiles)
|
||||
}
|
||||
|
||||
type stateFileInfo struct {
|
||||
path string
|
||||
identity string
|
||||
}
|
||||
|
||||
// findStateFiles locates all state.json files for agents.
|
||||
func (c *LifecycleHygieneCheck) findStateFiles(townRoot string) []stateFileInfo {
|
||||
var files []stateFileInfo
|
||||
|
||||
// Mayor state
|
||||
mayorState := filepath.Join(townRoot, "mayor", "state.json")
|
||||
if _, err := os.Stat(mayorState); err == nil {
|
||||
files = append(files, stateFileInfo{path: mayorState, identity: "mayor"})
|
||||
}
|
||||
|
||||
// Scan rigs for witness, refinery, and crew state files
|
||||
entries, err := os.ReadDir(townRoot)
|
||||
if err != nil {
|
||||
return files
|
||||
}
|
||||
|
||||
for _, entry := range entries {
|
||||
if !entry.IsDir() || strings.HasPrefix(entry.Name(), ".") || entry.Name() == "mayor" {
|
||||
continue
|
||||
}
|
||||
|
||||
rigName := entry.Name()
|
||||
rigPath := filepath.Join(townRoot, rigName)
|
||||
|
||||
// Witness state
|
||||
witnessState := filepath.Join(rigPath, "witness", "state.json")
|
||||
if _, err := os.Stat(witnessState); err == nil {
|
||||
files = append(files, stateFileInfo{
|
||||
path: witnessState,
|
||||
identity: rigName + "-witness",
|
||||
})
|
||||
}
|
||||
|
||||
// Refinery state
|
||||
refineryState := filepath.Join(rigPath, "refinery", "state.json")
|
||||
if _, err := os.Stat(refineryState); err == nil {
|
||||
files = append(files, stateFileInfo{
|
||||
path: refineryState,
|
||||
identity: rigName + "-refinery",
|
||||
})
|
||||
}
|
||||
|
||||
// Crew state files
|
||||
crewPath := filepath.Join(rigPath, "crew")
|
||||
crewEntries, err := os.ReadDir(crewPath)
|
||||
if err != nil {
|
||||
continue
|
||||
}
|
||||
for _, crew := range crewEntries {
|
||||
if !crew.IsDir() || strings.HasPrefix(crew.Name(), ".") {
|
||||
continue
|
||||
}
|
||||
crewState := filepath.Join(crewPath, crew.Name(), "state.json")
|
||||
if _, err := os.Stat(crewState); err == nil {
|
||||
files = append(files, stateFileInfo{
|
||||
path: crewState,
|
||||
identity: rigName + "-crew-" + crew.Name(),
|
||||
})
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return files
|
||||
}
|
||||
|
||||
// isSessionHealthy checks if the tmux session for this identity exists and is running.
|
||||
func (c *LifecycleHygieneCheck) isSessionHealthy(identity, _ string) bool {
|
||||
sessionName := identityToSessionName(identity)
|
||||
if sessionName == "" {
|
||||
return false
|
||||
}
|
||||
|
||||
// Check if session exists
|
||||
cmd := exec.Command("tmux", "has-session", "-t", sessionName)
|
||||
return cmd.Run() == nil
|
||||
}
|
||||
|
||||
// identityToSessionName converts an identity to its tmux session name.
|
||||
func identityToSessionName(identity string) string {
|
||||
switch identity {
|
||||
case "mayor":
|
||||
return session.MayorSessionName()
|
||||
default:
|
||||
if strings.HasSuffix(identity, "-witness") ||
|
||||
strings.HasSuffix(identity, "-refinery") ||
|
||||
strings.Contains(identity, "-crew-") {
|
||||
return "gt-" + identity
|
||||
}
|
||||
return ""
|
||||
}
|
||||
}
|
||||
|
||||
// Fix cleans up stale lifecycle state.
|
||||
// Fix cleans up stale lifecycle messages.
|
||||
func (c *LifecycleHygieneCheck) Fix(ctx *CheckContext) error {
|
||||
var errors []string
|
||||
|
||||
@@ -266,39 +105,8 @@ func (c *LifecycleHygieneCheck) Fix(ctx *CheckContext) error {
|
||||
}
|
||||
}
|
||||
|
||||
// Clear stuck requesting_* flags
|
||||
for _, stuck := range c.stuckStateFiles {
|
||||
if err := c.clearRequestingFlag(stuck); err != nil {
|
||||
errors = append(errors, fmt.Sprintf("failed to clear %s in %s: %v", stuck.flag, stuck.identity, err))
|
||||
}
|
||||
}
|
||||
|
||||
if len(errors) > 0 {
|
||||
return fmt.Errorf("%s", strings.Join(errors, "; "))
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// clearRequestingFlag removes the stuck requesting_* flag from a state file.
|
||||
func (c *LifecycleHygieneCheck) clearRequestingFlag(stuck stuckState) error {
|
||||
data, err := os.ReadFile(stuck.stateFile)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
var state map[string]interface{}
|
||||
if err := json.Unmarshal(data, &state); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
// Remove the requesting flag and any associated timestamp
|
||||
delete(state, stuck.flag)
|
||||
delete(state, "requesting_time")
|
||||
|
||||
newData, err := json.MarshalIndent(state, "", " ")
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
return os.WriteFile(stuck.stateFile, newData, 0644)
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user