woods 2.0.1 → 2.1.0

data/lib/woods/watch/supervisor.rb

@@ -25,7 +25,7 @@ module Woods
       def initialize(command:, root:, env: ENV.to_h, logger: $stderr, boot_timeout: 300,
                      shutdown_timeout: 10, retry_delays: RETRY_DELAYS)
         raise ArgumentError, 'WOODS_WATCH_IDLE_TIMEOUT must be unset for managed watching' unless
-          env['WOODS_WATCH_IDLE_TIMEOUT'].to_s.empty?
+          env['WOODS_WATCH_IDLE_TIMEOUT'].to_s.strip.empty?
 
         @command = command
         @root = File.realpath(root)

data/lib/woods/watch/tree_scan.rb

@@ -2,6 +2,7 @@
 
 require 'find'
 require 'set'
+require 'woods/source_path_encoding'
 
 module Woods
   module Watch
@@ -48,12 +49,15 @@ module Woods
       # @yieldparam path [String] absolute path to a regular file
       # @return [void]
       def each_file(root:, ignored:, visited: nil, &block)
-        base = root.to_s
+        base = SourcePathEncoding.utf8(root)
+        return unless base
+
         prefix = "#{base}/"
         visited = (visited || Set.new).dup
         visited << real_dir(base)
 
         Find.find(base) do |path|
+          path = decoded_entry(path)
           next if path == base
 
           skip = skip?(path.delete_prefix(prefix), ignored)
@@ -65,6 +69,11 @@ module Woods
         nil
       end
 
+      # Prune before any string matching or descent into an undecodable entry.
+      def decoded_entry(path)
+        SourcePathEncoding.utf8(path) || Find.prune
+      end
+
       # Classify one entry and act on it.
       #
       # @return [Symbol, nil] `:prune` when the caller should not descend
@@ -126,7 +135,7 @@ module Woods
 
       # @return [String, nil] resolved path, or nil for a broken/looping link
       def real_dir(path)
-        File.realpath(path)
+        SourcePathEncoding.utf8(File.realpath(path))
       rescue SystemCallError
         nil
       end
@@ -147,6 +156,9 @@ module Woods
       # @param relative [String] path relative to the watched root
       # @return [Boolean] whether any segment is a dotfile Woods does not read
       def hidden?(relative)
+        relative = SourcePathEncoding.utf8(relative)
+        return true unless relative
+
         relative.split(File::SEPARATOR).any? { |segment| hidden_segment?(segment) }
       end
 

data/plugin/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "woods-plugin",
   "description": "Woods user guides for Claude Code: set up, upgrade, configure MCP servers for, investigate codebases with, enable a repository's agents on, and diagnose the Woods Rails code-intelligence gem.",
-  "version": "2.3.54",
+  "version": "2.3.71",
   "author": {
     "name": "lost-in-the"
   },

data/plugin/hooks/adapters/normalize.rb

@@ -55,8 +55,9 @@ end
 
 if $PROGRAM_NAME == __FILE__
   begin
-    puts JSON.generate(WoodsHookEvent.normalize(ARGV.fetch(0), JSON.parse($stdin.read(1_048_577))))
-  rescue JSON::ParserError, ArgumentError, KeyError, TypeError
+    input = $stdin.binmode.read(1_048_577).force_encoding(Encoding::UTF_8)
+    puts JSON.generate(WoodsHookEvent.normalize(ARGV.fetch(0), JSON.parse(input)))
+  rescue JSON::ParserError, EncodingError, ArgumentError, KeyError, TypeError
     warn '[Woods hooks] Unsupported or malformed edit event; no refresh queued.'
     exit 1
   end

data/plugin/hooks/woods-input-rules.sh

@@ -111,6 +111,10 @@ woods_input_action() {
     case "$operation" in delete|move) printf full ;; *) printf incremental ;; esac
     return
   fi
+  if { { [[ "$path" != */tasks/* ]] && [[ "$path" != */generators/* ]] && { [[ "$path" == *.rb ]]; } && { [[ "$path" == lib/* ]]; }; }; }; then
+    case "$operation" in delete|move) printf full ;; *) printf incremental ;; esac
+    return
+  fi
   if { { { [[ "$path" == *.rake ]]; } && { [[ "$path" == lib/tasks/* ]]; }; }; }; then
     case "$operation" in delete|move) printf full ;; *) printf incremental ;; esac
     return

data/plugin/hooks/woods-refresh.sh

@@ -21,18 +21,26 @@ if command -v jq >/dev/null 2>&1; then
     printf '[Woods hooks] Unsupported or malformed edit event; no refresh queued.\n' >&2; exit 0;
   }
 elif command -v ruby >/dev/null 2>&1; then
-  payload="$(ruby "$hook_dir/adapters/normalize.rb" "$client" <"$input_file")" || exit 0
+  payload="$(ruby -EUTF-8:UTF-8 "$hook_dir/adapters/normalize.rb" "$client" <"$input_file")" || exit 0
 else
   printf '[Woods hooks] jq or Ruby is required; no refresh queued.\n' >&2
   exit 0
 fi
 rm -f "$input_file"
 trap - EXIT TERM INT
+# JSON and client paths are UTF-8 regardless of the host process locale.
+# Keep malformed queue/input diagnostics bounded and free of event contents.
+ruby_json() {
+  ruby -EUTF-8:UTF-8 "$@" 2>/dev/null || {
+    printf '[Woods hooks] Invalid UTF-8 or JSON data; refresh not confirmed.\n' >&2
+    return 1
+  }
+}
 field() {
   if command -v jq >/dev/null 2>&1; then
     printf '%s' "$payload" | jq -ej '.root'
   else
-    printf '%s' "$payload" | ruby -rjson -e 'print JSON.parse($stdin.read).fetch("root")'
+    printf '%s' "$payload" | ruby_json -rjson -e 'print JSON.parse($stdin.read).fetch("root")'
   fi
 }
 cwd="$(field; printf .)"
@@ -50,7 +58,7 @@ event_fields() {
   if command -v jq >/dev/null 2>&1; then
     printf '%s' "$payload" | jq -j '.events[] | .operation,"\u0000",.path,"\u0000"'
   else
-    printf '%s' "$payload" | ruby -rjson -e 'JSON.parse($stdin.read).fetch("events").each { |e| print e.fetch("operation"), "\0", e.fetch("path"), "\0" }'
+    printf '%s' "$payload" | ruby_json -rjson -e 'JSON.parse($stdin.read).fetch("events").each { |e| print e.fetch("operation"), "\0", e.fetch("path"), "\0" }'
   fi
 }
 relevant=0
@@ -83,7 +91,7 @@ enqueue() {
   if command -v jq >/dev/null 2>&1; then
     jq -nc --arg path "$1" '{path:$path,operation:"update"}' >"$event.tmp" || return 1
   else
-    ruby -rjson -e 'puts JSON.generate(path: ARGV[0], operation: "update")' -- "$1" >"$event.tmp" || return 1
+    ruby_json -rjson -e 'puts JSON.generate(path: ARGV[0], operation: "update")' -- "$1" >"$event.tmp" || return 1
   fi
   mv "$event.tmp" "$event.json"
 }
@@ -92,7 +100,7 @@ event="$queue/$$.$RANDOM.$RANDOM"
 if command -v jq >/dev/null 2>&1; then
   printf '%s' "$payload" | jq -c --arg root "$cwd" '.events | map(.path |= (if startswith($root + "/") then ltrimstr($root + "/") else . end))' >"$event.tmp" || exit 0
 else
-  printf '%s' "$payload" | ruby -rjson -e 'v = JSON.parse($stdin.read); puts JSON.generate(v.fetch("events").map { |e| e.merge("path" => e.fetch("path").delete_prefix(ARGV.fetch(0) + "/")) })' -- "$cwd" >"$event.tmp" || exit 0
+  printf '%s' "$payload" | ruby_json -rjson -e 'v = JSON.parse($stdin.read); puts JSON.generate(v.fetch("events").map { |e| e.merge("path" => e.fetch("path").delete_prefix(ARGV.fetch(0) + "/")) })' -- "$cwd" >"$event.tmp" || exit 0
 fi
 mv "$event.tmp" "$event.json" || exit 0
 
@@ -144,7 +152,7 @@ encode_batch() {
   if command -v jq >/dev/null 2>&1; then
     jq -sc --arg output "$configured_output" '{version:1,output:$output,events:(map(if type == "array" then . else [.] end) | add)} | @base64' "${batch[@]}" | tr -d '"\n'
   else
-    ruby -rjson -rbase64 -e '
+    ruby_json -rjson -rbase64 -e '
       output = ARGV.shift
       print Base64.strict_encode64(JSON.generate(version: 1, output: output,
         events: ARGV.flat_map { |path| value = JSON.parse(File.read(path)); value.is_a?(Array) ? value : [value] }))
@@ -237,7 +245,7 @@ while :; do
     if command -v jq >/dev/null 2>&1; then
       event_count="$(jq 'if type == "array" then length else 1 end' "$candidate")" || break
     else
-      event_count="$(ruby -rjson -e 'v = JSON.parse(File.read(ARGV[0])); puts(v.is_a?(Array) ? v.size : 1)' "$candidate")" || break
+      event_count="$(ruby_json -rjson -e 'v = JSON.parse(File.read(ARGV[0])); puts(v.is_a?(Array) ? v.size : 1)' "$candidate")" || break
     fi
     [ "$((batch_events + event_count))" -le 1000 ] || break
     [ "$((batch_bytes + event_bytes))" -le 49152 ] || break

data/plugin/hooks/woods-session-start.sh

@@ -11,9 +11,59 @@ field() {
   if command -v jq >/dev/null 2>&1; then
     printf '%s' "$1" | jq -r --arg name "$2" '.[$name] // empty' 2>/dev/null
   elif command -v ruby >/dev/null 2>&1; then
-    printf '%s' "$1" | ruby -rjson -e 'print JSON.parse($stdin.read).fetch(ARGV[0], "")' -- "$2" 2>/dev/null
+    printf '%s' "$1" | ruby -EUTF-8:UTF-8 -rjson -e 'print JSON.parse($stdin.read).fetch(ARGV[0], "")' -- "$2" 2>/dev/null
   fi
 }
+unavailable_summary() {
+  if command -v jq >/dev/null 2>&1; then
+    printf '%s' "$1" | jq -r '.unavailable | select(type == "object") |
+      select(.reason == "source_manifest_too_large" and (.size_bytes | type) == "number" and
+             (.limit_bytes | type) == "number") |
+      select(.size_bytes > .limit_bytes and .limit_bytes > 0) |
+      "source_manifest_too_large; \(.size_bytes) bytes; limit \(.limit_bytes)"' 2>/dev/null
+  elif command -v ruby >/dev/null 2>&1; then
+    printf '%s' "$1" | ruby -EUTF-8:UTF-8 -rjson -e '
+      value = JSON.parse($stdin.read)["unavailable"]
+      if value.is_a?(Hash) && value["reason"] == "source_manifest_too_large"
+        size, limit = value.values_at("size_bytes", "limit_bytes")
+        if [size, limit].all?(Integer) && size > limit && limit > 0
+          print "source_manifest_too_large; #{size} bytes; limit #{limit}"
+        end
+      end
+    ' 2>/dev/null
+  fi
+}
+source_advice() {
+  if command -v jq >/dev/null 2>&1; then
+    printf '%s' "$1" | jq -r 'if (.recommendations | type) == "array" then
+      .recommendations[] | select(. == "deep_check" or . == "fresh_capture" or . == "inspect_source_scan")
+      else empty end' 2>/dev/null
+  elif command -v ruby >/dev/null 2>&1; then
+    printf '%s' "$1" | ruby -EUTF-8:UTF-8 -rjson -e '
+      values = JSON.parse($stdin.read)["recommendations"]
+      allowed = %w[deep_check fresh_capture inspect_source_scan]
+      puts values.select { |value| allowed.include?(value) } if values.is_a?(Array)
+    ' 2>/dev/null
+  fi
+}
+unknown_message() {
+  local advice action
+  advice="$(source_advice "$1")"
+  if [ -z "$advice" ]; then
+    echo 'Woods source freshness is unknown: this generation lacks complete verified source evidence. Inspect woods_status; use woods-extract full for a fresh capture.'
+    return
+  fi
+  printf '%s' 'Woods source freshness is unknown: this check could not verify the current source contents.'
+  # Only the three literal identifiers above reach this word split.
+  for action in $advice; do
+    case "$action" in
+      deep_check) printf '%s' ' Request woods_status with source_check: deep; the quick reader reached its time limit.' ;;
+      fresh_capture) printf '%s' ' Fix the capture or boot evidence reported by woods_status, then run woods-extract full in the application environment.' ;;
+      inspect_source_scan) printf '%s' ' Inspect woods_status verification_reasons, source mapping, permissions and scan limits.' ;;
+    esac
+  done
+  printf '\n'
+}
 cwd="$(field "$payload" cwd)"
 [ -n "$cwd" ] && [ -d "$cwd" ] || exit 0
 configured_output="${WOODS_OUTPUT:-tmp/woods}"
@@ -25,7 +75,10 @@ esac
 if command -v jq >/dev/null 2>&1; then
   encoded="$(jq -nr --arg output "$configured_output" '{output:$output,mode:"quick"} | @base64')"
 elif command -v ruby >/dev/null 2>&1; then
-  encoded="$(ruby -rjson -rbase64 -e 'print Base64.strict_encode64(JSON.generate(output: ARGV[0], mode: "quick"))' -- "$configured_output")"
+  encoded="$(ruby -EUTF-8:UTF-8 -rjson -rbase64 -e 'print Base64.strict_encode64(JSON.generate(output: ARGV[0], mode: "quick"))' -- "$configured_output" 2>/dev/null)" || {
+    echo 'Woods source freshness is unknown: hook options could not be encoded. Inspect woods_status.'
+    exit 0
+  }
 else
   exit 0
 fi
@@ -63,7 +116,11 @@ status="$(tail -n 1 "$record")"
 state="$(field "$status" state)"
 case "$state" in
   current) ;;
+  unavailable)
+    detail="$(unavailable_summary "$status")"
+    printf 'Woods source freshness is unavailable: %s. The code index is usable; inspect woods_status. Repeating a full extraction will not reduce this size.\n' "${detail:-inspect woods_status for the unavailable evidence reason}"
+    ;;
   drifted) echo 'Woods source freshness is drifted: indexed application inputs differ from the working tree. Run woods-extract full, or inspect woods_status before choosing a targeted refresh.' ;;
-  *) echo 'Woods source freshness is unknown: this generation lacks complete verified source evidence. Inspect woods_status; use woods-extract full for a fresh capture.' ;;
+  *) unknown_message "$status" ;;
 esac
 exit 0