unity-open-mcp 1.2.3 → 1.4.0

package/dist/dialog-dismiss.d.ts

@@ -72,6 +72,8 @@ export interface DismissProbeOptions {
      * the modal is blocked (audit line, no click) unless this is set.
      */
     allowUnsavedSceneDismiss: boolean;
+    /** Allow continuing through a non-matching Editor warning. Off by default. */
+    allowVersionMismatch: boolean;
     /**
      * M12 — optional abort signal. When fired while a probe is in flight, the
      * underlying `execFile`/`execFileSync` child is killed so the loop can exit
@@ -108,8 +110,9 @@ export declare function tryDismissDialog(opts: DismissProbeOptions): Promise<Dis
  * Strategy:
  *   1. P/Invoke `EnumWindows` to enumerate every visible top-level window
  *      owned by `Unity.exe`.
- *   2. For each, normalize the title (alphanumeric lowercase) and match it
- *      against the embedded per-kind fragment table to classify it.
+ *   2. For each, skip the main Editor window ({@link isUnityEditorWindowTitle}),
+ *      normalize the title (alphanumeric lowercase) and match it against the
+ *      embedded per-kind fragment table to classify it.
  *   3. If the kind is in the blocked list → emit `blocked:<kind>` and skip
  *      the click. If the kind's token list is null → skip (not-found).
  *   4. Walk child windows with `EnumChildWindows`, find a Button whose
@@ -127,25 +130,94 @@ export declare function tryDismissDialog(opts: DismissProbeOptions): Promise<Dis
  * The token table is passed via stdin (one JSON blob) so the script body
  * stays constant and testable; embedding it inline would bloat the assertion.
  */
-export declare const WINDOWS_DISMISS_PS_SCRIPT = "\n$ErrorActionPreference = 'Stop'\ntry {\n  if (-not ([System.Management.Automation.PSTypeName]'UnityOpenMcp.Dialogs.Dismisser').Type) {\n    Add-Type -TypeDefinition @\"\nusing System;\nusing System.Collections.Generic;\nusing System.Runtime.InteropServices;\nusing System.Text;\nnamespace UnityOpenMcp.Dialogs {\n  public static class Dismisser {\n    public delegate bool EnumWindowsProc(IntPtr hWnd, IntPtr lParam);\n    [DllImport(\"user32.dll\")] static extern bool EnumWindows(EnumWindowsProc lpEnumFunc, IntPtr lParam);\n    [DllImport(\"user32.dll\", CharSet=CharSet.Unicode)] static extern int GetWindowTextW(IntPtr hWnd, StringBuilder lpString, int nMaxCount);\n    [DllImport(\"user32.dll\", CharSet=CharSet.Unicode)] static extern int GetClassNameW(IntPtr hWnd, StringBuilder lpClassName, int nMaxCount);\n    [DllImport(\"user32.dll\")] static extern bool IsWindowVisible(IntPtr hWnd);\n    [DllImport(\"user32.dll\")] static extern bool EnumChildWindows(IntPtr hWndParent, EnumWindowsProc lpEnumFunc, IntPtr lParam);\n    [DllImport(\"user32.dll\")] static extern uint GetWindowThreadProcessId(IntPtr hWnd, out uint lpdwProcessId);\n    [DllImport(\"user32.dll\", CharSet=CharSet.Unicode)] static extern IntPtr SendMessageW(IntPtr hWnd, uint Msg, IntPtr wParam, IntPtr lParam);\n    const uint BM_CLICK = 0x00F5;\n    static string Norm(string s) {\n      var sb = new StringBuilder(s.Length);\n      foreach (var ch in s) {\n        if ((ch >= '0' && ch <= '9') || (ch >= 'A' && ch <= 'Z') || (ch >= 'a' && ch <= 'z')) {\n          sb.Append(char.ToLowerInvariant(ch));\n        }\n      }\n      return sb.ToString();\n    }\n    public static string TryDismiss(int[] unityPids, string tableJson) {\n      var unityPidSet = new HashSet<uint>();\n      for (int i = 0; i < unityPids.Length; i++) unityPidSet.Add((uint)unityPids[i]);\n      var table = System.Text.Json.JsonSerializer.Deserialize<Types.Table>(tableJson);\n      var candidates = new List<IntPtr>();\n      var sb = new StringBuilder(512);\n      EnumWindows((hWnd, lParam) => {\n        if (!IsWindowVisible(hWnd)) return true;\n        uint procId; GetWindowThreadProcessId(hWnd, out procId);\n        if (!unityPidSet.Contains(procId)) return true;\n        sb.Length = 0;\n        GetWindowTextW(hWnd, sb, sb.Capacity);\n        candidates.Add(hWnd);\n        return true;\n      }, IntPtr.Zero);\n      foreach (var hWnd in candidates) {\n        sb.Length = 0;\n        GetWindowTextW(hWnd, sb, sb.Capacity);\n        var title = sb.ToString();\n        var norm = Norm(title);\n        string kind = null;\n        foreach (var kv in table.kinds) {\n          foreach (var frag in kv.Value.fragments) {\n            if (norm.Contains(frag)) { kind = kv.Key; break; }\n          }\n          if (kind != null) break;\n        }\n        string[] tokens = null;\n        if (kind != null) {\n          if (table.blocked.Contains(kind)) return \"blocked:\" + kind;\n          if (!table.kinds[kind].tokensSpecified) continue;\n          tokens = table.kinds[kind].tokens;\n        } else {\n          if (table.genericTokens.Length == 0) continue;\n          tokens = table.genericTokens;\n        }\n        IntPtr matchedButton = IntPtr.Zero;\n        string matchedText = null;\n        EnumChildWindows(hWnd, (hChild, lParam) => {\n          sb.Length = 0;\n          GetClassNameW(hChild, sb, sb.Capacity);\n          if (sb.ToString() != \"Button\") return true;\n          sb.Length = 0;\n          GetWindowTextW(hChild, sb, sb.Capacity);\n          var text = sb.ToString();\n          var textNorm = Norm(text);\n          foreach (var token in tokens) {\n            if (textNorm.Contains(token)) { matchedButton = hChild; matchedText = text; return false; }\n          }\n          return true;\n        }, IntPtr.Zero);\n        if (matchedButton == IntPtr.Zero) continue;\n        SendMessageW(matchedButton, BM_CLICK, IntPtr.Zero, IntPtr.Zero);\n        return \"dismissed:\" + matchedText + \":\" + (kind ?? \"unknown\");\n      }\n      return \"not-found\";\n    }\n  }\n}\n\"@\n  }\n  $unityPids = @(Get-Process -Name 'Unity' -ErrorAction SilentlyContinue | ForEach-Object { [int]$_.Id })\n  if ($unityPids.Count -eq 0) { Write-Output 'not-found'; return }\n  $tableJson = [Console]::In.ReadToEnd();\n  Write-Output ([UnityOpenMcp.Dialogs.Dismisser]::TryDismiss([int[]]$unityPids, $tableJson))\n} catch {\n  Write-Output ('error:' + $_.Exception.Message)\n}\n";
+export declare const WINDOWS_DISMISS_PS_SCRIPT = "\n$ErrorActionPreference = 'Stop'\ntry {\n  if (-not ([System.Management.Automation.PSTypeName]'UnityOpenMcp.Dialogs.Dismisser').Type) {\n    Add-Type -TypeDefinition @\"\nusing System;\nusing System.Collections.Generic;\nusing System.Runtime.InteropServices;\nusing System.Text;\nnamespace UnityOpenMcp.Dialogs {\n  public static class Dismisser {\n    public delegate bool EnumWindowsProc(IntPtr hWnd, IntPtr lParam);\n    [DllImport(\"user32.dll\")] static extern bool EnumWindows(EnumWindowsProc lpEnumFunc, IntPtr lParam);\n    [DllImport(\"user32.dll\", CharSet=CharSet.Unicode)] static extern int GetWindowTextW(IntPtr hWnd, StringBuilder lpString, int nMaxCount);\n    [DllImport(\"user32.dll\", CharSet=CharSet.Unicode)] static extern int GetClassNameW(IntPtr hWnd, StringBuilder lpClassName, int nMaxCount);\n    [DllImport(\"user32.dll\")] static extern bool IsWindowVisible(IntPtr hWnd);\n    [DllImport(\"user32.dll\")] static extern bool EnumChildWindows(IntPtr hWndParent, EnumWindowsProc lpEnumFunc, IntPtr lParam);\n    [DllImport(\"user32.dll\")] static extern uint GetWindowThreadProcessId(IntPtr hWnd, out uint lpdwProcessId);\n    [DllImport(\"user32.dll\", CharSet=CharSet.Unicode)] static extern IntPtr SendMessageW(IntPtr hWnd, uint Msg, IntPtr wParam, IntPtr lParam);\n    const uint BM_CLICK = 0x00F5;\n    static string Norm(string s) {\n      var sb = new StringBuilder(s.Length);\n      foreach (var ch in s) {\n        if ((ch >= '0' && ch <= '9') || (ch >= 'A' && ch <= 'Z') || (ch >= 'a' && ch <= 'z')) {\n          sb.Append(char.ToLowerInvariant(ch));\n        }\n      }\n      return sb.ToString();\n    }\n    public static string TryDismiss(int[] unityPids, string tableJson) {\n      var unityPidSet = new HashSet<uint>();\n      for (int i = 0; i < unityPids.Length; i++) unityPidSet.Add((uint)unityPids[i]);\n      var table = System.Text.Json.JsonSerializer.Deserialize<Types.Table>(tableJson);\n      var candidates = new List<IntPtr>();\n      var sb = new StringBuilder(512);\n      EnumWindows((hWnd, lParam) => {\n        if (!IsWindowVisible(hWnd)) return true;\n        uint procId; GetWindowThreadProcessId(hWnd, out procId);\n        if (!unityPidSet.Contains(procId)) return true;\n        sb.Length = 0;\n        GetWindowTextW(hWnd, sb, sb.Capacity);\n        candidates.Add(hWnd);\n        return true;\n      }, IntPtr.Zero);\n      foreach (var hWnd in candidates) {\n        sb.Length = 0;\n        GetWindowTextW(hWnd, sb, sb.Capacity);\n        var title = sb.ToString();\n        // The main Editor window title embeds project/scene names; never classify it.\n        if (title.IndexOf(\" - Unity \", StringComparison.OrdinalIgnoreCase) >= 0) continue;\n        var norm = Norm(title);\n        string kind = null;\n        foreach (var kv in table.kinds) {\n          foreach (var frag in kv.Value.fragments) {\n            if (norm.Contains(frag)) { kind = kv.Key; break; }\n          }\n          if (kind != null) break;\n        }\n        string[] tokens = null;\n        if (kind != null) {\n          if (table.blocked.Contains(kind)) return \"blocked:\" + kind;\n          if (!table.kinds[kind].tokensSpecified) continue;\n          tokens = table.kinds[kind].tokens;\n        } else {\n          if (table.genericTokens.Length == 0) continue;\n          tokens = table.genericTokens;\n        }\n        IntPtr matchedButton = IntPtr.Zero;\n        string matchedText = null;\n        EnumChildWindows(hWnd, (hChild, lParam) => {\n          sb.Length = 0;\n          GetClassNameW(hChild, sb, sb.Capacity);\n          if (sb.ToString() != \"Button\") return true;\n          sb.Length = 0;\n          GetWindowTextW(hChild, sb, sb.Capacity);\n          var text = sb.ToString();\n          var textNorm = Norm(text);\n          foreach (var token in tokens) {\n            if (textNorm.Contains(token)) { matchedButton = hChild; matchedText = text; return false; }\n          }\n          return true;\n        }, IntPtr.Zero);\n        if (matchedButton == IntPtr.Zero) continue;\n        SendMessageW(matchedButton, BM_CLICK, IntPtr.Zero, IntPtr.Zero);\n        return \"dismissed:\" + matchedText + \":\" + (kind ?? \"unknown\");\n      }\n      return \"not-found\";\n    }\n  }\n}\n\"@\n  }\n  $unityPids = @(Get-Process -Name 'Unity' -ErrorAction SilentlyContinue | ForEach-Object { [int]$_.Id })\n  if ($unityPids.Count -eq 0) { Write-Output 'not-found'; return }\n  $tableJson = [Console]::In.ReadToEnd();\n  Write-Output ([UnityOpenMcp.Dialogs.Dismisser]::TryDismiss([int[]]$unityPids, $tableJson))\n} catch {\n  Write-Output ('error:' + $_.Exception.Message)\n}\n";
 /**
  * The AppleScript template used for the macOS dismiss path. Exposed as a
- * function of the token table so tests can assert the script shape without
+ * function of the probe options so tests can assert the script shape without
  * launching osascript.
  *
- * Strategy: iterate every window of the Unity process, normalize its title,
- * classify it against the table, and click the first button whose normalized
- * label contains a policy token. If the kind is blocked, emit `blocked:<kind>`
- * without clicking. If Unity is not running OR no matching button exists, the
- * script reports `not-found`. Any AppleScript exception (e.g. Accessibility
- * permission not granted) is reported as `error:<msg>`.
+ * Strategy: iterate every window of the Unity process except the main Editor
+ * window (its title contains {@link UNITY_EDITOR_WINDOW_TITLE_MARKER}),
+ * classify it by its title normalized with `normLabel` (see
+ * {@link macosTitleMatch}), and act per kind as {@link focusDialogAction}
+ * decides from the shared policy table (see {@link macosDialogKindBlock}):
+ *
+ *   - `focus`   — the policy picks the kind's focused/default button
+ *                 ({@link DIALOG_FOCUSED_BUTTON_TOKENS}): press Return, but
+ *                 only when the window shows a button with that label.
+ *   - `named`   — the policy picks another button (e.g. Enter Safe Mode, or
+ *                 Quit/Cancel under `cancel`): click the first button whose
+ *                 normalized title equals a policy token, in priority order,
+ *                 or report `blocked:<kind>` when none is found. Return is
+ *                 never pressed, because it would click the focused button.
+ *   - `blocked` — the policy blocks the kind: report `blocked:<kind>`.
+ *   - `skip`    — the policy declines the kind: no block; the window falls
+ *                 through to `not-found` so polling continues.
+ *
+ * Project Upgrade is always reported as `blocked` (never clicked), and
+ * unsaved scene changes have their own named-button block behind the opt-in.
+ * If Unity is not running or nothing matches, the script reports `not-found`.
+ * Any AppleScript exception (e.g. Accessibility permission not granted) is
+ * reported as `error:<msg>`.
  *
  * Requires the Terminal / `node` binary to have been granted Accessibility
  * permission in System Settings → Privacy & Security → Accessibility.
- *
- * The token table is JSON-escaped into the script header.
  */
 export declare function macosDismissAppleScript(opts: DismissProbeOptions): string;
+/**
+ * AppleScript mirror of {@link normalizeDialogLabel}: keep ASCII letters and
+ * digits only. AppleScript string comparison (`is`, `contains`) ignores case
+ * by default, so the handler does not lowercase. Exported for tests.
+ */
+export declare const MACOS_NORM_LABEL_HANDLER = "\n-- Mirror of normalizeDialogLabel: keep ASCII letters and digits only.\n-- (String comparison is case-insensitive by default, so no lowercasing.)\non normLabel(s)\n  set out to \"\"\n  repeat with c in (characters of s)\n    set ch to contents of c\n    set cid to id of ch\n    -- \"id of\" yields a LIST for a character made of several code points\n    -- (decomposed accents, emoji); comparing a list numerically would raise\n    -- and abort the whole probe, so such characters are simply dropped.\n    if class of cid is integer then\n      if (cid >= 48 and cid <= 57) or (cid >= 65 and cid <= 90) or (cid >= 97 and cid <= 122) then set out to out & ch\n    end if\n  end repeat\n  return out\nend normLabel\n";
+/**
+ * How a focus-driven platform (macOS AppleScript, Linux xdotool) may act on a
+ * dialog of the given kind under the active policy. Derived from the shared
+ * policy table so every platform agrees:
+ *
+ *   - `blocked` — the policy blocks the kind (e.g. Project Upgrade without its
+ *                 opt-in). Report `blocked:<kind>`, never click.
+ *   - `skip`    — the policy declines the kind (no preference tokens). Do
+ *                 nothing; try the next kind.
+ *   - `focus`   — the policy's first choice is the kind's focused button
+ *                 ({@link DIALOG_FOCUSED_BUTTON_TOKENS}): pressing Return is
+ *                 safe.
+ *   - `named`   — the policy's first choice is some other button (e.g. Enter
+ *                 Safe Mode, Quit, Cancel). Return would press the focused
+ *                 button instead, so only a named-button click may act.
+ *
+ * Pure; exported for tests.
+ */
+export declare function focusDialogAction(kind: DialogKind, opts: Pick<DismissProbeOptions, "policy" | "allowProjectUpgrade" | "allowUnsavedSceneDismiss" | "allowVersionMismatch">): "blocked" | "skip" | "focus" | "named";
+/**
+ * AppleScript title condition per focus-classified kind, over `wn` — the
+ * window title normalized by `normLabel`. The fragments are normalized
+ * (letters and digits only), and AppleScript `contains` ignores case but not
+ * spaces or punctuation, so matching the raw title would never fire
+ * (`"Enter Safe Mode?" contains "safemode"` is false). Non-Matching Editor
+ * matches the short `nonmatchingeditor` stem so both the hyphenated and the
+ * spaced spelling qualify. Exported for tests.
+ */
+export declare function macosTitleMatch(kind: DialogKind): string;
+/**
+ * AppleScript block for one dialog kind inside the window loop (`w` / `wt`),
+ * shaped by {@link focusDialogAction}:
+ *
+ *   - `blocked` → `blocked:<kind>` when the title matches.
+ *   - `skip`    → a comment only; the window falls through.
+ *   - `focus`   → press Return on the focused button, when the window has a
+ *                 button whose normalized title equals the kind's focused
+ *                 token; otherwise fall through (no press).
+ *   - `named`   → for each policy token in priority order, click the first
+ *                 button whose normalized title equals it; if none matches,
+ *                 report `blocked:<kind>`. Never presses Return.
+ *
+ * Exact normalized equality (not substring) keeps a short token such as `no`
+ * from matching an unrelated label such as "Ignore".
+ *
+ * Pure; exported for tests.
+ */
+export declare function macosDialogKindBlock(kind: DialogKind, opts: Pick<DismissProbeOptions, "policy" | "allowProjectUpgrade" | "allowUnsavedSceneDismiss" | "allowVersionMismatch">): string;
 /**
  * Reset the cached `xdotool` presence flag. Test-only — production code
  * never needs to call this.
@@ -157,6 +229,26 @@ export declare function _resetXdotoolPresenceForTests(): void;
  * containing metacharacters would silently change the match semantics.
  */
 export declare function regexEscapeForXdotool(s: string): string;
+/**
+ * The `xdotool search --name` pattern for a normalized title fragment.
+ * Fragments are letters and digits only ("entersafemode"), while real titles
+ * carry spaces and punctuation ("Enter Safe Mode?"), so the pattern allows any
+ * run of non-alphanumerics between consecutive characters and spells each
+ * letter in both cases. That is the regex form of "the normalized title
+ * contains the fragment" and does not depend on xdotool's case handling.
+ * Any other character is escaped literally.
+ */
+export declare function xdotoolTitlePattern(fragment: string): string;
+/**
+ * What the Linux/X11 path does with a Unity-owned window of the given kind:
+ * {@link focusDialogAction} with `named` folded into `blocked`, because
+ * xdotool can only press Return on the focused (default) button — it cannot
+ * click a named one (e.g. `cancel` on the Safe Mode prompt, whose focused
+ * button is Ignore, is reported as blocked for a human).
+ *
+ * Pure; exported for tests.
+ */
+export declare function linuxFocusAction(kind: DialogKind, opts: Pick<DismissProbeOptions, "policy" | "allowProjectUpgrade" | "allowUnsavedSceneDismiss" | "allowVersionMismatch">): "blocked" | "skip" | "focus";
 /**
  * Parse the single-token contract every platform-specific dispatcher writes
  * to stdout. Inspects the LAST non-empty line of stdout, not the whole
@@ -206,6 +298,8 @@ export declare function armSigkillEscalation(child: {
  *   - `UNITY_OPEN_MCP_ALLOW_UNSAVED_SCENE_DISMISS=1` opts in to auto-dismissing
  *     the "Unsaved changes to scene" modal (destructive under every policy —
  *     off by default). See isUnsavedSceneBlocked.
+ *   - `UNITY_OPEN_MCP_ALLOW_VERSION_MISMATCH=1` opts in to opening a project
+ *     with a different Unity Editor version. Without it the dialog is blocked.
  *   - `UNITY_OPEN_MCP_DISMISS_TIMEOUT_MS` overrides the overall timeout.
  *   - `UNITY_OPEN_MCP_DISMISS_INTERVAL_MS` overrides the poll interval.
  *
@@ -219,6 +313,7 @@ export interface DismissConfig {
     policy: DialogPolicy;
     allowProjectUpgrade: boolean;
     allowUnsavedSceneDismiss: boolean;
+    allowVersionMismatch: boolean;
 }
 export declare function readDismissConfig(env?: NodeJS.ProcessEnv): DismissConfig;
 /**
@@ -233,6 +328,7 @@ export interface PollAndDismissOptions {
     policy: DialogPolicy;
     allowProjectUpgrade: boolean;
     allowUnsavedSceneDismiss: boolean;
+    allowVersionMismatch: boolean;
     /**
      * Override the platform — exposed for tests so the polling logic can be
      * exercised across all three OS branches without hopping process.platform.