fix: memory leak in lifeCycleMainService

[SimonSiefke]

Fixes a memory leak in lifeCycleMainService.

Details

It seems the way the event listener is setup, when an okChannel event is received, the listener for the cancelChannel is not removed and vice versa.

validatedIpcMain.once(okChannel, () => {
  /* 
   * this can run, but if we get into this branch, 
   * we never remove the cancelChannel listener 
   * from below and vice versa
   */
  resolve(false); // no veto 
});

validatedIpcMain.once(cancelChannel, () => {
  resolve(true); // veto
});

Change

The change tries to ensure that always both listeners are removed when either a okChannel event or a cancelChannel event is received.

const cleanup = (value: boolean) => {
  /* Always ensure to clean up both listeners */
  validatedIpcMain.removeListener(okChannel, okListener);
  validatedIpcMain.removeListener(cancelChannel, cancelListener);
  resolve(value);
};

const okListener = () => {
  cleanup(false); // no veto
};

const cancelListener = () => {
  cleanup(true); // veto
};

validatedIpcMain.on(okChannel, okListener);
validatedIpcMain.on(cancelChannel, cancelListener);

Before

When opening and closing a new window 17 times, the number of CodeWindow instances and some other functions seems to grow by 1 each time:

After

When opening and closing a new window 17 times, no more CodeWindow leak is detected. And overall less functions seem to be leaked:

[Copilot]

Pull request overview

This PR fixes a listener leak in the Electron main-process LifecycleMainService during the onBeforeUnload handshake with the renderer by ensuring both IPC listeners are always removed when either the “ok” or “cancel” reply arrives.

Changes:

• Replaces two independent validatedIpcMain.once(...) registrations with removable validatedIpcMain.on(...) listeners.
• Adds a shared cleanup function that removes both listeners and resolves the veto promise from either branch.

[deepak1556]

Thanks!