Remove misleading comment - correct NEWS

NEWS

@@ -7,8 +7,6 @@ Security updates:
 
 Various bugfixes:
  * Fixed various memory leaks, improving stability when running as a server
- * The 'Always on Top' setting is automatically disabled in fullscreen mode
-   (Win32 only)
  * Fix compilation with recent versions of FFmpeg
  * Correctly parses SAP announcements from MPEG-TS
  * Fix AAC resampling

modules/video_output/x11/xcommon.c

@@ -888,25 +888,6 @@ static int ManageVideo( vout_thread_t *p_vout )
         /* Update the object variable and trigger callback */
         val.b_bool = !p_vout->b_fullscreen;
 
-        /*
-         * FIXME FIXME FIXME FIXME: EXPLICIT HACK.
-         * On the one hand, we cannot hold the lock while triggering a
-         * callback, as it causes a deadlock with video-on-top handling.
-         * On the other hand, we have to lock while triggering the
-         * callback to:
-         *  1/ make sure video-on-top remains in sync with fullscreen
-         *    (i.e. unlocking creates a race condition if fullscreen is
-         *     switched on and off VERY FAST).
-         *  2/ avoid possible corruption bugs if another thread gets the
-         *     mutex and modifies our data in-between.
-         *
-         * This is obviously contradictory. Correct solutions may include:
-         *  - putting the fullscreen NAND video-on-top logic out of libvlc,
-         *    back into the video output plugins (ugly code duplication...),
-         *  - serializing fullscreen and video-on-top handling properly
-         *    instead of doing it via the fullscreen callback. That's got to
-         *    be the correct one.
-         */
 #ifdef MODULE_NAME_IS_xvmc
         xvmc_context_reader_unlock( &p_vout->p_sys->xvmc_lock );
 #endif