Well, the recording extension was definitely broken and disabled for a long time in Xorg. However, the inability to use separate control and data connections was caused by a bug in Xlib. Also, this report is not a bug report in the sense that you'd absolutely have to do fixes in order for things to work. I just wanted to notify you that the original functionality can be restored in the future.

If a distribution is using Xlib configured --without-xcb, it should be possible to use separate control and data connections without problems.

If Xlib is configured --with-xcb, which is the default, then the Xlib version needs to be higher than 1.3.3 for separate control and data connections to work (1.3.4 hasn't been released yet).

I don't know how to detect if Xlib was configured --with-xcb so maybe it would be best to just check if Xlib version is 1.3.3 or older in configure.ac and use same connection workaround in that case when building Xnee.

I haven't seen much discussion about recording problems in xorg-devel recently so I guess they have been fixed at least in 1.8 version of X server. I haven't checked this though. I guess the remaining problems were quite minor like sending events accidentally twice as core and extension events in some cases and things like that.

Hi,

A while back there were problems with multiple display connections in XNEE. Originally it was thought that it had something to do with broken record implementation in X server. It turns out that it was caused by a general synchronization problem in Xlib configured using --with-xcb option.

Reported bug: https://bugs.freedesktop.org/show_bug.cgi?id=27595
Patch: http://cgit.freedesktop.org/xorg/lib/libX11/commit/?id=75ea8c37935ccc911557d16a303ba595b8ab106b

When the fixed Xlib is taken into use in distributions you can fix XNEE to start using separate connections again:

  if ( xnee_record_from_data_display(xd))
    {
      /*
       * From X.org 1.6.0 to ????
       *
       *   there seem to be something strange about
       *   the XRecordCreateContext call
       *   which causes the XRecordEnableContextAsync
       *   to fail ... ugly fix, but it works
       *
       */
      fprintf(stderr, "Workaround: Creating context on data display instead of control \n");
      fprintf(stderr, "            You can ignore this message\n");
      context_display = xd->data ;
    }

-- Rami