cyberorg: we did some tests with compiz emmes, how are things :) at first i though i could trivially write a patch that would us all allow to be happy not perfectly sure atm anymore because mesa behaves a bit strangly regarding EXT_tfp exposal so we use wrapper script for 11.0? even if you definitely only have indirect rendering (like in DISPLAY=hostname:0 with hostname!= localhost) it still clamis that it doesn't support EXT_tfp except for when 1) LIBGL_ALWAYS_INDIRECT is set or(!) b) opening /dev/dri/* failed... :-/ * mhwork|gone is now known as mhutch i also tried to putenv() LIBGL_ALWAYS_INDIRECT that does work(!) unless i did some glx calls before if [ $($GLXINFO 2>/dev/null | grep GLX_EXT_texture_from_pixmap -c) -gt 2 ] ; then is how it is done in compiz-manager and I actually have to, because i want to detect whether this is necessary or not... yes, but then i cannot putenv() any more, because this doesn't change behavior any more i can restart compiz (like with the LD_PRELOAD case), but that is bad as well, because * poepil has quit ("sleeps") if compiz is the first client, the xserver will reset, and in case of Xgl (with -noreset) this will terminate the Xserver... * mmarek has quit ("Leaving.") :-((( actually, i'd like to have this logic all inside compiz, because otherwise people who upgrade from 10.2 will stumble at this point.