um sorry to bring up old news but a recent post in dbp board tricked me into revisiting this and
MAY have made some progression in narrowing this down, apologies for not sticking with this - my lifes been a bit of a drama these recent years so I completely forgot as I havent been using dbp much until recently
http://forum.thegamecreators.com/?m=forum_view&t=194537&b=1
first off I now use U7.5 and the code I used previously(which for some odd reason I never posted up) no longer creates the bug unless I ensure vsync is off in display mode command which also appears to overide graphics card control panel setting of force vsync on
the code
set display mode 1024,768,32,0
set window off
sync on
sync rate 60
make object sphere 1,1
do
sync
LOOP
secondly I cannot recreate the bug/issue unless my graphics control panel software is set to 4x AA as this is only sample that creates the issue, also when using the AA option in the panel while only using 4x AA "override application" is the only time the bug occurs
thirdly the bug isnt limited to fullscreen mode it will occur in window mode but not with the above code or just any project - look at the post I made in the link above, download resourcefuls project and alter it to what I posted on that thread - the application crashes rather than getting a black screen, yet while running the code I provided in this post in window mode, the application appears to run normal!!
With above settings and using "set display mode 1024,768,32,0,0,0" in the code above, while in fullscreen the app will fail in the same manner as I have just mentioned, however if the multisample value is anything other than zero it runs fine
Now in resourcefuls project with the code I posted on the same thread, it doesnt appear to matter what parameters I put in the display command vsync or aa, it will crash in window mode, however if I rem out the paste image command in the main loop, it doesnt crash and the black screen is with us once again lol
also worthy of note is the fact that with resourcefuls project the results occurring in window mode can be reproduced exactly the same in fullscreen mode too - the only difference between the two lots of code is a camera shader that automatically assigns the image number which is then taken control of by a command in the main loop
to sum up its late, Im tired, I dont know if this affects just a few of us or many of us but I do think it warrants further investigation as it breaks pretty much any project I am using camera shaders in that Ive tested, i have yet to test on "no shader" projects which I beleive
MAY run ok if I use all parameters in display command with multisample of 1, (not gonna test tonight though), oh one last thing my setup is the same as when I first made this thread except the card is now a 9600gt, so this issue is on two of my nvid cards..