PeekMessage(PM_NOREMOVE) + Fullscreen = fail?

Started by
2 comments, last by 21st Century Moose 12 years ago
Heya all

I'm following the DirectX tutorial here: [1]

I've found some strange behavior with the main loop... The tutorial uses the following

while(TRUE)
{
if(PeekMessage(&msg, NULL, 0, 0, PM_REMOVE))
{
TranslateMessage(&msg);
DispatchMessage(&msg);
if(msg.message == WM_QUIT)
break;
}
RenderFrame();
}


When you hit alt-enter to switch to fullscreen mode, it works flawlessly. However, if you change the main loop to this:

while(TRUE)
{
if(PeekMessage(&msg, NULL, 0, 0, PM_NOREMOVE))
{
GetMessage(&msg, NULL, 0, 0);
DispatchMessage(&msg);
if(msg.message == WM_QUIT)
break;
}
RenderFrame();
}



When you hit alt-enter, it starts to go fullscreen, then "dings" an error and returns. It seems like it has to do with the PM_NOREMOVE flag, as the following shows the same broken behavior as the latter:


while(TRUE)
{

if(PeekMessage(&msg, NULL, 0, 0, PM_NOREMOVE))
{

}
if(PeekMessage(&msg, NULL, 0, 0, PM_REMOVE))
{
DispatchMessage(&msg);
if(msg.message == WM_QUIT)
break;
}
RenderFrame();
}


The DirectX debugging output doesn't give any errors or warnings. Interestingly, it has to do with the context switching itself, not the fullscreen... if you change the line:

swapChainDesc.Windowed = false;

It starts fine in fullscreen, but fails to switch out of fullscreen on alt-enter. Any reason why PM_NOREMOVE would break fullscreen switching?


[1] http://www.directxtu.../BA4.aspx#still
Advertisement
If you use NOREMOVE then you'll never remove messages from the queue, and more and more messages will just continue to pile up. This is going to cause bad things to happen.

Either way that message loop is jacked up. You want to loop on PeekMessage until no messages are left, THEN render a frame. Otherwise you will only remove one message per frame, and messages could end up piling up and then getting dealt with slowly.
Sure, NOREMOVEchecks if there's a message on the queue and returns true if there is. Then, you use GetMessage to pull it off the queue and process it. You're right that looping on it until all messages have been processed is the better way to do it (even better: render in a separate thread, so you don't slow your render loop if processing tons of input/messages).

That said, it doesn't give any insight as to why the behavior is different. In both the first and second cases, the message queue is processed successfully; messages are caught, translated, and passed. Starting fullscreen or windowed works in both, but switching contexts via alt-enter doesn't work in the latter.
Are you handling your device reset when switching modes from WM_ACTIVATE? This is not a good idea for a number of reasons, not least because you're going to fail to catch all of the circumstances in which a lost device may occur. Better to check the HRESULT from your Present call and use TestCooperativeLevel.

Direct3D has need of instancing, but we do not. We have plenty of glVertexAttrib calls.

This topic is closed to new replies.

Advertisement