Ask for a context that is likely to be supported.

Most embedded GL chips support 24-bit depth with packed
8 bit stencil, but more rarely 32-bit. Asking for 32
bit means we often fail on other properties to, such as
multisampling.

Change-Id: Ib913d94af0635b09913ff15cff54cc694ba293fc
Reviewed-by: Samuel Rødal <samuel.rodal@nokia.com>

src/plugins/platforms/eglfs/qeglfsscreen.cpp

@@ -163,7 +163,8 @@ void QEglFSScreen::createAndSetPlatformContext()
         m_depth = 16;
         m_format = QImage::Format_RGB16;
     } else {
-        platformFormat.setDepthBufferSize(32);
+        platformFormat.setDepthBufferSize(24);
+        platformFormat.setStencilBufferSize(8);
         platformFormat.setRedBufferSize(8);
         platformFormat.setGreenBufferSize(8);
         platformFormat.setBlueBufferSize(8);