On Wed, 21 Sep 2011 18:21:32 +0100 Andre DRASZIK <andre.dras...@st.com> wrote:
> Hi Andi, > > It seems commits b368aac6b87cf760264d66cf5c0286cd1dbeb02d and > bc6c8a327919655986e02e9ec2e436d33f41538a (gfxcard: support clipping > for vertical and horizontal flips) (on branches master and > directfb-1.4) broke vertical and horizontal flipping when doing blits > that need to be clipped, i.e. they look very different than stretch > blits and also very different than software-only blits. > > I wrote a little test app (attached) to highlight this (it needs a > background.png in the same directory, mine is 1280x960). Run it in > 576p, 720p and 1080p, press return twice, and you'll see the issues. > With the above patch more or less reversed hardware accelerated blits > look identical to dfb:no-software. So I don't think it's a bug in my > driver or test app. > > What use case did your change fix? As I am tempted to revert the > patch. Ok I found out what it fixed and why it does not work anymore. My change in clip_blit_rotated() actually fixed clipping in the software driver for vertically and/or horizontaly flipped non-batch blits. The mentioned function is called from dfb_gfxcard_blit() and dfb_gfxcard_batchblit() indirectly through clip_blits(). For batch-blits it does not work somehow. The problem is that DirectFB only uses batchblits internally if the new Core-API is used (all this .flux and secure-fusion related stuff) I made a patch that enables the use of non-batchblits again if only one blit is performed. And voila clipping for flipped blits work again. What I do not understand is why my patches make anything worse for you. they should just fix vertically and horizontally flipped normal non- stretched non-batch blits for the software driver. _______________________________________________ directfb-dev mailing list directfb-dev@directfb.org http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev
