Claudio \"KLaN\" Ciccani said:
You can force df_xine to use a specific layer using the --layer option. Try df_xine --layer=2; this should work.
Hi Claudio,
after a long time I tried df_xine again, precisely, with this last weekend CVS of DirectFB, DFB++, DirectFB-extra, and xinelib-1.0-rc5. Now, specifying --layer=2 indeed forces df_xine to output the picture on CRTC2 (TV), but it's extremely slow, unlike on the BES (on the monitor). I also noticed that, if letting df_xine to choose itself the first layer, it finds that neither of the pixel formats UYVY, YUY2, I420 or YV12 is supported on the primary layer (which I set to 2 in directfbrc, to have freevo & mplayer output on TV, and they do work). Does this slowlyness have something to do with the pixelformat? How could I fix that?
There are two limitations currently:
1) the primary layer is excluded from all tests
2) if use the --layer option df_xine acts in stupid mode, it doesn't try to change layer configuration.
For now, you have to specify the pixelformat on the command line (for example --dfb:pixelformat=YV12).
I hope to fix the TV-OUT support soon.
I mention that mplayer-1.0-pre5 provided by current Gentoo portage tree works with current DirectFB CVS (although there is some jitter on MPEG2 files recorded from my PVR250 card, but DivX for example is fine). I would like to make at least df_xine work nicely with directfb on my TV, as it supports DVD navigation very nicely, and maybe I will be able to patch the "official" fbxine to use your unofficial DFB driver, as the original one simply doesn't work and is unmaintained. Is there a chance to accomplish this?
Have a look at the video_out_dfb.h header (located in interfaces/IDirectFBVideoProvider/video_out_dfb/), the video provider source and the sample2 to understand how to use the driver; it's not so different from other xine drivers.
You will also need to replace all framebuffer intialization and keyboard handling routines in fbxine with DirectFB routines.
Best regards, Lucian
Regards, Claudio
