Re: [Linux-fbdev-devel] [PATCH] Funny sync calculation correction

[James S. <jsi...@in...>]

> I've been digging around in the rivafb module to find out why some
> modes converted from XFree86 4.2 don't work on some display devices.
> Here's a funny part of the riva_load_video_mode function (Linux
> 2.4.21-pre2):
> 
> 	hStart = (hDisplaySize + video_mode->right_margin) / 8 + 2;
> 	hEnd   = (hDisplaySize + video_mode->right_margin +
> 		 video_mode->hsync_len) / 8 - 1;
> 
> Please notice how "hsync_len" does not actually represent the length
> of the sync. Rather, it's compensated with 24 pixels. The corresponding
> code in the XFree86 4.2 (nv_dac.c) contains no such compensation:
> 
>     int horizStart      = (mode->CrtcHSyncStart/8) - 1;
>     int horizEnd        = (mode->CrtcHSyncEnd/8)   - 1;
> 
> The difference seems to be almost insignificant for modern multisync
> screens, but devices such as TV:s get confused by the current version.
> Is there a reason for the difference in calculations?
> 
> The attached patch corrects the calculations in accordance with XFree86
> and works better with the modelines I've tested. 
> 
> Please give it a try. Thanks,

Works fine. I applied it to the 2.5.X tree.