Ah ha! >In my system video memory is double buffered. In this case >"df_andi" relies on this quirk. It attempts to create a >surface with triple buffering (DLBM_TRIPLE), and if it gets an >error code in response to its dfb->CreateSurface it reattempts >with double buffering. This "quirk" results in no error code >and df_andi fails.
That confirms the df_andi problems I spent more time trying to debug than desired a week ago or so on our Intel_CE_3100 drivers for our 1.0.1 derivative. I had ended up believing that the triple buffer request somehow failed silently because of leaked video memory from previous runs that failed/died, thus resulting in a failed attempt to create a double buffered surface without an style of memory identified (which falls back to a backsystem video memory surface being requested which we do not support). I later solved all/most of the multi-app failures so that no memory was being leaked and I never ended up seeing the df_andi issues again. This behavior/quirk should not be continued and depended upon IMHO, unless there is some really really major benefit to hiding an error and not using the preferred IDirectFBDisplayLayer interface with its TestConfiguation function. Regards, Timothy -- Timothy Strelchun CE Software Engineering Digital Home Group Intel Corporation The views expressed above are my own and not those of Intel >-----Original Message----- >From: directfb-dev-boun...@directfb.org >[mailto:directfb-dev-boun...@directfb.org] On Behalf Of Paul Morton >Sent: Monday, March 30, 2009 4:38 PM >To: Niels Roest >Cc: directfb-dev@directfb.org >Subject: Re: [directfb-dev] Bug in return code from >fbdev_ioctl (systems\fbdev\fbdev.c version 1.2.7) > >Niels said "I hope there is no code out there that relies on >this quirk". > >In my system video memory is double buffered. In this case >"df_andi" relies on this quirk. It attempts to create a >surface with triple buffering (DLBM_TRIPLE), and if it gets an >error code in response to its dfb->CreateSurface it reattempts >with double buffering. This "quirk" results in no error code >and df_andi fails. > > >-- > >.------------------------------------------. >| DirectFB - Hardware accelerated graphics | >| http://www.directfb.org/ | >"------------------------------------------" > > > >_______________________________________________ >directfb-dev mailing list >directfb-dev@directfb.org >http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev > _______________________________________________ directfb-dev mailing list directfb-dev@directfb.org http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev
