On Fri, 4 Aug 2006, cckuo wrote: > Thanks for your reply. > > I rmmod the fb module and try again, and I found the text blur still occur. > I will rebuild a new kernel without fb inside and try it again. > > By the way, I think I may give wrong description about #3. The "disappear" I > mean in #3 is the text blur will not happen while using "xinit' to activate > Xserver. And I don't mean the whole screen will disappear. I am sorry about > that. > > > > I also done some test about decreasing the size of memory I pass to > xf86InitFBManager function call. The text blur will be normal when the size > is below 4Mb. I know this size control the number of different slots created > by XAAInitPixmapCache. Since the driver didn't report any HW accelerations > about characters, I guess the text part of the picture may be taken as some > pixmaps and call HW bitblt function to move it from offscreen to onscreen. > If my assumption is correct, would you please tell me what functions in XAA > take charge of text such that I can trace the command made by that function > to check whether the content of offscreen pixmaps are already wrong or the > HW blur it during bitblt. Most simple text is rendered with the CPUToScreenColorExpand functions. Anti-aliased text or some other text used by the RENDER extension is rendered with the CPUToScreenAlphaTexture or CPUToScreenTexture functions. Mark. > > > > Thanks for your help very much, > > cckuo > > > > -----Original Message----- > > From: xfree86-admin@xxxxxxxxxxx [mailto:xfree86-admin@xxxxxxxxxxx] On Behalf > Of Mark Vojkovich > > Sent: Friday, August 04, 2006 1:12 AM > > To: cckuo > > Cc: xfree86@xxxxxxxxxxx > > Subject: RE: Will the behaviors of XFree86 be effected by > different kernel? > > > > Most likely, part of the offscreen video memory that the driver > > told XAA it could use for pixmap caching is not accessible so things > > placed there disappear. The order in which things are started > > determines the packing order of offscreen pixmaps so that's why > > #3 changes which things disappear. Most likely it is an interaction > > with a kernel framebuffer device. Namely, the kernel framebuffer > > device is setting some graphics card state that prevents all the > > videoram from being accessible and the XFree86 driver is unaware > > of this. This is unlikely to be an XAA issue. > > > > If the XFree86 driver is assuming the card is in the POSTed > > state, that may be a bad assumption in light of a kernel framebuffer > > device. You might want to check what the kernel framebuffer > > device is doing with regards to video memory partitions, etc... > > > > Mark. > > > > On Thu, 3 Aug 2006, cckuo wrote: > > > > > Sorry for late replying your email. > > > The situation is attached as the picture and we use the "rdesktop", a > remote > > > connect tool to control English Server 2003, the text part will be broken. > > > 1. After setting the option "XaaNoOffscreenPixmaps" or > > > "XaaNoScreenToScreenCopy" I can't reproduce the problem. I know these two > > > option will let the XAA determine whether it can call the call back > > > functions that driver registered in the beginning. > > > > > > 2. If I reduce the memory size reported to FBmanager, less than 4Mb, it > will > > > also disappear. I know this behavior will let the XAA generate less big > > > slots. > > > > > > 3. if I use xinit instead of windows manager to activate the Xserver, > > > execute "rdesktop" in xterm. It will disappear. > > > > > > 4. So far it only happens on the English Server 2003, and English XP is > OK. > > > > > > 5. use kernel 2.4.X it will disappear, but kernel 2.6.X will fail. > > > I am right now studying the source codes to realize the relationships > > > between 1 and 2. > > > But I really confuse why the 3, 4 and 5 will happen. > > > I am not quite sure whether it was a bug of XAA, the windows manager or > even > > > the driver's fault. But from 3, I guess it may not driver's fault. > > > > > > If you have any opinions about this bug, please let me know. > > > > > > Appreciate your help in advance. > > > Sincerely Yours, > > > cckuo > > > > > > > > > To: cckuo > > > Cc: xfree86@xxxxxxxxxxx > > > Subject: Re: Will the behaviors of XFree86 be effected by > > > different kernel? > > > > > > > > > I wouldn't expect the behavior of an XFree86 graphics driver to > > > change with a kernel except for, perhaps, something related to 3D > > > using the DRI. I suppose it's also possible that a kernel > > > framebuffer device could interfere with an xfree86 driver in some > > > way. The XFree86 sis driver has a "NoAccel" option you can use to > > > force software rendering. If there is still rendering corruption > > > with that option, that would suggest that it's not a driver > > > acceleration problem. > > > > > > Mark. > > > > > > > > > cckuo wrote: > > > > > > > Dear X friends: > > > > Recently I play the "rdesktop", a software like XP remote desktop, on > > > > XFree86 4.3.0. > > > > And the text will get blur only when I use Kernel 2.6.15.1 but it will > > > > disappear when I change to kernel 2.4.27. > > > > Since the accelerator in the XFree4.3.0 is XAA, are there some > > > acceleration > > > > in XAA will be effected by the kernel? > > > > Ps: My platform is sis741. > > > > Appreciate your help in advance. > > > > Sincerely Yours, > > > > cckuo > > > > > > > > _______________________________________________ > > > > XFree86 mailing list > > > > XFree86@xxxxxxxxxxx > > > > http://XFree86.Org/mailman/listinfo/xfree86 > > > > > > > > > _______________________________________________ > > XFree86 mailing list > > XFree86@xxxxxxxxxxx > > http://XFree86.Org/mailman/listinfo/xfree86 > > _______________________________________________ XFree86 mailing list XFree86@xxxxxxxxxxx http://XFree86.Org/mailman/listinfo/xfree86
