[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: FONT PROBLEMS! XF4-phase1 debs, v8 -> v11, mga, 16 bpp

>> "Thomas E. Vaughan" <vaughan@tensor.nssl.noaa.gov> writes:

 > Thank you for your efforts to help me understand what is happening.
 > Embarrassingly for me, I don't understand precisely what you are
 > saying.

 Oh, sorry, I tend to do that. :-(

 > My understanding is that there are
 >      (1) an mga kernel module and
 >      (2) an mga (X) server module, also known as the mga driver.

 There's a kernel drm module (in your case it's the mga module).
 There's a XFree86 driver (mga_drv) and there's a XFree86 dri module
 (mga_dri).  In order for your setup to work in direct rendering mode
 you need all of this, plus the agpgart kernel module, and you have to
 run the X server in 16 bpp mode (framebuffer).

 The kernel drm module and the XFree86 dri module are developed by a
 group of people at Presicion Insight (this is hosted at sourceforge,
 see http://dri.sourceforge.net/ if you want more information).  AFAIK
 the XFree86 driver is not developed by PI but by the XFree86 "itself"
 (Branden?).  PI's code was merged into the XFree86 code short before
 the 4.0.1 release.  XFree86 4.0.1's code was merged into PI short
 after the 4.0.1's release.  As far as I understand it, code merges
 between the two code bases don't happen that often.

 The problem you are seeing stems from the fact that Branden is
 applying patches over the 4.0.1 code in order to solve /other/
 problems, but these patches don't touch the DRI code, that means the
 DRI code is desynced with respect to the rest of the code.  Darryl's
 reply suggest they already saw/solved this problem, and he also
 hinted where the problem might be (planemasks).  I'll look into this
 over the weekend and try to produce a patch for Branden.

 And just to clear the mess in my previous email, I thought you were
 running DRI's code on top of Branden's packages, /because/ I hadn't
 seen this problem.  Until now, that is.  I just realized that when
 using Branden's packages my configuration wasn't running in direct
 mode.  Now I have fixed that and I get this problem, too.  Using
 DRI's code this doesn't happen.
 > Are you saying that I shouldn't be using the stock 2.4.0-test6
 > kernel module and should rather be using the kernel module from the
 > dri project's cvs repository?

 The current DRI's kernel code was merged in the kernel recently, so,
 if you are running 2.4.0-test6 it's the same if you use either one,
 it's exactly the same code.  If you use something like 2.3.99, it
 becomes an issue.

 > If so, then does Branden include in his debs the relevant kernel
 > module (either as compiled for some versions of the kernel or as
 > source code)?

 I was thinking about that.  This would be useful for people using pre
 2.4 kernels.  AFAICS, the code /should work/ even with 2.1.something
 kernels.  But in your case the code in the kernel source is ok.

 > Or are you saying that Branden's server module, the mga driver
 > mga_drv, is inconsistent with the rest of the patched XFree 4.0.1
 > that Branden distributes in his latest debs.

 That's it.  More specifically, it's incosistent with the sections of
 the code that are maintained by PI (the dri modules).

 Sorry again about the confusion,


Reply to: