Re: Building XServer

[der Mouse <mouse%Rodents-Montreal.ORG@localhost>]

> I thought the X server did poke in /dev/mem or similar, since it
> probes and plays with the hardware rather direct.

Only on badly-designed systems.  It should be using mmap() on the
device to map the regions of interest, rather than making stupid
assumptions about the machine's physical memory layout.

And it is, on the systems I use, except for one where the only working
X server I know of is XFree86 - it's architecturally a peecee, and I
mostly don't use X on peecees because it means XFree86, but this one
has enough other benefits to make me overlook it.  Fortunately I'm
lucky enough that it mostly Just Works.

> It also tries to find the PCI memory mapping (or the fact that there
> aren't any, as in this case).

That sounds like XFree86, and is the sort of thing that led me to call
it a crawling design horror.  Even discovering whether there _is_ a PCI
bus is way, way out of scope for userland; the server has no business
doing any kind of bus enumeration - that's what the kernel is for.  An
architecture that requires the server to do muck about with that stuff
is Just Plain Broken.  If X.org has been corrupted enough by XFree86
that it pulls that sort of shenanigans, I would say it's reason to
scrap it and use something with non-insane design.

> Oh, no doubt you *could* make it work.  My thought was only that
> noone have.

Likely.

I really need to dust off my VAXen.  Not sure what they've got for
framebuffers in them - I know my uV2 doesn't have a framebuffer in it,
but I think I've got a qbus fb somewhere.  Not sure about the desktop
VAXen.

                                        Mouse