Subject: Re: dlopen() for macppc XFree86 & enabling the 'nv' driver by default.
To: None <>
From: Chris Gilbert <>
List: tech-x11
Date: 08/23/2002 09:54:10
matthew green said:
> hi folks.
> to get X working on my G4 tower i had to enable the "nv" driver for the
> macppc (and more) XFree86.  this works OK for me, except that
> for the following files, i had to add a patch (copied from elsewhere in
> the xc tree) like:


Sounds good

> i also have tested this configuration with the dlopen() support for all
> "drivers" in XFree86.  this is a pretty huge difference:
> -rws--x--x  1 root  wheel  7738708 Aug 21 00:49
> /usr/X11R6.working/bin/XFree86* -rws--x--x  1 root  wheel  2042426 Aug
> 22 14:36 /usr/X11R6/bin/XFree86*
> with this as a new directory (in kilobytes):
> 	9434    /usr/X11R6/lib/modules

Excellent 8)

> i'd like to commit both of the changes inthe patch below, but i am much
> less sure about the MAXSHORT problem.  the only reason i've
> added the 3 line patch to the relevant files is that i found a
> handful of other files that do this - when in rome, as they say.
> comments?

Has xsrc been tagged for 1.6 yet? if it hasn't you'll also need to do the
package set changes and get them pulled up into the 1.6 branch, and at this
late stage I'm not sure if that'll happen or not.  IE talk to releng and
find out if they're happy to take such a change now, they might be as it
only effects macppc x sets.  Actually by adding the nv module does that
mean that i386 now produces an nv module, so it needs tweaking.

Actually what would be ideal would be just to have 2 package set lists for
pci based X servers, one for dynamic architectures eg macppc, i386, one for
monolithic, eg cats.  But that should wait for the xsets to get tidied up
(currently the X 4 ones seem to be a bit messy)