re: kern/41850: implicit agp dependency in (radeon)drm code

[matthew green <mrg%eterna.com.au@localhost>]

      | >How-To-Repeat:
      | Take GENERIC, comment out agp, enable radeondrm, try to compile.
    
    Are we (or you (generic you)) treating this kind of config error as a bug
    these days?
    
    If so, there are a gazillion ways it is possible to build a NetBSD
    kernel config file that config's without errors, but then generates
    either compilation errors, or more commonly as here, link errors, when
    the result is used to attempt to build a kernel.
    
    I'd always put this kind of thing down to operator error - just use a
    correct config file, and it all works.  I'd prefer it if all the 
dependencies
    could be automatic, but they never have been in the past.


this is true.

we'd like to make as many combos work and this one looks to be a
little invasive to fix properly, and possibly largely useless.
the radeon PCIE support depends upon the agp support, so turning
it off may not be a useful thing for any modern system.


.mrg.