On Tue, Feb 14, 2006 at 06:54:42PM +0300, Valeriy E. Ushakov wrote:
> On Tue, Feb 14, 2006 at 15:48:53 +0000, Nick Hudson wrote:
> 
> > That's not my understanding of Pavel's roblem...
> > 
> > On Saturday 11 February 2006 21:30, Pavel Cahyna wrote:
> > > In my case it won't, unless I add an explicit statement for .rodata:
> > >
> > >    .rodata    :
> > >    AT (LOADADDR(.text) + (ADDR(.rodata) - ADDR(.text)))
> > >    {
> > >      *(.rodata)
> > >    }
> > 
> > Are we mandating .rodata?
> 
> He said he needs to add .rodata to the the output .text section, so my
> (...)

I never said this... .rodata was in the output .text section in the
original Jason's script and I did not change it. What I said was that I
had to add .rodata separately, see above. (Now when I'm thinking about it,
why didn't they conflict?)

Now I looked if the .rodata is needed at all, so I commented out both places:
(...)
   .text (KERNEL_BASE_virt) :
   AT (KERNEL_BASE_phys)
   {
     *(.text)
     *(.text.*)
     *(.stub)
/*     *(.rodata) */
   } =0
/*   .rodata    : 
   AT (LOADADDR(.text) + (ADDR(.rodata) - ADDR(.text)))
   { 
     *(.rodata)
   } */
(...)

Now ld complains:
/home/pavel/scratch/current/sys/arch/i386/conf/kern.ldscript:49 non constant expression for load base

Given that the ldscript has 48 lines, reporting an error on line 49 seems
a bit strange, but well, nobody is perfect.

Pavel