Re: SDAREL output sections

[Geoff Keating <geoffk at geoffk dot org>]

> Date: Mon, 19 Feb 2001 10:18:02 +0100
> From: Momchil Velikov <velco@fadata.bg>
> X-Accept-Language: en,bg,ru
> CC: binutils@sourceware.cygnus.com
> 
> Geoff Keating wrote:
> > 
> > > From: Momchil Velikov <velco@fadata.bg>
> > > Date: 19 Feb 2001 01:14:30 +0200
> > 
> > > Geoff> Because this would prevent people from overriding the test by folding,
> > > Geoff> say, the .data section into .sdata.
> > >
> > > The test, as it stands now, prevents people (well, me) from folding the
> > > .sdata section into .data.
> > 
> > This is true, but why do you wish to do this?
> 
> For n + 1 reasons:
>   a) the n reasons (yet unknown to me) for which the people would
>      want to do the opposite (that's folding .data into .sdata) :-)

You mean, to ensure that no sdata relocs are in use?  That part
already works :-).

>   b) the PMAC OpenFirmware 1.0.5 recognizes only .text, .data and .bss
>      when loading a program. So my linker script looks like
>        .data { *(.data) *(.sdata) }
>        .bss  { *(.sbss) *(.bss) }

Hmmm.  Why doesn't it use the PHDR entries?  What does it do if there
are no section entries at all?

I think that for such special cases (broken loaders, vxworks, etc.),
it's best to consider them a different format from ELF.

-- 
- Geoffrey Keating <geoffk@geoffk.org>