linker bug 10515

[Alan Modra]

On Mon, Oct 05, 2009 at 01:01:46PM +0100, Nathan Sidwell wrote:
> Alan Modra wrote:
>
>>> 2) make it an error to have FILEHDRS and/or PHDRS in not-the-first  
>>> segment of the PHDRS statement.

I'd assumed here that you meant PT_LOAD header by "segment".

>> This is the best option IMO.
>
> This patch does that.  We allow FILEHDRS/PHDRS provided it's the first 
> segment, or the previous segment had such headers.  The phdrs.t testcase 
> has such an example.
>
> tested on powerpc-elf, ok?

No.  I think you should just check PT_LOAD headers.  That would allow
various non-load headers to appear first if anyone so desires
(eg. PT_PHDR, PT_INTERP).  It would also simplify the documentation
and error message.

-- 
Alan Modra
Australia Development Lab, IBM