RE: Exploring the possibility of linking objects with different relocationschemes.

RE: Exploring the possibility of linking objects with different relocationschemes.

[Prafulla Thakare]

Hi,

We have changed the ABI VERSION in elf header to identify the different 
relocation schemes. i.e. ABI = 1 stands for new (binutils 2.15 onwards) 
relocations while ABI = 0 stands for old (binutils 2.10) relocations. 

In linker, the relocations will be handled based on the ABI version. If 
the ABI version is "1", then relocation handling will remain same but 
if ABI version is "0", then relocations will be handled the way it is
done in 
binutils 2.10.1. 
For this purpose, we plan to re-use relocation handling source code of
2.10.1.

Please let us know, if this okay and also if anybody foresee any problem

with this implementation. 

Also any input on how above implementation will affect "-r" option to
linker?

>>Can anybody suggest how feasible is to
>> modify present linker to link objects with different relocation
>>scheme at the same time?

>This should be possible.  You may find that you cannot support the 
>--emit-relocs switch since you would not be able to label the output 
>file as containing both old and new ABI relocs.

-->This is okay.

Regards,
Prafulla Thakare
KPIT Cummins InfoSystems Ltd.
Pune, India

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Free download of GNU based tool-chains for Renesas' SH and H8 Series. 
The following site also offers free technical support to its users. 
Visit http://www.kpitgnutools.com for details. 
Latest versions of KPIT GNU tools are released on Oct 1, 2004. 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Re: Exploring the possibility of linking objects with different relocationschemes.

[Nick Clifton]

Hi Prafulla,

> We have changed the ABI VERSION in elf header to identify the different 
> relocation schemes. i.e. ABI = 1 stands for new (binutils 2.15 onwards) 
> relocations while ABI = 0 stands for old (binutils 2.10) relocations. 

Note - you should probably define constants for these values and put 
them into include/elf/sh.h.

> Please let us know, if this okay and also if anybody foresee any problem

This is OK.

> Also any input on how above implementation will affect "-r" option to
> linker?

It should not be a problem.  As long as the code correctly propagates 
the ABIVERSION value to the output bfd (which should happen 
automatically, you should not need to add any new code).

Cheers
   Nick