what's difference between "relocatable link" and "final link"?

what's difference between "relocatable link" and "final link"?

[Huang, Chunhao]

Greetings,

This is chunhao.

When I study the BFD library, I meet "relocatable link" and
"final link" several times, but I can NOT find out the difference
between them. My rough comprehension is that "relocatable link"
done by assembler, which is something like parsing the local
symbols in sections. while "final link" done by linker.

But I find such sentence in bfdint.text:
"Make sure you clearly understand the operations the linker must
perform during a relocatable link and during a final link." 
Which means that both "relocatable link" and "final link" are
performed by linker. So my orignal comprehension is not correct.

Is there any guy who can provide me the difference between them?
If you can provide me some simple examples in ARM or x86, that's
great.

Thanks a lot.
Best Regards

Chunhao

Re: what's difference between "relocatable link" and "final link"?

[Ian Lance Taylor]

"Huang, Chunhao" <Chunhao.Huang@sony.com.cn> writes:

> When I study the BFD library, I meet "relocatable link" and
> "final link" several times, but I can NOT find out the difference
> between them. My rough comprehension is that "relocatable link"
> done by assembler, which is something like parsing the local
> symbols in sections. while "final link" done by linker.

A relocateable link is what the linker does when you use the -r
option.

A final link is what the linker does when you do not use the -r
option.

Ian

Re: what's difference between "relocatable link" and "final link"?

[Nick Clifton]

Hi Chunhao,

> A relocateable link is what the linker does when you use the -r
> option.
> 
> A final link is what the linker does when you do not use the -r
> option.

To expand on Ian's answer, a relocatable link is one where only some of 
the relocations in the input files are resolved.  Any relocations that 
cannot be resolved are preserved and recorded in the output file.

A final link is one where all(1) of the relocations in the input files 
are resolved.  If a relocation cannot be resolved the linker will 
produce an error message.

The purpose of a relocatable link is to create an object file where some 
of the relocation work has already been done.  Thus when this file is 
included in a final link, the linker will take less time to complete its 
task.

Cheers
   Nick

(1) Really we are talking about static relocations here.  Dynamic 
relocations (against symbols in dynamic libraries) are not completely 
rresolved by the linker even during a final link.  They have to be 
resolved by the dynamic loader when the executable starts running.

Re: what's difference between "relocatable link" and "final link"?

[Simon Richter]

[-- Attachment #1: Type: text/plain, Size: 1318 bytes --]

Hello,

Huang, Chunhao schrieb:

> When I study the BFD library, I meet "relocatable link" and
> "final link" several times, but I can NOT find out the difference
> between them.

Basically, after a "relocatable" link, you have an object file that is 
similar to the input files, that is, it consists of multiple (in most 
cases zero-based) code blocks (sections) along with relocation (i.e. 
what to change if you move the code to another address) and symbol (i.e. 
how specific addresses are named) information. This format makes sense 
for further linking, but you would need a lot of postprocessing to 
actually get executable code. After a "final" link, the sections are 
concatenated into segments (this is what the loader cares about), and 
have absolute addresses in the virtual address space[1], and all symbols 
have been resolved and replaced by either their numerical value, or by a 
"dynamic symbol" if you have a (runtime) linker that supports shared 
libraries. After a final link, you can drop the symbol and relocation 
information at no harm, it is no longer needed (except for debugging).

Gee, somebody should document this. :-)

    Simon

[1] if there is no virtual address space, the output of a final link may 
still have relocation information, and the executable is relocated by 
the loader.

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 307 bytes --]

RE: what's difference between "relocatable link" and "final link"?

[Huang, Chunhao]

Dear Ian, Nick, and Simon:

Thank you very much for your information provided, I can catch the
difference between these 2 kinds of link now. :-)
If I need more help from you, I will send you next time.

Best Regards
Chunhao

> -----Original Message-----
> From: Simon Richter [mailto:Simon.Richter@hogyros.de]
> Sent: Monday, June 12, 2006 7:11 PM
> To: Huang, Chunhao
> Cc: binutils@sourceware.org
> Subject: Re: what's difference between "relocatable link" and "final
link"?
> 
> Hello,
> 
> Huang, Chunhao schrieb:
> 
> > When I study the BFD library, I meet "relocatable link" and
> > "final link" several times, but I can NOT find out the difference
> > between them.
> 
> Basically, after a "relocatable" link, you have an object file that is
> similar to the input files, that is, it consists of multiple (in most
> cases zero-based) code blocks (sections) along with relocation (i.e.
> what to change if you move the code to another address) and symbol
(i.e.
> how specific addresses are named) information. This format makes sense
> for further linking, but you would need a lot of postprocessing to
> actually get executable code. After a "final" link, the sections are
> concatenated into segments (this is what the loader cares about), and
> have absolute addresses in the virtual address space[1], and all
symbols
> have been resolved and replaced by either their numerical value, or by
a
> "dynamic symbol" if you have a (runtime) linker that supports shared
> libraries. After a final link, you can drop the symbol and relocation
> information at no harm, it is no longer needed (except for debugging).