Re: Does gold support deadcode elimination on relocatable files

["Neo Jia" <neojia at gmail dot com>]

On Fri, Dec 19, 2008 at 7:43 AM, Ian Lance Taylor <iant@google.com> wrote:
> "Neo Jia" <neojia@gmail.com> writes:
>
>>> There is work underway to do deadcode elimination, but not during a
>>> relocatable link.
>>
>> Could you show the reason why you don't have this for a relocatable file?
>
> To expand on what Cary said, garbage collection for a relocatable link
> is significantly different from garbage collection for a link of an
> executable.  When linking an executable you have a precisely defined
> root set for garbage collection: the entry point, any symbols
> specified with -u, any symbols in certain magic sections like .ctors.
> When doing a relocatable link there is no root set.  You wind up
> keeping everything.
>
> After the garbage collection patch for executables is committed, it
> would be perfectly reasonable to extend it by providing a mechanism
> for specifying an explicit root set.  At that time it would be
> reasonable to enable garbage collection for relocatable links when a
> root set is specified.  This work would not require any significant
> changes to the existing garbage collection work.

As most relocatable files are used in SDK, can we just use the SDK
header files to build the root set?

Thanks,
Neo

>
> Ian
>



-- 
I would remember that if researchers were not ambitious
probably today we haven't the technology we are using!