Re: strip vs. nm vs. readelf

[Erik Leunissen <e dot leunissen at hccnet dot nl>]

Indeed, the output of "nm -D" produced a lot of symbols, and inspection 
of readelf output listed a lot of symbols in the .dynsym section.

Thanks. This resolves my initial confusion regarding apparent 
contradictory output of these programs.

It also made me want to learn more about;
- the meaning of the various sections;
- which parts of source code can be found in which sections (related to 
loading, relocation and dynamic linking).

Could anybody recommend books or other sources of information from where 
I could start an exploration of these areas? My first interest is an 
introductory text, but more advanced material is also most welcome.


Thanks in advance,

Erik Leunissen
==============





> On Wed, Dec 14, 2005 at 10:24:11PM +0100, Erik Leunissen wrote:
>
> > L.S.
> >
> > I stumble across some apparent contradictory behaviour of the 
> > applications strip, nm and readelf.
> >
> > This occurs when processing a shared lib "myLib.so" which has been 
> > subjected to the strip application as follows:
> >
> >    strip --strip-unneeded myLib.so
>
>
>        --strip-unneeded
>            Remove all symbols that are not needed for relocation processing.
>
> ... but leaves the dynamic symbol table; that is always necessary.
>
>
> > Next, when I invoke:
> >
> >    nm -a myLib.so
> >
> > the output is (as expected):
> >
> >    nm: myLib.so: no symbols
>
>
> Try nm -D myLib.so.
>
>
> > Finally, when doing:
> >
> >    readelf -a myLib.so
> >
> > a lot of information is emitted, including the names of symbols that I 
> > believed to be removed by strip.
>
>
> See which section it says they're in.  They'll be in .dynsym, which is
> part of the executable image.