Re: simplest way to get the size of an output section

[Paul Brook <paul at codesourcery dot com>]

On Monday 04 October 2004 15:06, Etienne Lorrain wrote:
>  --- Daniel Jacobowitz <drow@false.org> a écrit?:
> > On Mon, Oct 04, 2004 at 11:51:01AM +0200, Etienne Lorrain wrote:
> > >   On a Linux ia32 system, I am trying to get the size of an output
> > >  section of an ELF file to use it automatically in a Makefile.
> > >
> > >  Ideally I would like to compile a small GCC program like:
> > > gcc -Wl,--defsym,linker_tells_size=SIZEOF(.text16) print_command_line.c
> > >  but that will not work because --defsym cannot get the SIZEOF().
> > >
> > >  So I have gone the objcopy way:
> > > objcopy --output-target binary --only-section .text16 vmlinux param.bin
> > > ./print_command_line ` cat param.bin | wc -c `
> > >  That does the job, but is there a way to not create the temporary file
> > >  "param.bin" (i.e. objcopy uses standard output) (because it is always
> > >  a problem in Makefiles: it may already exist and be write protected,
> > >  it needs to be cleaned by make clean, an interrupted make command
> > >  may leave the temporary file...) or another way, without using too
> > >  complex awk or sed scripts?
> >
> > Why not use objdump -h or readelf -S?
>
>   Basically because of the complexity of the sed or awk to get back the
>  hexadecimal number (in a Makefile):
>  readelf gives all the sections, so (not tested - not a real awk master)
> ./print_command_line ` readelf -S vmlinux | awk "/.text16/ {printf $...}" `
>  objdump - when I tried yesterday - wanted the number of the input
>  section and need quite a complex filtering of the standard output too.

How about "size -A"

Paul