Re: AR archive iteration broken?

[Richard Guenther <rguenther at suse dot de>]

On Sun, 1 Nov 2009, Roland McGrath wrote:

> Are you aware of a different libelf implementation (e.g. Solaris) where
> your test program works as you expect?

It works with libelf from Michael Riepe and I got the idiom from
some Solaris libelf tutorial (can't find it right now).

> The normal idiom for archive iteration is to use the return value from
> elf_next, rather than ignoring it.  It returns ELF_C_NULL when there is no
> later element.  You don't need to test that return value directly unless
> you want to.  The normal idiom is to pass that return value in to
> elf_begin, and ELF_C_NULL to elf_begin just returns NULL.  e.g.:
> 
>   Elf_Cmd cmd = ELF_C_READ;
>   while ((subelf = elf_begin (fd, cmd, elf)) != NULL)
>     {
>       Elf_Arhdr *arhdr = elf_getarhdr (subelf);
>       ... work on subelf ...
>       /* Get next archive element.  */
>       cmd = elf_next (subelf);
>       if (elf_end (subelf) != 0)
> 	... diagnose error ...;
>     }

Ok, I'll see if that works across the different libelfs I have access to.
Documentation seems to be a scarce resource here...

Thanks,
Richard.