another ill test?

[Alan Modra]

On Mon, Nov 20, 2006 at 10:38:01AM +0100, Jan Beulich wrote:
> ld-elf/warn2 uses symbol2w.s, which contains
> 
>         .section        .gnu.warning,"a",%progbits
> 	.global Foo
>         .type   Foo, %object
>         .size   Foo, 20
> Foo:
>         .string "function 'Foo' used"
> 
> As a result, warn2.d expects Foo to end up in .text, but to me placement
> of a symbol in a to be discarded section seems ill in the first place.

Well, yes, no one would write real code like this.  Typically, you'd
put a .gnu.warning section (with no symbols) in an object file
containing code/data for one function.

> I'd
> therefore like to move Foo into .data, and additionally relax warn2.d to
> not expect a specific section number.

No, leaving Foo there helps us test that symbols defined in discarded
sections are handled properly.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre