Re: [PATCH] Set the retain attribute on _elf_set_element if CC supports [BZ #27492]

["Fāng-ruì Sòng" <maskray@google.com>]

On Mon, Mar 15, 2021 at 12:28 AM Florian Weimer <fweimer@redhat.com> wrote:
>
> * Fangrui Song via Libc-alpha:
>
> > So that text_set_element/data_set_element/bss_set_element defined
> > variables will be retained by the linker.
> >
> > Note: 'used' and 'retain' are orthogonal: 'used' makes sure the variable
> > will not be optimized out; 'retain' prevents section garbage collection
> > if the linker support SHF_GNU_RETAIN.
>
> This needs recent-ish GCC 11, see commit 6347f4a0904fce17e ("Add retain
> attribute to place symbols in SHF_GNU_RETAIN section).
>
> > +#ifdef __has_attribute
> > +#if __has_attribute (retain)
> > +# ifdef SHARED
>
> Indentation is off there.

OK. I can increase the level of indentation...

> It's probably simpler to define a new macro, like
> ELEMENT_SECTION_ATTRIBUTES, and use that unconditionally, perhaps like
> this?
>
> #ifndef __ASSEMBLER__
> # if defined (__has_attribute) && __has_attribute (retain)
> #  define ELEMENT_SECTION_ATTRIBUTES used, retain
> # else
> #  define ELEMENT_SECTION_ATTRIBUTES used
> # endif
> #endif

`#ifdef __has_attribute` should be fine without #ifndef __ASSEMBLER__ ?

> Have you checked that __has_attribute (retain) is only true on GCC if
> the underlying linker has retain support and the attribute does not
> produce a warning?

There is a GCC bug. I've reported
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99587

>
> Thanks,
> Florian
>