Re: [RFC/PATCH] Fix `bfd_{get,set}_*' macros

[Doug Evans <dje at google dot com>]

On Wed, May 2, 2012 at 9:19 PM, Sergio Durigan Junior
<sergiodj@redhat.com> wrote:
> Hi,
>
> As a part of an effort to enable `-Wunused-variable' on GDB build
> system, I would like to propose this "fix" to the following macros
> declared in bfd-in2.h:
>
> ?bfd_get_section_name
> ?bfd_get_section_vma
> ?bfd_get_section_lma
> ?bfd_get_section_alignment
> ?bfd_section_name
> ?bfd_section_size
> ?bfd_get_section_flags
> ?bfd_get_section_userdata
>
> Those macros don't need the first argument (`bfd'), but they still
> require it. ?However, you can pass anything there and GCC won't complain
> about it, because the argument is unused after all. ?So, while fixing
> GDB to support -Wunused-variable, I saw that there was a declaration of
> a `bfd *', which is used as a first argument to those macros. ?GCC
> started to complain about it (unused argument). ?There were two paths
> that I could follow:
>
> 1) Delete the declaration of `bfd *', so this:
>
> ? bfd *abfd = objfile->abfd;
> ? bfd_get_section_name (abfd, sec);
>
> Would become:
>
> ? bfd_get_section_name (objfile->abfd, sec);
>
> or
>
> 2) Fix the macros, so that they use the first argument in a safe way
> like `(void) bfd', thus silencing GCC warnings.
>
> After having tried the first option, Pedro asked me to actually fix the
> macros, so here is the patch to do this, along with small fixes through
> the code in order to adapt it. ?I am labeling this patch as RFC because
> I'm not sure I fixed the code in the right way. ?It wasn't clear to me
> which `bfd' to use in each situation, so I used my best judgement (which
> may not be good).
>
> As you will notice, I did not changed every bfd_*_section_* macro
> because some of them are used as lvalue (left side of assignment).
>
> I regtested the patch on Fedora 16 x86_64 with GDB, without
> regressions. ?Ok to apply?

I never liked these macros, it's like they had two "this" arguments.

How much more work would it be to actually remove the bfd argument?