Re: [PATCH 2/9] Add __vfscanf_internal and __vfwscanf_internal with flags arguments.

[Florian Weimer <fweimer at redhat dot com>]

On 03/13/2018 01:35 PM, Adhemerval Zanella wrote:
> > @@ -1026,7 +1030,9 @@ compat_symbol (libc, __nldbl_vdprintf, vdprintf, GLIBC_2_0);
> >   compat_symbol (libc, __nldbl_vsnprintf, vsnprintf, GLIBC_2_0);
> >   compat_symbol (libc, __nldbl_vsprintf, vsprintf, GLIBC_2_0);
> >   compat_symbol (libc, __nldbl__IO_sscanf, _IO_sscanf, GLIBC_2_0);
> > +#if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_28)
> >   compat_symbol (libc, __nldbl__IO_vfscanf, _IO_vfscanf, GLIBC_2_0);
> > +#endif
> >   compat_symbol (libc, __nldbl___vfscanf, __vfscanf, GLIBC_2_0);
> >   compat_symbol (libc, __nldbl___vsscanf, __vsscanf, GLIBC_2_0);
> >   compat_symbol (libc, __nldbl_fscanf, fscanf, GLIBC_2_0);

> I am not sure that nesting SHLIB_COMPAT is correct way logically and stylisc here:
> the compat symbol will be create only for ABI which initial version of GLIBC_2_0
> (which might not be an issue with current ABIs that requires ldbl to dbl compat
> symbol).

Wouldn't GLIBC_2_0 be expanded to the relevant default ABI baseline for 
the port and do the right thing here?

Thanks,
Florian