Comment # 15 on bug 1135030 from Martin Li������ka

(In reply to Dario Faggioli from comment #14)
> (In reply to Martin Li������ka from comment #13)
> > > 
> > > just FTR, this is currently blocking the build of QEMU, on ppc64, in Factory
> > > and Tumbleweed...
> > 
> > I think it's a known issue:
> > https://github.com/fmtlib/fmt/issues/2794
> > 
> Ok, thanks for the pointer! :-)
> 
> The symptoms really looks the same, although, in that issue, the problem
> seems to come from the fact that the library and the tool wanting to use it
> are build with different compilers (gcc and clang).
> 
> This isn't true in our case (or at least, it doesn't look true to me). But
> maybe we are also building with incompatible options? Or could it be
> something else?

I think it's about a mixture of 2 long double formats

This comes from GCC 12 changes:

On POWER systems which support it, the -mabi=ieeelongdouble option now selects
the IEEE 128-bit floating point format for REAL(KIND=16). R16_IBM and R16_IEEE
have been added to the -fconvert option, the CONVERT specifyer of the OPEN
statement and the GFORTRAN_CONVERT_UNIT environment variable.

The 2 related options are: -mabi=ibmlongdouble, -mabi=ieeelongdouble and they
will be likely mixed.