Re: [Patch,Build,Quad,Fortran] PR 46520: Build only with languages=fortran; fix build with cross compiler

[Tobias Burnus <burnus at net-b dot de>]

Jakub Jelinek wrote:
> Why?  libquadmath ought to be usable not just from Fortran, so it should be
> built always when the target supports __float128.

I think that's at least the opinion of Joseph; I am not 100% sure about 
Ian, though. If there is consensus, one can also drop the change to 
gcc/fortran/* and keep building it on all _float128-supporting systems.

From #gcc:

<jsm28> libquadmath should build only for Fortran and only for the 
relevant architectures.
[16 hours earlier, recorded for me by "gccbot %seen"]

[Ian mentioned a failure with non-native builds; discussion moved to why 
(technical reason) it is always build:]
16:39 < iant> the question is when you put target-libquadmath in 
target_libraries in configure.ac
16:41 < iant> or maybe it is target_libs in gcc/fortran/config-lang.in
16:41 < xchat> shouldn't quad precision mathematics (binary128) be 
supported in C/C++ as well (as a GCC extension)?
16:42 < jsm28> Supporting for any other target would require a properly 
defined ABI for it first.
16:43 < jsm28> Supporting the "_Complex __float128" syntax is bug 32187, 
best fixed by making __float128 a keyword much like __int128.

Tobias