On Tue, Sep 26, 2017 at 11:03:09PM +0200, Kamil Rytarowski wrote: > On 26.09.2017 22:33, Joerg Sonnenberger wrote: > > On Tue, Sep 26, 2017 at 10:28:23PM +0200, Kamil Rytarowski wrote: > >> Can we just import as-is the FreeBSD version? It has support for pre-C11 > >> compilers. This has header to be supported in the clean C99 mode. > > > > I see absolutely no reason for wanting to support tgmath.h in pre-C11 > > mode and a lot of very good reasons why it should *not* work. > > > > C99 7.22 Type-generic math <tgmath.h>
That doesn't invalidate anything I said. Seriously, C99 doesn't provide the means to implement it sanely. The sizeof hacks are slow and fragile, completely inacceptable. > >> I propose to add for the completeness C++ support in this header as > >> defined by the C++ standard. > > > > tgmath.h is strictly a C-only header. Again, I see no reason why this > > trainwreck should be supported by anything but the strict minimum > > requirements. > > > > "For compatibility with the C standard library and the C Unicode TR, the > C ++ standard library provides the 25 C headers, as shown in Table 154." > and the table lists "<tgmath.h>". D.5 (Annex D) > > <ctgmath> has been introduced in C++11 and marked deprecated in C++2017 > (draft). Usage of the original C header in C++ code is marked deprecated > since C++2011 (D.5) > > Keeping C++ compat costs little, there is another option to include for > C++ >= 2011 the <ctgmath> header, but it depends on the proper C++ > runtime headers. This is why I propose to manually include <cmath> and > <complex>. It's been deprecated as soon as it has been added to C++. That's a good enough reason to not bother either. Joerg
