On Tue, Mar 07, 2023 at 07:51:03PM +0100, Bernhard Reutner-Fischer wrote:
> While it's a nice idea, there have been resentments towards (visible)
> C++ in the fortran frontend and especially the library, i think.
I thought libgfortran is written in C and Fortran and doesn't use gmp/mpfr,
so this doesn't apply to it (ok, intrinsics/trigd.inc uses mpfr_*
names macros if in library which do something different).
As for the FE, we don't need to change all places with manual
allocation/deallocation for those, though changing most of them
will help with maintainability as one doesn't have to care about leaks.
I only see 2 mpfr_init2 calls in Fortran FE though, so pressumably
everything else could use visually C like auto_mpfr var;
instead of mpfr_t var + mpfr_init + mpfr_clear.
Jakub
