On Fri, 2020-09-11 at 12:31 +0200, Andrea Corallo wrote: > Hi Dave, > > thanks for the review! > > David Malcolm <dmalc...@redhat.com> writes: > > [...] > > > Was there a reason for using reinterpret_cast here, rather than > > static_cast? > > Yes the reason is that apparently we can't use static_cast for that: > > "error: invalid ‘static_cast’ from type ‘gcc_jit_lvalue*’ to type > ‘gcc::jit::recording::global*’" > > Why is that is another question I fear I'm not very qualified to > answer. > If you feel is necessary and want to suggest some rework to improve > this > I'll be happy to put some effort in if you like.
It's because we're effectively downcasting from lvalue to global. This is safe because of the check for global->is_global (). A slightly cleaner way to do it would be to have a dyn_cast_global vfunc here rather than is_global, and to use the result - but I don't think it matters. [...] > > > After applying the suggestions and having tested it I've > > > installed > the > patch as 4ecc0061c40. > > Thanks! Thanks Dave
