On Mon, 29 Jul 2019 at 10:29, Paolo Bonzini <pbonz...@redhat.com> wrote: > > On 29/07/19 11:21, Peter Maydell wrote: > > On Mon, 29 Jul 2019 at 09:51, Paolo Bonzini <pbonz...@redhat.com> wrote: > >> There is no case where we're using per-.o file CFLAGS for anything other > >> than dependencies. > > > > disas/libvixl is a counterexample -- we use per-.o-file CFLAGS for: > > * suppressing warnings in third-party code we don't want to > > carry local modifications to > > * dealing with a NetBSD weirdness about header inclusion order > > The NetBSD thing could be worked around with a static library but > instead those -D options could be added as global C++ flags: > > __STDC_LIMIT_MACROS and __STDC_CONSTANT_MACROS are a workaround to > allow C++ programs to use stdint.h macros specified in the C99 > standard that aren't in the C++ standard. > > Likewise, the -Wno-sign-compare probably should be added to all files > for GCC <=4.6, but in fact we don't support anymore GCC 4.6 so it can go > away.
I think it's an indication that the mechanism in general is useful. Switching to a new build system worries me if we already find that it is lacking flexibility we're using with our current build system -- it suggests that there's likely to be missing stuff we're going to run into in future as well... thanks -- PMM
