On Thu, 5 Nov 2020 at 22:15, Richard Henderson <richard.hender...@linaro.org> wrote: > > On 11/5/20 1:23 PM, Peter Maydell wrote: > > +} __attribute__((aligned(16))); > > Hmph, 96 uses of the attribute directly, 20 uses of QEMU_ALIGNED. I suppose > we > should just remove the wrapper...
Oops, I forget about that. We're better at adhering to use of QEMU_SENTINEL and QEMU_NORETURN, at least. And a fair chunk of those 96 are in code-that's-not-ours like the headers imported from Linux or the pc-bios/s390-ccw code. I'm in two minds here -- the wrappers look less clunky than the __attribute__ syntax, but on the other hand "there is only one way this can be written" results in less inconsistency than "there are two ways". thanks -- PMM