On Fri, 9 Aug 2019 at 16:39, Eric Blake <ebl...@redhat.com> wrote:
> Also, would it be worth swapping out existing constants in the code base
> that should instead be using these macros, so that they actually have a
> use and so that we can see whether using them adds legibility?
>
> For example, block/nvme.c, block/qapi.c, block/sheepdog.c, blockdev.c,
> util/async.c, util/oslib-win32.c, util/qemu-thread-posix.c,
> util/qemu-timer.c all seem to be dealing with conversions between
> seconds and subdivisions thereof, where constants 1000000 or larger are
> in use and could be rewritten with these.

I'm not sure that it would be more readable to replace
1000000000LL with SI_G -- I would tend to assume the latter
would be 2^30. Using "1000LL * 1000 * 1000" inline at
the point of use, or better still abstracting any particular use
into something more semantically meaningful as we already
do with NANOSECONDS_PER_SECOND, would be my personal preference.

thanks
-- PMM

Reply via email to