Re: [Qemu-devel] [PATCH] dirty-bitmap: remove unnecessary return

[Changlong Xie]

On 07/01/2016 02:18 AM, John Snow wrote:
> On 06/30/2016 10:00 AM, Jeff Cody wrote:
> > On Thu, Jun 30, 2016 at 04:45:52PM +0800, Changlong Xie wrote:
> > > On 06/30/2016 04:25 PM, Fam Zheng wrote:
> > > > On Thu, 06/30 16:01, Changlong Xie wrote:
> > > > > Otherwise, we could never trigger assert(!bitmap->successor)
> > > > >
> > > > > Signed-off-by: Changlong Xie <address@hidden>
> > > > > ---
> > > > >   block/dirty-bitmap.c | 1 -
> > > > >   1 file changed, 1 deletion(-)
> > > > >
> > > > > diff --git a/block/dirty-bitmap.c b/block/dirty-bitmap.c
> > > > > index 4902ca5..e9df5ac 100644
> > > > > --- a/block/dirty-bitmap.c
> > > > > +++ b/block/dirty-bitmap.c
> > > > > @@ -131,7 +131,6 @@ int bdrv_dirty_bitmap_create_successor(BlockDriverState *bs,
> > > > >       if (bdrv_dirty_bitmap_frozen(bitmap)) {
> > > > >           error_setg(errp, "Cannot create a successor for a bitmap that is "
> > > > >                      "currently frozen");
> > > > > -        return -1;
> > > > >       }
> > > > >       assert(!bitmap->successor);
> > > >
> > > > This is wrong. Then we will always trigger assert for a frozen bitmap.
> > >
> > > IMO, when it's a frozen bitmap, we will always return -1. So
> > > "assert(!bitmap->successor)" is useless here, am i right?
> >
> > I don't see a path where the assert could trigger, so I would agree that the
> > assert itself, while harmless, is not necessary (although it could be argued
> > it is in place in case the code above it changes in a way that does not
> > check bitmap->successor).

Agree

> > That doesn't mean we want to try and trigger an assert, however! :) The
> > error return is the proper error handling -- we don't expect that asserts
> > should ever be encountered QEMU, if one happens that is a sign of a bug.

Got it

> > Jeff
>
> The assert was indeed added to ensure that if the valid states of the
> bitmap later expanded or changed, that the status checkers (e.g.
> bdrv_dirty_bitmap_frozen()) were changed to match.

Thanks for all explanations. Although my brain always forces to think 
it's a redundant execution path, but since it's harmless, let's keep it.

> .