Am 22.06.2017 um 15:54 hat Pavel Butsykin geschrieben: > On 22.06.2017 01:17, Max Reitz wrote: > >On 2017-06-13 14:16, Pavel Butsykin wrote: > >>The flag as additional precaution of data loss. Perhaps in the future the > >>operation shrink without this flag will be banned, but while we need to > >>maintain compatibility. > >> > >>Signed-off-by: Pavel Butsykin <pbutsy...@virtuozzo.com>
> >The functional changes look good to me; even though I'd rather have it > >an error for qcow2 now (even if that means having to check the image > >format in img_resize(), and being inconsistent because you wouldn't need > >--shrink for raw, but for qcow2 you would). But, well, I'm not going to > >stop this series over that. > > > > Why shrink for qcow2 image is dangerous, but for raw is not dangerous? I > think we should provide the same behavior for all formats. When --shrink > option will become necessary, it also should be the same for all image > formats. It is dangerous for both, but for raw we can't enforce the flag immediately without a deprecation period. With qcow2 we can (because it is new functionality), so we might as well enforce it from the start. Kevin