On 08/02/2017 08:55, Vladimir Sementsov-Ogievskiy wrote: > 07.02.2017 02:19, Eric Blake wrote: >> On 02/03/2017 09:47 AM, Vladimir Sementsov-Ogievskiy wrote: >>> Return 0 on success to simplify success checking. >>> >>> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsement...@virtuozzo.com> >>> --- >>> nbd/client.c | 35 +++++++++++++++++++---------------- >>> 1 file changed, 19 insertions(+), 16 deletions(-) >> I'm not sure that this simplifies anything. You have a net addition in >> lines of code, so unless some later patch is improved because of this, >> I'm inclined to say this is needless churn. >> > > I just dislike duplicating information like "drop_sync(ioc, 124) != > 124". In the code there is no place where positive and not equal to size > return value actually handled. But it is not so important, if you are > against i'll drop this, no problem.
I think I agree with Vladimir. > One note: I don't have good understanding of the following: actually > read can return positive value < queried size, which means that we > should read again. But it is not handled in the code (handled, but just > as an error), except drop_sync.. (With drop_sync it is side effect of > using limited buffer size, yes?). Is it all ok? It is handled in nbd_wr_syncv. Paolo
