Re: [PATCH 4/6] xfs: introduce xfs_bmapi_remap

[Christoph Hellwig <hch@lst.de>]

On Mon, Apr 10, 2017 at 12:50:04PM -0700, Darrick J. Wong wrote:
> Why change all this if the next patch removes the whole function except:
> 
> ip->i_d.di_nblocks += length;
> xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
> 
> Wouldn't it make more sense to have this patch add _bmapi_remap (as it
> appears in the next patch) and change the callers to use it; and then
> the next patch removes the old _bmap_remap_alloc and its callers?

I'd rather keep steps as small and self-explaining as possible,
so one is factoring out a new helper, and the other is removing a call
that's no needed.

> > +	else if (ip->i_d.di_format != XFS_DINODE_FMT_BTREE)
> > +		logflags &= ~XFS_ILOG_DBROOT;
> > +
> > +	if (logflags)
> > +		xfs_trans_log_inode(tp, ip, logflags);
> > +	if (cur) {
> > +		xfs_btree_del_cursor(cur,
> > +			error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
> 
> Double indent here?

Sure..