Re: [V9fs-developer] Re: [RFC][patch 3/7] v9fs: VFS inode operations (2.0-rc6)

["Ronald G. Minnich" <rminnich@lanl.gov>]

On Tue, 24 May 2005, Pekka Enberg wrote:

> > +static int
> > +v9fs_vfs_create(struct inode *inode, struct dentry *dentry, int perm,
> > +		struct nameidata *nd)
> > +{
> > +	int retval = -EPERM;
> > +	int open_mode = O_RDWR;
> > +
> > +	retval = v9fs_create(inode, dentry, perm, open_mode);
> > +
> > +	return retval;

> 
> Both local variables are redundant. Please just do:
> 
> 	return v9fs_create(inode, dentry, perm, O_RDWR);
> 
well, here is my first point of disagreement with the many good 
suggestions I have seen. 

I prefer the style as it is, and dislike the return vfs_create(...) 
change. 

First, we're all agreed here that this is a style, not efficiency thing, 
right? Makes no difference, compiler will do what is right. 

The reason I prefer the 'retval = blah blah; return retval;' usage is that
I frequently have run v9fs in UML, and debugging is just easier. Set the
breakpoint in v9fs_vfs_create, step through, maybe skip over that
v9fs_create function, look at retval. Also, should we get to the point
later that we have a debug print of some sort in that function, adding it
is trivial.

I realize there is a preferred style here, but is there room for some 
flexibility?

thanks

ron