Re: [PATCH][next] fs: hpfs: Avoid multiple -Wflex-array-member-not-at-end warnings

[Gustavo A. R. Silva]

On 05/02/25 23:33, Al Viro wrote:
> On Tue, Feb 04, 2025 at 02:13:37PM +1030, Gustavo A. R. Silva wrote:
>
> > +#define anode_btree	container_of(&anode->btree, struct bplus_header, __hdr)
> > +#define ranode_btree	container_of(&ranode->btree, struct bplus_header, __hdr)
> > +#define fnode_btree	container_of(&fnode->btree, struct bplus_header, __hdr)
> > +
> >   /* Find a sector in allocation tree */
> >   
> >   secno hpfs_bplus_lookup(struct super_block *s, struct inode *inode,
> > @@ -27,7 +31,7 @@ secno hpfs_bplus_lookup(struct super_block *s, struct inode *inode,
> >   				a = le32_to_cpu(btree->u.internal[i].down);
> >   				brelse(bh);
> >   				if (!(anode = hpfs_map_anode(s, a, &bh))) return -1;
> > -				btree = &anode->btree;
> > +				btree = anode_btree;
>
> Just for this - NAK.  And then you proceed to add dozens more of the same.
>
> If it looks like a variable name, it must not turn out to be a bloody macro;
> if a macro expansion depends upon a local variable with given name existing
> in scope of its use, make that name an explicit argument.  You manage to
> violate both.

Yeah, so the story is that I originally wrote this patch using container_of()
to directly replace all those instances of `&anode->btree`, `&ranoce->btree`
and `&fnode->btree`. But I really didn't like how it turned out, so I changed
it to this form at the last minute, and well, as you pointed out, that wasn't
the best decision.

OK, so the options I see are either using container_of() directly (as I
originally had it), or write a small wrapper to keep the lines shorter.
Probably something like this:

#define GET_BTREE(ptr) \
	container_of(ptr, struct bplus_header, __hdr);

and then of course just using it like this:

	btree = GET_BTREE(&anode->btree);

-Gustavo