Re: [PATCHv2 03/16] staging: rtl8192u: r8192U_core: add temporary variables to keep lines under 80 characters

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Thu, Sep 17, 2015 at 01:06:33AM -0400, Raphaël Beamonte wrote:
> 2015-09-17 0:57 GMT-04:00 Greg Kroah-Hartman <gregkh@linuxfoundation.org>:
> <SNIP>
> >> @@ -1748,8 +1755,9 @@ static short rtl8192_usb_initendpoints(struct net_device *dev)
> >>               oldaddr = priv->oldaddr;
> >>               align = ((long)oldaddr) & 3;
> >>               if (align) {
> >> -                     newaddr = oldaddr + 4 - align;
> >> -                     priv->rx_urb[16]->transfer_buffer_length = 16 - 4 + align;
> >> +                     align = 4 - align;
> >> +                     newaddr = oldaddr + align;
> >> +                     priv->rx_urb[16]->transfer_buffer_length = 16 - align;
> >
> > At a quick glance, this conversion looks wrong...
> 
> What is wrong with it?
> 
> oldaddr + 4 - align;
> can also be read:
> oldaddr + (4 - align);
> 
> as well as
> 16 - 4 + align;
> can also be read
> 16 - (4 - align);
> as when we remove the parenthesis, the - sign invert the parenthesis
> elements signs.
> 
> Calculating (4 - align) previously thus preserve the logic here!

Ugh, it's been a long day, yeah, ok, this is the same, you are right.

But step back please, what exactly is this trying to do?  I think it's a
round_up type function, perhaps that should be what we do instead (the
kernel has such functions already.)

> > And it's not what your changelog text said you were doing :(
> 
> It's true that I didn't add a new temporary variable here but instead
> re-used one that is not used after, but I thought it went in the same
> idea as the rest of this patch. Should I separate that as another
> patch?

It might be the same "idea", but it's not what you did, so don't try to
sneak it in.

greg k-h