Re: [Qemu-devel] [PATCH 2/5] coccinelle: use DIV_ROUND_UP

[Marc-André Lureau]

Hi

----- Original Message -----
> On 7 June 2017 at 08:46, Marc-André Lureau <address@hidden>
> wrote:
> > The coccinelle/round.cocci script doesn't catch hard coded values.
> >
> > I used the following script over qemu code base:
> >
> > (
> > - ((e1) + 3) / (4)
> > + DIV_ROUND_UP(e1,4)
> > |
> > - ((e1) + (3)) / (4)
> > + DIV_ROUND_UP(e1,4)
> 
> Why do we need both of these? Is it just "coccinelle is weird" ? :-)

I am total newbie to coccinnelle-land, but I think this one is useless 
duplication

> 
> > |
> > - ((e1) + 7) / (8)
> > + DIV_ROUND_UP(e1,8)
> > |
> > - ((e1) + (7)) / (8)
> > + DIV_ROUND_UP(e1,8)
> > |
> > - ((e1) + 15) / (16)
> > + DIV_ROUND_UP(e1,16)
> > |
> > - ((e1) + (15)) / (16)
> > + DIV_ROUND_UP(e1,16)
> > |
> > - ((e1) + 31) / (32)
> > + DIV_ROUND_UP(e1,32)
> > |
> > - ((e1) + (31)) / (32)
> > + DIV_ROUND_UP(e1,32)
> > )
> 
> > -                     next_op = op_pointer + ((oplen + 7) / 8);
> > +                     next_op = op_pointer + (DIV_ROUND_UP(oplen, 8));
> 
> I think there's a coccinelle trick for making it drop
> now-unnecessary brackets in substitutions like this, but I forget
> what it is. Maybe it's as simple as having substitutions for
> 

> > - (((e1) + 7) / (8))
> > + DIV_ROUND_UP(e1,8)
> 
> as well?


I think you need a second rule:
@@
expression e1;
expression e2;
@@
-(DIV_ROUND_UP(e1,e2))
+DIV_ROUND_UP(e1,e2)


I will fix it in second version.