Re: [PATCH 0/8] drop if around WARN_ON

[Julia Lawall]

On Sun, 4 Nov 2012, Sasha Levin wrote:

> On Sun, Nov 4, 2012 at 10:57 AM, Julia Lawall <julia.lawall@xxxxxxx> wrote:
> > On Sun, 4 Nov 2012, Sasha Levin wrote:
> >
> > > Hi Julia,
> > >
> > > On Sat, Nov 3, 2012 at 4:30 PM, Julia Lawall <Julia.Lawall@xxxxxxx> wrote:
> > > > These patches convert a conditional with a simple test expression and a
> > > > then branch that only calls WARN_ON(1) to just a call to WARN_ON, which
> > > > will test the condition.
> > > >
> > > > // <smpl>
> > > > @@
> > > > expression e;
> > > > @@
> > > >
> > > > (
> > > > if(<+...e(...)...+>) WARN_ON(1);
> > > > |
> > > > - if (e) WARN_ON(1);
> > > > + WARN_ON(e);
> > > > )// </smpl>
> > >
> > >
> > > So this deals with WARN_ON(), are you considering doing the same for
> > > the rest of it's friends?
> >
> >
> > I tried WARN_ON_ONCE, but the pattern never occurred.  Are there others that
> > are worth trying?
>
> Definitely!
>
> Here's the semantic patch I've got:
>
> @@
> expression e;
> @@
>
> (
> - if (e) WARN_ON(1);
> + WARN_ON(e);
> |
> - if (e) WARN_ON_ONCE(1);
> + WARN_ON_ONCE(e);
> |
> - if (e) WARN_ON_SMP(1);
> + WARN_ON_SMP(e);
> |
> - if (e) BUG();
> + BUG_ON(e);
> )
>
> This gave me a really huge patch output.
>
> I can send it out if you think the patch above looks good.

I didn't change any cases where the if test contains a function call.  The 
current definitions of WARN_ON seem to always evaluate the condition 
expression, but I was worried that that might not always be the case.  And 
calling a function (the ones I remember were some kinds of print 
functions) seems like something one might not want buried in the argument 
of a debugging macro.

WARN_ON_SMP is just WARN_ON if CONFIG_SMP is true, but it is just 0 
otherwise.  So in that case it seems important to check that one is not 
throwing away something important.

I remember working on the BUG_ON case several years ago, and other people 
worked on it too, but I guess some are still there...  The current 
definitions of BUG_ON seem to keep the condition, but there are quite a 
few specialized definitions, so someone at some point might make a version 
that does not have that property.

julia

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/