Re: [Outreachy kernel] Coccinelle Challenge Problem 2

[Julia Lawall <julia.lawall@lip6.fr>]

[-- Attachment #1: Type: text/plain, Size: 1617 bytes --]

> > Did you try, eg
> >
> > e =
> > - (e1 + e2)
> > + e1 + e2
> >
> > ?  This is not a good solution, but it could be good to think about why.
>
> I tried this after reading this mail :). I think it's not a good solution
> becauseof the false positives that it will give. Also for the cases where
> there's no
> space around '+' will be replaced as well which is not the intended use case
> for the patch. And all the cases where we have shifted the operands to avoid
> the 80 character limit, will all be moved on the same line.

Yes, it's all pretty sad :).  One conclusion is that it is better not to
remove code and then add it back.  When you add it back, Coccinelle takes
over the pretty printing, and Coccinelle is not super smart about pretty
printing binary operators.  It does try to do a good job for function
arguments, though.

But you probably noticed that it was changing things that didn't have
parentheses at all, which may be what you mean by false positives.  This is
because of the idea of an isomorphism.  Coccinelle considers that sometimes
people put parentheses and sometimes they don't.  And it is rather tiresome
to have to specify all of the possibilities by hand.  So if you put an
expression with parentheses that is either all annotated with - or that is
all unannotated, then it will consider the possibility that the parentheses
are not there as well.  You can see all of the patterns it is really trying
with spatch --parse-cocci foo.cocci.  But the way you wrote the rules, the
- was only on the (), so it considers that the parentheses are important,
so there is no problem.

julia