[Cocci] Matching format strings

[julia.lawall@lip6.fr (Julia Lawall)]

On Tue, 4 Jul 2017, Julia Lawall wrote:

> > @@
> > bindings
> > identifier x : script:python(some bindings, even from this rule) { code }
> > @@
> > a
> > -b
> > +new(stuff, x)
> > -c
>
> OK, maybe it could be
>
> identifier x = script:python(some bindings, but not from this rule) { code }
>
> Bindings from the same rule are problematic.  CTL is bottom up, so you
> don't know the bindings of the other variables when you are matching a
> given one.  I guess a possibility would be to ignore the scripts and
> collect all the bindings possible (ie any possible identifier for x above),
> and then check them at the end.

I'm sorry already for having suggested this.  It would be a mess, I think.
When there is ..., variables can have different values on different
control-flow paths.  I guess that their quantifiers could be pushed up
around the scripts that depend on them, but I'm not sure that the user
would find the result very understandable.

julia