On Thu, 29 Aug 2013, Nic Volanschi (R&D) wrote:
> On Thu, 2013-08-29 at 13:53 +0200, Julia Lawall wrote:
> > Perhaps to solve all of the problems, you could just put braces around all
> > if branches first? It seems likely that this will nto work in a generic
> > way, with a metavariable S, but you can make rules like
> >
> > @@
> > statement S;
> > @@
> >
> > if (...)
> > +{
> > return ...;
> > +}
> > else S
> >
> > To get all of the possibilities.
>
> I see what you mean. In fact, to solve the problem it would suffice to
> add braces only around "else" branches, so your rule would be rather:
>
> @@ statement S; expression E; @@
> if (...) S else
> +{
> \(return ...; \| E; \| <other-simple-stmt-patterns...> \)
> +}
>
> But the problematic case is exactly the one causing the problem, in
> which the "else" branch is an unbraced "if-then-else" statement. If we
> write the rule for that case:
>
> @@ statement s1, s2, s3; expression e; @@
> if(...)
> s1
> else
> +{
> if(...) s2 else s3
> +}
>
> It applies well when you have just one "else if", but not when there are
> more of them:
>
> --- test.c:
> void GetInfoDestTV(short sNoFo)
> {
> if(sNoFo == 1)
> printf("one");
> else if(sNoFo == 2)
> printf("two");
> else if(sNoFo == 3)
> printf("two");
> else
> printf("other");
> }
>
> In this case, I'll get the same error as in my initial rule:
>
> Fatal error: exception Failure("rule starting on line 1: already tagged
> fake token
> ")
OK, I agree.
julia
_______________________________________________
Cocci mailing list
[email protected]
https://systeme.lip6.fr/mailman/listinfo/cocci
