On Wed, 3 Nov 2010, Michael T wrote:
>
> Hello,
>
> I wanted to point out a couple of issues with your mini_null_ref2.cocci
> example.
> The first is that it is lacking a test:
>
> --- mini_null_ref2-2.cocci 2010-11-03 22:33:39.059472002 +0100
> +++ mini_null_ref2.cocci 2010-11-03 22:35:16.529472002 +0100
> @@ -9,13 +9,14 @@
>
> @@
> expression E,E1;
> -identifier f,f1,fld,fld1;
> +identifier f,fld,fld1;
> statement S1,S2;
> @@
>
> E->fld = f(...);
> ... when != E = E1
> when != E->fld1 = E1
> + when != f1(&E)
> if (
> - E
> + E->fld
>
> although that should also allow and number of arguments to f1 before and
> after &E,
Yes, this seems like a good idea. You can instead write f1(...,&E,...)
which allows &E to appear in any argument position. I will make the
change in the example on the web site.
> I just don't know the right coccinelle syntax. The second point is that the
> suggested
> fix is only one possible case - it may just be that someone is being too
> careful and
> testing for NULL although it is not needed. So the script should probably do
> a
> printf rather than suggesting a correction.
Normally, Coccinelle produces a patch, unless you give the -in_place or -o
option. The user should then look at this patch and see if it is
appropriate. In this case, I think it is appropriate often enough that it
is useful to have a patch already made, if one does want to apply it. If
it did a printf, one would have to make the adjustment manually, and could
make a mistake. On the other hand, if you prefer a printf, it is easy to
change (the following generates emacs org mode):
@r@
type T;
expression E;
identifier i,fld,f1;
position p,p1;
@@
T i...@p = E->fld;
when != i
when != f1(...,&E,...)
if (e...@p1 == NULL) { ... return ...; }
@script:python@
p << r.p;
p1 << r.p1;
@@
cocci.print_main("bad declaration",p)
cocci.print_secs("bad test",p1)
There is no trace, though of what the change could be, although one could
adjust the strings to give this information.
Thanks for the feedback.
julia_______________________________________________
Cocci mailing list
[email protected]
http://lists.diku.dk/mailman/listinfo/cocci
(Web access from inside DIKUs LAN only)
