On Sun, 20 Oct 2019, Joe Perches wrote: > On Sun, 2019-10-20 at 21:52 +0200, Julia Lawall wrote: > > On Sun, 20 Oct 2019, Joe Perches wrote: > [] > > > There's probably a generic cocci mechanism to check function > > > prototypes and then remove uses of unnecessary void pointer casts > > > in function calls. I'm not going to try to figure out that syntax. > > > > With the --recursive-includes option, perhaps: > > > > @r@ > > identifier f; > > parameter list[n] ps; > > type T; > > identifier i; > > @@ > > > > T f(ps, void *i, ...); > > > > @@ > > expression e; > > identifier r.f; > > expression list[r.n] es; > > @@ > > > > f(es, > > - (void *)(e) > > + e > > ,...) > > > > This of course only works for functions that have prototypes, and not for > > macros. It will also run slowly. > > You are not kidding about slow, but it doesn't seem to work > for mem<foo>, maybe because system includes aren't analyzed. No they are not. > Single file processing time on an XPS13 averages more than > 100 seconds per file. Not surprising. Actually, --include-headers-for-types should provide some benefit. That discards the header files after the type inference. > Also: > > expression e; > > could probably be better as: > > type T; > T *p; Good point. expression *e; would be sufficient. julia > > as some of the expressions cast to void are int or size_t > and it's probably better to restrict the conversions to > just pointer or array types. > > > _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
