On Tue, Nov 18, 2014 at 06:31:09PM +0300, Dan Carpenter wrote:
> On Mon, Nov 17, 2014 at 06:13:51PM -0800, Jörn Engel wrote:
>
> > - smatch gave no warning. The README file in the git tree is for
> > sparse, so I may have used it wrong. But with either default
> > parameters of -Wuninitialized there is no warning.
>
> Sorry for that. I've pushed some changes so hopefully the documentation
> to build and run is easier to find now. It's under Documentation/.
Thank you!
> > As a testcase, I can reduce the code down to nine lines:
> >
> > void callee(int *arg);
> > int caller(void)
> > {
> > int arg;
> > callee(&arg);
> > if (!arg)
> > return 1;
> > return 0;
> > }
> >
> > Is this something you could cook up a test for? I would like need two
> > weeks to work myself through the source code before I can even get
> > started.
>
> That test case will drive you nuts. You need to have cross function
> analysis for there to be a manageable number of false positives.
>
> Smatch does cross function analysis, but playing with this I see that I
> need to make some improvements here. There are still way way too many
> false positives. I have some ideas here.
>
> Anyway, just for laughs, I've attached the check I'm working on. You'll
> need to add it check_list.h and recompile.
Wow! That was far faster than I had hoped for. I'll see if this is
already useful somehow.
Personally I would even be happy to annotate call-by-reference
parameters with something like __set and __unset. If annotated with
__set, callers can assume the parameter is initialized on function
return and compilers can warn if the callee returns without initializing
it. __unset is the opposite, warn in the caller if the parameter wasn't
previously initialized.
But I assume that Linus would hate this. He's never been very fond of
annotating source code to make life easier on code checkers.
Jörn
--
Act only according to that maxim whereby you can, at the same time,
will that it should become a universal law.
-- Immanuel Kant
--
To unsubscribe from this list: send the line "unsubscribe smatch" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
