[cfe-dev] CLang Static Analysis -- false positives

[Kent Williams]

OK, well I guess what I need to do is submit a preprocessed source
file where I think there's a false positive.  Thanks for your help.

One thing I do like about scan-build is that a lot of the problems it
catches have implications beyond just bugs -- if I look at the code
and understand why it is complaining, many times it pushes me to come
up with a clearer and more robust flow of control within a given
function.

It also can fill one with dispair, if you process 3rd party libraries
that can't directly fix.  The world seems full of shoddy code that
accidentally works.

On Thu, Apr 26, 2012 at 11:54 AM, Jordy Rose <jediknil at belkadan.com> wrote:
> Sorry, what result are you expecting? There's no problem with the code you've posted. Are you /getting/ a warning that you don't expect?
>
> If you can find something that is definitively a bug, though, it's probably best to file it on our bug tracker at http://llvm.org/bugs/, though checking here to see if it's really a bug (or if it's a known problem) is perfectly okay.
>
> Best,
> Jordy
>
>
> On Apr 26, 2012, at 12:45, Kent Williams wrote:
>
>> I'm using scan-build (aka CLang Static Analysis) to check through ITK
>> (http://www.itk.org).
>>
>> There is a common 'single exit from function' pattern used several places:
>>
>> int
>> somefunc()
>> {
>>  int status = EXIT_SUCCESS;
>>  // do a bunch of stuff, and if there's a failure
>>  if(failure)
>>    {
>>    status = EXIT_FAILURE;
>>    }
>>  return status;
>> }
>>
>> I tried making a simple test case to flush out this false positive
>> error, but I can't seem to trigger it.  Is there some trick to this?
>> _______________________________________________
>> cfe-dev mailing list
>> cfe-dev at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>