[Bf-committers] PVS-Studio Static Analysis

[Erwin Coumans]

You are looking at the code that I fixed yesterday, see my commit message
for details here:
http://lists.blender.org/pipermail/bf-blender-cvs/2012-April/045090.html


>>V610 Undefined behavior. Check the shift operator '<<. The left operand
'(~0)' is negative. extern_bullet btquantizedbvh.h 82

~(0) is signed -1, negative, while ~(x&0) is positive unsigned, so it fixes
the undefined behavior.
Thanks,
Erwin



On 24 April 2012 18:01, Jason Wilkins <jason.a.wilkins at gmail.com> wrote:

> This line in btQuantizedBvh.h confuses me:
>
> int     getTriangleIndex() const
>        {
>                btAssert(isLeafNode());
>                unsigned int x=0;
>                unsigned int y = (~(x&0))<<(31-MAX_NUM_PARTS_IN_BITS);
>                // Get only the lower bits where the triangle index is
> stored
>                return (m_escapeIndexOrTriangleIndex&~(y));
>        }
>
> Ostensibly the error is that ~0 is -1 (unsigned) and shifting an
> unsigned is undefined or unspecified (forget which).
>
> But I'm actually confused why this code is written this way.
> x=0, then x&0 == 0, even if x wasn't 0.
>
> Only after that do we get to the undefined behavior.
>
> I guess I could puzzle through it, but somebody else might just know :-)
>
> On Tue, Apr 24, 2012 at 7:30 PM, Jason Wilkins
> <jason.a.wilkins at gmail.com> wrote:
> > I'm working through them and fixing what Clang did not catch.
> >
> > I was going to say the same thing as Nicholas, code analysis is such a
> > wide open field with virtually infinite number of things you can check
> > for, that having more than one tool is a good idea.
> >
> > On Tue, Apr 24, 2012 at 6:30 PM, Nicholas Bishop
> > <nicholasbishop at gmail.com> wrote:
> >> Not necessarily; different static analyzers can detect different types
> >> of potential problems. The PVS analysis contains some interesting
> >> things like the "Misprint in a homogeneous code block" section that I
> >> don't think clang's analyzer does.
> >>
> >> On Tue, Apr 24, 2012 at 7:20 PM, Tom M <letterrip at gmail.com> wrote:
> >>> We have a page with generated CLang static analysis,
> >>>
> >>> http://clang.blenderheads.org/trunk/
> >>>
> >>> It probably shows the same bugs as PVS-Studio does.
> >>>
> >>> LetterRip
> >>>
> >>> On Tue, Apr 24, 2012 at 4:15 PM, Nicholas Bishop
> >>> <nicholasbishop at gmail.com> wrote:
> >>>> I think some of these have been fixed already in recent commits from
> Campbell.
> >>>>
> >>>> On Tue, Apr 24, 2012 at 7:11 PM, Jason Wilkins
> >>>> <jason.a.wilkins at gmail.com> wrote:
> >>>>> http://www.viva64.com/en/b/0145/#ID0EO3BK
> >>>>>
> >>>>> It appears that Andrey Karpov has done an analysis of Blender source
> >>>>> code using his PVS-Studio tool.  He did this just yesterday, so I
> >>>>> assume the problems he found are still in the source.  He offers free
> >>>>> licenses to open source project members (3500 euro software
> otherwise,
> >>>>> wow).
> >>>>>
> >>>>> I was thinking of investigating the problems he found and seeing if I
> >>>>> could fix them.
> >>>>> _______________________________________________
> >>>>> Bf-committers mailing list
> >>>>> Bf-committers at blender.org
> >>>>> http://lists.blender.org/mailman/listinfo/bf-committers
> >>>> _______________________________________________
> >>>> Bf-committers mailing list
> >>>> Bf-committers at blender.org
> >>>> http://lists.blender.org/mailman/listinfo/bf-committers
> >>> _______________________________________________
> >>> Bf-committers mailing list
> >>> Bf-committers at blender.org
> >>> http://lists.blender.org/mailman/listinfo/bf-committers
> >> _______________________________________________
> >> Bf-committers mailing list
> >> Bf-committers at blender.org
> >> http://lists.blender.org/mailman/listinfo/bf-committers
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers
>