Re: clang memory sanitizer report unsigned int overflow

[Lorenzo Donati <lorenzodonatibz@...>]

In some context I may imagine an uint overflow could be a symptom of a 
bug even if it has a perfectly defined semantic in C. Maybe that tool 
can be configured to signal uint overflows even if they are perfectly 
defined in C?Just guessing.

-- Lorenzo

On 05/01/2023 15:05, Roberto Ierusalimschy wrote:
> > For reference that is this line marked with **:
> >
> >     |  unsigned int luaS_hash (const char *str, size_t l, unsigned int seed)
> > {
> >     |    unsigned int h = seed ^ cast_uint(l);
> >     |    for (; l > 0; l--)
> > **|       h ^= ((h<<5) + (h>>2) + cast_byte(str[l - 1]));
> >     |    return h;
> >     |  }
> >
> > Though not sure which part of that would be causing an add overflow because
> > h is an unsigned int which I believe should not trigger UB when it
> > overflows.
>
> The 'reason' itself states "Unsigned integer overflow", which does not
> exist in ISO C...
>
>      Process 626150 stopped
>      * thread #1, name = 'ufsrv', stop reason = Unsigned integer overflow
>                                        ??????????????????????????????????
>
> -- Roberto