[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [csmith-dev] math64 and longlong



This bug is now fixed in our development version.

In the near future Eric will send a message saying how to get access to it.

John



On 5/4/11 10:58 AM, Lee Taylor wrote:
Hi John,

That is certainly interesting.

Going off of that hint, I've reran configure with "-arch i386" in CXXFLAGS and rebuilt csmith.  It now seems to be behaving as expected.   It looks like there is definitely a bug here, but this should allow me to work around it for now.

Thanks,
Lee





On May 4, 2011, at 12:44 PM, John Regehr wrote:

Ok, I think I figured this out.  I couldn't reproduce this behavior because on a 32-bit machine this command:

csmith -s 1111004472  --no-math64 --no-longlong

includes this output:

struct S2 {
   int32_t  f0;
};

on a 64-bit machine the same invocation contains this:

struct S2 {
   int64_t  f0;
};

Utah people, what is going on here??  Csmith output must depend only on the random seed and other command line options.

John




On 05/04/2011 10:19 AM, John Regehr wrote:
Hi Lee,

I can't reproduce the behavior you're seeing, can you provide a full
command line including random seed?

If it is the stuff in safe_math.h that you are seeing -- as opposed to
code actually generated by Csmith -- then you can suppress it by
defining the NO_LONGLONG CPP symbol.

John



On 05/04/2011 09:29 AM, Lee Taylor wrote:
Hello,

I'm investigating the use of csmith for testing an embedded compiler
that doesn't have any 64 bit support, but I'm running into a little
trouble.

I've tried adding the "--no-math64" and "--no-longlong" options when
running csmith, but the generated output still seems to make heavy use
of the int64_t and uint64_t types. Is there some other option that I
also need to enable?

Thanks in advance for your help.

Best regards,
Lee