Re: [csmith-bugs] int64_t generated even if --no-math64 --no-longlong

[Xavier Leroy <Xavier.Leroy@inria.fr>]

Dear Xuejun,



2011/8/2 Xuejun Yang <jxyang@cs.utah.edu>:
> Hi Xavier,
>
> Thanks for the bug reporting. I wonder which version of Csmith were you
> running.

Right, I should have mentioned that I am experimenting with Csmith
2.0.0 as distributed on http://embed.cs.utah.edu/csmith/.   I didn't
notice that the GIT repository was available too, so maybe I should
switch to the development version of Csmith from the GIT repo.

Best wishes,

- Xavier


I can't reproduce the problem with the GIT head. And I have run
> this command 10,000 times without any problem:
>
> csmith --no-math64 --no-longlong | grep int64
>
> thanks,
> -Xuejun
>
>>
>> Dear Csmith developers,
>>
>> Before finding the --ccomp option, I was fiddling with the --no-math64
>> and --no-longlong options to try to prevent Csmith from ever
>> generating anything involving 64-bit ints.  I was surprised that even
>> with these two options, Csmith generates code that uses int64_t and
> uint64_t:
>>
>> xleroy@jollyjumper:~/tmp/csmith-2.0.0$ src/csmith --seed 123 --no-math64
>> --no-longlong | grep int64
>>    int64_t  f3;
>>    int64_t  f0;
>>    int64_t  f7;
>>    int64_t  f3;
>>    const uint64_t  f9;
>>
>> Is this intended?  I feel that it would be useful to have options that
>> guarantee that no 64-bit integers are generated, ever.
>>
>> Best wishes,
>>
>> - Xavier
>
>