[cfe-dev] Found a bug - maybe?

[Clark Gaebel]

Here's a reduced test case. Strangely enough, if I change buffer to be
of size 1, this code works fine.

    clark at clark-server ~/test $ clang++ --version
    clang version 1.1 (branches/release_27)
    Target: x86_64-pc-linux-gnu
    Thread model: posix
    clark at clark-server ~/test $ cat foo.cpp
    #include <cstdio> // For printf.

    int main()
    {
            const char buffer[] = {
                    0x00, 0x11
            };

            if(buffer > (buffer - 4))
                    printf("SUCCESS!\n");
            else
                    printf("FAILURE!\n");
    }

    clark at clark-server ~/test $ clang++ -O0 foo.cpp && ./a.out
    SUCCESS!
    clark at clark-server ~/test $ clang++ -O2 foo.cpp && ./a.out
    FAILURE!



On 06/07/10 00:03, Clark Gaebel wrote:
> A picture is worth a thousand words (and by picture, I mean shell
> output). This is with
>
>     clark at clark-server ~/test $ clang++ --version
>     clang version 1.1 (branches/release_27)
>     Target: x86_64-pc-linux-gnu
>     Thread model: posix
>     clark at clark-server ~/test $ cat foo.cpp
>     #include <cstdlib>
>     #include <cstdio>
>
>     struct ParseFailed {};
>
>     int main()
>     {
>             bool succeeded = false;
>
>             const char buffer[] = {
>                     0x00, 0x11
>             };
>
>             try {
>                     if(buffer > (buffer - 4))
>                             throw ParseFailed();
>
>                     succeeded = false;
>             } catch(ParseFailed) {
>                     succeeded = true;
>             }
>
>             if(succeeded)
>                     printf("SUCCESS!\n");
>             else
>                     printf("FAILURE!\n");
>     }
>
>     clark at clark-server ~/test $ clang++ -O0 foo.cpp && ./a.out
>     SUCCESS!
>     clark at clark-server ~/test $ clang++ -O2 foo.cpp && ./a.out
>     FAILURE!
>
> What should I do? This little oddity popped up in one of my unit tests
> while running my codebase through clang.
>
>   

-- 
Regards,
-Clark

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20100607/33c8dbd0/attachment.html>