Long long problem

ROOT
1

Hi,

I want to use the 64 bit integer type (long long or Long64_t) in my macro but I get funny results when I run it in ROOT or CINT. This is a code that demonstrates it:

//
 unsigned long long  func(){
   unsigned long long  m=4294967295ULL;
   unsigned long long  x=79ULL;
   unsigned long long  r=x%m;
   return r;
}

void test(){
   cout<<func()<<endl;
   cout<<func()<<endl;
   cout<<func()<<endl;
   for(int i=0;i<3;i++){
      cout<<func()<<endl;
   }
}

//

The output is:
79
79
79
0
0
0

i. e. I get wrong result when I call the function func() from the for-loop.
When I use m=4294967296ULL which is 2^32, ROOT crashes with “floating point exception”. When I compile the code everything seems to be ok.

My system:
ROOT version 5.10/00
cint version 5.16.8
Linux version 2.6.12-1-686 (gcc version 4.0.2 20050917 (prerelease) (Debian 4.0.1-8 ))

I tried this on 2 more PCs with other ROOT versions. I get similar results with:
ROOT version 4.04/02, cint 5.15.169,
Linux version 2.4.25 (gcc version 2.95.3 20010315 (release))

BUT NOT with:
ROOT 4.03/02, cint 5.15.159
Linux version 2.4.29-z1 (gcc version 2.95.3 20010315 (release))
-here it seems to be o.k.

Do you think it is a CINT bug?

Cheers,
Juraj

2

Hi,

This is problem in the CINT bytecode compiler you can work around the problem either by disabling it (CINT command: .O 0) or by compiling your code. We will correct this issue as soon as possible (but this might be after the current overhaul being done on CINT).

Thanks for reporting this issue.
Philippe

3

Thanks for replay.
I noticed that compiling the code while loading the file with "++"
works fine. (compiling always takes some time though…)

Cheers,
Juraj

4

Hi,

If you use just “+”, it will only recompile if you have changed the source file.

Cheers,
Philippe