Long64_t and root 5.26.00b

Hi, I think it is a cint problem, because when compiled it disappears.
I send a macro and a library.
The result of the macro is:
test:
2.55529e+10 25552929944 25552929944
test passed OK!
in the loop:
—>>> 1.48663e+14 * 0.000171885 = 2.55529e+10 or 2147483648
NOT OK! why?

basically when doing casting Long64_t insida a look I get max of 32-bit integer. Outside the loop it is fine.

Cheers, Mariusz

PS. Library is 32 bit intel. But I cannot send it…
mac64.C (1.33 KB)

Hi,

This is indeed a but on 32 bit platform. To work-around the problem you can either disable the CINT optimization (.O 0) or use ACLiC (.L mac64.C+).

I added a savannah report for your issue at: savannah.cern.ch/bugs/index.php?64723

Cheers,
Philippe.

This problem is fixed in revision 32822 on the root trunk.

Thank you for your report!