Based on a post by Alexander Hasselhuhn in the forum (http://www.nikhef.nl/~form/forum/viewtopic.php?f=3&t=194). The following example gives a corrupt result for an expression GG, containing coeff_ and pi_, while a mathematically equivalent expression FF gives the correct result:
CF rat;
PolyRatFun rat;
S cw,sw,e;
*S MZ,sp12; * <- This fixes the problem.
S sp12,MZ;
L F = cw * sw * e * rat(- MZ, 2 * sp12 - 1 * MZ);
L G = 2 * cw * sw * e * rat(- MZ, 4 * sp12 - 2 * MZ);
.sort
L FF = F^2;
L GG = G^2;
P +s;
.end
Output:
FORM 4.1 (Sep 21 2016, v4.1-20131025-245-g104d5d8) 64-bits Run: Sat Oct 1 14:21:33 2016
F =
+ cw*sw*e*rat( - MZ,2*sp12 - MZ)
;
G =
+ cw*sw*e*rat(MZ, - 2*sp12 + MZ)
;
FF =
+ cw^2*sw^2*e^2*rat(MZ^2,4*sp12^2 - 4*sp12*MZ + MZ^2)
;
GG =
+ cw^2*coeff_*pi_*rat(MZ^2,4*sp12^2 - 4*sp12*MZ + MZ^2)
;
Note that F and G are equivalent but have different representations. Sounds like it's related to the normalization of rat functions, which somehow gives an incorrect memory overwriting. So inserting the following code in an adequate place would be a workaround:
.sort
PolyRatFun rat; * redeclared; renormalize rat forcely
.sort
It seems that this bug was introduced by ffabb37261.
tueda
commented
7 years ago
Based on a post by Alexander Hasselhuhn in the forum (http://www.nikhef.nl/~form/forum/viewtopic.php?f=3&t=194). The following example gives a corrupt result for an expression
GG, containingcoeff_andpi_, while a mathematically equivalent expressionFFgives the correct result:Output:
Note that
FandGare equivalent but have different representations. Sounds like it's related to the normalization ofratfunctions, which somehow gives an incorrect memory overwriting. So inserting the following code in an adequate place would be a workaround:It seems that this bug was introduced by ffabb37261.