Strange bug with antisymmetric function

[crmafra]

I found a bug which manifests itself by using the antisymmetric specification for a function.

That is, in the code below the correct result (zero) is obtained if

CFunction bmnp;

is used. However, if I use

CFunction bmnp(a);

it does not work and the result is not zero. The only difference is adding the (a) antisymmetric specification to the function.

The following code shows the bug.

   Indices m,n,p,q,r;
    Cfunction gmn, cdelta, g1100, tmpF;

•    CFunction bmnp;
  CFunction bmnp(a);
  
  L zero =
  + gmn(m,n)*bmnp(p,q,r)
  - tmpF(m,n,p,q,r)
  ;
  
  id gmn(m?,n?)*bmnp(p?,q?,r?) = g1100(p,q,r)*cdelta(m,n);
  id tmpF(m?,n?,p?,q?,r?) = g1100(p,q,r)*cdelta(m,n);
  
  print +s -f zero;
  .end 
  

When 'Cfunction bmnp' is commented out and 'Cfunction bmnp(a)' is used, instead of zero I get

    zero =
   - cdelta(m,n)*g1100(p,q,r)
   + cdelta(m,n)*g1100(r,p,q)
  ;

which shows that the arguments of bmnp were permuted even though they should not.

This bug happens in the latest code from github and the bug was also present in 2017.

[jodavies]

Here the antisymmetry (or symmetry) of bmnp affects how the wildcards are assigned, in the LHS of your id statement, and this also depends on the order in which p,q,r are defined.

I am not sure if this is a bug or not. The symmetry properties are accounted for in the pattern matching such that the following matches both terms, for eg:

CFunction fun, asfun(a);
Index a,b;

Local test =
   + fun(a,b)*asfun(a,b)
   + fun(b,a)*asfun(a,b)
   ;

If (Match( fun(a?,b?)*asfun(a?,b?) ));
   Print "match: %t";
EndIf;

Print +s;
.end