Puzzling behavior of factarg (freezes etc.)

[vsht]

Perhaps it is my lack of understanding of factarg, but I'm wondering
why the following simple code

CF f;
S p1,p2;

L exp = f(-p1-p2);

factarg,(0),f;
print;
.end

makes FORM freeze? Modifying the argument slightly

CF f;
S p1,p2;

L exp = f(p1+p2);

factarg,(0),f;

print;

.end

outputs

   exp =
      162000*f*pi_*p1;

which is even more weird. An if p1 and p2 are promoted to vectors, then with

CF f;
V p1,p2;

L exp = f(p1+2*p2);

factarg,(0),f;

print;

.end

I get

!!!This $ variation has not been implemented yet!!!
Called from execarg
Program terminating at /home/vs/Downloads/FORM-test-script.frm 
Line 8 --> 

All these examples work properly if I employ normalize,(0),f; instead of factarg,(0),f; which is also sufficient
for what I'm interested in. Still, since the documentation (p.86) says

`(0) Eliminates the coefficient of the term in the argument. Similar to Normalize,(0),....

I though that both statements should produce similar results.

Cheers
Vladyslav

[jodavies]

For me your example simply hangs, with nothing useful from valgrind.

Edit: the Vector example, however, does produce some useful valgrind output which is perhaps a clue.

==27642== Conditional jump or move depends on uninitialised value(s)
==27642==    at 0x45DF66: Normalize (normal.c:534)
==27642==    by 0x40B8F7: execarg (argument.c:1704)
==27642==    by 0x4BEDB8: Generator (proces.c:3445)
==27642==    by 0x4BFAA0: Generator (proces.c:3824)
==27642==    by 0x4BFF0A: Generator (proces.c:3992)
==27642==    by 0x4C148A: Processor (proces.c:405)
==27642==    by 0x43B2B4: DoExecute (execute.c:843)
==27642==    by 0x454A84: ExecModule (module.c:274)
==27642==    by 0x4B71F3: PreProcessor (pre.c:967)
==27642==    by 0x4F1C99: main (startup.c:1619)
==27642==
==27642== Conditional jump or move depends on uninitialised value(s)
==27642==    at 0x45DF6B: Normalize (normal.c:542)
==27642==    by 0x40B8F7: execarg (argument.c:1704)
==27642==    by 0x4BEDB8: Generator (proces.c:3445)
==27642==    by 0x4BFAA0: Generator (proces.c:3824)
==27642==    by 0x4BFF0A: Generator (proces.c:3992)
==27642==    by 0x4C148A: Processor (proces.c:405)
==27642==    by 0x43B2B4: DoExecute (execute.c:843)
==27642==    by 0x454A84: ExecModule (module.c:274)
==27642==    by 0x4B71F3: PreProcessor (pre.c:967)
==27642==    by 0x4F1C99: main (startup.c:1619)
==27642==
==27642== Conditional jump or move depends on uninitialised value(s)
==27642==    at 0x45DF70: Normalize (normal.c:578)
==27642==    by 0x40B8F7: execarg (argument.c:1704)
==27642==    by 0x4BEDB8: Generator (proces.c:3445)
==27642==    by 0x4BFAA0: Generator (proces.c:3824)
==27642==    by 0x4BFF0A: Generator (proces.c:3992)
==27642==    by 0x4C148A: Processor (proces.c:405)
==27642==    by 0x43B2B4: DoExecute (execute.c:843)
==27642==    by 0x454A84: ExecModule (module.c:274)
==27642==    by 0x4B71F3: PreProcessor (pre.c:967)
==27642==    by 0x4F1C99: main (startup.c:1619)
==27642==
==27642== Conditional jump or move depends on uninitialised value(s)
==27642==    at 0x45DF79: Normalize (normal.c:587)
==27642==    by 0x40B8F7: execarg (argument.c:1704)
==27642==    by 0x4BEDB8: Generator (proces.c:3445)
==27642==    by 0x4BFAA0: Generator (proces.c:3824)
==27642==    by 0x4BFF0A: Generator (proces.c:3992)
==27642==    by 0x4C148A: Processor (proces.c:405)
==27642==    by 0x43B2B4: DoExecute (execute.c:843)
==27642==    by 0x454A84: ExecModule (module.c:274)
==27642==    by 0x4B71F3: PreProcessor (pre.c:967)
==27642==    by 0x4F1C99: main (startup.c:1619)
==27642==
!!!This $ variation has not been implemented yet!!!
Called from execarg

[tueda]

Your second example (weird pi_) sounds like #140.

[tueda]

Actually, the oldest version form 4.0.0 (FORM 4.0 (Mar 29 2012)) that I can test gives the same results for your 3 examples (freeze, 162000*f*pi_*p1 and !!!This $ variation has not been implemented yet!!!).

[vermaseren]

It seems that the(0) option had never been tested on the case that no factor needed to be divided out. The latest commit should take care of it.

[vsht]

Great, many thanks!

[tueda]

I'm confused; to me, any of the above examples still doesn't work and behaves erroneously, though something seemed to be done in 9b0b89b... For example:

$ ./vorm test.frm
FORM 4.2.1 (Jul 15 2020, v4.2.1-9-g9b0b89b) 64-bits  Run: Fri Jul 17 18:20:17 2020
    CF f;
    S p1,p2;

    L exp = f(p1+p2);

    factarg,(0),f;

    print;

    .end

Time =       0.00 sec    Generated terms =          1
             exp         Terms in output =          1
                         Bytes used      =         56

   exp =
      162000*f*pi_*p1;

  0.00 sec out of 0.00 sec

[vsht]

Sorry, I didn't have time to check it yesterday and I assumed that if this specific issue
was mentioned then the examples will for sure work.

I agree with Takahiro, it seems that all of the original code samples still do not work
as they should.

[tueda]

Now it seems to work.