Proc returning a variable causes weird error

[fingolfin]

The following is unexpected:

gap> LoadPackage( "SingularInterface", false );;
gap> Singular("int x=42;");
true
gap> SingularValueOfVar("x");
42
gap> Singular("proc HACK(){return(x);}");
true
gap> SI_CallProc("HACK", []);
Error, IsCopyableObjSingular: unsupported singtype 1
not in any function at line 5 of *stdin*

A singtype of 1 should never occur. Huh?

This might also be related to the crash in #47

[fingolfin]

Uhh, note this:

gap> Singular("HACK();");
...parse error
error occurred in or before STDIN line 1: `return(foo);`
leaving STDIN
false

Something is fishy?!

[fingolfin]

Even more concise:

gap> LoadPackage( "SingularInterface", false );;
gap> Singular("int foo=42;proc HACK(){return(foo);};HACK();");
...parse error
error occurred in or before STDIN line 1: `return(foo);`
leaving STDIN
false

The exact same command line works in the Singular interpreter:

                     SINGULAR                                 /  Development
 A Computer Algebra System for Polynomial Computations       /   version 4.0.1
                                                           0<
 by: W. Decker, G.-M. Greuel, G. Pfister, H. Schoenemann     \   Sep 2014
FB Mathematik der Universitaet, D-67653 Kaiserslautern        \
> int foo=42;proc HACK(){return(foo);};HACK();
42
>

So clearly we are doing something very wrong here.

This particular issue goes away if I remove the line myynest = 1; from Func_SI_EVALUATE in src/calls.cc. However, then other things fail. Anyway, I never understood that line anyway... we need to revise it, and this time, make sure whatever replace it, gets documented at the same time :).

[hannes14]

myynest is the nesting level of the Singular interpreter and is mainly important for the definition of variables. Variables which are defined at myynest==0 are global (i.e. visible at any level), other are only visible in the same level. If the level was entered like a procedure (via iiMakeproc/iipStart/iiAllStart) all variable of the current level will be removed when leaving that level.

[fingolfin]

Some more notes on consequences of removing the myynest = 1; (causing myynest to be 0):

• Now Singular("foo"); causes a hang (because the Singular interpreter takes over and uses readline to wait for input from the use). So I guess that explains why we had the myynest: So that we could get the Singular interpreter to return control to us.
• When running tst/testall.g, then tst/homalg_Purity.tst produces test failures, starting in the line which computes homalg_variable_9. Interestingly, if one instead runs tst/homalg_Purity.tst on its own (i.e. in a fresh GAP instance, directly after loading SingularInterface), it works just fine. So, something changes global state.

[fingolfin]

@hannes14 Thanks for your explanation, much appreciated! That fits with what I "discovered" via trial and error. Any idea why then, in the current code in the repo (where myynest ist set to 1), this code fails:

Singular("int foo=42;proc HACK(){return(foo);};HACK();");

Naively, it sets the variable foo on level 1, then defines a function on the same level, and calls it. Why doesn't that work?

[hannes14]

because foo is defined at level 1, and the statement "return(foo)" searches for the variable foo at level 2 (each procedure call increases myynest at its start and decreases it at the end)

[fingolfin]

Ah, and at level 1, the variable is not global, I see. Thus this works:

Singular("int foo=42;export foo;proc HACK(){return(foo);};HACK();");

Thanks, Hans!