Why do I get different outputs at different run's?

Hi, there's something very odd happening with my code. I'm doing three times the same calculations for three different input values. I.E.:

x1=a;
x2=b;
x3=c;
y=x^2;
plot

Well, this should always give the same answer for x1, x2 and x3. My routine, a bit larger, does not any fuzzy nor random calculus. When I execute the program I get an answer and if I execute it a second time I get another one. Since second execution, I always get the same answer. If I change a, b and c I will get the same answer as the last execution, and I need to execute it a second time.

I added clear x1; clear x2; and clear x3; trying to get rid of old memory data but nothing changes.

The part of the code where it occurs is:

global C1 C2 C3;
C1=7.996328e4;
C2=4.859647e4;
C3=4.965676e4;
clear C;
C=C1;
[ resu1 ] = nInt( fun , ab , tspan, titol, rotulacio,C );
clear C;
C=C2;
[ resu2 ] = nInt( fun , ab , tspan, titol, rotulacio,C );
clear C;
C=C3;
[ resu3 ] = nInt( fun , ab , tspan, titol, rotulacio,C );
rD1=resu1(2,:)./d;
rpD1=resu1(4,:)./d;
rD2=resu2(2,:)./d;
rpD2=resu2(4,:)./d;
rD3=resu3(2,:)./d;
rpD3=resu3(4,:)./d;
plot (rD1,rpD1,rD2,rpD2,rD3,rpD3)
title('rp/d vs r/d');
xlabel('r/d');
ylabel('rp/d');

nInt is a script proven to work correctly. I'm 99% sure the failure must be in this part of the code. MATLAB 2014.

[Brendan Hamm]

Does nInt use any sort of random number generation?

Furthermore, why are you using global variables?

[Muthu Annamalai]

I don't see a definition for variable d. Is it a global? Does its value somehow change?

[Felix Lauwaert]

d is global. I use global variables because I use several constant values into functions that go through 2 subroutines. nInt does not use any random number gerneration nor does it's subroutines.

[Brendan Hamm]

This is a poor reason for using global variables. Consider re-writing your code to just accept these variables as input and I would not be surprised if your problem vanishes.

[Brendan Hamm]

See: Global Variable. "They are usually considered bad practice precisely because of their non-locality: a global variable can potentially be modified from anywhere."

[Brendan Hamm]

The last thing I would point out is just because C1,C2,C3 are global, this does not make C global. But other than this, there is not enough information to know with certainty what is wrong.

[John D'Errico]

I like the statement that you are 99% sure the problem is in this code, and ONLY in that code, since you are POSITIVE that nInt is written correctly. Of course, we can't see what is in that code, so how can we possibly disprove your statement? For that matter, how can we know what you have done wrong? We are not given the ability to see fun either. So any guess on our part is just random.

That nInt has been proven to work correctly merely means that you have not seen it fail. It may just mean that your testing was insufficient.

As for the use of global variables, they are a bad thing. They make debugging a mess, as I think you are finding out.

[Stephen23]

Using globals comes in second place on Doug's list here:

http://blogs.mathworks.com/videos/2010/03/08/top-10-matlab-code-practices-that-make-me-cry/

Answer 1

[Felix Lauwaert]

Thanks to everybody for your comments. I see global variables are a mess. I finally solved the problem this way: fun goes from the main script to nInt and from nInt to ode78, and so do their variables. I just put fun (and company) into nInt and called nInt without fun & co. input. I'll take care next time and I hope I become better in the future. I'm just starting with Matlab and programming in general.