CoCoA » CoCoA-5

I think there are 7 places to revise...
Just search for LinKer and see which ones actually are printed.

John Abbott wrote:

tut-CoCoLa2.cocoa5
prints out the result of LinKer which may vary from call to call.

Change the test so that it does not print out values which may vary from run to run!

[2022-11-11: Recd by email from Ulrich]

I'm not sure this is the same issue I alluded to in my email.

The problem I meant in my email is triggered by line 181 in tut-CoCoLA2.cocoa5:

eigenfactors(A3, z); --> two eigenfactors

This boils down to

f := x^4 - 20*x^2 + 36;
factor(f).factors;

The expected output (as given in tut-CoCoLA2.out) is:
[x^2 -2, x^2 -18]

I get the reversed order
[x^2 -18, x^2 -2]
randomly (about every second time two).

Note: I have observed this bug on OpenBSD only, never on a GNU/Linux system.

PS: Sorry for the late clarification -- I wasn't aware that this was posted here and only saw it by chance.

Ulrich von der Ohe wrote:

Note: I have observed this bug on OpenBSD only, never on a GNU/Linux system.

The problem also appears on other systems.

For example, this is literal input/output from a CoCoA-5 session (running on GNU/Linux, but now I don't think that it matters):

# use R ::= QQ[x];
# f := (x^2 - 2) * (x^2 - 1000);
# f;
x^4 -1002*x^2 +2000
# factor(f);
record[RemainingFactor := 1, factors := [x^2 -2,  x^2 -1000], multiplicities := [1,  1]]
# factor(f);
record[RemainingFactor := 1, factors := [x^2 -1000,  x^2 -2], multiplicities := [1,  1]]

Reminder for me: check both LinKer and eigenfactors

John Abbott wrote:

What do you think?

I think it would be preferable to have deterministic output (i.e. literally the same strings). Generally this is much easier to work with. But if that is not doable then changing the test as you suggest is at least a way to avoid the test failing due to this unexpected presentation of the result. (By the way, the test still fails with CoCoALib-0.99850 with apparently a 50% chance.)

Ah, oops!
Since Anna did not object, I have just made the change hinted at in my comment 8 above. This should avoid the non-deterministic printed result -- Ulrich is right that we should avoid printed values whose printed string form may vary.