Project

General

Profile

Bug #1062

IsRadical bug?

Added by John Abbott almost 7 years ago. Updated over 6 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Anna Maria Bigatti
Category:
Incomplete function
Target version:
CoCoA-5.2.2
Start date:
09 May 2017
Due date:
% Done:

100%

Estimated time:
Spent time:
4.45 h

Description

I have an example where IsRadical produces an error (when applying a RingHom)
0-dimensional bug: related with MinPolyModular, moved to #1064

And this triggers SEGV (non-0-dim bug)

use R ::= QQ[X0,X1,X2,X3,X4,X5,X6];
I := ideal(X3^3 +X0*X3*X4 +X4^3,
  X0*X1*X3 +X3*X5^2,
  -X0^6*X1*X3 +X3*X6^7,
  X0 -X1 -X2 -X3 -X5 -X6,
  -X2 +X5 +X6,
  -X0 -X5 -X6,
  X1 -X2 +X3 -X6,
  X1 -X2 +X3 -X5);
radical(I);

Related issues

Related to CoCoALib - Design #1063: Catching an (expected) errorClosed2017-05-09

Related to CoCoALib - Bug #1064: Bug in MinPolyModular (ugly prime)Closed2017-05-10

History

#1 Updated by John Abbott almost 7 years ago

This example has simpler coeffs:

ideal(X3^3 +X0*X3*X4 +X4^3,  8*X0*X1*X3 +4*X0*X2*X3 +6*X0*X4^2 -6*X1*X4^2 +3*X0*X3*X5 +X3*X5^2,  5*X0^6*X1*X3 -8*X0^6*X2*X3 +6*X0^6*X4^2 -36*X0^5*X2*X4^2 +90*X0^4*X2^2*X4^2 -120*X0^3*X2^3*X4^2 +90*X0^2*X2^4*X4^2 -36*X0*X2^5*X4^2 +6*X2^6*X4^2 +4*X0^6*X3*X6 +X3*X6^7,  4*X0 -2*X1 +X2 -2*X3 +3*X4 +8*X5 -8*X6,  -X0 +7*X1 -8*X2 +8*X3 -7*X5 -X6,  -X0 -5*X1 -7*X2 -X3 -6*X4 -9*X5 -9*X6,  4*X0 -3*X1 +7*X2 -9*X3 -X4 -3*X5 -2*X6,  4*X0 -6*X1 -X2 -2*X3 +5*X4 -4*X5 +9*X6)

#2 Updated by John Abbott almost 7 years ago

Aaargh!! Now I have a SEGV :-(

use R ::= QQ[X0,X1,X2,X3,X4,X5,X6];

define RndCoeff()
  return random(-1,1);
enddefine; -- RndCoeff

StartTime := CpuTime();
for i := 1 to 10000 do

J := ideal(
  X0 * X3 * X4 + X3^3 + X4^3,
  X3 * X5^2 + RndCoeff() * X0 * X3 * X5 + RndCoeff() * X0 * X1 * X3 + RndCoeff() * X0 * X2 * X3 + RndCoeff() * (X1 - X0) * X4^2,
  X3 * X6^7 + RndCoeff() * X0^6 * X3 * X6 + RndCoeff() * X0^6 * X1 * X3 + RndCoeff() * X0^6 * X2 * X3 +RndCoeff() * (X2 - X0)^6 * X4^2,
           sum([RndCoeff()*x | x in indets(R)]),
           sum([RndCoeff()*x | x in indets(R)]),
           sum([RndCoeff()*x | x in indets(R)]),
           sum([RndCoeff()*x | x in indets(R)]),
           sum([RndCoeff()*x | x in indets(R)])
);

  if IsRadical(J) then println "RADICAL: ", J; endif;
endfor;
EndTime := CpuTime();
println "Loop time: ", FloatStr(EndTime-StartTime);

#3 Updated by John Abbott almost 7 years ago

This example triggers SEGV:

ideal(X3^3 +X0*X3*X4 +X4^3,  X0*X1*X3 +X3*X5^2,  -X0^6*X1*X3 +X3*X6^7,  X0 -X1 -X2 -X3 -X5 -X6,  -X2 +X5 +X6,  -X0 -X5 -X6,  X1 -X2 +X3 -X6,  X1 -X2 +X3 -X5);

#4 Updated by Anna Maria Bigatti almost 7 years ago

I'm not that surprised. Radical for non zero-dimensional ideals is not that robust.
I'm investigating

#5 Updated by Anna Maria Bigatti almost 7 years ago

  • % Done changed from 0 to 10

Smaller example

use R ::= QQ[X1,X3,X4];
J := ideal(X1 +X3,  X3^3 +X4^3);
radical(J);

#6 Updated by Anna Maria Bigatti almost 7 years ago

  • Assignee set to Anna Maria Bigatti
  • % Done changed from 10 to 30

The two bugs (0-dim and non-0-dim) are quite distinct.

Debugging the zero-dimensional example in the description of this issue was relatively easy: this is the first time we actually stumble on an ugly prime!!

So I just had to put the apply of the partial homomorphism inside a try/catch.
Such an easy thing to say and such a tricky thing to do.

I would like to have a standard way to catch some expected error (I'll put this in a related issue).

#7 Updated by Anna Maria Bigatti almost 7 years ago

  • Related to Design #1063: Catching an (expected) error added

#8 Updated by John Abbott almost 7 years ago

  • Status changed from New to In Progress

When this is fixed remember to add some new tests to exbugs.cocoa5!

#9 Updated by Anna Maria Bigatti almost 7 years ago

  • Related to Bug #1064: Bug in MinPolyModular (ugly prime) added

#10 Updated by Anna Maria Bigatti almost 7 years ago

  • Description updated (diff)

#11 Updated by Anna Maria Bigatti almost 7 years ago

  • % Done changed from 30 to 40

Bug (radical >0 dim) trapped by Elisa Palezzato

/**/ saturate(ideal(x -y), ideal(one(CurrentRing))); 

Process cocoa5 segmentation fault: 11

#12 Updated by Anna Maria Bigatti almost 7 years ago

Found it in TmpGOperations.C:
Saturation by ideal(f) calls factor(f) and wrongly assumes that the factor list contains at list one element.
Fixed it by intercepting IsOne(f)

UPDATE fixed better by using IsInvertible o/w saturate(ideal(x-y, ideal(2*one(P))) gives SEGV

#13 Updated by Anna Maria Bigatti almost 7 years ago

  • Description updated (diff)

#14 Updated by Anna Maria Bigatti almost 7 years ago

  • Status changed from In Progress to Feedback
  • % Done changed from 40 to 90

added tests. CVS-ed

#15 Updated by John Abbott almost 7 years ago

Does your test of saturate in anna.cocoa5 compute a saturation with twin-float coeffs??
I think you might want to insert a use command a line 143. Right?

#16 Updated by Anna Maria Bigatti almost 7 years ago

John Abbott wrote:

Does your test of saturate in anna.cocoa5 compute a saturation with twin-float coeffs??

It was not intentional, but why not?

#17 Updated by John Abbott over 6 years ago

  • Status changed from Feedback to Closed
  • % Done changed from 90 to 100

Also available in: Atom PDF