Segfault for boolean evaluation of expression with assumptions

[sagetrac-tmonteil]

As reported on this ask question:

sage: x, y = var('x, y')
sage: assume(x>0)
sage: assume(y>0)
sage: bool(y*(x-y)==0)

leads to (on my computer 8.9.rc1) a sequence of:

;;;
;;; Detected access to protected memory, also kwown as 'bus or segmentation fault'.
;;; Jumping to the outermost toplevel prompt
;;;

followed by a Segmentation fault crash of Sage.

Or (as reported, on 8.8):

RuntimeError: ECL says: C-STACK overflow at size 1048576. Stack can probably be resized. Proceed with caution.

Exchanging x and y works correctly:

sage: x, y = var('x, y')
sage: assume(x>0)
sage: assume(y>0)
sage: bool(x*(y-x)==0)
False

Upstream ticket: https://sourceforge.net/p/maxima/bugs/3583/

Depends on #30063

Upstream: Fixed upstream, in a later stable release.

CC: @rwst @kcrisman

Component: symbolics

Author: Thierry Monteil

Branch/Commit: cb70171

Reviewer: Matthias Koeppe

Issue created by migration from https://trac.sagemath.org/ticket/28538

[EmmanuelCharpentier]

comment:1

On 8.9.rc1+#28534 (Python 3-based), I get a lot of

;;;
;;; Detected access to protected memory, also kwown as 'bus or segmentation fault'.
;;; Jumping to the outermost toplevel prompt
;;;

and a Sage crash:

Process Sage erreur de segmentation

Nice one...

[kcrisman]

comment:4

I'm so sorry I just don't have time any more to track down as many of these (though once in a while I somehow make the time). But I think the best thing to do is to do whatever bool does in Maxima-in-sage sage -maxima and then load exactly the packages preloaded by Sage - the complex domain is usually the most suspicious one on these fronts, though I have to say this is really puzzling. I imagine bool calls a comparison with zero at some point in Maxima, though I don't recall any more because I wasn't involved with the comparison-with-zero code much.

[nbruin]

comment:5

In maxima:

domain: complex;
assume(x>0,y>0);
is(equal(y*(x-y),0));

replicates the crash. That's sufficient to report upstream. Perhaps they can fix it.

[sagetrac-tmonteil]

Upstream: Reported upstream. No feedback yet.

[sagetrac-tmonteil]

comment:6

Thanks for tracking, this is now tracked upstream as https://sourceforge.net/p/maxima/bugs/3583/

[sagetrac-tmonteil]

Changed upstream from Reported upstream. No feedback yet. to Fixed upstream, in a later stable release.

[embray]

comment:8

Ticket retargeted after milestone closed

[sagetrac-tmonteil]

Dependencies: #30063

[sagetrac-tmonteil]

comment:10

Once #30063 will be merged, i will make a patch to doctest that ticket.

[EmmanuelCharpentier]

comment:11

Replying to @sagetrac-tmonteil:

Once #30063 will be merged, i will make a patch to doctest that ticket.

The original problem seems fixed by #30063 :

sage: x, y = var('x, y') 
....: assume(x>0) 
....: assume(y>0) 
....: bool(y*(x-y)==0)                                                          
False

Testing it properly might be a bit tricky, tough...

[sagetrac-tmonteil]

comment:12

Replying to @EmmanuelCharpentier:

Replying to @sagetrac-tmonteil:

Once #30063 will be merged, i will make a patch to doctest that ticket.

The original problem seems fixed by #30063 :

This is why i am waiting #30063 to be merged to add a doctest.

sage: x, y = var('x, y') 
....: assume(x>0) 
....: assume(y>0) 
....: bool(y*(x-y)==0)                                                          
False

Testing it properly might be a bit tricky, tough...

What is wrong with using the raw example ?

[EmmanuelCharpentier]

comment:13

Replying to @sagetrac-tmonteil:

[ Snip... ]

Testing it properly might be a bit tricky, tough...

What is wrong with using the raw example ?

Too special case... I am not sure what the original problem was.

[sagetrac-tmonteil]

Branch: u/tmonteil/segfault_for_boolean_evaluation_of_expression_with_assumptions

[sagetrac-tmonteil]

Commit: fe71c17

[sagetrac-tmonteil]

New commits:

fe71c17

#28538 : add doctest for #28538

[sagetrac-tmonteil]

Author: Thierry Monteil

[sagetrac-tmonteil]

comment:16

Replying to @EmmanuelCharpentier:

Replying to @sagetrac-tmonteil:

[ Snip... ]

Testing it properly might be a bit tricky, tough...

What is wrong with using the raw example ?

Too special case... I am not sure what the original problem was.

This is a Maxima bug, it was reported and fixed upstream. I do not have the skill to inspect further within Maxima source code, so this doctest is the best i can provide, and it corresponds to the reported bug. If someone could provide more doctests to surround the original problem more securely, i am all for it.

[kcrisman]

comment:17

I think that testing the original problem no longer leads to a segfault is fine. I agree that this is all we can do if Maxima upstream fixed it and we don't really know what the issue was. (Though it looks like it was, again, our use of domain:complex that triggered it.)

[seblabbe]

comment:18

Typo: Chack -> Check

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

cb70171

#28538 : typo

[sagetrac-git]

Changed commit from fe71c17 to cb70171

[sagetrac-tmonteil]

comment:20

Fixed, thanks for pointing this !

[mkoeppe]

Reviewer: Matthias Koeppe

[vbraun]

Changed branch from u/tmonteil/segfault_for_boolean_evaluation_of_expression_with_assumptions to cb70171