a warning for methods which are known to return wrong results.

26 views
Skip to first unread message

Nathann Cohen

unread,
Feb 24, 2013, 12:04:09 PM2/24/13
to Sage devel, Nicolas M. Thiery, Florent Hivert
Helloooooooooo everybody !

I ... Well, I wondered if we could do something about methods that are known to return wrong results. Something which would be easy and quick to do, when where writing a bugfix can take some time.

This kind of things happen from time to time. For instance, I have to confess that I do not know how to fix #13744 as I have very few power over the upstream code and a hard time identifying the bug itself. And some methods that patch #14138 touched explicitely say that they sometimes return bad answers (as warnings in the code, or in the docstring).

It has been proved that such situations can last for very (too) long. And while I am thinking that it may be wiser to remove Graph.modular_decompositon from Sage users may use that method and get into trouble.

SOooooooooooooooo what could we do ? All I can think of is to add yet another warning, a "A-bug-has-been-reported-about-this-method-Warning" which we could add in the method when we find a bug, but there are two problems with this :

1) It means that a patch has to be created and reviewed, just to add a warning
2) It means that all doctests that use this method (or use it as a subroutine) would have to be changed, as (I believe) that this is how the deprecation warning works right now. Is there any workaround, by the way ?

3) It means that you have to upgrade your version of Sage in order to know that a method may be wrong. But of course "downloading a list of methods that are known to be wrong" would be a it complicated to setup too.

Well, what do you think ?

I hate wrong results. I would sleep better if I knew that nobody can trust a method which I know shouldn't be trusted.

Have fuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuunnn !!

Nathann

Volker Braun

unread,
Feb 24, 2013, 12:12:51 PM2/24/13
to sage-...@googlegroups.com, Nicolas M. Thiery, Florent Hivert

Nathann Cohen

unread,
Feb 24, 2013, 12:18:41 PM2/24/13
to sage-...@googlegroups.com, Nicolas M. Thiery, Florent Hivert
> See http://www.sagemath.org/doc/developer/trac.html#stopgaps

Cooooooooooooooooooooooooooooooooooooooooooll !!!

It still requires to create a ticket, but at least it is already available... I will create one for modular decomposition immediately ! :-)

Thaaaaaaanks ! And sorry for the noise :-)

Nathann
Reply all
Reply to author
Forward
0 new messages