On Mar 11, 8:46 pm, Starx <jst...@gmail.com> wrote: > The patchbot is reporting a doctest failure in a patch of > mine:http://trac.sagemath.org/sage_trac/ticket/12630#comment:10 > > The failure itself is completely benign, "a + b" is being returned > when "b + a" is expected, but the addition is commutative anyway. So > I'm not so concerned with the particular error as I am confused about > what one does when patchbot is getting different results than I am > because I just built a brand new copy of 5.0.beta7 and doctested my > patch and got no errors (OSX 10.6.8 if that matters). > > Should I change the doctest so that it it will pass for patchbot even > though it fails for me? Or should I assume something is funky with > patchbot and set the ticket back to needs_review with a comment about > the issue?
If two closely related builds of sage already produce different output (as strings), I think it shows that the code that decides the order of the arguments in the sum is not deterministic. Fixing the doctest so that the particular configuration that the patchbot is using will be very fragile. There's a good chance that the release manager will already have trouble (and he'll have no specific knowledge of your code, so he has to go with "do all doctests pass?"). Clearly, the order of the returned sum is not important to you. So your doctest should not test it either. Would it be bad to change the test to sage: <expression yielding a+b> - (a+b) 0 or would that be too obscure? Or perhaps (so that the general output format is documented but that only the essential property of the result is tested): sage: result = <expression yielding a+b>; result # random a + b sage: result == (a + b) true -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
