On 20 December 2012 08:41, P Purkayastha <ppu...@gmail.com> wrote:
> On 12/20/2012 04:19 PM, Jeroen Demeyer wrote:
>>
>> On 2012-12-20 01:50, P Purkayastha wrote:
>>>
>>> #12798 is another ticket which gave trouble on solaris and was working
>>> just fine otherwise. I just gave up on that ticket. Perhaps you can try
>>> and see if it works with OpenIndiana. If it works on OpenIndiana, this
>>> would make it the second ticket which gives trouble in solaris.
>>
>>
>> I disagree that the problem with #12798 is Solaris. Here the true bug is
>> that you're creating a 3D surface with NaNs in some coordinates of
>> vertices. That this works on other systems is more chance than anything
>> else.
>>
>
> I have explained in the last post in the ticket why NaNs or Infs are
> required. I don't know of any other way to represent "invalid" points on a
> *square* grid.

I think you have just provided an excellent example of why it is good
to keep Solaris.

Your patch is a hack which adds NaN to represent what you consider
invalid points because you can't think of another way. If I understand
the problem correctly from what Jeroen wrote on the ticket, you are
plotting arbitrary 3D figures given by faces and vertices's. There is
no reason for them to lie on a square grid, but the underlying code
makes the assumption they do.

I suspect the behaviour of the code you are calling with this hack is
not defined, which is why it behaves differently on different
operating systems. It might well behave differently if another version
of matplotlib was used. Adding NaN is a poor hack.

I'd also point out that contrary to what you think on that ticket,
Solaris is not an architecture. Solaris is an operating system which
runs on two architectures - SPARC and x86.

Dave

-- 
You received this message because you are subscribed to the Google Groups 
"sage-devel" group.
To post to this group, send email to sage-devel@googlegroups.com.
To unsubscribe from this group, send email to 
sage-devel+unsubscr...@googlegroups.com.
Visit this group at http://groups.google.com/group/sage-devel?hl=en.


Reply via email to