[Numpy-discussion] Warning on http://scipy.org/ about binary incompatibility ?

David Cournapeau david@silveregg.co...
Thu Jan 28 18:58:31 CST 2010


Charles R Harris wrote:
> 
> 
> On Thu, Jan 28, 2010 at 12:33 AM, David Cournapeau 
> <david@silveregg.co.jp <mailto:david@silveregg.co.jp>> wrote:

> 
>     Because Travis was against it when it was suggested last september or
>     so. And removing in 1.4.x a feature introduced in 1.4.0 is weird.
> 
> 
> But wasn't that decision based on the premiss that the datetime work 
> wouldn't break the ABI?

Well, this and because Travis is the BFDL of NumPy as far as I am 
concerned :) So I think it should be his decision whether to remove it 
or not.

> I don't see anything weird about making 1.4 work 
> with existing binaries.

Indeed, but that's not really what I am saying :) I am saying there is a 
tradeoff between breaking people's code (for people using datetime) and 
keeping a compatible ABI.

So the decision depends quite a bit on how many people use the datetime 
code.

> If we are going to break the ABI, and it looks 
> like we will, then it would be better if the word went out early so that 
> projects that depend on numpy can be prepared for the change. So my 
> preference would be to remove the incompatibility in 1.4 and introduce 
> it in 1.5.

Assuming not many people depend on datetime in 1.4.0, that would be my 
preference as well.

cheers,

David


More information about the NumPy-Discussion mailing list