Closed kynan closed 10 years ago
Fixed in c2415d3
Turns out this isn't quite fixed yet. For reasons I can't quite tell the build on 2.6 still fails.
I sort of don't care about this. 2.7 is not /that/ new. Is there a compelling reason to tie ourselves in knots to support 2.6? I'd be more worried about forward compatibility to 3.4.
There's certainly no reason to care about 2.6 other than deploying to clusters. I agree that we should rather start worrying about forward compatibility.
I believe the future for deployment on clusters and supercomputers is via python platforms such as Anaconda. This is resolving many ongoing python/numpy/scipy issues for us on HECToR/ARCHER. In addition, Andy Terrel and Continuum are working with Cray to sort out scalability etc. so this approach is gaining momentum.
ie - we really don’t care about 2.6 etc.
Yes, if we could have a recent Anaconda distribution on HPC clusters that would be fantastic. With that 2.6 would not be a concern, but we should keep forward compatibility in mind.
Closing this, since we've decided we don't care enough about 2.6
The
core_types
extension module is not compatible with Python 2.6 due to the use of capsules, which were introduced in 2.7.A possible workaround would be introducing a compatibility header that redefines PyCapsule C API calls in terms of PyCObject.