Wednesday, June 12, 2013

Py3k status update #11

This is the 11th status update about our work on the py3k branch, which we
can work on thanks to all of the people who donated to the py3k proposal.

Here's some highlights of the progress made since the previous update:

  • PyPy py3k now matches CPython 3's hash code for
  • Various outstanding unicode identifier related issues were
    resolved. E.g. test_importlib/pep263/ucn/unicode all now fully pass. Various
    usage of identifiers (in particular type and module names) have been fixed to
    handle non-ascii names -- mostly around display of reprs and exception
  • The unicodedata database has been upgraded to 6.0.0.
  • Windows support has greatly improved, though it could still use some more
    help (but so does the default branch to a certain degree).
  • Probably the last of the parsing related bugs/features have been taken care
  • Of course various other smaller miscellaneous fixes

This leaves the branch w/ only about 5 outstanding failures of the stdlib test

  • test_float

    1 failing test about containment of floats in collections.

  • test_memoryview

    Various failures: requires some bytes/str changes among other things (Manuel
    Jacob's has some progress on this on the py3k-memoryview branch)

  • test_multiprocessing

    1 or more tests deadlock on some platforms

  • test_sys and test_threading

    2 failing tests for the New GIL's new API

Probably the biggest feature left to tackle is the New GIL.

We're now pretty close to pushing an initial release. We had planned for one
around PyCon, but having missed that we've put some more effort into the branch
to provide a more fully-fledged initial release.

Thanks to the following for their contributions: Manuel Jacob, Amaury Forgeot
d'Arc, Karl Ramm, Jason Chu and Christian Hudon.



Anonymous said...

In my new project I'm using Python3.
I can't when I will run it with PyPy.

Thanks for your work!

Unknown said...

I just donated and found this post :) Great work!

Paul Jaros said...

The "new GIL" picked my curiosity. Was is it? Is it related to the STM or is it a separate thing?

Also, thanks for the update.

Philip Jenvey said...

The new GIL is briefly explained here:

Additionally, David Beazly has done a couple talks/blog posts about the problems of the old GIL and how the new GIL has improved over the old design.

Paul Jaros said...

Thanks for the link

randomlessly said...

kkk @Tom Li

Unknown said...

Will the pre-release already be optimized?

Tony said...

This is cool!