Log file Viewer

[03:20:45] <jimbaker> just wanted to mention that i have a branch of pip working against a branch of jython 2.7

[03:20:54] <jimbaker> (which is pretty cool)

[03:22:22] <jimbaker> the branch for pip is small, although i have only tested this to a minimal point. it actually requires a fix of html5lib so that isolated unicode surrogates are not directly in unicode literals, which jython due to its utf-16 encoding, doesn't support

[03:22:57] <jimbaker> the branch for jython is substantially larger, since it completely reworks socket/select/ssl support. but we will handle that part

[03:23:58] <jimbaker> so here it is - https://github.com/jimbaker/pip

[03:25:24] <jimbaker> Arfrever, ^^^

[03:27:42] <jimbaker> looking forward to figuring out how to upstream what should be a very minimal change!

[05:18:29] <dstufft> jimbaker: the only thing needed is a fix to html5lib?

[05:41:08] <dstufft> jimbaker: so I'm a pip dev :) We don't modify the vendored libraries at the moment. So the ideal path forward would be to get html5lib to accept your changes. Then it's rather trivial for us to upgrade to a new version of html5lib, and then figure out how to turn Jython on in our test matrix.

[11:16:42] <jimbaker> dstufft, right, that's the path i assume would be necessary - and ideal in terms of improving other packages with a html5lib dependency

[12:48:58] <dstufft> jimbaker: :D

[13:00:41] <dstufft> jimbaker: When the jython branch is merged, I assume it'll be part of http://hg.python.org/jython default branch?

[15:59:27] <jimbaker> dstufft, yes, that's the plan!

[15:59:31] <jimbaker> (sorry, i'm traveling today - at denver airport this moment - so intermittent)

[15:59:48] <jimbaker> it's very important for jython that it fully works with the overall ecosystem. clearly support for pypa work is a key enabler of that!

[16:00:37] <dstufft> jimbaker: no worries :) Just thinking, should be easy to run our test infra on then. I added pyenv support to travis-ci which is just waiting on travis-ci to deploy it

[16:00:49] <dstufft> so unless it takes forever we should be able to test against jython default

[16:00:55] <dstufft> takes forever to build*

[16:01:17] <jimbaker> dstufft, maybe 2 minutes to build from source

[16:01:26] <dstufft> yea that's reasonable

[16:01:32] <jimbaker> i haven't tried that w/ cpython recently, but i'm sure that's comparable

[16:01:57] <dstufft> we don't test against CPython default right now (though I want too) but yea it's roughly comparable

[16:02:07] <dstufft> I was just making sure it wasn't like PyPy

[16:02:13] <dstufft> which takes like a billion hours

[16:02:14] <dstufft> roughly

[16:02:30] <jimbaker> right, that's closer to building a jvm from scratch :)

[16:02:38] <jimbaker> first, assume java

[16:04:05] <dstufft> :)

[16:04:10] <dstufft> This'd be cool though

[16:04:23] <dstufft> Apparently pip has been broken on Jython since 1.3 :(

[16:04:29] <dstufft> when it switched to SSL

[16:05:02] <jimbaker> correct, so jython-socket-reboot basically adds ssl support, which completely rebuilding our socket/select implementation in the process

[16:06:02] <jimbaker> socket-reboot basically bases the underlying low-level networking on netty 4

[16:07:32] <jimbaker> eventually this should make our networking performance potentially much better, especially for tooling like trollius or twisted where there can be fewer layers. TBD. what matters is it works :)

[16:10:25] <dstufft> awesome

[16:29:33] <jimbaker> so i have been running so more informal tests, to see what works and what doesn't. (remember, i just got this going end-to-end yesterday evening. so i decided that was good enough to stop :)

[16:29:47] <jimbaker> 1. wheel format has some problem, probably minor

[16:31:24] <jimbaker> move_wheel_files looks like it might be the culprit

[16:32:12] <jimbaker> 2. tox wants to use virtualenv. as i suspected, virtualenv has some current problems on jython. most likely minor

[16:33:17] <jimbaker> so those are problems in running pip unit tests. i will look at vendor libs next

[16:36:06] <dstufft> jimbaker: virtualenv is kind of a hack, so it doesn't surprise me there are problems

[16:36:27] <jimbaker> dstufft, right, i know it worked in the past, but probably some recent delta

[16:36:43] <dstufft> also virtualenv has it's own pip and setuptools bundled inside of it (which use wheels) so it might be that

[16:37:47] <jimbaker> dstufft, got it. i will take a look in more detail

[16:37:52] <jimbaker> but first - got a flight to board!

[16:37:56] <dstufft> have fun!

[16:38:12] <jimbaker> dstufft, thanks! looking forward to making some quick progress on pip/virtualenv!