Log file Viewer

[07:34:16] <pombreda> !logs

[07:34:16] <pmxbot> http://chat-logs.dcpython.org/channel/pypa

[14:11:15] <dstufft> nedbat: what does coverage.py's C extension do? Is it optional?

[14:11:18] <dstufft> is it just speed ups?

[14:20:30] <nedbat> dstufft: "just" speedups, but many features don't work without it

[14:20:45] <dstufft> nedbat: gotcha

[15:48:18] <ionelmc> nedbat: what features don't work?

[15:48:39] <nedbat> ionelmc: plugins

[15:48:54] <ionelmc> nedbat: interesting, why?

[15:49:54] <nedbat> ionelmc: because i didn't want to implement it twice, and all the conditionals would slow down non-plugin use even more.

[15:50:33] <nedbat> ionelmc: and --concurrency

[15:54:08] <ionelmc> nedbat: what's special about --c?

[15:54:29] <nedbat> ionelmc: nothing, just more code i didn't want to duplicate, and slow down the python tracer

[15:55:28] <ionelmc> well

[15:55:42] <ionelmc> might as well drop the python tracer altogether :)

[15:57:08] <nedbat> ionelmc: perhaps

[16:10:26] <ronny> nedbat: what about pypy tho?

[16:10:46] <ronny> (would writing the tracer in python and supplementing a cython type spec do any good?

[16:11:07] <nedbat> ronny: yeah pypy is stuck with the python tracer

[16:11:31] <nedbat> stefan behnel gave me a cython implementation of the tracer. I didn't have the cycles to understand what to do with it.

[16:27:21] <dstufft> nedbat: does that mean PyPy doesn't get plugins?

[16:27:27] <dstufft> or concurrency?

[16:29:31] <nedbat> right. i haven't heard complaints yet.

[16:29:35] <nedbat> not sure why

[16:29:58] <nedbat> well, --concurrency is for non-thread concurrency: gevent, greenlet, multiprocessing

[20:08:24] <nicksloan> Was just periodically reminded to thank all of you packaging contributors for not being NPM.

[21:00:44] <dstufft> nicksloan: the fs thing?

[21:01:08] <nicksloan> Yeah. I know they aren't entirely to blame.

[21:01:52] <dstufft> nicksloan: heh yea, it's likely something similar could happen with PyPI, except we're too bogged down to answer support requests in a reasonable time frame (better UX through slacking?)

[21:02:02] <nicksloan> haha

[21:02:36] <dstufft> That's what NPM gets for trying too hard

[21:06:41] <ngoldbaum> http://status.npmjs.org/incidents/dw8cr1lwxkcr

[21:06:42] <ngoldbaum> heh

[21:07:05] <ngoldbaum> More detail: the "fs" package is a non-functional package. It simply logs the word "I am fs" and exits. There is no reason it should be included in any modules. However, something like 1000 packages *do* mistakenly depend on "fs", probably because they were trying to use a built-in node module called "fs".

[21:07:10] <ngoldbaum> lol

[21:07:10] <ngoldbaum> javascript everyone

[21:33:14] <nicksloan> Yeah. That's the amazing thing. Not too put too much on NPM's shoulders, but they certainly could have seen that trend and taken action to remedy the situation in advance.

[21:33:34] <nicksloan> s/Not too/Not to/

[22:11:24] <nanonyme> Infinite deprecation policy is pretty dumb though

[22:11:37] <nanonyme> What's the point of deprecating something if you're not going to remove it?

[23:19:56] <[Tritium]> putting fs on npm is like putting os on pypi

[23:21:31] <[Tritium]> nanonyme: npm has the concept of flagging a package as deprecated - presumably to tell people to move to a different package. Just like doesnt generally remove releases, npm doesnt either (usually... the exceptions make tech news.)

[23:22:24] <tdsmith> i wonder if i can register that

[23:23:03] <[Tritium]> try?

[23:23:52] <[Tritium]> simplehttpserver or some such

[23:24:00] <[Tritium]> oh!

[23:24:09] <[Tritium]> uh... netrc

[23:25:10] <tdsmith> hmm https://pypi.python.org/pypi/telnetlib

[23:26:29] <tdsmith> so "yes" probably :p

[23:27:53] <[Tritium]> project for when I get home from work: go through and find all the standard library modules that that done exist on pypi... register them and create an sdist that intentionally fails with print("You dont need to pip install the standard library")

[23:28:44] <tdsmith> dstufft: do you have opinions wrt whether pypi should allow new registrations that conflict with stdlib modules :p

[23:28:45] <ngoldbaum> heh

[23:28:46] <ngoldbaum> https://pypi.python.org/pypi/RE

[23:29:13] <[Tritium]> there is a backports usecase

[23:29:19] <[Tritium]> anyways, work

[23:29:47] <dstufft> Backports, also pulling something into the stdlib that was previously a third party thing, also new things being added to the stdlib that shadow stuff on PyPI

[23:29:55] <dstufft> not sure how If eel about it in general though