pmxbot IRC Log Viewer

[00:11:07] <glyph> ionelmc: "readme_renderer" you mean?

[00:11:45] <ionelmc> glyph: no, i'm not using that, i'm using "readme"

[00:12:23] <ionelmc> "readme" now fails to install and asks me to install something else, "readme_renderer"

[00:12:27] <ionelmc> makes no sense

[00:12:42] <glyph> ionelmc: it's the same package

[00:13:07] <glyph> "an installation issue that could prevent it from being installed on some systems" is not super clear, it's true

[00:13:32] <carljm> ionelmc: The name-change makes sense and was necessary, but yeah, releasing a new version of "readme" that just fails to install is a pretty harsh way to deliver the deprecation.

[00:13:54] <ionelmc> carljm: necessary?

[00:14:00] <carljm> (PSA: always pin all your dependencies.)

[00:14:07] <glyph> https://github.com/pypa/readme_renderer/issues/26

[00:14:25] <glyph> carljm: only in requirements.txt! never in setup.py

[00:14:40] <glyph> but yeah 'readme' should have gone with a warning not an error

[00:15:30] <carljm> glyph: truth.

[00:15:44] <glyph> carljm: also hey how have you been, I don't hang out in your corner of the internet too often :)

[00:15:51] <ionelmc> carljm: it would generate me an endless and unnecessary stream of churn to pin down development tools

[00:16:14] <ionelmc> test deps, yes, pin, dev tools no

[00:16:15] <glyph> ionelmc: but you can get that endless stream of churn as a web service! https://requires.io

[00:16:18] <carljm> glyph: i'm well! hope you are too. packaging was the corner of the internet we shared, and then i mostly left it :-)

[00:16:31] <glyph> ionelmc: after tox broke all my builds 3 times in a row, I pin all dev dependencies as well.

[00:16:44] <glyph> carljm: I do as little of it as I can ;)

[00:16:49] <carljm> (I would actually love to spend more time in the Twisted corner of the internet, but people keep wanting to pay me to do Django work, so...)

[00:16:51] <ionelmc> glyph: yeah, sadly requires.io lack in the automation department

[00:17:02] <ionelmc> so there is still monkey work to do

[00:17:16] <glyph> carljm: https://hendrix.readthedocs.org/en/latest/ :)

[00:18:15] <ionelmc> anyway

[00:18:27] <ionelmc> my point is that the release can be done properly

[00:18:49] <ionelmc> with an alias package that don't put any files in site-packages but depends on readme_renderer

[00:20:07] <carljm> yup, that'd be a better plan. It can even still output the setup.py warning

[00:20:41] <carljm> glyph: yeah, it's on the list of things to play with!

[00:27:51] <dstufft> setup.py output is hidden unless the build fails

[00:28:01] <dstufft> and we had to break the API to change the import name.

[00:46:09] <ionelmc> dstufft: what's wrong it an alias

[00:46:25] <ionelmc> s/it/with/

[00:47:15] <dstufft> didn't think of it

[00:49:53] <ionelmc> i've opened an issue with the setup.py trick to produce an empty wheel

[00:52:49] <agronholm> dstufft: are there any concrete plans on implementing those nice looking warehouse features like changelog or test status/test coverage?

[00:53:30] <dstufft> agronholm: right now the plans are roughly "think up something clever to make it work"

[00:53:34] <dstufft> so not really :(

[00:53:47] <dstufft> they'll probably get disabled before we go live, and then we'll work on re-enabling them over time

[00:53:50] <agronholm> aww :(

[00:54:39] <agronholm> dstufft: maybe build a library for interfacing with warehouse and then submit the statistics in a travis on_success handler like I do with coveralls now

[00:55:02] <dstufft> yea, there's a good chance it'll be something like that

[00:55:20] <ionelmc> dstufft: don't we need new metadat for that tests status stuff?

[00:55:28] <dstufft> unless we get something like cpan testers

[00:55:39] <dstufft> ionelmc: depends on how it's implemented!

[00:55:44] <agronholm> what I would really like to have is 1) bug tracker url, 2) documentation url

[00:55:48] <dstufft> possibly.

[00:56:03] <dstufft> agronholm: those exist today in a way, at least as far as PyPI is concerned

[00:56:08] <dstufft> just setuptools doesn't expose them

[00:56:23] <agronholm> dstufft: and documentation url isn't user configurable

[00:56:39] <dstufft> specifically, PyPI has the ability to have arbitrary key -> URL mappings

[00:57:04] <dstufft> probably we'll end up with some well known names like "Documentation"

[09:51:26] <jezdez> glad everyone was having a talk about the release here, while I’m here

[09:51:56] <jezdez> you know the way ionelmc gave feedback first on twitter about the release

[09:52:04] <jezdez> then apparently here, without talking to me

[09:52:06] <jezdez> then an issue

[09:53:07] <jezdez> I’m done with that passive-aggressiveness and won’t do any pypa work because of it

[09:53:14] <jezdez> I hope you’re happy now

[10:08:22] <xafer> :-/

[11:19:21] <ionelmc> I'm bad

[18:04:17] <buck1> i dont understand the complaint

[18:16:08] <umeshksingla_> Hello guys, do we have only these commands for pip:

[18:16:25] <umeshksingla_> InstallCommand, DownloadCommand, UninstallCommand, FreezeCommand, ListCommand, ShowCommand, SearchCommand, WheelCommand, HashCommand, HelpCommand ?

[19:52:38] <matclayton> Hey guys, got a strange pip install problem (7.1.2), using the following requirements.txt it fails

[19:52:59] <matclayton> http://dpaste.com/0S7F3M5

[19:53:17] <matclayton> fairly sure this is because scipy needs numpy to be installed before the wheel can be built, any suggestions?

[19:56:37] <dstufft> matclayton: install numpy first I think is your only option.

[19:56:55] <matclayton> hmmmm its not really an option in our CI system

[19:57:05] <matclayton> it could be done, but its not exactly easy

Log file Viewer

Help | Karma | Search:

#pypa logs for Tuesday the 5th of January, 2016