Log file Viewer

[00:27:55] <Whisket> Is it possible to include static files when installing with setuptools from source? I tried this http://stackoverflow.com/questions/11848030/how-include-static-files-to-setuptools-python-package but it doesn't add the files to the egg or site-packages although the static files are listed in SOURCES.txt

[00:33:49] <Whisket> nvm, issue solved. Needed to add include_package_data=True to setup()

[06:12:08] <mgedmin> wanted: a pypi proxy/cache solution as simple to use as sudo apt-get install squid-deb-proxy-client

[07:20:49] <ronny> mgedmin: sounds a bit like you want someone to package devpi + a configfile

[07:21:02] <mgedmin> maybe?

[07:21:09] <mgedmin> I want someone (not me) to figure it all out

[07:21:47] <ronny> i installed devpi in my home and made a systemd unit for it

[07:27:20] <mgedmin> hm

[07:27:49] <mgedmin> what bugs me: buildout installs inside a vagrant VM take a long time/become impossible when pypi gets flaky like yesterday

[07:27:58] <mgedmin> also, maybe less importantly, jenkins jobs

[07:28:29] <mgedmin> I'm not sure about using caches for jenkins jobs: I want to know right away that my build is broken if someone deletes a package from pypi

[07:28:48] <mgedmin> and then there are production deploys that currently also talk to pypi directly

[07:29:24] <mgedmin> I'd love to switch to wheel-based deployments, but zc.buildout :(

[07:31:00] <ronny> i started writing a pep to fix the problem with pip/virtualenv in a nice way giving a tool like buildout

[07:31:06] <ronny> but its tricky

[07:31:13] <mgedmin> "the problem"?

[07:31:47] <ronny> the core solution is simple, adapting the tooling and writing the pep is whats hard ^^

[07:32:20] <ronny> and it would even be distro and de-vendoring friendly

[07:32:26] <mgedmin> how do you define "the problem"?

[07:33:46] <ronny> mgedmin: lack of multi-version lib usage (split per process), lack of sharing installed wheels, lack of a smart way to import those

[07:36:01] <ronny> mgedmin: imho pip install should a mattter of subseconds in 90% of the cases, just symlinking packages and creating scripts

[07:38:47] <mgedmin> buildout's shared egg cache was awesome for my dev machine

[07:39:41] <mgedmin> but lack of it isn't my #1 issue with pip/virtualenv

[07:57:26] <ronny> mgedmin: without shared/cache based installs getting many things quick and compact is impossible

[07:57:43] <mgedmin> yes

[07:57:49] <ronny> mgedmin: network-less install is tricky in any case as well

[07:58:19] <mgedmin> what prevents me from ditching buildout is recipes other than zc.recipe.egg :/

[07:59:29] <ronny> which ones are you using?

[08:03:49] <mgedmin> zc.zdaemonrecipe, zc.zope3recipes, zc.zodbrecipes, zc.recipe.testrunner

[08:04:24] <mgedmin> and in another project djangorecipe, gp.recipe.node, z3c.recipe.usercrontab (augh! no! evil! killitwithfire!)

[08:04:59] <mgedmin> isn't it fun when a test build creates /var/spool/cron/jenkins that spams my mail every 5 minutes?

[08:14:00] <ronny> ouch

[08:47:10] <fladd> Dear PyPi maintainers. I am having issues with my package Expyriment. The total downloads count is not working. The user dstufft in here told me that downloads via the JSON API should show on average after an hour due to caching. However, this is not true for our package. The counter is stuck at 1328 for days. I have downloaded the wheel fole several times in the last 72 hours but they are not counted. It would be great if someone could look into this and/or conta

[09:13:18] <Terminus-> hello. i was told to ask my question here. why do some packages not show up in pip search or pypi but i can install them? an example would be aldryn-boilerplates. there's a page for it on pypi when i enter the url manually though.

[13:26:40] <ionelmc> argh!

[13:26:44] <ionelmc> ronny: how do i fix this https://travis-ci.org/pytest-dev/pytest-cov/jobs/77499462#L135

[13:35:03] <ronny> ionelmc: by not taking a tarball but ding a pip install for git+htps or so

[13:35:18] <ionelmc> well i open an issue

[13:35:19] <ionelmc> :)

[13:35:37] <ronny> ionelmc: its a wontfix, git tarballs are bad

[13:35:57] <ionelmc> just put some placeholder value in the version

[13:36:04] <ronny> nope ^^

[13:36:09] <ionelmc> and emit big fat warning instead

[13:36:11] <ronny> that speicfically broke builds in strange ways

[13:36:23] <ionelmc> how exactly?

[13:36:33] <ronny> which is why it now bails out early so people dont get broken things

[13:36:54] <ionelmc> you mean if the package actually tries to use the version?

[13:37:06] <ionelmc> setuptools_scm seems broken xD

[13:37:08] <ronny> bascially the version value was wrong, that broke expectations in various places

[13:37:25] <ionelmc> it should be able to deal with this graciously

[13:37:26] <ronny> ionelmc: well, git suks, its entriely gits fault hte tarballs dont have metadata in files

[13:37:35] <ionelmc> sure sure :)

[13:37:49] <ronny> ionelmc: its impossible to deal with it gracious without putting completely wrong data into installs/wheels/whatever

[13:38:03] <ronny> so its either bail out or FML for someone else

[13:38:24] <ronny> do a pip install git+http

[13:39:27] <ionelmc> yeah i'm doing that now but i'm still surprised and angry xD

[13:40:04] <ronny> ionelmc: should i make a better error?

[13:40:17] <ronny> dont be angry at me, the error is the bet i can do with git

[13:40:23] <ionelmc> ronny: you could try adding the correct url, like an example

[13:40:45] <ionelmc> cause i always forget the git+bullshit?alskjdfhdsakfjh#something=otherthing and have to check pip's docs

[13:40:49] <ronny> ionelmc: pull request?

[13:41:05] <ronny> (because its tricky, and i have not really much opensource time atm

[13:43:27] <ionelmc> ok

[13:45:34] <The-Compiler> ronny: s/git/GitHub/g ;)

[13:47:19] <ionelmc> lol

[13:47:32] <ionelmc> The-Compiler: gitweb don't have that either :)

[13:47:32] <The-Compiler> for "git archive" tarballs, you can at least get the commit ID - maybe more, not sure

[13:47:41] <ionelmc> ronny: https://github.com/pypa/setuptools_scm/pull/47

[13:52:02] <ronny> ionelmc: whops, i didnt evne think of just adding the error, i was thinking of creating a correct url instead, thanks

[13:52:51] <ronny> ionelmc: release will be either today or next week ^^

[19:13:56] <itissid> Hey lets say we have a project /tmp/foo and another called /tmp/bar. How do I "install" foo in develop mode( "http://setup.py/ develop" ) into bar?

[19:15:14] <ronny> itissid: you develop install into a python environent, not into a project

[19:15:29] <nedbat> itissid: use a virtualenv. then: pip install -e .

[19:21:03] <itissid> ronny: Sorry paraphrased it a bit. Yes I have virtual env for both

[19:21:14] <itissid> nedbat: So pip install -e ."

[19:21:32] <nedbat> itissid: just this: pip install -e .

[19:21:53] <itissid> nedbat: So I do that in the bar environment?

[19:22:06] <nedbat> itissid: I think you will want to do it in both

[19:22:37] <nedbat> itissid: that installs the project in the current directory into the current virtualenv, in a way that the files in the current directory *are* the installed files.

[19:22:45] <nedbat> so you can continue to edit them, and see the effect immediately

[19:23:30] <itissid> Ah ok so with the bar virtual env active I go to the /tmp/foo and just say "pip install -e ." Right?

[19:23:44] <itissid> nedbat: Let me try that

[19:31:32] <ronny> hi nedbat

[19:31:38] <nedbat> hi

[19:32:48] <ronny> nedbat: it seems like coverage report can completely swap io freeze a linux computer when failing to ignore large autogenerated files

[19:33:05] <nedbat> ronny: "failing to ignore" ?

[19:33:34] <ronny> nedbat: coverage reporting would run out of memory when trying to run reports with those files as part of the report

[19:34:10] <nedbat> ronny: how large are we talking about?

[19:35:04] <ronny> between 10 and 15 mb

[19:35:19] <nedbat> in one file!? yikes

[19:36:02] <ronny> the problem was fixed by teaching someone about json ^^ but coverage killing the system in a hard freeze was still no fun

[19:36:29] <nedbat> ronny: is there a copy of this file I can see?

[19:36:46] <ronny> no its cmopany internal, and im back at home

[19:37:23] <nedbat> ronny: when you get back to work, you still can't send it?

[19:38:20] <ronny> nedbat: bacially think of a file starting with declaring a immense list of dicttionaries and then a tiny loop invoking a factoryboy fctory passing those as kwargs

[19:39:10] <nedbat> ronny: was this coverage 4.0?

[19:39:38] <ronny> nedbat: i tried both 3.7.1 and 4.0something

[19:43:32] <nedbat> ronny: this file: https://github.com/justinwp/croplands/blob/83874e8efd532d2efe324231ec8875377cd9759a/gfsad/utils/countries.py reported a crash against 3.7.1

[19:49:44] <ronny> nedbat: problem is, its not really a crash, its a swap io deadlock

[19:50:17] <ronny> (i.e. swap io and waiting will completely grind everything to halt unless a limiting cgroup is used)

[19:53:07] <nedbat> ronny: right, I know, I'm wondering why your larger file doesn't crash 3.7.1

[20:00:59] <ronny> nedbat: i suspect its entirely rlated to the system grinding to a halt before

[20:02:22] <nedbat> ronny: anything you can do to help me reproduce it will be appreciated

[21:42:59] <elarson> is pypi getting restarted or anything?

[22:06:41] <dstufft> elarson: Nothing should have restarted any processes or anything atm

[22:07:59] <elarson> dstufft: ok, just checking. I was getting some errors for a bit, but it seemed to sort itself out