pmxbot IRC Log Viewer

[07:57:53] <mgedmin> pypi down for maintenance?

[07:58:04] <mgedmin> https://status.python.org/ says nothing

[08:11:06] <mgedmin> oh, it's back up

[08:49:48] <indistylo> My goal is to build real time notification , Django realtime: : How to solve the NameError:name 'root' is not defined, Read problem details at http://stackoverflow.com/questions/32944158/django-realtime-how-to-solve-the-nameerrorname-root-is-not-defined

[09:31:45] <ronny> indistylo: wrong channel, and you never defined root

[09:33:47] <indistylo> thanks, I fixed that.

[10:15:38] <mgedmin> I can'

[15:14:40] <elarson> if I have a requirements.txt that includes foo >= 1.0 where the latest package foo is 1.5. If I then install foo == 1.3, will 1.3 become the "active" version in the virtualenv?

[15:18:14] <elarson> looks like pip uninstalls the later version and installs the 1.3

[15:19:51] <Ivo> elarson: yep, you basically answered yourself :)

[15:19:57] <Ivo> *previous version

[15:20:11] <Ivo> oh mm

[15:20:22] <Ivo> you can also use ~=

[15:20:28] <Ivo> pretty hand

[15:20:30] <Ivo> *handy

[15:20:55] <Ivo> but if you tell it to install an exact version itll even uninstall a newer one, yes

[15:25:12] <elarson> Ivo: perfect, thanks for confirming!

[15:25:16] <gthank> I'm sure I'm missing something obvious, and will feel like a dolt when you guys answer this, but: how do I put a dependency on a VCS version of a lib in my requirements file? I'm trying `uWSGI==git+https://github.com/unbit/uwsgi.git#egg=uWSGI` with no success

[15:26:14] <gthank> The error is https://gist.github.com/gthank/6f803612c9a26ff863d1

[15:26:43] <gthank> But when I use that same URL to install manually, it works (as far as I can tell, at least)

[15:26:46] <doismellburning> gthank: drop the `uWSGI==`

[15:26:49] <tdsmith> i think you'll want to drop the prefix and just use the URL

[15:27:09] <gthank> …and I was right about the "something obvious"

[15:28:35] <Ivo> gthank: any particular reason you want master and not 2.0.11.1?

[15:28:58] <gthank> Because their compile flags are a little confused

[15:29:06] <gthank> And they haven't cut a new release.

[15:29:18] <gthank> Under El Cap, they use an undefined symbol

[15:29:51] <Ivo> righto

[15:29:58] <gthank> Because their feature detection used to only pick up some feature under Linux, but El Cap added the feature, but the symbol exported by socket.h on Cap is named differently

[15:30:13] <gthank> Ah the joys of living in the fast lane :-D

[15:30:16] <Ivo> #justapplecompilerthings

[15:30:45] <gthank> Given that it only used to compile under GCC on Apple, I'm inclined to say #justpeoplewhodontwriteportablecode things

[15:31:05] <Ivo> definitely apple compiler thigns

[15:31:05] <gthank> (I may have an irrational hatred of GCC that stems from really old and busted versions 15+ years ago)

[15:31:13] <Ivo> can't remember whether it was this or last year

[15:31:17] <Ivo> they removed a flag

[15:31:25] <Ivo> but didn't rebuild their python with that flag removed

[15:31:37] <gthank> Oh I bet that was tons of fun

[15:31:48] <Ivo> so python distutils was then trying to build every C extension

[15:32:00] <Ivo> with a non-existent flag

[15:32:05] <Ivo> ofc that's pip's fault....

[15:32:14] <gthank> of course!

[15:32:16] <Ivo> =_=

[15:32:34] <gthank> Any success with getting Debian to stop breaking pip?

[15:33:54] <gthank> I've got a ton of respect for the word you do despite the amount of random breakage and hostile people you deal with

[15:34:15] <Ivo> https://stackoverflow.com/questions/22313407/clang-error-unknown-argument-mno-fused-madd-python-package-installation-fa

[15:35:32] <Ivo> Until pip/python changes its whole install methodology to work like node's

[15:36:11] <Ivo> i.e pip's depedencies all live under a tree under pip and don't affect anything else

[15:36:34] <Ivo> it's pretty much debian's policy to accidentally break pip by de-vendoring

[15:37:17] <gthank> Can you call it "accidentally"?

[15:37:38] <gthank> It's not like they haven't been notified/begged/told/cursed-at multiple times

[15:37:50] <Ivo> I'm sure they don't actually conscientously mean to, is what I mean by "accidentally"

[15:38:34] <Ivo> also you can pretty much not break things by using virtualenvs/venvs always & everywhere

[15:39:51] <Ivo> I'd consider the coincidence of encouraging that behaviour a nice side-effect :D

[15:39:55] <gthank> lol

[15:45:15] <ionelmc> what i don't understand is why people bother with installing pip from debian/ubuntu's repos

[15:45:45] <ionelmc> breakage is one thing

[15:46:03] <ionelmc> out of date releases is another thing

[15:51:02] <ronny> ionelmc: well, they are unaware the distro isnt doing its job well

[15:54:22] <elarson> ionelmc: operators often assume a system package provides a better supported system.

[15:55:49] <ionelmc> the pervasive opinion in #python is to use just that

[15:55:54] <ionelmc> boggles my mind

[15:56:28] <dstufft> fwiw I am working with Debian to try and resolve the conflict and figure out a reasonable path forward.

[15:57:17] <ionelmc> dstufft: what's the status on that?

[15:57:18] <Wooble> What, their opinion isn't that pip needs to die and everyone should install everything for the system python using apt? :)

[15:59:39] <dstufft> ionelmc: they're working on a thing called "dirtbike" which will enable them to rebundle pip based on the versions of bundled software that exist in the Debian apt repo

[15:59:59] <dstufft> so pip will still be bundled, they'll just rebundle it at python-pip build time

[16:00:16] <elarson> dstufft: ah interesting.

[16:00:29] <elarson> does that mean the deps still get unvendored?

[16:00:59] <dstufft> Wooble: There are some folks who think that pip needs to die, but some folks (particularly Barry Warsaw, who is awesome) who don't agree and are pushing forward a lot of the work to make pip and Debian play nicely together

[16:01:10] <ionelmc> dstufft: where can i read about this dirtbikeshed thing? :)

[16:01:12] <elarson> or is it like unvendoring, and then revendoring with deb's package code

[16:01:25] <dstufft> elarson: sort of? they'll remove the copies that pip has vendored, but revendor with Debian's copy of requests, cachecontrol, etc

[16:01:29] <elarson> ionelmc: dirtbikeshed++ :)

[16:01:46] <elarson> dstufft: that is a pretty reasonable compromise

[16:01:49] <ionelmc> could not resist the pun

[16:01:51] <elarson> !m debian

[16:01:51] <pmxbot> you're doing good work, debian!

[16:01:56] <dstufft> https://github.com/paulproteus/dirtbike

[16:02:28] <dstufft> elarson: I like to think of it as static linking for Python

[16:02:54] <ionelmc> what's going on here https://github.com/paulproteus/dirtbike/blob/master/dirtbike/__init__.py#L173

[16:03:44] <dstufft> dunno, I haven't actually touched or worke don it yet, my involvement is mostly advisory

[16:04:01] <elarson> the comment suggests that will be laughing, which always seems like a good thing

[16:04:57] <ionelmc> dear god, it's calling the bdist_wheel command directly, with mock objects

[16:05:35] <ionelmc> i guess there's no wheel spec

[16:05:48] <ionelmc> make sense right? xD

[16:06:03] <elarson> ionelmc: there is a comment at the beginning of the file suggesting that is a known bug

[16:06:19] <dstufft> right now it's largely in "let's make it work" phase

[16:06:56] <ionelmc> comments in code == all is forgiven

[16:08:10] <ionelmc> it's hard to be serious with all the crazy packaging stuff going on, no hard feelings :)

[16:08:37] <dstufft> ionelmc: probably a bug suggesting they directly build a wheel wouldnt' be a bad idea :)

[16:09:00] <elarson> ionelmc: exactly :) I comment all my code that way.

[16:09:29] <dstufft> # I'm sorry

[16:09:30] <elarson> # NOTE: I know this looks like it won't work, but it will. We're all good.

[16:09:32] <dstufft> best comment

[16:09:55] <dstufft> (I use the # I'm Sorry comment a lot :D)

[16:18:18] <gthank> If I just want pip to download the archives, but not install (because I have to manually bundle some stuff up and then ship it all over a big tarball for use with --find-links), isn't `pip install --download $MY_DIR` the way to do it?

[16:19:45] <dstufft> gthank: yea

[16:22:37] <gthank> Weird. I'm seeing FileNotFoundError: [Errno 2] No such file or directory: 'pip_cache/PyYAML-3.11.zip'

[16:22:57] <gthank> Where pip_cache is the directory I told it to download to, and PyYAML-3.11 is one of the dependencies I want it to download

[16:28:56] <dstufft> gthank: does pip_cache exist?

[16:29:26] <gthank> It does, but now that you mention it, the permissions are probably borked from earlier futzing about

[16:30:01] <gthank> *sigh* Yes.

[16:30:06] <gthank> I'm on FIRE today, people

[16:39:44] <Ivo> dstufft: what do you think of virtualenv looking in .pipcache for setuptools/pip/wheel wheels?

[16:39:59] <dstufft> Ivo: to what end?

[16:40:13] <Ivo> for newer versions of those, to install, than what it has in virtualenv_support

[16:40:45] <dstufft> why not just remove the --no-index flag and let it pull from PyPI and let pip's built in caching handle it?

[16:41:21] <Ivo> ....that could be cool

[16:42:09] <Ivo> only issue is if we care about creating longer 1st use times

[16:42:34] <Ivo> I guess that might add time to e.g travis like setups

[16:42:51] <Ivo> maybe add a flag to optionally turn it off would be enough

[16:43:09] <Ivo> *'add --no-index back in'

[16:45:23] <dstufft> we used to have --download and --no-download, but it used an insecure method of downloading

[16:45:47] <dstufft> I think re-adding those flags (defaulting to --download) which only controls whether or not --no-index is passed would make sense

[20:40:09] <ghadis> can the requirements file support a git repo with a path into the git repo? I see the options for branch/tag

[20:49:45] <paulproteus> dstufft + ionelmc: (-:

[20:58:54] <Ivo> ghadis: yes, there is a subdirectory query param you can specify

[21:00:07] <Ivo> ghadis: https://pip.pypa.io/en/stable/reference/pip_install/?highlight=subdirectory#examples

[21:01:31] <ghadis> thanks Ivo, missed that

Log file Viewer

Help | Karma | Search:

#pypa logs for Monday the 5th of October, 2015