pmxbot IRC Log Viewer

[05:33:24] <mgedmin> dreaming about the world where pip handles setup_requires natively

[05:33:45] <mgedmin> (or where ssl downloads from pypi don't terminate prematurely on Rackspace)

[07:09:23] <ronny> mgedmin: robert collings has prepared something like that for combination with pbr, also i recently createdissues for solving that

[07:09:40] <mgedmin> tbh pbr scares me

[07:10:01] <ronny> mgedmin: then maybe help with the issues wrt opting out of implicit setuptools ^^

[07:11:36] <ronny> s/setuptools/easy_install

[07:27:54] <ronny> dstufft: btw, any oppinionon supporting including unpacked wheels within .eggs as part of a sdist (when the total size is small enough)

[08:56:32] <ionelmc> ronny: what does RC's pip patches have to do with pbr?

[09:11:32] <ronny> ionelmc: setup_requires in setup.cfg interpreted by pip?

[09:12:51] <ionelmc> ronny: how does that "combine" with pbr?

[09:13:18] <ionelmc> afaik, pbr would just parse that, and pass it to setuptools

[09:15:18] <ronny> ionelmc: yes, but setuptools will always opt into setup_requires, so if pip was parsing and doing them first, it wouldnt trigger easy_install in setup_requires

[09:16:10] <ionelmc> so what does all that have to do with pbr???

[09:16:11] <ionelmc> :)

[09:18:14] <ronny> ionelmc: askt robert colins, personally i think the startwith opt-out path is better

[13:21:55] <isthisreallife> how to fix it? http://pastie.org/10120341

[13:23:03] <doismellburning> isthisreallife: https://packaging.python.org/en/latest/single_source_version.html#single-sourcing-the-version

[13:30:35] <Monkeyget> I looked at how I retrieved the version in my setup

[13:31:03] <Monkeyget> I did import foo followed by set(version=foo.__version__)

[13:31:05] <Monkeyget> oops!

[13:56:17] <ionelmc> ronny: why don't ya add it on that page?

[13:56:49] <tonythomas> Hello. I am getting this error " No distributions at all found for html-cleaner " when trying to install django-blog using pip install git+git://github.com/ilblackdragon/django-blogs.git

[13:56:57] <tonythomas> does such a package ever exist ?

[13:57:18] <ionelmc> tonythomas: doesn't seem so

[13:57:44] <tonythomas> ionelmc: I wanted a blog feature in my django website - I was following https://pypi.python.org/pypi/django-blogs/0.2.0

[13:57:53] <tonythomas> thats okey, right ?

[13:58:05] <ionelmc> tonythomas: oh, he use dependency_links

[13:58:14] <ionelmc> not use if that works anymore

[13:58:30] <tonythomas> ionelmc: can you provide some more 'exact' link ?

[13:58:33] <ionelmc> tonythomas: install directly from https://github.com/ProstoKSI/html-cleaner/archive/master.zip ?

[13:58:47] <tonythomas> okey. will try that

[13:59:00] <ionelmc> tonythomas: make sure you fill a bugreport for missing package on pypi

[13:59:28] <tonythomas> ionelmc: here ? https://bitbucket.org/pypa/pypi/issues

[13:59:50] <ionelmc> tonythomas: why would pypa fix other people's projects?

[14:00:08] <ionelmc> tonythomas: here https://github.com/ProstoKSI/html-cleaner/issues

[14:00:20] <tonythomas> ionelmc: will do :)

[14:00:23] <doismellburning> tonythomas: so, PyPA doesn't work like (e.g.) Debian packaging

[14:00:48] <doismellburning> tonythomas: the packages on PyPI are user-supplied; the PyPA isn't a "packaging authority" like, say, Debian packaging maintainers

[14:01:09] <doismellburning> tonythomas: more like the PyPA maintain the packaging _toolchain_

[14:01:23] <doismellburning> ionelmc: and that is a reason why people might think thusly

[14:01:39] <tonythomas> doismellburning: okey - so its like user managed. cool. I should be finding more of it soon

[14:02:01] <atal> hey guys, has anyone had success uploading large packages via twine?

[14:02:25] <atal> large as in, bigger than the available memory on the host

[14:02:37] <tonythomas> ionelmc: done. https://github.com/ProstoKSI/html-cleaner/issues/1

[14:02:52] <ionelmc> doismellburning: so if i make a debian bugreport for some firefox issue the debian maintainers report it upstream?!

[14:03:58] <doismellburning> ionelmc: for how firefox is packaged in Debian? sure

[14:04:06] <doismellburning> er, sorry, let me rephrase

[14:04:23] <doismellburning> ionelmc: no, and I have limited patience for flamebait / etc.

[14:04:45] <doismellburning> the issue was with the packaging of some code

[14:05:34] <dstufft> atal: are you uploading this to PyPI? Or to something internal

[14:06:02] <atal> something internal

[14:07:24] <dstufft> atal: Ok, was going to say that PyPI has a limit on the size of something you can upload :) Right now twine loads the whole file into memory, but not for any particular reason other than it's what I did when I wrote it

[14:07:40] <atal> yeah, that fp.read() line is the worry :)

[14:08:02] <ronny> ionelmc: i suppose i should

[14:08:09] <dstufft> It wouldn't be very difficult to modify it to not do that, you'll need to read it once to generate a MD5 hash, and then read it again and pass the iterator to requests to upload

[14:09:01] <dstufft> atal: Can you open an issue on github.com/pypa/twine? And if you feel so inclined a PR would be great too :)

[14:09:25] <atal> heh, I'll see what I can do :) issue at least inc

[14:10:24] <ronny> dstufft: ping? do you see any technical problem with treating wheels like zip-unsafe eggs?

[14:15:09] <dstufft> ronny: currently? No. but the wheel spec isn't beholden to what eggs used to do, so it's quite possible the layout can change to where that wouldn't be valid anymore.

[14:15:57] <ronny> dstufft: as far as i can tell in terms of import mechanisms and activation, being able to unpack, and activate wheels that fit the platform should work

[14:17:04] <ionelmc> doismellburning: flamebait? not very charitable today eh? :-)

[14:19:46] <doismellburning> ionelmc: not really, sorry

[15:06:27] <atal> dstufft: https://github.com/pypa/twine/pull/105 i think that should do, have to double check requests implementation there yet

[15:07:26] <atal> that change on L225 is unreleated, but it bugged on me during testing so it's fixed too. is_redirect is a property tho

[15:10:09] <atal> sonofa. https://github.com/kennethreitz/requests/blob/master/requests/models.py#L150

[18:55:24] <buck1> is there any way to reference site-packages from a whl?

[18:58:07] <aclark> buck1: not sure what you mean

[18:58:46] <aclark> buck1: you can check __file__ if that helps, but that has nothing to do with whl as far as I know

[19:27:13] <ronny> buck1: what do you even mean by that, a whl gets unpacked into site-packges, end

[20:31:17] <buck1> aclark: ronny: I'm specifically talking about datafiles

[20:31:26] <buck1> they seem to get unpacked to the prefix

[20:31:43] <buck1> and i don't see any reasonable way to target site-packages with datafiles

[20:32:41] <ronny> buck1: ah, data files get moved according tothe fsh, put them in as package data if you want them in a module

[20:33:04] <buck1> fsh?

[20:33:10] <ronny> buck1: also im still not 100% sure what you mean, an example would be nice

[20:33:18] <ronny> buck1: whops eh FHS

[20:33:20] <buck1> i'm trying to package a pth hack :(

[20:33:31] <buck1> which means a non-py file directly in site-packages

[20:33:41] <buck1> I've tried many methods, none straightforward, none reliable

[20:34:59] <buck1> i did finally put the datafile to lib/python<X.Y>/site-packages, but it makes problems when th whl is marked as "py2" pure python

[20:35:03] <buck1> doesn't in fact work on all python2

[20:35:14] <buck1> because of the x.y encoded into the whl

[20:36:09] <ronny> buck1: for what reason do you do that?

[20:36:20] <ronny> (i know how to do it, but its evil)

[20:36:40] <buck1> cov-core does it to enable to have a startup hook to turn on coverage

[20:38:14] <ronny> buck1: i ee, see how its done in the pdbpp package, it ships a pth file in wheels and on install

[20:38:23] <buck1> thanks

[20:41:01] <buck1> ronny: how do they avoid my above whl problem?

[20:41:08] <buck1> or is it not a pure-python package

[20:41:42] <ronny> buck1: i personally subclssed the install command and wrote a pth file to a purelib location

[20:41:59] <buck1> ah i missed the purelib bit

[20:42:28] <ronny> the purelib bits go straight to site-packages, the platlib ones get moved

[20:43:57] <buck1> ronny: I don't see how install_pth_hack.install_dir gets set

[20:44:06] <buck1> that's the purelib location though?

[20:44:49] <ronny> buck1: yes, ah hmm, you might be in trouble because of the install command

[20:45:33] <ronny> i forgot about purelib vs platlib

[20:45:39] <ronny> (pdbpp is pure python)

[20:46:00] <ronny> you can either make a extra package for the hook or read what the install commands to (its a bit of a mess)

[22:10:08] <buck1> ronny: two things: the pth is left behind on uninstall

[22:10:25] <buck1> and, the whole thing crashes with setuptools 3.4

[22:10:34] <buck1> i'm not in a good position to upgrade our virtualenv/setuptools

[22:11:20] <buck1> ronny: this is the traceback http://paste.pound-python.org/show/msBfg2TTq64y7XHQuTSJ/

[22:15:10] <buck1> ronny: my application is also pure python, so that bit's fine

[22:29:51] <ionelmc> buck1: you want a pth?

[22:30:14] <buck1> yes, packaged, in a whl

[22:30:40] <ionelmc> buck1: look no further https://github.com/ionelmc/python-hunter/blob/master/setup.py

[22:30:44] <ionelmc> it works everywhere :)

[22:30:51] <buck1> <3 trying now

[22:31:12] <ionelmc> sdist, wheel, develop, you name it :P

[22:31:47] <buck1> ionelmc: do you know the difference between this and ronny's?

[22:32:16] <buck1> ionelmc: https://bitbucket.org/antocuni/pdb/src/f6d8011e48ddf92ea76ba8d02d11bce4ddd65bae/setup.py?at=default

[22:32:45] <ionelmc> buck1: wost way to do it imo

[22:33:00] <buck1> ok but i dont understand what the difference is

[22:33:04] <ionelmc> broken develop, broken whl, prolly broken easy_install

[22:33:11] <buck1> my distutils-fu is weak

[22:33:30] <ionelmc> buck1: i overwrote the build command, the install will install whatever is built

[22:34:03] <buck1> ok

[22:34:05] <ionelmc> if you override the build then mostly ever4ything works out of the box: install, wheel, other bdist dists

[22:34:34] <ionelmc> the easy_install/develop are for the damn egg/egg-link installs

[22:34:58] <buck1> ionelmc: python-hunter needs more badges

[22:35:00] <buck1> imo

[22:35:07] <ionelmc> where the build command "has no say"

[22:35:20] <ionelmc> buck1: how many? i could add 3 more :P

[22:36:33] <ionelmc> it's a trick actually, the badges are there to deter you from looking at the low star count haha

Log file Viewer

Help | Karma | Search:

#pypa logs for Wednesday the 29th of April, 2015