Log file Viewer

[00:28:46] <torham> I am installing a package using `pip install -e <url>`, but if one of its requirements changes I need to resolve the new requirement manually, is it possible to have pip do this?

[00:30:12] <torham> (it is a git url)

[12:32:37] <apollo13> dstufft: around? getting unable to execute 'x86_64-linux-gnu-gcc': No such file or directory -- but apparently only via ansible, not when I try build psycopg2 in the venv directly, ever seen that?

[12:33:01] <apollo13> which makes no sense since it is in /usr/bin *sigh*

[12:46:22] <apollo13> mhm, can I disable the creation of wheels? ie disable bdist_wheel?

[12:48:24] <apollo13> ok, also errors out with --no-binary :all:

[12:48:48] <ronny> apollo13: why not push prebuilt wheels fot your target?

[12:48:57] <apollo13> ronny: cause …

[12:49:04] <ronny> (as far as i can tell your target has a broken installation base)

[12:49:10] <apollo13> ronny: I still would like to debug the issue :D

[12:49:21] <apollo13> ronny: if I do it manually it works fine, so the install base should be okay

[12:49:32] <ronny> apollo13: so it happens only in ansible?

[12:49:35] <apollo13> yes

[12:49:46] <ronny> are there ans shell scripts that do setup you might not get with ansible?

[12:49:50] <apollo13> to be fair I do not know the exact command yet that it execute

[12:50:02] <apollo13> no, all I do is source the venv and pip install

[12:50:27] <ronny> apollo13: ssh login vs ansible is different

[12:50:36] <ronny> ansible wont run your shell init

[12:50:44] <ronny> so please take a look at the paths

[12:51:01] <apollo13> I did not put anything into the path, this is a newly setup ubuntu 16.04

[12:51:10] <apollo13> lets see if I can get the paths somehow

[12:53:33] <apollo13> ronny: lol, wrong machine I was on :/

[12:54:25] <ronny> clad my pverly pedantic inquiry still helped you find detail problems

[12:54:29] <ronny> *glad

[12:54:54] <ronny> apollo13: systems tend to be a pain ^^

[12:55:31] <gsnedders> Is there any way to get a `pip freeze` list after running `pip install -t foo six`?

[12:56:34] <ronny> gsnedders: not really, you can point a pkg_resources working set at the folder and iterate the distributions * print name/version

[12:57:10] <gsnedders> ronny: k, thx. Just want some way of getting a fixed list following vendoring updates, etc.

[13:00:19] <ronny> gsnedders: care to turn that into a pypi package?

[13:21:11] <gsnedders> ronny: sure

[13:22:52] <ronny> \o/

[13:23:25] <ronny> gsnedders: if you want to overdo it, teach it to patch imports to vendored packages in other vendored packages, then pip/pytest and others can automate

[13:23:54] <gsnedders> ronny: that was absolutely something I was intending on

[13:24:04] <gsnedders> ronny: per the discussion in here with you before :)

[13:24:27] <ronny> gsnedders: i need that for pytest, it would also be an alternate slution to my vendoring problem

[13:24:50] <ronny> gsnedders: please use pyproject.toml for the configfile ^^

[13:26:15] <gsnedders> ronny: my first version will probably be terrible :)

[13:26:36] <ronny> gsnedders: thats fine, i jsu dont want to start, im happy to help later on

[13:26:44] <ronny> gsnedders: release rarly, releae often

[13:27:04] <gsnedders> ronny: indeed

[13:29:20] <ronny> ok, my typo rate is trough the roof, brb getting sugar

[13:40:34] <gsnedders> ronny: so what is the advantage of using pyproject.toml v. a requirements.txt?

[13:41:18] <ronny> gsnedders: a) its a standard

[13:41:47] <ronny> b, you are not handlign requirement there, you are handleing a set of packages that is to be vendored and may need code editing as specified

[13:48:04] <gp> I am having an issue with get-pip.py and vagrant using a proxy on Ubuntu 14.04. In my provision script I create a virtualenv and use get-pip.py to update the virtualenv pip. Then I install requests[security] with pip to fix ssl issues. However, when the provisioner script runs again the get-pip.py script seems to ignore the installed requests[security] packages and uses the insecure version bundled with it. Is this expected or am I doing

[13:48:04] <gp> something wrong?

[13:49:11] <gp> The warning observed indicates it is not using the virtualenv security updates: /tmp/tmpEN33qJ/pip.zip/pip/_vendor/requests/packages/urllib3/util/ssl_.py:122: InsecurePlatformWarning

[13:53:41] <gp> ignore that

[13:54:11] <gp> always seems like it takes asking a question in public to find a typo =P

[13:54:49] <ronny> gp: yes, in case of doubt write it to a rubber duck first ^^

[13:55:03] <gp> ronny: haha

[13:59:07] <ronny> for morbid effect, print it with a rubber duck as background

[14:49:38] <gp> Is there a simple/quick bash test I can use to determine if https requests can be secure or not? I'd like to omit '--index-url=http://pypi.python.org/simple/ --trusted-host pypi.python.org' in my provision script if ssl packages are installed on ubuntu 14.04

[14:49:57] <gp> but before they are installed i need that to work through my proxy

[15:12:41] <nanonyme> Why do you think proxy would matter at all?

[15:13:28] <nanonyme> Oh, never mind, maybe I misunderstood

[17:47:10] <gp> nanonyme: perhaps the proxy is irrelevent. it isn't really important for the fix. but when I try to let pip get downloads from pypi over https before installing the requests[security] package it fails to download and warns about insecure and lack of sni support. i figured something was getting munged by the proxy but maybe it is irrelevant. I am just wondering if there is any way i can "test" for the security packages being applied

[17:47:11] <gp> for pip's https requests so I can omit the index override that forces http downloads

[17:48:43] <gp> nanoyme: pseudocode http://dpaste.com/3M8V8HP

[17:48:56] <gp> Looking for the test that gives me SSL_IS_NOT_SECURE

[17:49:57] <gp> figured since pip throws warnings there might be an easy way to test for it

[18:48:54] <nanonyme> gp, yeah, sure, HEAD to https://pypi.python.org would probably be enough

[18:50:13] <gp> nanonyme: ah... like manually make a request in python and check for warning text? that would work I guess. I was thinking there might be a simple way to do it with pip

[18:50:41] <nanonyme> I find it really unfortunate the SSL improvements were backported to 2.x in a way that broke backwards-compatibility

[18:50:49] <gp> yeah =/

[18:51:19] <nanonyme> It was totally possible to have the best out of it IOW give new ciphers, give SNI, give *optional* certificate validation

[18:51:42] <nanonyme> gp, how about pip search?

[18:53:37] <nanonyme> It probably hits the backend so it should fail if connections can't be done

[18:55:06] <gp> nanonyme: that would work. I will go that route if it ends up taking too much effort to determine without making a request. was hoping there was a way to inspect the installation quickly to determine if request[security

[18:55:14] <gp> hit enter instead of bracket lol

[18:55:43] <gp> was installed or if there was a flag exposed I could use

[18:55:58] <nanonyme> Well, *I* don't know whether there is

[18:56:43] <gp> pip search is probably going to be the fastest =) was just in a different direction than I was thinking. thanks for the idea

[18:57:03] <gp> *fastest meaning path of least resistance

[19:16:34] <mitsuhiko> does anyone know if there are docker images available for a manylinux build that have clang and other modern compilers on?

[19:16:40] <mitsuhiko> very recent clang in fact

[19:17:51] <dstufft> can you use a recent clang with an old glibc

[19:17:57] <dstufft> (I have no idea, I suck at compilers)

[19:24:22] <ngoldbaum> mitsuhiko: per pep-513, i think you *need* to use old compilers, where "old" is whatever's on centos 5

[19:24:49] <mitsuhiko> ngoldbaum: yeah. i already decided to "fuck everything"

[19:24:57] <mitsuhiko> none of what i try to do works

[19:25:07] <mitsuhiko> because someone decided to break all the stuff on the way to newer c++

[19:25:21] <mitsuhiko> linux binary compatibility is such a shitshow :(

[19:25:36] <ngoldbaum> i always nope out whenever i see a python project depends on new c++ stuff

[19:25:41] <ngoldbaum> since i know it's going to get painful

[19:26:17] <ngoldbaum> see also when a python project needs npm to build :(

[19:29:47] <mitsuhiko> ngoldbaum: i literally compile llvm into a library here

[19:29:49] <mitsuhiko> so i don't have much choice

[19:32:47] <agronholm> to be fair, none of that binary compatibility shitshow is python's fault

[19:32:57] <ngoldbaum> true

[19:33:10] <ngoldbaum> many of python's build headaches are really c or C++ build headaches

[19:33:19] <agronholm> c++ has some fundamental problems

[19:33:27] <dstufft> computers have some fundamental problems

[19:33:44] <agronholm> ofc, but not all languages suffer from problems like c++'s

[19:34:24] <agronholm> personally I won't touch it if I can possibly avoid it

[19:34:38] <agronholm> I took some time years ago to try and learn it

[19:34:45] <agronholm> the only language I've ever given up on learning

[19:37:20] <DanielWHolth2> Hello compiler fans