Log file Viewer
#pypa-dev logs for Tuesday the 16th of June, 2015
[00:04:40] <lifeless> (I don't think we need to / don't need to support it. But I think we should perhaps be clear about that)
[00:38:58] <lifeless> tchaypo: Nakato: also we have some work to do with test-requires: http://logs.openstack.org/07/192007/1/check/check-tempest-dsvm-full/2096399/logs/devstacklog.txt.gz#_2015-06-15_23_20_45_443
[00:39:43] <lifeless> I don't know if thats 'changes in pbr' or 'changes in how we mark up test requirements' or 'changes in setuptools'
[01:46:09] <Nakato> Probably reasonable to add/modify a test to make sure it doesn't break down the line though.
[06:11:18] <tchaypo> Nakato: same as above - if safe_config_parser handles them sanely that’s great; if it doesn’t we’ll need to document it
[06:11:33] <tchaypo> from what testing I’ve done so far I haven’t seen any errors that look like s_c_p isn’t handling them sanely
[06:12:25] <tchaypo> For backwards compatibility, only ; starts an inline comment, while # does not.
[06:13:03] <tchaypo> so document that the file is parsed by safe_config_parser, hence its rules apply - # comments an entire line, if it appears before any non-whitespace characters, but ; is used for in-line comments
[06:45:16] <lifeless> I may find that I can't write out the config file with comments in those lines
[06:52:27] <lifeless> Nakato: I was just giving an example of how the sync code will (hopefully) be producing things
[06:53:11] <lifeless> Nakato: I find that gist hard to evaluate? are they global options? nested blocks like extras are? something else?
[06:54:51] <Nakato> lifeless: So with configparser in py2, if a line starts with whitespace the line is not treated as a comment.
[06:55:08] <lifeless> ('test', "\nhacking>=0.10.0,<0.11\ncoverage>=3.6\ndiscover\ndoc8 # Apache-2.0\nf...
[07:11:15] <lifeless> what we put into the generated setup_kwargs obviously has to match setuptools needs
[07:21:34] <lifeless> https://docs.python.org/3/library/configparser.html#customizing-parser-behaviour is worth a read
[15:11:19] <MathewJennings> Hi all, I am a new contributor who just introduced himself in an email to the Dev mailing list. My name is Mathew Jennings and I am a summer intern at Foursquare. I'm trying to implement an option for the 'pip wheel' command and was wondering if anyone could help me properly test it.
[15:24:53] <Wooble> MathewJennings: I assume the timeout you're seeing in travis is from that pytest.set_trace() you introduce. It's waiting for input.
[15:27:03] <MathewJennings> Ah, yes, I had not realized that I missed that when I committed my changes. Thank you. I also found it difficult to run specific tests locally on my machine. How do you recommend doing this? When I try to test, it always executes all of the functional and all of the unit tests, which takes ~10 minutes to finish.
[15:45:23] <dstufft> well I didn't write the whole CSS, it's using bourbon instead of bootstrap. No real reason other than to try it.
[15:46:49] <maxking> dstufft: I have sent a pull request: https://github.com/pypa/warehouse/pull/536
[17:02:22] <MathewJennings> So I wrote a functional test in test_wheel.py, called test_pip_wheel_specify_architecture. I placed two wheels in test/data/packages: emptywheel-1.0-cp27-none-linux_x86_64.whl and tests/data/packages/emptywheel-1.0-cp27-none-macosx_10_9_x86_64.whl. My test's body looks like this:
[17:05:15] <MathewJennings> Er, well, not sure if there's a good way t copy multiple line of code into here. Sorry, I'm quite new to this. Essentially, I run result = script.pip('wheel', '--no-index', '--find-links', data.find_links, '--arch', 'linux)x86_64', 'emptywheel',) and make sure that the the linux-only wheel I mentioned is listed in result.files_created. I also do the same for the mac wheel, specifying the masosx_10_9_x86_64 architecture.
[17:06:04] <MathewJennings> This test runs perfectly locally. But it seems to be failing on Travis CI: https://travis-ci.org/pypa/pip/builds/67060242
[17:14:43] <MathewJennings> I'm wondering if adding a couple of test wheels to tests/data/packages for my test didn't translate to the Travis CI build for some reason, and it just can't find the 'emptywheel's I'm asking to search for.
[18:09:50] <MathewJennings> Does anyone know if there is a "correct" way to add test data to tests/data/packages, or is it sufficient to just add files to that directory and push them upstream?
[20:04:08] <dstufft> lifeless: what's your email. for some reason it's not coming up in my mail client
[20:12:29] <dstufft> we don't want to know what version of pip is installed, we want to know what version of pip is running because there are many common cases where you run pip without installing it (and I'd like to use the same system for virtualenv, which is even more likely to be ran without being installed, since that's a supported configuration)
[20:14:09] <dstufft> or have any runtime dependencies besides the stdlib other than the fact it shells out to pip via the bundled pip .whl
[20:18:04] <lifeless> so we can either write something that looks very similar to pkg resources metadata, or vendor a package_resources, or a mini-can-only-get-version-impl in this thing in some form. I'm not proposing answers yet, just sketching the space
[20:18:28] <dstufft> I don't need a fnacy version object or anything though fwiw, it could even be as a simple as open("pip/_version", "wb").write("7.1")
[20:19:17] <dstufft> lifeless: oh I forgot the other thing, is we'd need to support all of the versions that pip itself supports, so 2.6, 2.7, 3.2+ currently
[20:22:09] <dstufft> I dind't propose that yet, but last time we proposed dropping 3.2 nobody really cared
[20:23:36] <dstufft> https://caremad.io/images/a-year-of-pypi-downloads/stacked-py3-pct.png (from https://caremad.io/2015/04/a-year-of-pypi-downloads/)
[20:24:10] <maxking> dstufft: Can we talk a little about what we want in User Profile page? referring to https://github.com/pypa/warehouse/issues/403
[20:25:41] <dstufft> maxking: https://warehouse.python.org/user/dstufft/ that's what I did before. https://rubygems.org/profiles/aaronp shows what rubygems does, https://www.npmjs.com/~substack shows what npm does
[20:26:07] <sigmavirus24> I'm also just teasing a bit because I hate some versions of Python (2.6, and 3.2 among them)
[20:28:15] <dstufft> lifeless: anyways, 3.2 isn't a big deal to support for pip, it's not super hard but if nobody is really using it (like for real, I don't really care so much about maintaing supprot for something everyone is testing but nobody is really using) i'd rather not waste the mental cycles making sure things don't break 3.2
[20:28:33] <dstufft> but if there's a big enough contigent of people actually using it we can keep it around
[20:29:20] <lifeless> I'd like to specifically reach out to a few communities before pulling the 3.2 rug on testing-cabal stuff
[20:30:19] <dstufft> I'll probably post to distutils-sig before we actually remove it to make sure folks who care get a chance to comment
[20:32:39] <maxking> dstufft: I am a little confused, what is actually up at warehouse.python.org is something different than what is there in the repo?
[20:33:12] <dstufft> basically I started writing warehouse using werkzeug and my own little custom framework
[20:33:40] <dstufft> and that nobody but me had ever seen it before starting to work on Warehouse and yada yada yada
[20:34:27] <dstufft> warehouse.python.org is still running the old werkzeug code because I hadn't switched the deployment stuff over to use the Pyramid code yet
[20:35:32] <dstufft> older version used bootstrap yes, new version uses bourbon done in a half assed sort of way. My intent is to try and trick a real designer to take ownership of the design
[20:36:07] <dstufft> maxking: the currently deployed code is in the "werkzeug" branch of the repo fwiw
[20:38:16] <lifeless> dstufft: have you been offered early-access to the github fine grained perms thing ?
[20:38:44] <maxking> dstufft: I guess, I have an understanding of the codebase now, WIll try to tackle some hard isssues now. You have anyone specific in mind or I should just look into bug tracker.
[20:39:29] <dstufft> maxking: there is a milestone for things we need to become pypi.python.org, but other than that I'd just grab any issue that seems exciting to you
[20:40:20] <straycat> dstufft, i've written an extension for a tool that converts packages on pypi into packages that can be used with our distro (baserock), to do that i needed to be able to get a list of runtime dependencies out of pip
[20:43:20] <dstufft> straycat: is there a reason you're getting a list of runtime dependencies out of pip instead of just getting it out of the package itself?
[20:45:52] <straycat> dstufft, i need the requirements set for a package, not just its "direct" dependencies
[20:47:08] <dstufft> straycat: That's kind of confusing to me, but I may be misunderstanding what you're doing :) I figured your package system would be able to handle the transitive depednencies
[20:48:41] <straycat> the tool has no dependency resolution, and i figured that to add the correct resolution i'd end up duplicating part of pip anyhow
[20:51:40] <straycat> but to simplify the addition of tools that are already packaged, with rubygems, pypi or whatever, we've written a tool that allows you to "import" packages from these systems into baserock
[20:55:26] <straycat> looks like someone just commented on that pr, seems i'm supposed to have used pkg_resources for this
[20:57:24] <dstufft> I think this is probably something that doesn't make sense to have in pip itself, though i recognize that it may make your tool harder. Long term what I think we want to do is start pulling out some of pip's functionality into reusable libraries with like, actual APIs for cases like this-- but until then I think that your best bet would be to either reimplement or to use pip's internals with the knowledge they are internal and may have
[21:24:49] <maxking> dstufft: I don't see a many to many relationship between 'Package' and 'User' models in warehouse. Is it for a specific reason or I can just add it?
[21:33:37] <dstufft> maxking: request.db.query(Project).filter(Project.name.in_(db.query(Role.package_name).filter(Role.user == request.user))).all()
[21:34:07] <maxking> dstufft: Yeah, I did made up that query, but why not just add one more attribute?
[21:35:18] <dstufft> maxking: because having two sources for the same information tends to lead to one of those sources getting out of date
[21:36:20] <maxking> dstufft: But how? Both Project and User can have a many to many relationship *using* the "Role" model?
[21:37:15] <dstufft> oh you mean mapping it so that there a SQLAlchemy M2M through the Role model? yea idc about that as long as it doesn't involve modifying the DB itself
[21:52:20] <dstufft> if an easy way to figure out some relation at the Python level hasn't been added yet, it probably either means I didn't need it yet, or i got confused trying to add it and gave up :)
[22:00:57] <maxking> dstufft: I have done quite some work with relationships using sqlalchemy in Mailman.
[22:01:37] <maxking> Its too complex to write a query, when everything can easily be done with a M2M relationship very easily.
[22:26:20] <maxking> dstufft: If I change anything in sass files, are they compiled on the fly or do I need to restart everything?