[00:02:14] <dstufft> it should give you a debugger session
[00:02:31] <Guddu> Thanks dstufft I will try that in a local environment now
[00:03:16] <Guddu> Once i get to debugger session is there anything specific i should do?
[00:03:31] <dstufft> try importing setuptools and see what happens
[00:07:41] <Guddu> dstufft, Unfortunately I am not able to log in to VPN . . I will continue this from office the first thing tomorrow. I will do all this from command prompt then and not setup.py of reportlab right?
[13:48:12] <SamuelMarks> Handling multiple repositories, each with their own dependencies. Why isn't p2—which is a dependency of p1, which is a dependency of p0—being installed [via github]? - stackoverflow.com/q/28540839
[13:54:19] <ronny> SamuelMarks: did you list them in setup.py? the requirements files are pip only
[13:55:29] <ronny> SamuelMarks: requiremens files are good for pinning versions or finalr projects, they have no use in libraries unless you parse them in each setup.py (which is kinda tricky)
[13:56:21] <SamuelMarks> Yeah, I tried putting them in install_requires and dependency_links; but that failed to find the package
[13:56:31] <SamuelMarks> (see my git commit history for what I tried)
[13:56:37] <mgedmin> can you link to the git repos?
[13:57:09] <mgedmin> the SO question links to github.com/bettertutors, which has four repos, and I don't know where I should look
[13:58:00] <SamuelMarks> ronny: All my stuff is open source, so will eventually put it on pypi proper. For now though, everything is too much in an alpha dev stage.
[13:58:05] <mgedmin> when you say "failed to find the package", what package do you mean?
[13:58:19] <mgedmin> ideally pastebin the console session where you tried to pip install something and got an error
[13:58:30] <ronny> SamuelMarks: you can spin up a devpi instance even on your local computer
[13:59:25] <ronny> SamuelMarks: for development it might also make sense to just make a virtualenv in a parent directory, and pip install -e the repos in order
[14:00:18] <SamuelMarks> ronny: Maybe, but at the end of the day I am deploying through a CD system, so everything should be automated + standards driven rather than through custom scripts
[14:00:28] <SamuelMarks> (e.g.: I deploy to Heroku, OpenShift and others)
[14:01:56] <SamuelMarks> mgedmin / ronny : I have to run, if you figure out where I went wrong I'd appreciate it if you could reply on that stackoverflow post or comment on that github gist. Thanks
[14:02:31] <ronny> SamuelMarks: then go for a devpi server and also use devpi test ^^
[15:37:48] <doismellburning> is there a way to push arbitrary-project-A to arbitrary-pypi-B purely by running a command, with no ~/.pypirc or similar hackery?
[15:43:06] <doismellburning> (because I would like to not have to hardcode my PyPI credentials in a file in my repo :()
[15:58:43] <ronny> doismellburning: maybe with twine? also devpi has such tooling
[16:07:19] <pf_moore> PEP 440 - am I missing something, how do you specify "I want version X.Y.Z or U.V.W or A.B.C"? Comma means "and", what means "or"?
[16:23:33] <ronny> dstufft: is there a reason why twine cant just be passed a directory?
[16:24:20] <doismellburning> ronny: FYI no, not with twine
[16:26:06] <ronny> doismellburning: if you want to have sets of internal pypis i think devpi is the perfect option
[16:26:22] <ronny> it supports uploads and pushing things between indexes
[16:28:56] <doismellburning> I don't quite get it - using arbitrary PyPI is easy for down (`pip install -i $URL`) - why the hoops on the way up?
[16:29:02] <ronny> doismellburning: the devpi client can just be configured with credentials via the devpi login command, then it stores a signed auth cookie that it will use for the next 10 hours
[16:29:15] <ronny> doismellburning: and when using devpi, just use devpi install to use the selected index
[16:29:31] <ronny> so the worflow is devpi use someindex: devpi install somepackage
[16:30:06] <ronny> doismellburning: one of the reasons is that pypi uploads where implmeented in distutils, thata waaaaay off from pip
[16:30:11] <doismellburning> ronny: sure, thanks, I've seen _how_, I'm failing to understand why
[17:09:44] <xafer> well I'm at loss, here... You could try to upgrade your pip, or manually rm -rf dns and dnspython-1.12.0-py2.7.egg-info to have the same behavior as pip
[17:10:02] <ThomasV> xafer: btw, what is the reason behind having some packages in /usr/lib and some in /usr/local/lib ?
[17:11:03] <ThomasV> well, manyally removing works.. but I am considering adding dnspython as a dependency in my project, and I would prefer if it was able to uninstall
[18:07:40] <Wooble> 2.7.8 should have it, actually, it was committed in june.
[18:11:02] <[Tritium]> doismellburning: do i even have to ask if you just did python -V
[18:18:09] <doismellburning> [Tritium]: not really
[18:21:26] <doismellburning> is there anything I can do to facilitate a release of twine perchance? I've just fallen foul of github.com/pypa/twine/issues/8
[18:24:59] <doismellburning> ...my apologies, misread the dates, thought that issue showed unreleased changes from 2013, oops
[18:55:51] <ronny> jaraco: ping, i finished modularization of hgdistver, it now discovers all metadata extractors based on hooks and supports plugins cappable of adding new scm's easyly
[19:13:39] <ronny> ah, i forgot, its still required for distutils style usage
[19:15:16] <ronny> jaraco: but bascially now all one needs to do to add another scm is register at 2 entrypoints with the names of the control directories
[19:17:54] <ronny> jaraco: could you perhaps review the readme and skim over the modules? i'd like to get some feedback from someone not used to the codebase, then do a initial release
[19:22:29] <ronny> jaraco: btw, file finders are not configurable enough, unlike with the distribution configuration i cant pass extra arguments
[19:24:11] <ronny> jaraco: whats the policy on the setuptools repo? is pushing to named branches for proposals/experiments ok ir should a separate one be used
[19:43:09] <jaraco> ronny, I'm happy to review. I'm not sure when I'll get to it. I'll be travelling Thursday and Monday, so I might pull a clone and review in the plane.
[19:44:06] <ronny> jaraco: the codebase is rather small, and i only need a skimming, before the first release undere the new name, i suspect you can do it in 5-10 minutes
[19:44:11] <jaraco> ronny, as for setuptools, there's no policy per-se. I personally prefer unnamed branches (typically managed with bookmarks) for small-scale commits, following the Mercurial guidance for creating named branches.
[19:45:49] <jaraco> So if you submit a small PR to a named branch, I might rebase it. But I might also keep it in the named branch, especially if that name is specific and descriptive.
[19:46:36] <ronny> jaraco: i prefer to use namd branches with feature names that get closed *before* merge]
[19:48:19] <ronny> i'll use the branch name experiment/configurable-file-finder
[22:30:53] <SamuelMarks> apollo13: I'm deploying to Heroku, where I don't have control over the pip command. Can I set those options somewhere else? - Also, how does that explain 2/3 of the packages being installed?
[22:31:37] <apollo13> SamuelMarks: as the example shows: in the requirements file