[00:07:41] <tos9> Which means I'm not immediately wrong at least :)
[00:09:23] <tos9> zouhair: So the setup.py there says that it's version 5.2.0 in fact, but yeah that sdist says it isn't. File a bug with feedparser :)
[04:48:00] <ztane> I'd want to download the *whole* pypi changelog, any ideas? the pypi interface doesn't make it easy :D
[04:49:23] <ztane> xmlrpc only has the "since" parameter, I am not too sure it is safe to do "since=0"
[04:53:26] <ztane> dstufft: do you have a db dump?
[06:56:03] <ronny> nicksloan: trpot a bug to the github one, since its not installable in a sane manner if it is not on pypi and on pypi there already is a package with the same name
[08:18:05] <ztane> was doing analysis on pypi data, 50 packages removed within 2 weeks or so
[14:07:36] <ionelmc> just noticed that pip doesn't really respect the order packages from a requirements file
[14:47:39] <nedbat> when installing from git repos, is there a way to suppress the message, "Could not find a tag or branch '42cb1bb1dc680a60c6452e8bb2b843c2a0382c90', assuming commit."
[14:50:04] <ronny> ionelmc: i dont recall it everdoing that
[14:50:07] <nedbat> tos9: that's an interesting question: 6.0.8, which tells me to upgrade to 6.1.1, which I do successfully, but then pip is still 6.0.8
[14:50:24] <tos9> nedbat: 6.1 has lots more silencing IIRc.
[14:50:32] <tos9> nedbat: So I'd solve that problem and then your first one I think will go away
[14:50:52] <nedbat> tos9: good to know, I will try "pip install --upgrade-really-no-i-mean-it pip" and see if that helps :)
[14:51:35] <nedbat> tos9: oh: my requirements.txt has "pip==6.0.8" in it....
[14:53:14] <nedbat> i seem to remember there was a problem with pip and something else being out of sync, and pinning helped it? Maybe we don't need the pin any more.
[14:56:53] <nedbat> tos9: hmm, 6.1.1 still yammers about commits
[15:05:50] <tos9> nedbat: can you send me something I Can reproduce? I tried installing from a repo and didn't see the message
[15:06:12] <tos9> I believe you obviously so probably that won't help you :) -- if it's not a warnings.Warning I would guess the answer is going to be "no" now
[15:07:14] <nedbat> tos9: are you using a particular commit, like this line in reqs.txt: -e git+https://github.com/edx/django-staticfiles.git@d89aae2a82f2b#egg=django-staticfiles
[15:07:51] <tos9> nedbat: No, I was using a branch name, which I guess explains why I don't see a message saying "could not find a branch".
[19:48:54] <flippmoke> so its a hackish way I can distribute binary wheels for linux?
[19:49:39] <flippmoke> wheels with linux binaries*
[19:49:53] <DanielHolth> hackish or explicitly designed to work that way, take your pick, but watch out for linux's complete lack of binary compatibility between distributions
[19:50:01] <ronny> flippmoke: atm its next to impossible to ensure binary compatibility
[19:52:10] <flippmoke> DanielHolth: ronny: yeah I know its next to impossible
[19:52:22] <DanielHolth> But if you can live with that go for it!
[19:52:40] <DanielHolth> Write your extensions in rust and link with MUSL perhaps.
[19:53:18] <flippmoke> this is bindings for a very large C++ library
[19:53:56] <ronny> flippmoke: do you need it published, or can you use a internal pypi?
[19:54:11] <flippmoke> published if possible its an open source project
[19:54:48] <flippmoke> we have a way to download all the other prebuilt binaries for the library and then build the bindings with a sdist (its rather complicated)
[19:54:56] <flippmoke> since we basically built our own distribution system
[19:55:23] <ronny> flippmoke: problem is that wheels dont yet include the distribution as abi flag
[19:55:33] <ronny> flippmoke: do you build for all distributions as of now?
[19:55:54] <DanielHolth> you would just have to put it somewhere that is distribution-specific and let people add that index to pip, if they are using that Ubuntu. Get you pretty far.
[19:56:35] <flippmoke> long story is we just pulled the python bindings out of our core C++ library, we actually have node bindings as well
[19:56:58] <flippmoke> we currently have builds for linux(ubuntu 12.04)/osx/windows
[19:57:21] <flippmoke> but we try to bundle everything so it relies only on libc
[19:59:01] <ronny> flippmoke: glibc symbol versioning is a major source of OMG ^^
[19:59:59] <ronny> flippmoke: i suspect that if you figured a way to compile without symbol versions, things might be better
[20:00:42] <flippmoke> lol, if we had the bindings build "everything" the script would take forever
[20:01:18] <flippmoke> It takes us about 2 hours on an 8 core machine to build all dependencies with and then build core library + time to build bindings
[20:05:28] <flippmoke> For a python distribution, it will include everything like a png, tiff, gdal (another massive project), sqlite, boost libs, webp, etc. etc.
[20:05:39] <flippmoke> all bundled into a nice binary
[20:06:10] <flippmoke> trying to make it easier for people who are just python users
[20:06:21] <flippmoke> to use our bindings, rather then just very brave C++ devs
[20:19:11] <DanielHolth> It's a great idea, I'd just recommend providing a small package index with just your binary wheel "for last Ubuntu LTS release" and don't worry about trying to somehow conquer libc versioning
[20:19:35] <DanielHolth> If you want to hack on packaging, figure out how to put the OS release in a truncated SHA256 hash instead of the third part of the wheel tag, and add support to pip and bdist_wheel.
[21:51:29] <linovia> has the cert option in the pip configuration file been removed ? It doesn't seem to work
[21:57:42] <linovia> Amazing, I renewed my ssl certificate used by my private mirror and pip won't work any more under ubuntu / Python 2.7 arguing the certificate verify failed. It works on Python 2.7 / OS X and ubuntu / Python 3.4