PMXBOT Log file Viewer

Help | Karma | Search:

#pypa logs for Tuesday the 4th of October, 2016

(Back to #pypa overview) (Back to channel listing) (Animate logs)
[08:05:35] <stor> what's the proper way to use requirements files if i want to install some packages from pypi and some from local filesystem?
[08:05:51] <stor> e.g. if i have a fork of a package also available in pypi
[08:06:14] <stor> and i don't want to install that one, but my patched version
[08:58:57] <brechtm> stor: you can add a filesystem path to the fork, but I'm not sure about the syntax
[09:00:37] <stor> i guess if i don't care if pip does a pypi lookup, i can add a -foo to the my fork's package version and require that version
[09:03:15] <brechtm> stor: you should be able to just add the path to your local fok, see https://pip.pypa.io/en/stable/reference/pip_install/#example-requirements-file
[09:04:20] <brechtm> stor: the path can point to a package (whl, sdist) but also a directory with a setup.py, I think
[09:05:20] <stor> hmm, i know... maybe my problem is how install_requires in setup.py and requirement files interact
[09:06:22] <stor> maybe i specify a pkgx==1.2-foo in setup.py and add a path to requirements.txt that allows pip to find this version
[09:07:26] <stor> ...and i'll not care about a lookup to pypi
[09:08:00] <brechtm> ah, now I understand the problem
[09:08:38] <brechtm> it doesn't work if you place the requirement at the top of requirements.txt?
[09:09:53] <brechtm> if the dependency is already installed, pip shouldn't touch it
[09:10:35] <mgedmin> PEP-440 suggests +local1 as a version suffix
[09:10:49] <mgedmin> for packages with local modifications
[09:11:50] <stor> too bad https://packaging.python.org/patching/ is a fixme :/
[09:12:01] <mgedmin> it's what I'd do -- -f path/to/local/packages and pkg==1.2+local1
[09:13:10] <stor> yep. this does a pypi lookup, right?
[09:14:35] <stor> would be nice to have --no-index in requirements.txt, then some (local) packages, then --index and pypi packages
[09:14:47] <brechtm> Is all that necessary when specifying the fork's local path in requirements.txt?
[09:20:54] <stor> i think not
[16:08:49] <narfology> I am trying to update PIP but I get this error: https://gist.github.com/l3m/0c78244634dd96119aede0ab5e6e125a
[16:08:53] <narfology> Any ideas?
[16:10:57] <xafer> What does "python -m pip --version" give you ?
[16:11:22] <narfology> pip 7.1.0 from /usr/lib/python2.7/site-packages (python 2.7)
[16:12:10] <narfology> xafer: The funny thing is, I even get the upgrade message on the call to pip install --upgrade pip
[16:12:18] <narfology> argh
[16:13:03] <xafer> What does `python -c "from pkg_resources import Environment;print(Environment()['pip'])"` print ?
[16:16:17] <narfology> [pip 7.1.0 (/usr/lib/python2.7/site-packages)]
[16:17:51] <xafer> and "pip install --upgrade pip --verbose" ?
[16:18:43] <narfology> ok nice
[16:18:49] <narfology> xafer: Could not fetch URL https://pypi.python.org/simple/pip/: connection error: [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed (_ssl.c:765) - skipping
[16:19:25] <narfology> thanks a lot
[16:19:32] <xafer> you're welcome :)
[16:22:32] <narfology> argh
[16:22:35] <narfology> now I get a hash fail
[16:22:43] <narfology> Hash of the package https://pypi.python.org/packages/9c/32/004ce0852e0a127f07f358b715015763273799bd798956fa930814b60f39/pip-8.1.2-py2.py3-none-any.whl#md5=0570520434c5b600d89ec95393b2650b (from https://pypi.python.org/simple/pip/) (62d0eaa70fcc9f0580596d2144573954) doesn't match the expected hash 0570520434c5b600d89ec95393b2650b
[16:26:51] <narfology> I assume this is our proxy interfering. I added our custom ca-cert to CentOS, but how can I make it known to pip? Looks like pip doesn't use the system ca-certs
[16:28:10] <ngoldbaum> narfology: if you're not getting the same md5 hash as what's listed on pypi, the file you've downloaded is corrupted somehow
[16:28:38] <tdsmith> dstufft: does pip just use certifi now?
[16:30:24] <tdsmith> narfology: you can pass --cert path/to/certificates to have pip use an alternative set
[16:31:33] <narfology> tdsmith: doesn't seem to work, I still get [SSL: CERTIFICATE_VERIFY_FAILED] even if I point it to the same cert that I needed to make yum work...
[16:35:04] <tdsmith> yes, https://github.com/pypa/pip/commit/c77d4ab55ed412a3b72d0b73f504e8ddec918683#diff-f21e4e94baf6a2d2f98faab0451e418a was the change i was looking for
[16:35:33] <xafer> since 8.0.2
[16:36:30] <narfology> so I'm on 7.1.0 and can't upgrade to 8.x
[16:36:38] <narfology> yay
[16:36:49] <xafer> You can still use get-pip.py
[16:37:10] <narfology> but even if I download the pip file manually (with wget), I get a different value if I md5sum
[16:37:20] <xafer> or download pip wheel and install it
[16:37:28] <xafer> well that's strange...
[16:38:05] <tdsmith> i see 0570520434c5b600d89ec95393b2650b here
[16:38:16] <narfology> I do know my corporate proxy is evil, but this evil that it even modifies harmless pip downloads?
[16:38:17] <narfology> urgh
[16:38:33] <tdsmith> if you manually inspect the downloaded file do you see anything?
[16:38:54] <narfology> yes
[16:38:56] <narfology> argh
[16:39:07] <narfology> an html page telling me it's blocked
[16:39:11] <narfology> thanks
[19:15:48] <bowlofeggs> hello! i think the recent change to make exclude faster might have broken the ability to include subpackages: https://github.com/pypa/setuptools/issues/808
[19:16:58] <bowlofeggs> by "exclude", i mean the exclude argument to find_packages() ☺
[19:19:08] <bowlofeggs> in my case, i can workaround it, but i think the new behavior is a little surprising
[19:19:26] <bowlofeggs> and the changelog did not mention that the include behavior changed as well
[20:43:22] <dstufft> tdsmith: it bundles it's own certificate, or uses certifi if it's available