Closed clone1018 closed 9 years ago
Blech. I looked at this thinking it might be a simple Python thing but [Errno 71] Protocol error
looks more virtualization-related.
TeamX
Let's try this? https://github.com/pypa/virtualenv/pull/409
Can someone on a failing windows version give the --always-copy
flag a shot?
@patcon
For me it works. Without --always-copy
flag:
vagrant@vagrant-ubuntu-trusty-64:/vagrant$ virtualenv --no-site-packages env
New python executable in /vagrant/env/bin/python
Traceback (most recent call last):
File "/usr/bin/virtualenv", line 3, in <module>
virtualenv.main()
File "/usr/lib/python2.7/dist-packages/virtualenv.py", line 825, in main
symlink=options.symlink)
File "/usr/lib/python2.7/dist-packages/virtualenv.py", line 985, in create_environment
site_packages=site_packages, clear=clear, symlink=symlink))
File "/usr/lib/python2.7/dist-packages/virtualenv.py", line 1405, in install_python
os.symlink(py_executable_base, full_pth)
OSError: [Errno 71] Protocol error
With --always-copy
flag:
vagrant@vagrant-ubuntu-trusty-64:/vagrant$ virtualenv --no-site-packages env --always-copy
New python executable in env/bin/python
Installing setuptools, pip...done.
Windows 7 64-bit, vagrant (Ubuntu 14.04 64-bit)
Thanks @lukaszbanasiak! So where do we need to add --always-copy
? To the Makefile? Want to give us a PR for that?
Just run into this. I think I'll be able to tackle this. The problem is likely related to permissions for files mounted as VirtualBox share. Git also chokes on them, because all files are executable, permissions are immutable etc.
Tracing the chain so far. This is invoked from /vagrant
directory.
[vagrant provision] -> [scripts/vagrant-setup.sh] -> [make env] ->
-> [virtualenv-1.9.1.py ...]
virtualenv-1.9.1
calls failing copy tree two times until fault:
shutil.copytree(src, dst, symlinks=False):
==> default: Copying /usr/lib/python2.7/lib-dynload ./env/lib/python2.7/lib-dynload
==> default: Copying /usr/lib/python2.7/config ./env/lib/python2.7/config
/usr/lib/python2.7/config
appears to be a directory:
drwxr-xr-x 2 root root 4096 Apr 14 2014 .
drwxr-xr-x 26 root root 24576 Apr 14 2014 ..
...
-rw-r--r-- 1 root root 18965158 Feb 27 2014 libpython2.7-pic.a
lrwxrwxrwx 1 root root 23 Feb 27 2014 libpython2.7.so -> ../../libpython2.7.so.1
...
This is a symlink that can not be symlinked to /vagrant
share, because it is mounted from Windows which doesn't support this.
virtualenv 1.11.6
doesn't fix this. --always-copy
was added in 1.10
, so we need to upgrade virtualenv anyway.
@techtonik If you want help ping @joeyespo. He's into Vagrant + Windows.
Don't need it yet, but I'll keep that in mind. Thanks. The only help needed right now is to merge some virtualenv
pull requests.
Filled (pypa/virtualenv#663). Going to add --always-copy
for now and upgrade virtualenv to 1.11.6 to see how it goes.
Didn't get far. Hit mitchellh/vagrant#4745 on the way. Shot myself in the foot with this one, actually.
Next fail with virtualenv 1.11.6:
...
==> default: Cannot find a wheel for setuptools
==> default: Cannot find a wheel for pip
==> default: Installing setuptools, pip...
==> default: Complete output from command /vagrant/env/bin/python2.7 -c "import sys, pip; sys...d\"] + sys.argv[1:]))" setuptools pip:
==> default: Traceback (most recent call last):
==> default: File "<string>", line 1, in <module>
==> default: ImportError: No module named pip
==> default: ----------------------------------------
==> default: ...Installing setuptools, pip...done.
==> default: Traceback (most recent call last):
==> default: File "./vendor/virtualenv.py", line 2338, in <module>
==> default: main()
==> default: File "./vendor/virtualenv.py", line 824, in main
==> default: symlink=options.symlink)
==> default: File "./vendor/virtualenv.py", line 992, in create_environment
==> default: install_wheel(to_install, py_executable, search_dirs)
==> default: File "./vendor/virtualenv.py", line 960, in install_wheel
==> default: 'PIP_NO_INDEX': '1'
==> default: File "./vendor/virtualenv.py", line 902, in call_subprocess
==> default: % (cmd_desc, proc.returncode))
==> default: OSError: Command /vagrant/env/bin/python2.7 -c "import sys, pip; sys...d\"] + sys.argv[1:]))" setuptools pip failed with error code 1
==> default: make: *** [env] Error 1
Looks like pip upgrade is needed.
Downloading pip
and setuptools
wheels helped. Going to commit later.
Well, another problem:
...
File "distribute_setup.py", line 129, in _do_download
_build_egg(egg, tarball, to_dir)
File "distribute_setup.py", line 120, in _build_egg
raise IOError('Could not build the egg.')
IOError: Could not build the egg.
RE: Vagrant + Windows
Honestly, even after fixing the quirks, working within a Vagrant environment hasn't been a pleasant dev experience. It's terribly slow. And getting set up is still quite tedious (which is the whole point of Vagrant). This has kind of pushed me away from participating. I don't feel productive because of how much time it takes to get from "idea I want to test" or "issue I want to resolve" to "localhost".
Instead, I'd much rather work within a "native" Python and JavaScript dev environment since these technologies are already cross-platform. Striving for a single, cross-platform run
step (and ideally, a single, cross-platform install
step too) is a much better goal than "let's fix Vagrant."
I've applied this approach to other websites (both monolithic and microservices), and it's made both onboarding and day-to-day development much smoother. Contributors can continue to use the tools and the environments that make them maximally productive. And there's no time wasted with installation hand-holding. Requiring each machine to be exactly the same in order to make high-level changes shuts out a lot of people. (As for the dev/prod-parity as a bug-catching mechanism, which usually comes up, auto-running tests in a prod-like CI tool before merging PRs has been equivalently effective.) I'd be happy to talk more about the approaches I've used to make webapps more cross-platform if there's any interest here, but unfortunately, I don't think I'll be much help with Vagrant issues anymore.
Thanks for checking in, @joeyespo! So are you saying Vagrant in general is a buzzkill for you lately, or Vagrant as we've implemented it? I think I'm hearing you say the former.
@whit537 Sure! Glad things are going strong. I also really like the rename, it's catchy :-)
Yeah, the former.
@joeyespo So what's your secret formula? Let me guess: use SQLite? ;-)
IOError: Could not build the egg.
is caused by old distribute
conflicting with new setuptools.
https://bitbucket.org/pypa/setuptools/issue/18/packages-that-use_setuptools-from-older
This one is blocked by #2886 through #2867 ATM..
@whit537 Haha, yeah, that's one option. A decent one too, since it's built-in to Python and is supported by popular ORMs.
The general idea is use cross-platform auto-installable components for dev, and production-optimized components for prod, staging, and tests running on CI.
Since Python/JS packages are already auto-installable and cross-platform, it's a natural fit to use them in place of the prod-ready counterparts. Of course, not every prod-ready component has a pure-Python implementation. In those cases, you have to substitute higher up. ORMs are a good option because the ability to plug in different backends is already built for you.
As for postgres.py, I'm not sure what the best approach would be. It highly couples your app to Postgres instead of to an object model. But I'm willing to talk through some ideas. Let's move the discussion to a new issue.
Bootstrap fails with DB error:
==> basebox: env/bin/honcho -e defaults.env,local.env run env/bin/fake_data fake_data
==> basebox: Traceback (most recent call last):
==> basebox: File "env/bin/fake_data", line 9, in <module>
==> basebox: load_entry_point('gratipay==1736.post27', 'console_scripts', 'fake_data')()
...
==> basebox: psycopg2.IntegrityError: new row for relation "transfers" violates check constraint "positive"
==> basebox: DETAIL: Failing row contains (578395790, 2014-08-20 12:00:37+00, Dominicbdp, Katharina92q, 0.00, tip).
==> basebox: make:
==> basebox: *** [data] Error 1
Moved to #2950
Please test new vagrant up
and create tickets for any issues you may run into. I think now it is good on Windows.
No idea where to even start with this one. Fresh clone with fresh vagrant box.
Here's the weird part, at home my Windows 8.1 x64 computer experiences this problem but my Windows 7 x64 computer at work doesn't. This also happens to @abnor