"--user" fixes #511

qwcode · 2012-04-16T07:17:31Z

Pull request to fix various "--user" issues

493: pip shouldn't allow --user installs in virtualenvs (w/o --system-site-packages)
440: pip install --user tries to uninstall system package
494: can't uninstall --user "develop" install
495: user site should be "local" in virtualenvs with --system-site-packages (mostly so we can test)

Tests are passing for py25, py26, py27, and py32 (against rebased branch from 4/15)

I'm happy to trim out or alter parts that may cause concern in the interest of getting as many of these issues resolved.

As mentioned to Carl over email, I'm concerned about the level of testing we can achieve for the "--user" feature, given that the test framework uses virtualenvs, and that full coverage would require real python installs.

I'd like to still add some mock/patch tests for the significantly altered util.egg_link_path, and more tests in general.

carljm · 2012-04-16T22:31:55Z

pip/commands/install.py

@@ -184,6 +185,8 @@ def run(self, options, args):
        options.src_dir = os.path.abspath(options.src_dir)
        install_options = options.install_options or []
        if options.use_user_site:
+            if running_under_virtualenv() and virtualenv_no_global():


why check both here, since virtualenv_no_global implies running_under_virtualenv?

early on in the commit history, "virtualenv_no_global" didn't imply "running_under_virtualenv". the former was just checking for the global file. later I changed "virtualenv_no_global", but didn't update these if statments.
I can change this.

well, even when "virtualenv_no_global" was only checking for the existence of a "no-global-site-packages.txt" in the right place, that's still pretty conclusive. not sure really why I ever had both checks in there : )

carljm · 2012-04-16T22:54:33Z

Thanks for working on this, generally looks great! More tests would be excellent, though I'd probably merge it with the current level of testing, pending comments above.

qwcode · 2012-04-17T05:10:41Z

ok, cool, I'll post some updates in a day or two.

qwcode · 2012-05-05T17:16:42Z

I think i've addressed everything that's been mentioned

questions:

whether we should "fix" issue 495 (see comments in that issue). in this branch, it's "fixed"
whether to use my new "path_in_dir" function or just to fall back to the old code

next weekend (when I have some more time) I plan on getting my environment straight once and for all, so I can pass all the tests.
currently, I always have some stragglers that fail. I'll feel more confident after that.

tests I've been running:
--linux: py24, py25, py26, py27, py32
--windows: py27

pnasrat · 2012-05-13T19:55:59Z

pip/backwardcompat.py

@@ -112,6 +113,14 @@ def fwrite(f, s):

 from distutils.sysconfig import get_python_lib, get_python_version

+def get_user_site():
+    "return site.USER_SITE or None if not implemented in earlier version of python"


Nit we should follow http://www.python.org/dev/peps/pep-0257/#one-line-docstrings and make it a sentence ie capitalize Return and end in a period.

pnasrat · 2012-05-13T19:58:30Z

lgtm - if you're happy with your testing

qwcode · 2012-05-13T21:15:18Z

ok, I worked on my environments, and ran tests again for
linux: py24, py25, py26, py27, py32
windows: py27

for linux testing, down to 2 unrelated failures:

1 bz2 test failing for my non-py26 installs, due to it not being compiled in those py versions
test_vcs_backends.test_git_with_tag_name_and_update fails across the board due to my git version having a different format for that kind of error

on windows:

various failures related to my svn installation that I didn't take the time to sort out. I'm pretty out of touch with svn config on windows.
a few failures because I don't have bzr installed

pnasrat · 2012-05-13T21:42:27Z

Github is complaining about mergability. I can try merge and thus cause a travis CI run on my personal develop branch, or maybe we can wait a little longer and get the test on pull request working see Issue #527

qwcode · 2012-05-13T23:17:09Z

I'm willing to rebase my userscheme branch on my locally updated develop branch (and retest) and then force push userscheme to my origin. github shouldn't complain then about you merging userscheme, but I guess that would blow away all the comment references in this pull request. so probably not desired?

pnasrat · 2012-05-13T23:19:22Z

Are you on #pip on freenode? May be easier to discuss there.

Conflicts: pip/util.py

qwcode · 2012-05-14T06:27:25Z

ok, I think I did what we talked about on #pip
there's was just a minor conflict in util.py in the import lines.
tests still look good.

qwcode · 2012-05-23T00:08:52Z

Carl/Paul:

I fear the scope of this branch has gotten too large for anyone to feel comfortable merging it.
I'm willing break this up into a series of smaller pull requests that build up to the same result.
The commit history in the smaller pulls would be much cleaner a 2nd time through, now that I've sorted thru all the issues.

let me know, if you think that would help?

pnasrat · 2012-05-23T00:14:59Z

If you feel willing and more comfortable doing that - that's fine with me.

However I'm probably happy enough to merge this shortly - I was hoping to fix #530 first so we have a good green build to depend on.

qwcode · 2012-05-23T00:37:21Z

hmm...It's hard to decide. I know I'm more making more work for myself. : )
I'm inclined to be really conservative knowing pip is so widely used, and honestly I'm pretty new to contributing to a major project like this.
A series of smaller pulls with clear/concise commit histories would probably serve the history of the project better.
I have time this weekend, so I could start on it.

OTOH, if that sounds like more work for you, I understand.

pnasrat · 2012-05-23T00:38:55Z

That works for me, thanks for all your work on this! I'm travelling back home this weekend so I wouldn't expect responsiveness, but I'll try address next week.

carljm · 2012-05-23T02:43:37Z

I'm with Paul - I wouldn't have asked you to do that extra work, but I do find more smaller pull requests easier to deal with when possible.

qwcode · 2012-05-29T04:32:40Z

here's my plan for breaking this up
https://gist.github.com/2822510

I'll reference the plan when I do requests.

qwcode · 2012-07-08T22:43:12Z

closing, as this request is being handled in smaller parts:
https://gist.github.com/2822510

qwcode added 12 commits April 15, 2012 16:36

--user option fixes

dcdc338

work around pypi_intercept.pth early init of pkg_resources.working_set

da81a5c

user installs should be considered 'local'

cfee18f

remove abbreviations from test suite

3d59a7d

handle when user_base() is None

564f88e

fall back to using hack to create virtualenv with system site pkgs

c86dde0

add system_site_packages=True to old python test

ba09c5e

handle when develop_egg_link is None

7ad7eb2

don't look in original site-pkgs for egg link even when on sys.path

90c29c2

make virtualenv_no_global() more explicit

c49b9df

spaces, comments, and another --user uninstall test

55cdb9c

comment and put back test that didn't migrated

e5a5ff2

carljm reviewed Apr 16, 2012
View reviewed changes

qwcode added 4 commits April 19, 2012 22:14

remove unnecessary condition; better error message

5ffdaf0

remove unnecessary conditions

fd4a9e0

is_local() was not handling the case of a no-global virtualenv correctly

d4ffdad

updated error message in test

30aeb68

pnasrat reviewed May 13, 2012
View reviewed changes

pep 257 docstrings

310d770

Merge branch 'develop' into userscheme (pull request #511)

94c9ac0

Conflicts: pip/util.py

carljm mentioned this pull request May 15, 2012

pip should be able to uninstall --user #535

Closed

This was referenced May 31, 2012

--user fixes part 1 #548

Merged

--user fixes part 2 #553

Merged

pnasrat mentioned this pull request Jun 2, 2012

Added #subdirectory tag specify a relative subdirectory inside a repo… #526

Closed

This was referenced Jun 6, 2012

--user fixes pt 3 Issue #493 #567

Merged

--user fixes part 4, issue #440 #574

Merged

qwcode mentioned this pull request Jun 13, 2012

--user fixes part5: reset_env option for adding patches to sitecustomize #576

Merged

qwcode mentioned this pull request Jun 30, 2012

--user fixes part 6 #573 #585

Merged

qwcode closed this Jul 8, 2012

pnasrat mentioned this pull request Nov 12, 2012

Rework command handling v2 #721

Merged

lock bot added the auto-locked Outdated issues that have been locked by automation label Jun 5, 2019

lock bot locked as resolved and limited conversation to collaborators Jun 5, 2019

"--user" fixes #511

"--user" fixes #511

Uh oh!

Conversation

qwcode commented Apr 16, 2012

Uh oh!

carljm Apr 16, 2012

Choose a reason for hiding this comment

Uh oh!

qwcode Apr 17, 2012

Choose a reason for hiding this comment

Uh oh!

qwcode Apr 17, 2012

Choose a reason for hiding this comment

Uh oh!

carljm commented Apr 16, 2012

Uh oh!

qwcode commented Apr 17, 2012

Uh oh!

qwcode commented May 5, 2012

Uh oh!

pnasrat May 13, 2012

Choose a reason for hiding this comment

Uh oh!

pnasrat commented May 13, 2012

Uh oh!

qwcode commented May 13, 2012

Uh oh!

pnasrat commented May 13, 2012

Uh oh!

qwcode commented May 13, 2012

Uh oh!

pnasrat commented May 13, 2012

Uh oh!

qwcode commented May 14, 2012

Uh oh!

qwcode commented May 23, 2012

Uh oh!

pnasrat commented May 23, 2012

Uh oh!

qwcode commented May 23, 2012

Uh oh!

pnasrat commented May 23, 2012

Uh oh!

carljm commented May 23, 2012

Uh oh!

qwcode commented May 29, 2012

Uh oh!

qwcode commented Jul 8, 2012

Uh oh!

Uh oh!