Infinite loop on pip when lockfile can't acquire a lock

[minchinweb]

• Pip version: 8.0.3
• Python version: 3.5.1
• Operating System: Windows 10x64

Description:

• running pip list -o with Sphinx 1.3.6 (the current version) causes pip to go into some sort of infinite loop. Nothing is printed to the console.
• trying to install Sphinx 1.3.6 from PyPI start an infinite loop and the package is never installed.
• tyring to update Sphinx (pip install sphinx -U) starts an infinite loop and nothing seems to happen
• Sphinx can be installed from a downloaded wheel (from PyPI) without issue.
• other pip commands seem to work without issue.
• the issue doesn't seem to be limited to Sphinx 1.3.6, but I can't seem to narrow that down any further

What I've run:

[xavfernandez]

@minchinweb Does adding --no-cache-dir to pip list -o --no-cache-dir avoid the issue ?

[minchinweb]

@xavfernandez , yes that seems to solve it. both pip list -o --no-cache-dir and pip install sphinx -U --no-cache-dir work

[xavfernandez]

So this looks like a duplicate of #3245.

Could you investigate with a step-by-step debugger to check where pip is hanging ?

[minchinweb]

Running with verbose mode on (in a new virtualenv) provides the following:

> pip install sphinx -vvv

Collecting sphinx
  1 location(s) to search for versions of sphinx:
  * https://pypi.python.org/simple/sphinx/
  Getting page https://pypi.python.org/simple/sphinx/
  Looking up "https://pypi.python.org/simple/sphinx/" in the cache
  No cache entry available
  Starting new HTTPS connection (1): pypi.python.org
  "GET /simple/sphinx/ HTTP/1.1" 200 5450
  Updating cache with response from "https://pypi.python.org/simple/sphinx/"
  Caching b/c date exists and max-age > 0

[ it stalls here, so I end it with CTRL+C ]

Cleaning up...

Operation cancelled by user

Exception information:
Traceback (most recent call last):
  File "c:\tmp\env\lib\site-packages\pip\_vendor\lockfile\mkdirlockfile.py", line 40, in acquire
    os.mkdir(self.lock_file)
FileExistsError: [WinError 183] Cannot create a file when that file already exists: 'c:\\users\\william\\appdata\\local\\pip\\cache\\http\\6\\3\\c\\4\\2\\63c426d150091b9663b3d4923999b229a56ede976ab93677382d9562.lock'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "c:\tmp\env\lib\site-packages\pip\basecommand.py", line 209, in main
    status = self.run(options, args)
  File "c:\tmp\env\lib\site-packages\pip\commands\install.py", line 310, in run
    wb.build(autobuilding=True)
  File "c:\tmp\env\lib\site-packages\pip\wheel.py", line 747, in build
    self.requirement_set.prepare_files(self.finder)
  File "c:\tmp\env\lib\site-packages\pip\req\req_set.py", line 359, in prepare_files
    ignore_dependencies=self.ignore_dependencies))
  File "c:\tmp\env\lib\site-packages\pip\req\req_set.py", line 511, in _prepare_file
    finder, self.upgrade, require_hashes)
  File "c:\tmp\env\lib\site-packages\pip\req\req_install.py", line 277, in populate_link
    self.link = finder.find_requirement(self, upgrade)
  File "c:\tmp\env\lib\site-packages\pip\index.py", line 436, in find_requirement
    all_candidates = self.find_all_candidates(req.name)
  File "c:\tmp\env\lib\site-packages\pip\index.py", line 394, in find_all_candidates
    for page in self._get_pages(url_locations, project_name):
  File "c:\tmp\env\lib\site-packages\pip\index.py", line 539, in _get_pages
    page = self._get_page(location)
  File "c:\tmp\env\lib\site-packages\pip\index.py", line 642, in _get_page
    return HTMLPage.get_page(link, session=self.session)
  File "c:\tmp\env\lib\site-packages\pip\index.py", line 751, in get_page
    "Cache-Control": "max-age=600",
  File "c:\tmp\env\lib\site-packages\pip\_vendor\requests\sessions.py", line 480, in get
    return self.request('GET', url, **kwargs)
  File "c:\tmp\env\lib\site-packages\pip\download.py", line 377, in request
    return super(PipSession, self).request(method, url, *args, **kwargs)
  File "c:\tmp\env\lib\site-packages\pip\_vendor\requests\sessions.py", line 468, in request
    resp = self.send(prep, **send_kwargs)
  File "c:\tmp\env\lib\site-packages\pip\_vendor\requests\sessions.py", line 608, in send
    r.content
  File "c:\tmp\env\lib\site-packages\pip\_vendor\requests\models.py", line 737, in content
    self._content = bytes().join(self.iter_content(CONTENT_CHUNK_SIZE)) or bytes()
  File "c:\tmp\env\lib\site-packages\pip\_vendor\requests\models.py", line 660, in generate
    for chunk in self.raw.stream(chunk_size, decode_content=True):
  File "c:\tmp\env\lib\site-packages\pip\_vendor\requests\packages\urllib3\response.py", line 344, in stream
    data = self.read(amt=amt, decode_content=decode_content)
  File "c:\tmp\env\lib\site-packages\pip\_vendor\requests\packages\urllib3\response.py", line 301, in read
    data = self._fp.read(amt)
  File "c:\tmp\env\lib\site-packages\pip\_vendor\cachecontrol\filewrapper.py", line 54, in read
    self.__callback(self.__buf.getvalue())
  File "c:\tmp\env\lib\site-packages\pip\_vendor\cachecontrol\controller.py", line 297, in cache_response
    self.serializer.dumps(request, response, body=body),
  File "c:\tmp\env\lib\site-packages\pip\download.py", line 280, in set
    return super(SafeFileCache, self).set(*args, **kwargs)
  File "c:\tmp\env\lib\site-packages\pip\_vendor\cachecontrol\caches\file_cache.py", line 99, in set
    with self.lock_class(name) as lock:
  File "c:\tmp\env\lib\site-packages\pip\_vendor\lockfile\__init__.py", line 197, in __enter__
    self.acquire()
  File "c:\tmp\env\lib\site-packages\pip\_vendor\lockfile\mkdirlockfile.py", line 57, in acquire
    time.sleep(wait)
KeyboardInterrupt

So if I read that right, it can't create the lock file because one already exists, so it's waiting for the lock folder to disappear (probably to try again). My guess is this is the infinite loop it's stuck in.

Deleting the lock folder in question doesn't seem to help running processes, but does allow a new process to install Sphinx.

Update: Turns out there was another rogue lock folder (f9daaa0b90d3f26ba99d3a4afed1be31abd9eddd0acd969de27db871.lock). Deleting that allows running processes to keep going and finish as expected.

Does there need to be a limit applied to how long it will wait? Should some procedure be put in place to clear out old lock folders? Or at least tell the user which lock folder is holding up pip? As a minimum, if you break out of the pip (without using the verbose options), it should tell you what lock folder it's hung up on.

[minchinweb]

@xavfernandez : Have I provided enough information to fix the issue?

[alanhamlett]

Also seeing this infinite loop issue with pip v8.1.1 after installing any package.

[Reston]

I had the same issue so to bypass it I had to delete the cache .lock of pip as @minchinweb has said or I had to put a return after line 56 in mkdirlockfile.py to avoid the time.sleep function

[jatindhankhar]

Hi,
Sorry for interrupting
I had the similar issue with locking (not same but related) with both pip2 and pip3 on Arch Linux
pip --version [22:27:05] pip 8.1.2 from /usr/lib/python3.5/site-packages (python 3.5)
pip2 --version [22:27:11] pip 8.1.2 from /usr/lib/python2.7/site-packages (python 2.7)

and the same solution as done by @Reston, fixes the issues. Before this every pip install got stuck after the successfully installed message

[ColdHeat]

I'm running into this issue as well and it basically boils down to mkdirlockfile getting stuck because the timeout is None.

You get stuck hitting the time.sleep() and while loop endlessly here:
https://github.com/openstack/pylockfile/blob/master/lockfile/mkdirlockfile.py#L57

This happens to me on OSX Sierra using python and pip installed via brew.

❯ pip --version
pip 9.0.1 from /usr/local/lib/python2.7/site-packages (python 2.7)

[ColdHeat]

Perhaps the simplest is to add self.release() after time.sleep(wait) on line 58 of mkdirlockfile.py