How to fix Python ValueError:bad marshal data?

Question

Running flexget Python script in Ubuntu, I get an error:

$ flexget series forget "Orange is the new black" s03e01
Traceback (most recent call last):
File "/usr/local/bin/flexget", line 7, in <module>
from flexget import main
File "/usr/local/lib/python2.7/dist-packages/flexget/__init__.py", line 11, in <module>
from flexget.manager import Manager
File "/usr/local/lib/python2.7/dist-packages/flexget/manager.py", line 21, in <module>
from sqlalchemy.ext.declarative import declarative_base
File "/usr/local/lib/python2.7/dist-packages/sqlalchemy/ext/declarative/__init__.py", line 8, in <module>
from .api import declarative_base, synonym_for, comparable_using, \
File "/usr/local/lib/python2.7/dist-packages/sqlalchemy/ext/declarative/api.py", line 11, in <module>
from ...orm import synonym as _orm_synonym, \
File "/usr/local/lib/python2.7/dist-packages/sqlalchemy/orm/__init__.py", line 17, in <module>
from .mapper import (
File "/usr/local/lib/python2.7/dist-packages/sqlalchemy/orm/mapper.py", line 27, in <module>
from . import properties
ValueError: bad marshal data (unknown type code)

Answer 1

If the accepted answer does not work, try the following steps:

1. Read the stack trace from the bottom to the top to find the first installed package causing the issue.
2. Uninstall and then reinstall the problematic package.

For example, in the stack trace below:

$ flexget series forget "Orange is the new black" s03e01
Traceback (most recent call last):
  File "/usr/local/bin/flexget", line 7, in <module>
    from flexget import main
  File "/usr/local/lib/python2.7/dist-packages/flexget/__init__.py", line 11, in <module>
    from flexget.manager import Manager
  File "/usr/local/lib/python2.7/dist-packages/flexget/manager.py", line 21, in <module>
    from sqlalchemy.ext.declarative import declarative_base
  File "/usr/local/lib/python2.7/dist-packages/sqlalchemy/ext/declarative/__init__.py", line 8, in <module>
    from .api import declarative_base, synonym_for, comparable_using, \
  File "/usr/local/lib/python2.7/dist-packages/sqlalchemy/ext/declarative/api.py", line 11, in <module>
    from ...orm import synonym as _orm_synonym, \
  File "/usr/local/lib/python2.7/dist-packages/sqlalchemy/orm/__init__.py", line 17, in <module>
    from .mapper import (
  File "/usr/local/lib/python2.7/dist-packages/sqlalchemy/orm/mapper.py", line 27, in <module>
    from . import properties
ValueError: bad marshal data (unknown type code)

The first problematic package appears to be sqlalchemy in from sqlalchemy.ext.declarative import declarative_base, as all subsequent imports are relative that happen within sqlalchemy package

Answer 2

I had the same error in a conda environment which traced back to importing the matplotlib package.

simply pip uninstall matplotlib and then pip install matplotlib solved the problem.

Answer 3

It may be because of the damage of the library. Try re-install the package.

Answer 4

I get this error in Ubuntu 18.04 Raspberry Pi 3 when I trying update my system typing sudo apt-get update and solve this error just typing:

sudo find /usr -name '*.pyc' -delete

This is remove all .pyc file in my system. Now I typing again sudo apt-get update && sudo apt-get upgrade and I get my update without thie error marshal-data

Answer 5

I solved this problem by the following procedure :

In the error code message, you can see from sqlalchemy.ext.declarative import declarative_base cause this error.

So just pip uninstall sqlalchemy and pip install sqlalchemy, problem solved.

Answer 6

I also got this problem in Windows environment(win 10).

I fixed it by going to the Settings and repairing Python 3.7 with its installer. Everything works fine since then.

As far as I could recall, I had kept a dash server running when my computer went to hibernation. Maybe the damage was done in the hibernating process somehow.

Answer 7

There also appears to have been some sort of regression in setuptools with use with python 3.7. See for an example - https://github.com/pypa/setuptools/issues/1257

Forcing reinstallation of setuptools fixed this issue for me.

sudo pip3 install --upgrade --force-reinstall setuptools

Answer 8

I resolved a similar error by un-installing and re-installing the Python application I was using, and all dependencies, using the system package manager.

In my case I was using awscli on Debian 9 and the error was "ValueError: bad marshal data (set size out of range)".

I ran as root:

apt-get purge awscli
apt-get autoremove
apt-get install awscli

And then the error was fixed.

I could imagine cases where the broken package might not get removed (for example because it was marked as manually installed, or was a dependency of another application still installed), in those cases this action may not resolve the error. However I thought I should try this way before manually deleting .pyc files the system installed, and I got lucky.

Answer 9

This can happen if you have Python 2.7 .pyc files and you try to load them using Python 3.5. In my case this was a third-party tarball that erroneously included pre-compiled Python 2.7 .pyc files along with the source code.

Answer 10

Just delete

/usr/local/lib/python2.7/dist-packages/sqlalchemy/orm/properties.pyc

it is corrupt as the text indicates. You'll probably have to do so as root.

After that start (again as root) run python (2.7):

/usr/bin/python -c "import sqlalchemy.orm.properties"

to recreate this .pyc file.

If you don't recreate the .pyc file, your program starts slower than necessary as the .py file takes longer to load than the .pyc (and a normal user cannot write the .pyc file).

Answer 11

If you get that error, the compiled version of the Python module (the .pyc file) is corrupt probably. Gentoo Linux provides python-updater, but in Debian the easier way to fix: just delete the .pyc file. If you don't know the pyc, just delete all of them (as root):

find /usr -name '*.pyc' -delete