python:3.12.0 name not defined error

[CyberQin]

scipy stats module report name 'obj' not defined

Context information (for bug reports)

• Output of pyinstaller --version: 6.0.0

• Version of Python: 3.12.0

• Platform: Windows (zh-CN)

• How you installed Python: conda

• Did you also try this on another platform? Does it work there?: have tried ubuntu22.04, the same error reported

  • start with clean installation
  • use the latest development version
  • Run your frozen program from a command window (shell) — instead of double-clicking on it
  • Package your program in --onedir mode
  • Package without UPX, say: use the option --noupx or set upx=False in your .spec-file
  • Repackage you application in verbose/debug mode. For this, pass the option --debug to pyi-makespec or pyinstaller or use EXE(..., debug=1, ...) in your .spec file.

A minimal example program which shows the error

from scipy.stats import weibull_min


if __name__ == '__main__':
    cc = weibull_min(c=2.5, loc=0, scale=1)
    print(cc.cdf(0.001))

Stacktrace / full error message

Traceback (most recent call last):
  File "main.pyw", line 15, in <module>
  File "<frozen importlib._bootstrap>", line 1354, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1325, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 929, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 391, in exec_module
  File "scipy\stats\__init__.py", line 608, in <module>
  File "<frozen importlib._bootstrap>", line 1354, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1325, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 929, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 391, in exec_module
  File "scipy\stats\_stats_py.py", line 46, in <module>
  File "<frozen importlib._bootstrap>", line 1354, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1325, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 929, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 391, in exec_module
  File "scipy\stats\distributions.py", line 8, in <module>
  File "<frozen importlib._bootstrap>", line 1354, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1325, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 929, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 391, in exec_module
  File "scipy\stats\_distn_infrastructure.py", line 360, in <module>
NameError: name 'obj' is not defined
[26600] Failed to execute script 'main' due to unhandled exception!

I read the source code in scipy, the 'obj' located at scipy.stats._distn_infrastructure.py line 360. 'obj' is not defined before but only in some comprehensive list.

It totally ok when python vesion=3.11, but when i change to python 3.12, it report error. I was wondering if its a python 3.12 import system problem. Seems pyinstaller + python 3.11 won't import this 'del obj', but pyinstaller + python 3.12 will do.

# clean up all the separate docstring elements, we do not need them anymore
for obj in [s for s in dir() if s.startswith('_doc_')]:
    exec('del ' + obj)
del obj

[rokm]

Hmm, this definitely looks like a bug, although I will need to dig further to see if it is ours or python's (or some weird combination of both).

[rokm]

The behavior is definitely bizarre, as demonstrated by following minimal program:

import sys

if len(sys.argv) != 2:
    print(f"usage: test <dir|locals|globals>")
    sys.exit(1)
mode = sys.argv[1]

# The comprehension must use same variable name as the code that attempts `del`.
_allvalues = ''.join([myobj for myobj in ['a', 'b', 'c']])

myobj = None  # for del below
if mode == 'dir':
    print("DIR():", dir())
elif mode == 'locals':
    print("LOCALS():", locals())
elif mode == 'globals':
    print("GLOBALS():", globals())

del myobj

The codepaths that call either dir() or locals() end up with myobj undefined when we try to delete it.

So far, all evidence points towards PyInstaller.building.utils.strip_paths_in_code being the culprit. Making that no-op makes both the above example and an import scipy.stats program work. So there's probably some incompatibility with 3.12 in there that will need to be investiagted and fixed.

[CyberQin]

https://docs.python.org/3.12/whatsnew/3.12.html#pep-709-comprehension-inlining

maybe python3.12 pep709 affect it?

Comprehension iteration variables remain isolated and don’t overwrite a variable of the same name in the outer scope, nor are they visible after the comprehension. Inlining does result in a few visible behavior changes:

There is no longer a separate frame for the comprehension in tracebacks, and tracing/profiling no longer shows the comprehension as a function call.

The [symtable](https://docs.python.org/3.12/library/symtable.html#module-symtable) module will no longer produce child symbol tables for each comprehension; instead, the comprehension’s locals will be included in the parent function’s symbol table.

Calling [locals()](https://docs.python.org/3.12/library/functions.html#locals) inside a comprehension now includes variables from outside the comprehension, and no longer includes the synthetic .0 variable for the comprehension “argument”.

A comprehension iterating directly over locals() (e.g. [k for k in locals()]) may see “RuntimeError: dictionary changed size during iteration” when run under tracing (e.g. code coverage measurement). This is the same behavior already seen in e.g. for k in locals():. To avoid the error, first create a list of keys to iterate over: keys = list(locals()); [k for k in keys].

[rokm]

Looks like a bug in python's code.replace(), which we use in strip_paths_in_code. I've opened a bug report with python.

[CyberQin]

although it's not pyinstaller's bug, but should pyinstaller treat python 3.12.0 as a not full compability py version.

[bwoodsend]

We can add a sentence saying as much in the readme but that's all we can do. It's impossible to withdraw support for a Python version once we've marked one release as supporting it. I'll pin the issue for now – that tends to keep people from raising duplicate issues.

[jeffheaton]

FYI, Running into this one as well. Only see it in Pyinstaller too, only in 3.12. Will revert Python versions.

[noahyoah]

Good day everyone!

I am reaching out to determine if anyone has found a work around for this issue. I am attempting create an executable using auto-py-to-exe and unfortunately it is failing due to 'obj' not defined.

Additionally, I believe that this same issue is happening with the module statsmodels.api as well.

Any information you provide would be greatly appreciated.

[rokm]

I am reaching out to determine if anyone has found a work around for this issue

The obvious work-around would be not to use python 3.12 until they fix code.replace(), would it not?

Other than that, you could also:

• patch your local copy of scipy and comment that attempt at removing obj (here)
• patch your local copy of PyInstaller to make strip_paths_in_code no-op, i.e., immediately return original co (if non-anonymized paths are not a problem for you)

[noahyoah]

Rokm,

I appreciate your response. I used an older version of python and was able to create the executable. Thank you for pointing me in the correct direction.

[CyberQin]

Good News, A bug fix have been merged to python:main and python:3.12 branch 9 hours ago, maybe 3.12.1 will sove this problem! python/cpython#111866 .

[CyberQin]

python 3.12.1 released , this bug has been fixed ,pls upgrade your python version to avoid this issue.

Actually pyinstaller not work well only with python=3.12.0