Discussion:
[issue25585] Bad path leads to: ImportError: DLL load failed: %1 is not a valid Win32 application.
Laura Creighton
2015-11-08 17:36:01 UTC
Permalink
New submission from Laura Creighton:

Somebody reported this today:

File "C:/Python27/kivyhello.py", line 4, in <module>
from kivy.app import App
File "C:/Python27\kivy\app.py", line 316, in <module>
from kivy.base import runTouchApp, stopTouchApp
File "C:/Python27\kivy\base.py", line 30, in <module>
from kivy.event import EventDispatcher
File "C:/Python27\kivy\event.py", line 8, in <module>
import kivy._event
ImportError: DLL load failed: %1 is not a valid Win32 application.

----------------

The problem was that they needed to set their PATH properly so kivy could be found.

Couldn't we generate a more informative error message here?
Is the %1 even correct?

----------
messages: 254349
nosy: lac
priority: normal
severity: normal
status: open
title: Bad path leads to: ImportError: DLL load failed: %1 is not a valid Win32 application.
versions: Python 2.7

_______________________________________
Python tracker <***@bugs.python.org>
<http://bugs.python.org/issue25585>
_______________________________________
SilentGhost
2015-11-08 17:40:23 UTC
Permalink
Changes by SilentGhost <***@gmail.com>:


----------
components: +Windows
nosy: +paul.moore, steve.dower, tim.golden, zach.ware

_______________________________________
Python tracker <***@bugs.python.org>
<http://bugs.python.org/issue25585>
_______________________________________
R. David Murray
2015-11-08 17:50:21 UTC
Permalink
R. David Murray added the comment:

I'm guessing that's the error we got from the OS. Maybe there's additional info we could add, though, I don't know. It would be interesting to know if it does the same thing in python3...and unfortunately it isn't as easy to change things in the python2 import system as it is in python3.

----------
nosy: +r.david.murray

_______________________________________
Python tracker <***@bugs.python.org>
<http://bugs.python.org/issue25585>
_______________________________________
eryksun
2015-11-09 06:02:01 UTC
Permalink
eryksun added the comment:

The "DLL load failed" message is from Python, but the rest is the text for the Windows error code, ERROR_BAD_EXE_FORMAT (0x00c1) [1]. The "%1" in the string can be expanded as the first element of the Arguments array parameter of FormatMessage [2]. But currently the code in Python/dynload_win.c uses FORMAT_MESSAGE_IGNORE_INSERTS and does no post-processing to replace the "%1".

I don't know why the Windows loader reported ERROR_BAD_EXE_FORMAT instead of ERROR_MOD_NOT_FOUND. Possibly it found another version of a dependent DLL that was corrupt or for a different architecture.

Note that the setup in this case is odd in that the package is installed in C:\Python27 instead of in the site-packages directory.

[1]: https://msdn.microsoft.com/en-us/library/ms681382#ERROR_BAD_EXE_FORMAT
[2]: https://msdn.microsoft.com/en-us/library/ms679351

----------
nosy: +eryksun

_______________________________________
Python tracker <***@bugs.python.org>
<http://bugs.python.org/issue25585>
_______________________________________
R. David Murray
2015-11-09 16:07:38 UTC
Permalink
R. David Murray added the comment:

OK, so it sounds like there are improvements we could make here if someone wants to work on it.

----------

_______________________________________
Python tracker <***@bugs.python.org>
<http://bugs.python.org/issue25585>
_______________________________________
STINNER Victor
2015-11-09 16:19:10 UTC
Permalink
STINNER Victor added the comment:

Does Windows provide a function to format a string using %1, %2, etc.?

Note: Python 3.6 uses the same code than Python 2.7.

----------
nosy: +haypo

_______________________________________
Python tracker <***@bugs.python.org>
<http://bugs.python.org/issue25585>
_______________________________________
Steve Dower
2015-11-09 17:12:09 UTC
Permalink
Steve Dower added the comment:

It does, but you need to identify the error code precisely and use that to provide the parameters when obtaining the message. It's more about localization than a general way of obtaining error text. Better off just using our own message in this case (which is probably a breaking change for 2.7...)

----------

_______________________________________
Python tracker <***@bugs.python.org>
<http://bugs.python.org/issue25585>
_______________________________________
eryksun
2015-11-09 21:15:08 UTC
Permalink
eryksun added the comment:

Steve, do you think it's OK to abandon localization for exception messages? If so, should we be using English for all FormatMessage calls? It's a bit ugly that Python's exceptions and the CRT error messages are in English, but then whenever we call FormatMessage with LANG_NEUTRAL/SUBLANG_DEFAULT we're getting localized error text. For example, the import error in the following case has a mix or Russian and English:

import io, sys, ctypes
kernel32 = ctypes.WinDLL('kernel32')

MUI_LANGUAGE_NAME = 8
kernel32.SetThreadPreferredUILanguages(MUI_LANGUAGE_NAME, u'ru-RU\0', None)
kernel32.SetConsoleOutputCP(1251)
sys.stderr = io.TextIOWrapper(open(r'\\.\con', 'wb'), encoding='1251')
open('blah.pyd', 'w').close()
import blah
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ImportError: DLL load failed: %1 не является приложением Win32.

----------

_______________________________________
Python tracker <***@bugs.python.org>
<http://bugs.python.org/issue25585>
_______________________________________
Steve Dower
2015-11-10 04:29:11 UTC
Permalink
Steve Dower added the comment:

For 3.6 I think it's fine, but people are almost certainly relying on the current message on their system and changing it at all for any existing release is unnecessary pain (3.5.1 *might* be okay, but only because it's so recent).

It would be nice to invest in some better diagnostics here, though with the ABI tag in 3.5 you're now unlikely to get an incompatible binary. File not found will be the most common message.

----------

_______________________________________
Python tracker <***@bugs.python.org>
<http://bugs.python.org/issue25585>
_______________________________________
Steve Dower
2015-11-10 04:30:15 UTC
Permalink
Steve Dower added the comment:

Correction, *import* not found, and the only way to resolve it is to do a second scan of sys.path for the sole purpose of diagnostics. Probably more harmful than helpful.

----------

_______________________________________
Python tracker <***@bugs.python.org>
<http://bugs.python.org/issue25585>
_______________________________________

Loading...