The traceback I get when running an r1609-compiled instance of my code
is:

Process SendeventProcess-1:
Traceback (most recent call last):
  File "C:\Users\b\p\tst\d\pyinstaller-trunk-r1609\gui\build
\pyi.win32\gui\outPY
Z1.pyz\multiprocessing.process", line 232, in _bootstrap
  File "utils.pyx", line 2855, in utils.SendeventProcess.run (utils.c:
57485)
  File "C:\Users\b\p\tst\d\pyinstaller-trunk-r1609\gui\build
\pyi.win32\gui\outPY
Z1.pyz\multiprocessing.queues", line 75, in put
WindowsError: [Error 5] Access is denied

Line 2855 in utils.pyx reads:
self.resultsQueue.put((o, e))

SendeventProcess is a multiprocessing.Process.

I tried building the code using both hook-cv.py and, alternately,
"import numpy". With hook-cv.py, I saw the above traceback, as I did
with "import numpy", but I also, on other occasions, saw:
Traceback (most recent call last):
  File "C:\Users\b\p\tst\d\pyinstaller-trunk-r1609\gui\build
\pyi.win32\gui\outPY
Z1.pyz\multiprocessing.queues", line 238, in _feed
IOError: [Errno 6] The handle is invalid

and on other times I didn't see either of the two errors, but an
error, I don't know what, was caught within the code. My guess is that
I would have also seen these other two errors with the "import numpy"
approach had I run it enough times. The code invariably fails with
r1609 regardless of whether hook-cv.py or "import cv" is used.

To compile the code with r1609, I call:
python pyinstaller.py --onedir --icon=path/to/icon1.ico <moduleString>

I don't see any errors when running the code with 1.5.1 on Windows 7.
(Remember, it was broken with 1.5.1 on XP). To get the code to run
with 1.5.1, I add the line:
import numpy.core.multiarray
and the "import numpy" line was removed. That is the only difference
in the source code.

To compile it with 1.5.1, I call:
python Configure.py
python Makespec.py --onedir --icon=path/to/icon1.ico <moduleString>
python Build.py gui/gui.spec

I do have multiprocessing.freeze_support() at the very start of my
code, but I don't have any multiprocessing.Manager's.


On Oct 5, 6:04 pm, Martin Zibricky <[email protected]> wrote:
> Brian píše v St 05. 10. 2011 v 15:04 -0700:
>
>
>
> > Would hook-cv.py currently look like this?:
>
> > hiddenimports = ["numpy"]
>
> Yes, that's exact.
>
>
>
> > On another note, I have not researched this yet, but I've encountered
> > a regression and PyInstaller is one of the view things that has
> > changed between versions. Is it possible that a recent change could
> > result in something strange going on with multiprocessing, especially
> > multiprocessing queues? Don't research this just yet, because I have
> > yet to confirm that pyinstaller is the only variable.
>
> Are you experiencing issues like:
>
> http://www.pyinstaller.org/ticket/182http://www.pyinstaller.org/ticket/146

-- 
You received this message because you are subscribed to the Google Groups 
"PyInstaller" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/pyinstaller?hl=en.

Reply via email to