Fix #5687: Don't pass unicode to QApplication()

[pankajp]

PySide QApplication constructor cannot handle unicode text
(segfaults), so we instead pass an empty list to the constructor.

Fixes #5687

[jdmarch]

Confirmed fixes segfault on Mac and Linux

[tacaswell]

Thing change went in in 378e37d so this is definitily a regression from 1.4.

Sorry, I could have sworn I tested this against PySide when this went in.

attn @mfitzp

[tacaswell]

tracking down what is going on here:

http://doc.qt.io/qt-5/qapplication.html#QApplication is underlying c++, says must be at least one as does http://pyqt.sourceforge.net/Docs/PyQt4/qapplication.html#QApplication where as pyside does not explicitly note this https://srinikom.github.io/pyside-docs/PySide/QtGui/QApplication.html#PySide.QtGui.QApplication

[pankajp]

@tacaswell Thanks for pointing to the documentation which states there must be atleast one string in the list. I was not aware of that. In fact i tested using a empty list and it works fine in all combinations of PySide/PyQt4/PyQt5 X python2/python3
I guess I should just change it back to a normal str with space then as it was before that commit.

[pankajp]

I have reverted the QApplication() args to the way it was before commit 378e37d broke it.

[tacaswell]

If an empty list works, I would assume the documentation is wrong 😈

Is this only a pyside + LPy issue? I think it would be better to have conditional code to special case that rather than have incorrect unicode handling.

[pankajp]

I don't think passing string (in py2) is incorrect behavior because the underlying C++ method accepts char**, and also because passing unicode segfaults pyside. It was changed to unicode in a misguided attempt to unify text type. Using unicode here has no correctness benefit simply because we are passing in a constant, and not text which can be set by user and/or the OS.

[tacaswell]

Can you also file a bug/is there a bug upstream with pyside?

I guess this is ok. The path to dropping py2 is going to include a sed s/six.text_type/str/g operation anyway and as you point out we do not take user input here so we can not be surprised by unicode input.

Test failures look un-related (and my fault 😞 )

[tacaswell]

@pankajp Can you please follow up with upstream (pyside) on this?

[pankajp]

Reported upstream: https://bugreports.qt.io/browse/PYSIDE-306. Should be easy to fix if someone gets the time.

[tacaswell]

Thanks !

On Tue, Dec 22, 2015, 00:11 Pankaj Pandey notifications@github.com wrote:

Reported upstream: https://bugreports.qt.io/browse/PYSIDE-306. Should be
easy to fix if someone gets the time.

—
Reply to this email directly or view it on GitHub
#5690 (comment)
.

[tacaswell]

Backported as 1241122