On 03/04/13 01:17, Phil Thompson wrote:
On Wed, 03 Apr 2013 00:16:49 +0100, Baz Walter <baz...@ftml.net> wrote:
Hello
For Qt4, QtCore.qInstallMsgHandler installs a Qt message handler with
the following signature:
void myMsgHandler(QtMsgType, const char *);
Using PyQt4 with Qt4, this correctly results in the second argument
being passed as a python bytes object (for both the v1 and v2 apis).
However, for Qt5 it only does this when using Python3 with the v2 api.
For Python3 with the v1 api, it passes a QString, and for Python2 it
either passes a QString (v1 api) or a python unicode object (v2 api).
I realize that QtCore.qInstallMsgHandler has been deprecated in Qt5, but
it would still be great if this could be fixed at some point.
Simple test script:
import sys, sip
sip.setapi('QString', int(sys.argv[1]) if sys.argv[1:] else 1)
from PyQt4 import QtGui, QtCore
app = QtGui.QApplication([])
def handler(kind, msg):
print('API v%s: %s, %s' % (
sip.getapi('QString'), type(kind), type(msg)))
QtCore.qInstallMsgHandler(handler)
app.desktop().screenGeometry(None)
As the Qt5 message handler is passed a QString rather than the Qt4 char *
I think the current behaviour is the right one. Qt5 is not Qt4 so PyQt4
built against Qt5 is not going to have the same API as PyQt4 built against
Qt4.
No, the Qt5 message handler is also passed a const char *.
There are two versions of the Qt message handler in Qt5:
void myMessageHandler(QtMsgType, const QMessageLogContext &, const
QString &);
and
void myMsgHandler(QtMsgType, const char *);
The second one is deprecated, and I would argue that there is not much
point in supporting transitional apis like this if they don't match the
original signature. If they do match, it makes it a lot easier to
maintain a single (transitional) PyQt4 code-base that runs on both Qt4
and Qt5.
But in any case, PyQt's current treatment of this api is inconsistent:
for Qt5, it passes a bytes object with the Python3 v2 api, and otherwise
passes either a QString or unicode object. So there is something amiss
with the current behaviour, whichever way you look at it.
--
Regards
Baz Walter
_______________________________________________
PyQt mailing list PyQt@riverbankcomputing.com
http://www.riverbankcomputing.com/mailman/listinfo/pyqt
