On Thu, 30 May 2013, Phil Thompson wrote:
On Thu, 30 May 2013 22:49:24 +0100 (BST), John Lee <j...@pobox.com> wrote:
Hi
Is it possible to guarantee that a C++ destructor of a QObject wrapped by
PyQt is never called?
[...]
If I can't guarantee that from PyQt, can I write a simple sip wrapper that
will achieve the same end, without copying and pasting the Qt code?
sip.setdestroyonexit() might do what you want.
Thanks, that sounds promising. It does trigger some followup questions:
If I call sip.setdestroyonexit(False) before I construct any QProcess
Python instances, can ~QProcess still be called for any reason, even if I
ensure that:
1. no Qt code deletes the C++ object, and
2. my code does not call sip.delete(), and
3. I create a module-global reference (my module) to the wrapping
QtCore.QProcess Python instance which I never explicitly unbind (but for
all I know might get unbound by some part of the PyQt/sip/Python shutdown
process)?
Other than the usual QObject parent/child hierarchy, are there any cases
where Qt deletes things, that might surprise me?
Since setdestroyonexit affects all cleanup and not only that of QProcess:
Are there other notable side effects that you or others know of that I
might notice / be surprised by from calling sip.setdestroyonexit(False),
as a result of the lack of the usual cleanup associated with particular Qt
classses? In other words, what cases are known / can be guessed at where
my program might start causing a leak of OS resources like temporary
files, or badly terminated network connections, etc?
John
_______________________________________________
PyQt mailing list PyQt@riverbankcomputing.com
http://www.riverbankcomputing.com/mailman/listinfo/pyqt
