On 11/29/12 4:06 PM, Nik Melchior wrote: > On Mon, Nov 26, 2012 at 01:49:37PM -0500, John Ehresman wrote: >> On 11/15/12 5:52 PM, John Ehresman wrote: >>> I'm trying to fix a bug that causes segfaults if QApplication.style() and >>> QApplication.setStyle() are used repeatedly. I think the issue is: > > John, > > My application has random crashes due to QStyle objects as well. I had some > trouble tracking down the problem, and I originally thought it was related to > a bug that reported QNetworkReply objects being randomly returned. I added my > analysis to PySide bug #16, but I haven't seen any activity there.
I think this a symptom of the same object lifecycle bug -- PySide doesn't always detect when a C++ object is no longer valid and when that happens a bad entry gets left in the hash table mapping C++ object addresses to Python wrapper objects. When another C++ object is created at the same address, PySide will find the bad entry in the hash table and reuse the old wrapper. A fix went into shiboken in June that fixes a few cases that can lead to this with event objects, but I don't know if there's been a release since then. Your bug with eventFilters sounds a lot like a bug I was seeing and the shiboken fix did seem to address them. I also think this problem might be solved more generically; I have a fix for QObject derived classes that I think might work. Other classes will need a different approach, but I do think something can be done for them. Cheers, John _______________________________________________ PySide mailing list [email protected] http://lists.qt-project.org/mailman/listinfo/pyside
