https://bugs.kde.org/show_bug.cgi?id=483163
--- Comment #23 from Jakob Petsovits <jpe...@petsovits.com> --- (In reply to Jakob Petsovits from comment #22) > (In reply to Jakob Petsovits from comment #21) > > And after the PAM authentication workers have started, we've got two invalid > > reads with an actually useful backtrace for which I just created > > https://bugreports.qt.io/browse/QTBUG-123878. Either I'm stupid and can't > > read code / valgrind output, or this is actually a non-trivial bug in > > central Qt code. > > Digging a little deeper, it looks like Qt is not broken and I was in the > wrong. Bwah. I wonder whether that means we have an issue in KSvg though? Alright, so I added lots of debug output and tracked down the valgrind warnings past initial app instantiation down to a memory read just slightly past the end of the SVG text that we're decompressing within KSvg. The invalid read is performed by libpcre from within QRegularExpression, reproducible without KDE code entirely, and https://bugreports.qt.io/browse/QTBUG-123892 determined that despite the valgrind warning there is no bug. So we still need to find the real culprit. On the memory side of things, there's only one more valgrind warning left apart from the above, in libxcb as used on app startup by Qt's X11 backend in QXcbConnection::initializeScreensFromMonitor(). *Maybe* that's related. But I've had two duds so far, so don't hold out your hopes too high. -- You are receiving this mail because: You are watching all bug changes.
