----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://git.reviewboard.kde.org/r/126993/ -----------------------------------------------------------
Review request for KDE Software on Mac OS X and Plasma. Repository: plasma-workspace Description ------- While looking to build DrKonqi for OS X/MacPorts I came across a few issues that appear to be syntax/coding errors that, surprisingly, were never caught before given that they also caused compile failures when I used the same configure/build/packaging script on Linux. The code uses methods that no longer exist in Qt 5.5+ : `QString::fromAscii()`, `QHeaderView::setResizeMod()` (replaced with `QHeaderView::setSectionResizeMode()` and `QInputDialog::getInteger()` (instead of `QInputDialog::getInt()`). Those are the easy and straightforward ones fixed by this patch. The handling of file dialogs (in 2 locations) was more problematic. Both Apple's clang "6.0.0" and gcc 5.3 (on Linux) refused statements like `QWeakPointer<QFileDialog> dlg = new QFileDialog(parent,defname);` . My initial attempt was to follow the documentation for QWeakPointer, and do `QWeakPointer<QFileDialog> dlg = QSharedPointer<QFileDialog>(new QFileDialog(parent,defname));` but that caused a crash as soon as I clicked on the "save backtrace as" button - on both platforms. It turned out that `QSharedPointer<QFileDialog>()` returned a valid (non-null) shared pointer, but after creating a QWeakPointer from it, `dlg.data()` returned NULL -- immediately. I had a look around in the 5.17.0 frameworks codebase, and could find no relevant examples of using QWeakPointer; some expressions appear to be equivalent to the original DrKonqi code, in other locations QWeakPointer instances are indeed created "going through" a QSharedPointer. In the end I tested using QFileDialog (and `QFileDialog::selectedUrls()` as kate5 does, and that works (on OS X; haven't had the chance to test on Linux yet). That's what the attached patch does: no QFileDialog pointer at all, just an instance. This RR is intended to fuel a discussion on the subject, as suggested by Aleix; therefore I left the comment about the possibility of invalidating/deleting the file dialog e.g. through a DBus call. Is that actually a possibility, and why wouldn't it be in Kate? Are there no other ways to prevent it, or check for its occurrence? This patch also marks drkonqi as a nongui executable, a requirement on OS X and which shouldn't cause any regressions elsewhere AFAIK. Diffs ----- drkonqi/CMakeLists.txt deb8c40 drkonqi/bugzillaintegration/bugzillalib.cpp 802c5fb drkonqi/bugzillaintegration/reportassistantpages_bugzilla.cpp e60fb14 drkonqi/bugzillaintegration/reportassistantpages_bugzilla_duplicates.cpp 4f8f4ea drkonqi/drkonqi.cpp b12c118 Diff: https://git.reviewboard.kde.org/r/126993/diff/ Testing ------- On OS X 10.9.1 and Kubuntu 14.04, both with Qt 5.5.1 and KF5 5.17.0 installed into /opt/local Thanks, René J.V. Bertin
_______________________________________________ Plasma-devel mailing list Plasma-devel@kde.org https://mail.kde.org/mailman/listinfo/plasma-devel
