> On June 9, 2011, 9:49 a.m., Aaron J. Seigo wrote: > > my biggest question with this change would be performance. QFileInfo is not > > fast, and to make matters "worse" this code in KStandardDirs influences > > start up speed. have you measured the performance after this change > > relative to before? > > > > with Qt5 this may become a moot issue with the changes coming in the QFile > > backends ... but i don't think it makes sense to introduce possible > > performance regressions on code that currently works. > > > > is there a platform for which the current code fails?
I did the patch solely for simplicity and portability reasons, and perhaps (or hopefully) also in the Platform 11 spirit. :) After all, "outsourcing" platform-specific details to Qt wherever we can seems like a good idea to me, such that we can take advantage of new platforms much quicker. That said, I couldn't measure any performance difference when logging into a plasma session. In both cases, it takes ~13 secs for the wallpaper to appear, and ~30 secs until the KWallet password dialog appears. - Bernhard ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/101554/#review3799 ----------------------------------------------------------- On June 9, 2011, 7:06 a.m., Bernhard Beschow wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://git.reviewboard.kde.org/r/101554/ > ----------------------------------------------------------- > > (Updated June 9, 2011, 7:06 a.m.) > > > Review request for kdelibs, kdewin and David Faure. > > > Summary > ------- > > Increase portability of kdecore a tiny bit by making KStandardDirs use > QDirIterator rater than platform-specific code. > > > Diffs > ----- > > kdecore/kernel/kstandarddirs.cpp ba90270 > > Diff: http://git.reviewboard.kde.org/r/101554/diff > > > Testing > ------- > > I'm running my platform 4.6 with this patch and haven't noticed any > regressions yet. > > > Thanks, > > Bernhard > >
