> On Dec. 6, 2012, 9:15 p.m., Mark Gaiser wrote: > > Do i also need to push this to the frameworks branch? If so, how? > > git checkout KDE/frameworks > > git pull > > git checkout <myFixBranch> > > git rebase KDE/frameworks > > git checkout KDE/frameworks > > git merge <myFixBranch> > > git push > > > > Just verifying if these are the steps i should do if i need to merge it to > > frameworks branch as well. > > A lot of steps btw. > > David Faure wrote: > No. We use git merge between 4.10 and frameworks, so if you want to merge > this just do: > git fetch > git checkout frameworks > git merge origin/KDE/4.10 > <solve conflicts> > git push > > If you end up with too many unrelated conflicts you don't know how to > fix, cancel everything and ask me to do it :-) > > Mark Gaiser wrote: > I regret to ask if you could do it... > I'm getting some merge conflicts where i don't quite know what the > intention is.. (deleted files, but modified in head.. and some other merge > conflicts).
What does the code do, if the QTime wraps around? Maybe it needs QElapsedTimer, see also bug 306186. - Christoph ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/107520/#review23096 ----------------------------------------------------------- On Dec. 5, 2012, 6:54 p.m., Mark Gaiser wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://git.reviewboard.kde.org/r/107520/ > ----------------------------------------------------------- > > (Updated Dec. 5, 2012, 6:54 p.m.) > > > Review request for kdelibs and David Faure. > > > Description > ------- > > listDir is batching only in a vague manner. It required to have details set > to something bigger then 0 and required to know the amount of files that > where going to be send. Even then it was fixed to broadcast only the first > 100 entries followed by _everything_ that comes after it. For more details, > look at the diff. If you did happen to do a request with details set at 0 > (thus only the filename and filetype are stored) then a different code path > was taken where it wasn't even possible to batch. > > I changed this to do time based batching. Now it broadcasts all the entries > it currently has after 300ms have been passed. In normal situations on the > local file system that means that you will never see batching. You will start > to see them if you hit folders with 10.000+ entries in then (for a SSD that > is). For the slower HDD drives you probably see more batching. > > In the worst case scenario you could find yourself in the situation that 1 > entry is broadcasted every 300ms. That is only on very slow external devices > or internet places. > > The advantage of batching (just repeating that here since it's currently not > working in any KDE app) is that you can start showing results to the user > without all the results being in. This gives the user a "feeling" that it's > fast even though KIO might still be working hard to get all entries in. > Dolphin is one app in particular that can really make use of this. Currently > it doesn't do that. IT waits till all the results are in. > > While working on this, i also updated the style of the FileProtocol::listDir > function and greatly simplified how it lists the directory. Now the actual > listing just follows one code path instead of one for details 0 and one for > the rest. > > I'd like to get this in kdelibs for KDE 4.10 because it is a bugfix even > though there is no bug report. > > > Diffs > ----- > > kio/kio/slavebase.h 0bdb146 > kio/kio/slavebase.cpp 8f8ff03 > kioslave/file/file_unix.cpp 2b47b7c > > Diff: http://git.reviewboard.kde.org/r/107520/diff/ > > > Testing > ------- > > I tested this in various ways: > (SSD) 100.000 entries in a folder with details set to 0: 1 signal with all > entries > (SSD) 100.000 entries in a folder without details set (all info): 2 or 3 > signals each with between 20.000 - 50.000 entries > (slow ftp) ftp://ftp.mirror.aarnet.edu.au/pub/archlinux/extra/os/x86_64/ > results start slowly with 20 - 50 per signal and quickly climb up. > Ran the kio test cases. Apparently i have a "failed" for kio-krununittest, a > testcase that David Faure has fixed if i'm right. > Updated the documentation to not mention the total files anymore. KioSlaves > can still set it, but it's just unused for the listDir now. > Last, running with this now and i haven't seen any issues so far. > > > Thanks, > > Mark Gaiser > >
