[RFC] Use QFileSystemWatcher as another method for KDirWatch?

Andreas Pakulat apaku at gmx.de
Wed Dec 5 22:33:19 GMT 2007


On 05.12.07 23:11:53, Christian Ehrlicher wrote:
> Andreas Pakulat schrieb:
> > On 05.12.07 20:24:27, Christian Ehrlicher wrote:
> >> Christian Ehrlicher schrieb:
> >>> Hi,
> >>>
> >>> Using kdirwatch on windows is very cpu consuming (kded sometimes use 40% cpu on my duron 1300). Therefore I want to implement QFileSystemWatcher as an optional method. Is there a reason why this should not work? Should it be an optional or a hard dependency?
> >>>
> >> Ok, here the patch.
> >>
> >> I enabled it only for windows to not get in trouble with 4.0.0. There's
> >> a bug (task-tracker id 170021) which prevents watching directories on a
> >> fat32 partition. It's fixed in 4.4 but it would be nice when this patch
> >> could be applied to qt-copy.
> > 
> > Does this support all features of KDirWatch? Last time I checked
> > QFSWatcher had some limitations.
> > 
> There's one major difference between KDirWatch and QFileSystemWatcher.
> KDirWatch can watch files/dirs which do not exist yet.
> Any other difference you're aware of?

No, and now that you mention it, I think thats exactly what I was
thinking about. I'm not sure though how much that feature is used...

> > And if it supports all features, I think KDirWatch could simply drop its
> > kqueue and inotify implementations and use QFSWatcher for that as well.
> > Which would mean using QFSWatcher as default on Win32 and wherever
> > kqueue is used (I think BSD) and using FAM on linux or falling back to
> > QFSWatcher if fam doesn't work/is not installed.
> > 
> Yes, but we should wait for 4.1 until it's used on non-windows.

Of course, I wasn't meaning anything else :)

Andreas

-- 
Artistic ventures highlighted.  Rob a museum.




More information about the kde-core-devel mailing list