Do we want to mark the "old" methods as deprecated?

Update the docu?

Yes, if this is confirmed to work we can deprecate stopDirScan/restartDirScan.

It's kind of harmless i guess, but when doing a rename via F2 in dolphin we do stopDirScan on the url twice, one in CopyJobPrivate::startRenameJob and another in CopyJobPrivate::createNextDir
that means you get the following warning
org.kde.kcoreaddons: doesn't know "/home/tsdgeos/bcnfs.github.io/sponsoring/"

Do we care?

In fact, CopyJob has no way to know whether a view somewhere is watching that directory. When an app other than Dolphin uses CopyJob, there isn't going to be any watching, possibly.
So I think we need to remove the qDebug in KDirWatch so that removeDir is silent if the dir wasn't watched [this is faster than checking before removing, from the outside].

At the same time, this shows that the CopyJob handling of m_parentDirs can be improved a little bit. Posting an update.

In D5856#109721, @dfaure wrote:

So I think we need to remove the qDebug in KDirWatch so that removeDir is silent if the dir wasn't watched [this is faster than checking before removing, from the outside].

+1

I know what the problem is with this.
Previously, if nobody was watching the directory, stopDirScan/restartDirScan would basically do nothing.
Now, after running a kio job in that directory, KDirWatch will start watching it, for no purpose, due to the unconditional addDir().

I guess this needs checks with KDirWatch::contains() before removeDir to only call addDir on dirs that were actually watched in the first place...

ping @dfaure

Is it still valid ?

Probably still valid, yes. Needs investigation, and a unittest....