ralf.habacker at freenet.de
Thu May 17 11:59:15 CEST 2007
Andreas Pakulat schrieb:
> On 15.05.07 00:44:12, Ralf Habacker wrote:
>> Ralf Habacker schrieb:
>>> What about kio_file. Does it work ? The last time I tried to port
>>> kioslaves kio_file works and kio_http in non ssl mode. Are there any
>>> error messages printed ? You can run kdebugdialog to activate all messages.
>> I tried recent kio_file and it looks there is a problem in the slave
>> communication. klauncher returns an error after launching a slave.
>>  kio (KLauncher): Slave launched, pid = 3416
>> trying to load 'kio_file'
>>  kio (KLauncher): KLauncher: launching new slave kioslave with
>> protocol=file args=("kio_file", "file", "", "P5")
>>  kio (KLauncher): kioslave (pid 3532) up and running.
>> request->status 2
>>  kio (KLauncher): Slave launched, pid = 3532
>> trying to load 'kio_file'
>>  kio_file: Starting 3532
>>  kio (kioslave): SlaveBase: failed to connect to P5
>>  kio (kioslave): Reason: name lookup has failed file P5
> I didn't even get that far, I'm getting a "couldn't create slave : cannot
> talk to klauncher: The name org.kde.klauncher was not provided by any
> .service files".
> After moving klauncher from lib/kde4/libexec to bin/ that one is gone (I
> guess on windows there shouldn't be anything installed into libexec??)
> and I get the same message as you, saying that a file slave is created.
> One thing that is weird is that the path is something like
> file:///C:/Dokument 4096nd
There may be KUrl/Qurl fixes required see
http://lists.kde.org/?l=kde-windows&m=116663586908938&w=2, but this
issues should be fixed in the second step. More important is that the
socket communications works.
> And tehn I'm getting a kio (Slave): slave failed to connect to
> application pid=1244 protocol=file.
> I checked the kio_file slave and there didn't seem to be any changes
> that could have explained this.
The most big change is that Thiago migrated parts of the KDE socket code
to QSocket and friends. May be that there are some issues or migrations
required to QSocket based code and Thiago can help in this area.
One of the reason why Thiago migrated to QSocket is that Windows Sockets
are unreliable in some areas and using QSocket makes it easier for us,
because Trolltechs supports this socket code.
> I'll now try to find the base classes of
> kio_file slave and check wether I find any changes in that.
See kio\kio\slavebase.cpp. On unix the master opens a unix domain socket
to which the client connects. On windows the socket should be tcp and
the message above indicates that the client could not connect to the
socket the master has opened.
> If that doesn't bring anything, how do I debug kio slaves under win32?
The problem is that debugging kioslaves is very dynamic because of
process starting and socket connection timeouts.
As I debugged it the last time i had used a tcp monitor (search google
for tdimon and/or windump http://www.winpcap.org/windump) to inspect the
states of the socket communications. Additional I placed debug messages
in important places of the kio slave code to see in which state the
BTW: I used kioslavetest to check the kio communication.
More information about the Kde-windows