From kde-core-devel  Thu Jan 31 22:21:36 2002
From: Julian Seward <Julian_Seward () muraroa ! demon ! co ! uk>
Date: Thu, 31 Jan 2002 22:21:36 +0000
To: kde-core-devel
Subject: use of uninitialised field in KIO::Scheduler::JobData
X-MARC-Message: https://marc.info/?l=kde-core-devel&m=101255868808425


The following message pertains to the cvs head as of Thu 31 Jan at about
20:00 GMT and in particular to kdelibs/kio/kio/scheduler.cpp rev 1.83.

I'm seeing the following complaint:

  Use of uninitialised CPU condition code
     at 0x40C7BE39: KIO::Scheduler::findIdleSlave(KIO::Scheduler::ProtocolInfo *, 
                       KIO::SimpleJob *, bool &) (scheduler.cpp:516)
     by 0x40C7A9FB: KIO::Scheduler::startJobDirect(void) (scheduler.cpp:414)
     by 0x40C796BD: KIO::Scheduler::startStep(void) (scheduler.cpp:264)
     by 0x40C8046C: KIO::Scheduler::qt_invoke(int, QUObject *) (scheduler.moc:154)

scheduler.cpp:516 is:

516: if (jobData->checkOnHold)
     {
        slave = Slave::holdSlave(jobData->protocol, job->url());
        if (slave)
           return slave;
     }

from which I guess that the KIO::Scheduler::JobData::checkOnHold field
is sometimes not initialised.  Indeed I can't see a constructor for
JobData at all, so it may well be not always initialised.

How to reproduce: 
   start   kate   with no arguments
   the problem is reported during kate startup; by the time kate is
      up and running it has already happened.

I'm not claiming that this problem is particularly kate-specific;
I'm just using kate as a relatively simple test program to shake out
bugs in the libraries.

This is probably pretty straightforward for a suitably knowledgable
person to track down.

J