Fixing krunner's threading

[Andreas Hartmetz]

On Monday 22 February 2010 20:23:45 Ingo Klöcker wrote:
> On Monday 22 February 2010, Andreas Hartmetz wrote:
> > Hi,
> > 
> > Now that many of the more prominent crashes in krunner have been
> > fixed I think it's time to attack something more "theoretic":
> > runners run in threads and often don't pay any attention to
> > thread-safety of API they use. Two weeks ago I've added a warning to
> > KIO::Scheduler() to warn about its use from non-main threads, but so
> > far nobody paid attention. The offending runner here is the places
> > runner.
> > This is a real problem, and it has to be fixed if krunner is ever
> > supposed to become *really* stable.
> > So I'd like to hear from the Plasma team where they can fix things,
> > and maybe some things can be made thread-safe in kdelibs and friends
> > without too much effort. But most of all we have to talk about the
> > problem instead of ignoring it. I'm not even sure who to talk to
> > because runners are made by different people.
> > I get the impression that krunner and/or most runners take the "put
> > things into threads now, ask questions later" approach. It may very
> > well be that runner authors don't even know that there is a problem -
> > or the problems are being ignored.
> > I'm a big fan of the krunner concept, but krunner has to stop
> > crashing so much.
> > 
> > Cheers,
> > Andreas
>
> Hi Andreas,
> 
> I think plasma-devel is a much better place for this discussion. Thread-
> safety surely is a general problem, but I guess the runner authors are
> more likely to be found on plasma-devel.
> 
> 
> Regards,
> Ingo
> 
(I took the liberty to reorder to un-top-post)
I decided against posting to plasma-devel for several reasons:
- While the problems come from plasma, they could possibly be solved in other
  code
- Most technical experts are on k-c-d and not on plasma-devel, and multi-
  threading is a very technical topic. I'm not very good at it myself.
- This is a topic of general interest
- I don't want to subscribe to yet another mailing list