Re: [Gtk2hs-users] segmentation fault on 'disconnect'

[Michal P. <al...@v-...>]

Thanks a lot!

I would have had less trouble if I used GHC 6.4, maybe it's time to
switch.

I have one more question: Is it possible to create a callback that
disconnects itself? I can't do it in a straightforward way because the
callback can't use the handle that is returned from the callback setting
function:
con <- onClicked button ({- can't use con here -})

I would be able to use IORef, but I want to keep the amount of global
state minimal. Or maybe disconnecting callbacks is a bad practice and
I'm missing some design pattern here?

By the way, why do you use CVS instead of Darcs used by most
Haskell-oriented projects?

Regrads,
Michal

On Sat, 2005-07-23 at 02:11 +0100, Duncan Coutts wrote:
> On Thu, 2005-07-21 at 13:53 +0200, Michal Palka wrote:
> > Hello,
> > 
> > My last email probably didn't make it to the list, I'm resending it.
> 
> Actually it did but we were probably all on holiday! :-)
> 
> > When I run the following program it crashes with a segmentation fault.
> > It runs fine when I remove the line with 'disconnect con'. The machine
> > is i386, I'm using GHC 6.2.2 and gtk2hs 0.9.8.
> > 
> > Is using disconnect the correct way of removing callbacks?
> 
> You know I've never actually written a program that uses it! But yes it
> is supposed to be the right thing.
> 
> On GHC 6.4 instead of segfaulting it reports:
> a.out: schedule: re-entered unsafely.
>    Perhaps a 'foreign import unsafe' should be 'safe'?
> 
> Which is almost certainly a correct diagnosis of the problem. In other
> words it's our fault! :-)
> 
> When I change the disconnect function to be a safe foreign call your
> example prog does indeed work.
> 
> I have to say I find that very confusing since I can't see what Haskell
> code is being called when we call disconnect. I've had a similar report
> about timeoutRemove/idleRemove where it's also not clear that there
> should be any callback into Haskell land.
> 
> Anyway, it's fixed in cvs and will be included in the 0.9.8.1 bugfix
> release.
> 
> I just wish I understood what was going on here! :-)
> 
> Duncan
>