Abdelrazak Younes wrote: > Peter Kümmel wrote: >> Abdelrazak Younes wrote: >>> Edwin Leuven wrote: >>>> Abdelrazak Younes wrote: >>>>> This patch solves it. Committing now. >>>> i still see a crash... >>> Could you set a breakpoint at breakpoint at the beginning of >>> Dialogs::disconnect(string const & name) in frontends/Dialogs.C and >>> continue step by step until you see the crash? >>> >>> Abdel. >>> >>> >> >> This hack avoids the crash: >> >> Index: frontends/controllers/Dialog.C >> =================================================================== >> --- frontends/controllers/Dialog.C (revision 14889) >> +++ frontends/controllers/Dialog.C (working copy) >> @@ -115,7 +115,7 @@ >> if (!view().isVisible()) >> return; >> >> - controller().clearParams(); >> +// controller().clearParams(); > > That means that controller() points to nothing real. That's quite weird > considering that it is a boost::scoped_ptr and that it is set in > Dialog::setController(Controller * i) and never release after that. > > Could you put a breakpoint in this method (setController) before opening > the dialog in order to verify that controller_ptr_ is correctly reset? > > Georg, Lars, did you do the BOOST upgrade to 1.34RC at the end? This > problem might be related to that. > > Abdel. > >
I have done look at the boost update: nothing smart pointer related. And I don't think that boost's smart pointer are broken. It is one of the simplest code in boost, it is the origin of TR1 (as far as I know), I also assume it is one of the most used part of boost, so it is the stablest and most tested code of boot. -- Peter Kümmel
