On Sun, Aug 16, 2009 at 12:49:28PM +0200, Peter Kümmel wrote: > Abdelrazak Younes wrote: > >> Is it ok when I update without a branch? > > > > Yes. > > > > OK, when I have commit rights I will update and test on Linux. > > >> This means I will commit without testing on Linux before commiting, > >> only Windows. (But I don't see why it should not work on Linux) > >> > >> Does such a update fix the problems with GCC 4.4? > >> > > > > Yes. > > > > Peter, could you please write down somewhere in the wiki the recipe to > > update boost? > > > > Yes, I write down the recipe, but it isn't that complicated. > > > But why do we still use boost? > > - signals: > * could be replaced by Andre's classes > there are only several boost::signal<void()> signals and some other > signatures > Two signals would be enought: > 1. signal/connect without any arguments > 2. a templated 'T const&' which we internally cast to > 'void*' to be usable with Qt mechanism and without moccing > Other arguments could be wrapped in a struct.
It should use just plain Qt signals/slots instead of those wrappers. We use autoconf, uic, gettext, perl, python and probably other tools to generate code. Singling out moc was and is ridiculous. > - regex: could be replaced by std::tr1 or Qt > - smart pointers: std::tr1 (self made scoped_pointer) > - bind: std::tr1 > > What were the reasons at the last meeting not to drop boost? Using tr1 turned out not to be a drop-in replacement. Andre'
