On Wed, Aug 29, 2012 at 10:22 PM, Greg Ercolano <[email protected]> wrote:
> On 08/29/12 14:15, Greg Ercolano wrote: > > On 08/29/12 06:01, Bogdan Popescu wrote: > >> Hello everyone, > >> > >> I've recently migrated/rewrote our small communication application from > >> Perl to C++ and used FLTK for the GUI part, the idea was to be able to > >> start using it on Win/Mac rather than giving Linux boxes to users... > >> > >> As it's made of 2 pieces, the receiver (which is a daemon/service > running > >> under the user that is logged in) and the sender (a small app that > search > >> for and connect to the daemon/service running on the receiver's side), I > >> needed the daemon to keep the FLTK code and make sure the window is > >> displayed with the sender details... everything seems to work fine under > >> Linux/X11, though on the Mac side the code compiles fine but some > settings, > >> mainly FL_Window::show(), fail under fork() and the child segfaults (and > >> the window is not displayed), if I run the code without forking > anything it > >> seems to work fine... any idea how I could work around this problem or > did > >> I miss anything? > > > > Oh man, don't get me started on Macs and its limitations with > fork() > > and window manager stuff, esp. in the context of daemons. > > <mad rant> > > BTW, sorry -- not ranting at you OP, I'm PO'ed at APPL's > OS designers. > > The TN2083 thing is a giant PITA especially in the context of > distributed computing. > > Perhaps doing one of their 'agent' approaches recommended in the > above tech note is best for your situation. Just be aware that this > means you can loose parent/child relationships, so you'll have to > do > things like detect if one client/server have disappeared as part of > your protocol between them. > > Or perhaps you can avoid this all with threads. > I'm not that good with threads... I gave it a go under Linux and it seems to work fine for the first time it renders the window, after the second or third try it might segfault (when calling Fl_Window::show() or Fl::run()) and you cannot close the window opened by FLTK, also the window is redrawn when a new message is received instead of spawning a new one... I'm thinking the way FLTK works with threads is different, right now I have all my FLTK code under the last called function, do you think I need to/should initialize fltk in main() and then spawn windows from the running threads? Right now I replaced the fork calls with pthread (adding the pthread_detach/join calls where needed), and stick Fl::lock/unlock in the function that deals with FLTK... I assume there must be a way of dealing with this without going for the agent approach but thank you for your replies! Regards, Bogdan P. _______________________________________________ fltk mailing list [email protected] http://lists.easysw.com/mailman/listinfo/fltk
