Hi Antonio, I have been thinking about ipc, implemented as a (blocking) worker thread and how to communicate with the main GUI thread when a message is received. I found the 2008/12 thread with IupAsync and had a look at how that was implemented... and then I started looking intently at WM_COPYDATA... which I managed to convince myself would be a pretty good windows-only solution... and then I started looking at the mentioned g_idle_add, and realised how non-thread-safe it would be to try and get to that via IupSetFunction (because of the iupTableRemove()/iupTableSetFunc()) and then finally I looked at iupdrvSetIdleFunction() in both iupgtk_loop.c and iupwin_loop.c
It appears that the calls to iupdrvSetIdleFunction() are thread safe*. I can see that quite clearly in iupwin_loop.c but I'm not at all familiar with g_idle_add etc as used in iupgtk_loop.c (ditto XtRemoveWorkProc/XtAppAddWorkProc in iupmot_loop.c), and judging from the way those xxx_idle_cb are casually set to NULL in said files, would hazard table updates may not be necessary anyway? Lastly I realised that if you had to go via idle_cb on Linux, but WM_COPYDATA on Windows, then best programming practice would be "invoke idle_cb from COPYDATA_CB", so why not go direct? After all that, it seems we just need to export iupdrvSetIdleFunction() as IupSetIdleFunction()? Three other quick points: I think it would be fair to say it is the programmer's responsibility to put any message somewhere idle_cb can find it, thread-safely (and nothing to do with IUP). It may also be the case that idle_cb exiting and nullifying xxx_idle_cb may overlap with another thread setting it, and again it would be reasonable to require the programmer to start a timer on exit from idle_cb to re-check things (once) say 0.5s later (also nothing to do with IUP). *By thread safe I meant as in two threads setting a static int. They may over-clobber such that a reader never gets to see both, but otherwise nothing catastrophic should happen. In this case I am thinking two threads simultaneously setting a static var with the same value anyway. Update: a quick google suggests g_idle_add and XtAppAddWorkProc are possibly thread safe. As usual, however, the deeper I dig the less clear things become. I'll just add a couple of references, which may or may not be relevant: https://mail.gnome.org/archives/gtk-devel-list/2007-November/msg00125.html https://developer.gimp.org/api/2.0/gdk/gdk-Threads.html Pete ------------------------------------------------------------------------------ _______________________________________________ Iup-users mailing list Iup-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/iup-users
