Hi, > Since yesterday I am busy with intensive experiments > applying various combinations of the hbQT destructor code > and GC and have come out with following : > > Link hbIDE ( any other application as well ) with -hbnocppmm as > c:\>hbmk2 hbide.hbp -hbnocppmm and GPF's will disappear. > > The reason is: as we cannot handle how and when Qt applies > destructions in an object's life-cycle, we must not allocate those > pointers via our own memory managers. But still we need to inform > Qt that an object is no longer in use if variable holding it > goes out of scope. > > Other all aspects are still the same minus GPF. > Please test on various OS's/Systems and report on the list.
Quite strange conclusion. QT should work no matter what sort of memory allocation subsystem we use, so I assume this is wrong conclusion and just some luck on your part not getting errors with default allocator in your specific tests. Please note that we had tons of similar problems (heavy GPFs and leaks) even before introducing memory allocator override. In fact our own memory allocator was the tool to reveal some of the grave problems, which were just hidden before. So IMO this advice is nothing more than a new dead road "towards" the solution. IMO there is no "magic", "one click" solution to this problem, but only to do the hard work, understanding the details of QT object handling and fixing the problems for real. IOW: hiding the problems won't solve the problems. Brgds, Viktor _______________________________________________ Harbour mailing list (attachment size limit: 40KB) [email protected] http://lists.harbour-project.org/mailman/listinfo/harbour
