On 2007-06-01 at 07:44:08 [+0200], Ingo Weinhold <bonefish@xxxxxxxxxxxxxxx> wrote: > > On 2007-05-31 at 22:09:12 [+0200], Oliver Tappe <zooey@xxxxxxxxxxxxxxx> wrote: > > Ingo, I am pretty sure that the problem you have described is precisely the > > reason why I had to introduce some "specialty" to gcc-2.95.3-beos that would > > fix crashing problems related to type_info of BDirectWindow (by providing a > > separate correct type_info function in a special object file). > > I figured the problem would be related, but I still seem to miss the link. > The BDirectWindow type info function in libgame.so looks actually quite OK. > libgame.so also contains the type info functions and type_infos for all base > classes. I didn't check all, but I suspect they are OK, so that the > BDirectWindow type info function should actually return a properly > initialized type_info. > > The only problem I can imagine would occur when an executable containing a > BDirectWindow subclass would contain a BDirectWindow type_info, but no > corresponding type info function. Then the subclass type info function would > use the uninitialized type_info. But since gcc seems to be configured to > neither produce type_info nor type info function, I don't see where things go > wrong. Do you have an example how to reproduce it? Sure, just edit the compiler's specs file (.../tools/gnupro/lib/gcc-lib/i586-pc-beos/2.95.3-beos-060710/specs) and remove this from it: %{!no-beos-fixes: fix_bdirectwin_typeinfo.o%s} If you then do a 'make clean && make' on GLTeapot and start it, you should be able to observe the crash. At least it "worked" for me, when I tested it just now... cheers, Oliver