Am I correct in assuming that also means main?
I've never encountered that case, where libfoo.so referenced a function quux()
and the body was supplied by the main program. Usually, it's done in some
way where libfoo calls through a function pointer, and the address of quux() is
passed as a parameter when registering the callback....
I'm rapidly arriving at the conclusion I may have to restructure my code
to register these callbacks with my own libraries. Windows was strike
one, OpenVMS strike two, and now AIX is strike three.
I think I need to take option a - in my case libfoo is being dynamically
loaded vi g_module_open.
I would have thought that libtool would have included that option for
AIX when it was told to -export-dynamic?
Check what version of libtool - many older ones (and possibly new ones) will
get weirded out by the bizarre semantics of the AIX linker.....
Also, read the documentation on -brtl as well - there's some gotchas. One biggie
is that just adding -brtl to the final link of the main program doesn't actually
do much good - you need to rebuild the glib and gtk shared library objects
with -brtl as well, because if the .so doesn't have some stuff tagged in the
header, the main program's runtime linking won't look at it. And then once you
do that, trying to link a main program to it and omitting -brtl will cause
odd failures....
Joy. When I tried to recompile glib (I'm not - at least not yet -
trying to use gtk) the IBM 7.0 compiler went belly-up with an internal
error. And, I probably cannot require folks recompile glib and the
other dependencies of netperf4 (eg libxml2) - that sounds like it would
rapidly devolve to "since glib was recompiled, now all the other stuff
on the box using glib must be recompiled..." hell.
You may find that recoding things to pass a function pointer is the more sane
way to work around it. ;)
Yep :(
rick jones
_______________________________________________
gtk-list@xxxxxxxxx
http://mail.gnome.org/mailman/listinfo/gtk-list
