From: Pavel Mayer (pavel++at++ARTCOM.DE)
Date: 12/19/2001 02:00:41
Hi there,
I did not see the dynamic_cast malfunction you described with my system,
but performer/kernel/libc/compiler compatibility problems seem to be
worse in 2.5 than in 2.4.
My problem is that exceptions did not work in peformer 2.5 libpf linked
code (they worked for me with 2.4, but other reported problems); this
could be related to your dynamic_cast problem.
I could fix it by statically linking every performer executable and
library with libgcc (-lgcc); it might also help with the dynamic_cast
problem.
I once had a dynamic_cast problem with performer that was somehow
related to the shared memory/multiprocessing/dynamic linking mechanism.
Somehow dynamic_cast used the wrong vtable adress from a dynamically
linked object instead of the original one, which resulted in a segfault
because the offset to access the vtable was way too large. This problem
was resolved by loading the dynamic object earlier by creating an early
instance of the desired object. (before pfConfig())
Please let the list know if something of the above works for you.
Pavel Mayer
P.S.: It is really annoying not have the sources of something you want
to run with linux; the NVIDIA drivers and performer are the best
examples for it; they cause about 90% of our stability and compatibility
problems.
-----Ursprüngliche Nachricht-----
Von: Allen Bierbaum [mailto:allenb++at++vrac.iastate.edu]
Gesendet: Dienstag, 18. Dezember 2001 21:03
An: info-performer++at++sgi.com
Betreff: Re: using dynamic_cast with Performer lib
Has this bug been resolved with the latest release of Performer for
Linux?
In other words, are the binaries in the latest release compiled with the
most recent C++ ABIs or will it still seg fault when running
applications that use dynamic_cast<>?
-Allen
[lots of citation deleted]
This archive was generated by hypermail 2b29 : Wed Dec 19 2001 - 01:59:20 PST