[omniORB] Solaris 2.7 and libthread

[Sébastien Zimmermann]

Hi,

I had a problem last week which I finally managed to solve.

When running my client on a Sun U60 with Solaris 2.7 and omniORB 2.7.1,
compiled with SUN WS C++ 4.2 on a Solaris 2.5.1 system, it would display
around 30 times "Bad tag" then print a libthread panic ("no mem").

The libthread panic and "Bad tag" message would not appear if I put
libthread.so from a Solaris 2.5.1 first in the LD_LIBRARY_PATH. This would
make my client and my server connect properly, but then, no call to the user
objects could be performed (client and server would hang up).

The problem was solved using the original libthread at runtime and linking
the one from Solaris 2.5.1 last at compilation time.

I find it strange also, that I have to mess around with the order of
libraries. Maybe I'm missing a compilation option?

Also, a strange thing is that the echo examples, compiled on the same
platform, would run just fine with the U60 platform, but would print a
segmentation fault on the 2.5.1 platform.

There is probably an explanation, but I am puzzled by those strange Sun
Systems behaviors.

Have a nice week.

Sebastien Zimmermann.
zimmese@gemse.fr