SIGUSR1 using omniORB2.4 with egcs-1.0.1 and Redhat 5.0

Sai-Lai Lo S.Lo@orl.co.uk
Fri, 6 Mar 1998 16:46:59 GMT


Firstly, one shouldn't use egcs-1.0.1 without the thread-safe exception
patch to compile a multithreaded C++ program. It will fail in an apparently
random fashion.

The latest egcs snapshot is purported to support the thread-safe exception
handling stuff. However, it seems to me that it still generates the wrong
code. The symptom being that eg3_impl and eg3_clt hang. Looking at the
programs under gdb, it hangs after a throw and never reach the catch in the
outer scope.

I also found that, with the latest egcs snapshot (980302) and with static
library only and by explicitly choose setjmp/longjmp exceptions
(-fsjlj-exceptions), the code generated seems to run fine.

I have not been able to get to the bottom of this. So unless you are a egcs
compiler expert, I would suggest stay away from it.

In fact, with the problem with c++ on x86 Redhat 5.0 still unresolved, I
wouldn't suggest one move from Redhat 4.2 to 5.0 either. 

I hope someone can shed some light on the egcs problem (and how to resolve
this). 

Sai-Lai