Puzzle about the _fini function in ia64 Linux

[Gewj]

hello,Andreas, thank for you answer,but I still have one question:
The same code (it is IA64 CodeClean) run quite well under ia32 linux 
system (such as redhat Kernel 2.4.20-8 gcc-3.2.2-5).

would it be a ia32/ia64 linux difference?


Andreas Schwab wrote:
> Gewj <geweijin@sinosoft.com.cn> writes:
> 
> 
>>I'm working on a ia64 linux,(kernel 2.4.18,gcc3-3.0.4-1)system.
>>I redefine the _fini function in a dll library with the -nostartfiles in
>>its Makefile.
> 
> 
> Don't do that.  If you need a finalizer use __attribute__((destructor)).
> 
> Andreas.
> 


----------------original message------------------------------------
Hi,

I'm working on a ia64 linux,(kernel 2.4.18,gcc3-3.0.4-1)system.
I redefine the _fini function in a dll library with the -nostartfiles in
its Makefile.
then I write a execute process which have many threads, all those
threads  linked with this library in its Makefile, the puzzle thing is
in some case, it seems that in some threads, the _fini function of the
library have been called even since the library was been used by other
threads.

Does the above thing may happen(gcc bug in ia64) or it turn out to be my
fault?

Thank you for any help.

Regards
--
Ivod.