[omniORB] Building on Alpha Linux 2.0

Sai-Lai Lo S.Lo@uk.research.att.com
01 Apr 1999 11:10:05 +0100


Hi! I'd like to see omniORB2 runs on Alpha Linux. Unfortunately, the last
time I look (end of 98) the egcs compiler + linker on the alpha is still
not doing the job properly.

Indeed the thread safe exception handling seems to have sorted.

But now the executables died when it is running the ctor of global
variables, i.e. before it hits main().

In particular, it core dump when executing the ctor of omni_thread::init_t,
which call pthread_key_create() etc. We use omni_thread::init_t to ensure
that the omnithread library is initialised before main().  Now this works
on every other platform/compiler combination but on alpha linux this causes
problem. I guess the ctor is executed before the pthread library has been
initialised.

It looks to me this is a problem with the linker.

With some egcs releases, I managed to get the examples to run but then
when you past anything like a reference to a struct in a function call, the
stack frame is screwed up.

All these problems have been reported to the egcs maintainers. I do not
know if anyone is looking into it. 

Sai-Lai


>>>>> Brent Fulgham writes:

> I am maintain the omniorb binary distribution for the Debian GNU/Linux
> distribution.  I have been very successful working with omniORB running on
> the i386 architecture.

> Recently we received a donated alpha machine to help speed up our porting to
> the alpha architecture.  I have been trying to build omniORB on this alpha
> machine with little success.  The build proceeds properly, however the final
> executable for omniNames dies on startup, prior to any configuration file
> parsing, etc.  As near as I can tell, it seems to die when linking to the
> pthreads library, although Sai-Lai's thread test program (posted to this
> list a few months back) built and ran properly.

> I wonder if anyone on this list with an alpha-based Linux distribution could
> indicate whether they have been successful building omniORB.  Any pointers
> would be greatly appreciated.






-- 
Sai-Lai Lo                                   S.Lo@uk.research.att.com
AT&T Laboratories Cambridge           WWW:   http://www.uk.research.att.com 
24a Trumpington Street                Tel:   +44 223 343000
Cambridge CB2 1QA                     Fax:   +44 223 313542
ENGLAND