[omniORB] AIX: pure virtual method called

Renzo Tomaselli renzo.tomaselli at tecnotp.it
Wed Nov 12 17:11:55 GMT 2003


Gary,
    our issue looks pretty different. We can see no problems in catching
exceptions even across shlibs. The point is that they *disappear* (or at
least their virtual table is invalidated) after a while, e.g. after being
caught. There is certainly a time factor involved, since waiting in a catch
for a couple of seconds grants such failure for sure (which appears as
random otherwise).
If we are able to marshal the exception back (or to rethrow a similar fresh
one), there's no problem at all. That's why I suspect about a concurrent
thread, although I can't figure out how such thread might interfere with
another's one stack, unless there is a big big mess in AIX thread library.
Or unless there's some delayed trouble in stack handling while
throwing/catching between two different shlibs.
Thanks,

Renzo Tomaselli


----- Original Message -----
From: "Gary Duzan" <gduzan at bbn.com>
To: "Renzo Tomaselli" <renzo.tomaselli at tecnotp.it>
Cc: "Omniorb list" <omniorb-list at omniorb-support.com>
Sent: Wednesday, November 12, 2003 4:21 PM
Subject: Re: [omniORB] AIX: pure virtual method called


> In Message <025401c3a90f$44f66d70$250f0aa2 at Renzo> ,
>    "Renzo Tomaselli" <renzo.tomaselli at tecnotp.it> wrote:
>
> =>Hi,
> =>    we just ported our architecture to AIX 5.2 through g++ 3.3.1, while
> =>OmniORB version is 4.02, standard sources. The same runs fine on Windows
> =>(msvc 6.0), Solaris (g++ 3.2.3), Linux (3.2.0).
> =>Apparently everything was ok on AIX too, but then we got random failures
> =>after throwing user exceptions. In short, an user exception is thrown
from a
> =>servant (say throw NotFound()) and caught by the caller in the same
process
> =>(although in a different shlib).
>
>    gcc on AIX has had problems with throwing exceptions out of shared
> libraries in the past, and it looks like this one popped up again:
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11111
>
> The bug is marked as resolved, so you might try a newer version of the
> compiler and see if it helps.
>
> Gary Duzan
> BBN Technologies
> A Verizon Company
>
>
>
> _______________________________________________
> omniORB-list mailing list
> omniORB-list at omniorb-support.com
> http://www.omniorb-support.com/mailman/listinfo/omniorb-list
>




More information about the omniORB-list mailing list