libstdc++/9556: problem in dynamic casting with gcc 3.2.1

[lchandra at pactolus dot com]

>Number:         9556
>Category:       libstdc++
>Synopsis:       problem in dynamic casting with gcc 3.2.1
>Confidential:   no
>Severity:       critical
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Feb 03 20:36:00 UTC 2003
>Closed-Date:
>Last-Modified:
>Originator:     Lokesh Chandra
>Release:        gcc 3.2.1
>Organization:
>Environment:
Sun OS 5.8
>Description:
We are having run time problems with the code built in gcc version 3.2.1 on solaris systems. However the same source code works fine with gcc version 2.95.3 and other compilers like VC++.


Brief synopsis about the code:-

There are two dynamically loaded libraries. An object ( PS_MyMsg ) is created in one of the library (my_msg_lib.cpp) and "smart" pointer ( PCA_Ptr<IPSMessage> ) to one of object's interfaces is passed on to the other library(my_pac_lib.cpp). The second library (my_pac_lib.cpp) tries to cast this pointer to another interface ( PCA_Ptr<IPSMsgParameters> ) of that object and fails.

There is a commented section in my_main.cpp (class RuntimeSite). This section does the same casting as the second library(my_pac_lib.cpp). If this section is uncommented, casting succeeds here and in such case, casting succeeds in the second library (my_pac_lib.cpp) too. 

If this section is commented then the test fails to cast in the second library (my_pac_lib.cpp).
>How-To-Repeat:
1. unzip the files to any directory.
2. run make_me. This will generate two .so files and my_test executable.
3. run my_test and it fails to do dynamic casting.
4. The same set of files could be built with gcc version 2.95.3 and run successfully.
>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted:
----gnatsweb-attachment----
Content-Type: application/x-gzip-compressed; name="lib_test_mod.tar.gz"
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename="lib_test_mod.tar.gz"
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