Compiling/running Policy Agent 2.2 for Apache (Linux)

Question

Hi,
I know that running the policy agent for Linux is not supported by Sun on other platforms than Red Hat Enterprise Linux, but anyway I'm qurious to see if others have looked into this.
I've done some testing on my Ubuntu Dapper Linux, using the precompiled version for Red Hat Enterprise, and I kind of made it work. Only problem: I have to start apache using "strace" to have it running. If I run /usr/sbin/apache2 I get "Segmentation Fault", but if I run "strace /usr/sbin/apache2" it runs... I'm able to create a core-dump, but to get something out of it I guess I have to compile the policy agent myself, so I've tried that as well.
To compile I've checked out the opensso package by CVS and installed libxml2-2.6.23, nss-3.11, nspr-4.6.1 and apache-2.0.59, sort of like what it says in the Readme for compiling under Red Hat Enterprise Linux. Result from running "make BUILD_DEBUG=optimize BUILD_AGENT=apache" is:
hash_table.h: In member function �typename smi::HashTable<Element>::EntryType smi::HashTable<Element>::findEntry(const std::string&)�:
hash_table.h:319: error: expected �;� before �__null�
hash_table.h:319: warning: statement has no effect
The test with the precompiled version was done on Ubuntu Dapper Linux using the standard apache 2.0.55 package that comes with Ubuntu. As I said: I've managed to get it running (doing SSO login through Federation Manager running on the same machine, with Access Manager running on another Solaris server) but I would prefer to have a setup that doesn't involve using "strace" to have apache whith the policy agent module running... Anyone else done something like this?
In the end I guess I would like to have some kind of release of the policy agent that doesn't have to be packaged as RPMs just for Red Hat Enterprise servers. It doesn't have to be flagged as "supported by Sun" but more like "you're on your own this release". ;-) That goes for the Federation Manager as well. I've managed to have the FM running on Ubuntu Dapper as well, so I know it's possible...
- Anders

Answer 1

The notes in this thread date from about October of 2006.
Does anyone know why current versions of the gcc compiler refuse to compile the statement that leads to the reported error?
The statement that is failing is:
    if (entry && entry->getExpirationTime() < PR_Now()) {
     return (HashTable<Element>::EntryType)NULL;with the error
[exec] hash_table.h:320: error: expected �;� before �__null�Is this a problem with the compiler or with the definition of the NULL macro?