[AGENT++] MibTable and subagents

[Razvan Cojocaru]

>> Unfotunately when I try to snmpwalk the table, nothing shows up. It's as
> 
> Have you checked that the table has been registered successfully at
> the master agent?

Not as such, but I have checked the subagent logs and there were no
errors I could see in there.

Also, I did check that update() _is_ being called on the
MibTable-derived class instance I'm using. I've checked that in the
debugger, and add_row() was called, the indices I'm using check out, and
the MibTableRow pointer that add_row() receives is not NULL. Still, the
table appears empty to snmpwalk (and snmpd, apparently). Again, none of
this stuff happens to simple MibLeaves, and trap sending works fine.


> The NET-SNMP master agent does not handle search ranges correctly.
> The NET-SNMP subagent does not handle search ranges correctly either,
> but I have implemented a workaround for that in the AgentX++ master.
> Thus, the better working combination might be AgentX++ master and
> NET-SNMP subagent - although this depends on your system and
> requirements.

Well, my requirements are "universal subagent", so a subagent that does
not work with a particular major master agent is not acceptable. I'm not
interested in using AgentX++ as a master, only in subagents and
standalone agents. I'm willing to debug the problem until it's solved if
you think that helps - just let me know what I can do besides whatever
obvious stuff I might think of. For now, I think I'll try to use
AgentXSharedTable and see if it works, and if not, I'll use an AgentX
master agent and see if it works like that, at least that will narrow
the problem area down.


> By theory, the table implementation should be the same for
> a subagent and stand-alone agent. If does not work with
> NET-SNMP counter-part, the concrete reasons for that need to
> be debugged and solved.

First thing tomorrow morning. I'll report back the results.
Thanks.