Experimental Physics and
Industrial Control System

Benjamin Franksen <[email protected]> · Fri, 21 Sep 2007 12:30:18 +0200

On Thursday 20 September 2007 16:16, Eric Norum wrote:
> It appears to be faulting inside the RPC library clnttcp_create
> routine.  Are there any diagnostics you can enable for that library?
> Do you have the source so you could check what's happening in the
> library code?

Having a hard time to believe that the RPC implementation is at fault here, 
I pursued a different path and with success. The backtrace is misleading 
(or should I say, wrong?), since the point where the IOC crashes is really 
in clnt_destroy, not in clnttcp_create.

The fault, it seems, is in the routine vxiConnectPort (in 
asyn/vxi11/drvVxi11.c). vxiConnectPort calls clnttcp_create and afterwards 
calls vxiBusStatus. If vxiBusStatus fails, vxiConnectPort calls 
ctln_destroy. However, if vxiBusStatus fails it has already called 
clnt_destroy, since it is implemented using clntCall, which calls 
vxiDisconnectPort on failure, which calls clnt_destoy.

Thus, clnt_destroy gets called a second time, causing the crash.

I have veryfied all this multiple times by adding enough breakpoints (all 
call sites of clnt_create (1) and clnt_destroy (7) in 
asyn/vxi11/drvVxi11.c).

I have not produced a fix for this problem. I guess you (Eric) know better 
how to approach this.

Cheers
Ben

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System