Re: Asyn/VXI-11 problems

[Eric Norum <norume@aps.anl.gov>]

On Sep 21, 2007, at 8:36 AM, Benjamin Franksen wrote:

> On Friday 21 September 2007 12:30, Benjamin Franksen wrote:
> > I have not produced a fix for this problem. I guess you (Eric) know
> > better how to approach this.
>
> Well, my current, very crude, solution is to replace all (seven)
> clnt_destroy calls by
>
>     if (pvxiPort->rpcClient) {
>         clnt_destroy(pvxiPort->rpcClient);
>         pvxiPort->rpcClient = 0;
>     } else {
>         if(pasynUser) asynPrint(pasynUser,ASYN_TRACE_ERROR,
>             "%s internal error: RPC handle already
> destroyed\n",pvxiPort->portName);
>     }

Would an alternative fix be to change all the clnt_destroy  calls in  
vxiConnectPort to:
@@ -933,7 +934,8 @@
                     "%s vxiConnectPort neither system controller nor "
                     "controller in charge -- initialization aborted 
\n",
                     pvxiPort->portName);
-                clnt_destroy(pvxiPort->rpcClient);
+                if (pvxiPort->server.connected)
+                    vxiDisconnectPort(pvxiPort);
                 return asynError;
             }
         }

> In fact I can see the above 'internal error' messages appearing  
> when I run
> our test setup.
>
> There is yet another problem: the error checking after calls to  
> clnt_call si
> wrong. The code is
>
>     if(stat!=RPC_SUCCESS || errno!=0) {
>     	/* error handling */
> 	}
>
> which leads to pseudo errors; errno should be checked only if the  
> return
> value is abnormal. It should be just
>
>     if(stat!=RPC_SUCCESS) {
>     	/* error handling */
> 	}

Both the places where this code appears are preceded by assignments  
of 0 to errno.   The fact that this was done makes me think that the  
code was placed there for a reason.  There appear to be several  
substandard versions of RPC libraries out there.

> Also, the error message should not output the numerical value of  
> errno, but
> the number converted to a string (via strerror) as is done in all  
> other
> parts of the file.

  I agree and will change all the appropriate code to use strerror 
(errno) rather than just errno in the error messages.

> Cheers
> Ben

--
Eric Norum <norume@aps.anl.gov>
Advanced Photon Source
Argonne National Laboratory
(630) 252-4793