Re: fanout processing and timed-out records

[Andrew Johnson <anj@aps.anl.gov>]

Emmanuel Mayssat wrote:
> I have noticed something strange with the fanout record (selective
> processing record).
> My fanout has for input 6 asynchronous records.

The fanout record type doesn't have any input links, and the dfanout 
record type only has one â did you mean to ask about the sequence record 
type (seq)?

> If one of the record timeout, the fanout stops processing any further.
> So if one of the first records times out, the following ones aren't
> processed at all. 

This is the relevant code from the process() routine in fanoutRecord.c, 
which has not changed since Version 3.14.2:

    switch (pfanout->selm){
    case (fanoutSELM_All):
        plink=&(pfanout->lnk1);
        state=pfanout->seln;
        for ( i=0; i<6; i++, state>>=1, plink++) {
            if(plink->type!=CONSTANT) dbScanFwdLink(plink);
        }
        break;

When SELM="All" there are no conditions, it always runs through all six 
forward links and scans them all (the link type will never be CONSTANT 
if you have connected that link to another record).  Looking at it the 
'state' variable is not actually needed at all, but doesn't affect what 
the record does at all, which is to process all its forward links.

> Now if you have a fanout of fanout records, you may end-up only having a
> few record triggered.
>
> Has anybody seen this ?
> Is there a way to force the fanout to process all of its INP records?
> Is there another reliable way to do selective processing?

- Andrew
--
When a distinguished but elderly scientist states that something is
possible, he is almost certainly right.  When he states that something
is impossible, he is very probably wrong.  -- Arthur C. Clarke