Experimental Physics and
Industrial Control System

Ben Franksen <[email protected]> · Mon, 5 May 2008 21:46:35 +0200

On Montag, 5. Mai 2008, J. Lewis Muir wrote:
> I would be happy to submit a 
> patch to make errlog.c only print to stderr.  However, I'm not sure of
> the intent of this logging facility.  Obviously the name would indicate
> that everything it logs is an error.  If this is truly the intent, then
> I think it's reasonable to change it to only print to stderr.  However,
> if the name is kind of out of date and it's really meant to be a general
> logging facility, not necessarily just for logging errors, then I'm not
> sure how others would feel if all messages went to stderr.  I'd be fine
> with it though.

The name 'errlog' is historical; the errlog facility can and should be used 
for any kind of message that should appear on the console (IOC shell) /and/ 
be logged (if logging is turned on). It is not only for errors, witness the 
errlogSevPrintf procedure which takes a severity as argument, the lowest of 
which is 'info' i.e. not even a warning, just informational content. (This 
procedure is, btw, seriously underused, IMHO.)

As regards the output channel messages should be sent to (in addition to the 
log server), this should IMO be configurable; from the ioc shell and also 
directly from code. However, as default I'd strongly prefer stderr over 
stdout. Note that stderr is traditionally used not only for error messages, 
but for all kinds 'meta messages', i.e. information /about the process/ of 
computing some result (that itself should go to stdout) or producing the 
desired effect. This includes debug messages (if turned on), progress 
report, warnings, errors, etc. The intention is that computed output (if 
any) doesn't become polluted with such messages and thus can be redirected 
to a file or some other process without extra effort to separate it from 
such messages.

Cheers
Ben

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System