Experimental Physics and
Industrial Control System

"J. Lewis Muir via Tech-talk" <tech-talk at aps.anl.gov> · Fri, 27 Mar 2020 10:28:27 -0500

On 03/27, Konrad, Martin wrote:
> My recommendation would be to print all messages to the same stream.

I suppose some other options could be:

1. Disable stdout and stderr buffering on the IOC console.

2. Add an fflush call after any legitimate printf/fprintf in EPICS and
   EPICS-related code.

3. Add special console_{printf,fprintf} functions (or whatever they
   should be named) that will do the printf/fprintf, and then do an
   fflush.  Then say that all code that would have used printf/fprintf
   should use these instead.

4. A thread-safe option (If the only callers of printf/fprintf are
   guaranteed to be running in a single-process/single-threaded context,
   then this is unnecessary): Add special console_{printf,fprintf}
   functions (or whatever they should be named) that are thread-safe
   (i.e., can be called concurrently from two or more processes or
   threads) that will acquire a console mutex (or whatever it should
   be named), do the printf/fprintf, and then do an fflush.  Then say
   that all code that would have used printf/fprintf should use these
   instead.

5. Come up with another logging mechanism for the legitimate uses of
   printf/fprintf that is like the errlog facility and will guarantee
   that output will not be garbled, but that should be used instead
   of printf/fprintf for the cases where using the errlog facility is
   considered inappropriate.  Then say that all code that would have
   used printf/fprintf should use these instead.

Lewis

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System