Experimental Physics and
Industrial Control System

Benjamin Franksen <[email protected]> · Tue, 4 Jun 2013 14:45:02 +0200

Hi Emmanuel

On Monday, June 03, 2013 15:36:33 Emmanuel Mayssat wrote:
> I upgraded to the latest version of seq (2-1-12) from (2-0-12) thinking that
> the version I use could be out of dateTurns out, seqLog is not part of the
> libseq.so library anymore.... is seqLog obsolete?

Interesting: you are the first user who mentions that.

There are two reasons I remove it in version 2.1. The first: it takes a
(program-wide) lock during the actual write to the log file. This was an
acceptable implementation back when SNL programs were single threaded (seq
versions 1.9.x). But in a multi-threaded program this is a very bad idea,
because writing to a file can take an arbitrarily long amount of time (think
of VxWorks using rsh file access). So each call to seqLog potentially blocks
the whole program i.e. all its state sets.

The second reason is that the standard EPICS errlog facility can be used to
much the same effect (and note that it uses a much saner locking policy), i.e.
a simple wrapper around errlogPrintf, perhaps using a CPP macro to add the
state set and/or program name to the message. Admittedly, you cannot log
different program instances into different files, but if you incorporate a
suitably unique string into the messages then filtering them out of the IOC-
wide log file can be done easily with standard unix command line tools such as
grep.

Cheers
--
Ben Franksen
()  ascii ribbon campaign - against html e-mail
/\  www.asciiribbon.org   - against proprietary attachm€nts

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System