> > What happens if we keep calling syslog in this situation? If we keep calling syslog, the vlog module will keep throwing error to stderr saying 'No handlers could be found for logger "syslog"'
>That is, I > can see that this is an optimization, but I don't know what prompted it. The background is as follows: * I added a python unit test with the commit d69d61c7c (vlog: Ability to override the default log facility). * The unit test would pass on local development machine but would fail in chroot environment (where VMware's build systems run). On investigation, I found that chroot environment does not have access to "/dev/log". So this commit provided a way to avoid test failure when there is no "/dev/log" So far (before my commit through which a new function add_syslog_handler() got added), python vlog handled lack of syslog logging (along with any other issues while initializing 'file' or 'console' logging) by increasing the severity level for that destination to "EMER". That way, it won't throw any errors for repeated attempt at logging (unless the logging was done with the severity level "EMER".). I think it is a clever trick. Now that I think of it, I should probably abandon this change and follow the same strategy to the newly added function too. > > Linux uses /dev/log, what about *BSD? "/dev/log" has been used in python vlog for more than 2 years now. Since BSD tests have been passing, I think they work fine there. _______________________________________________ dev mailing list [email protected] http://openvswitch.org/mailman/listinfo/dev
