I just committed a change to the trace levels:

TRACE_STOP is gone, because it's the same as TRACE_FATAL; I can't
imagine why we needed two such levels. The place I might have broken
something is in main.c, where "no such user" was calling TRACE_STOP in
order to give the sysexit 75 code. It was a horrible way to code that.

TRACE_MESSAGE / TRACE_WARNING switch places, jump around TRACE_ERROR. So
it should work like this:

 o TRACE_FATAL is hard to silence because your process just died
   and you should know why.
 o TRACE_WARNING is a serious problem that will probably lead to
   things breaking pretty soon.
 o TRACE_ERROR is an error that we've worked around. No biggie.
 o TRACE_MESSAGE is something you might want to know about, e.g. that
   a connection is being handled, that a message was delivered, etc.
 o TRACE_INFO is something you don't care to know about.
 o TRACE_DEBUG is just too much information.

There are plenty of places where calls are made using the wrong levels,
and probably a big range of opinions on where we draw the line between
MESSAGE and INFO.

Aaron


Reply via email to