// $Id$
//
-// Copyright (C) 2007
-// Fraunhofer Institut fuer offene Kommunikationssysteme (FOKUS)
-// Kompetenzzentrum fuer Satelitenkommunikation (SatCom)
+// Copyright (C) 2007
+// Fraunhofer Institute for Open Communication Systems (FOKUS)
+// Competence Center NETwork research (NET), St. Augustin, GERMANY
// Stefan Bund <g0dil@berlios.de>
//
// This program is free software; you can redistribute it and/or modify
/** \file
\brief Config public header */
-#ifndef HH_Config_
-#define HH_Config_ 1
+#ifndef HH_SENF_Utils_Logger_Config_
+#define HH_SENF_Utils_Logger_Config_ 1
// Custom includes
#include "Levels.hh"
<em>Runtime</em> configuration on the other hand deals with routing all those messages, which
are enabled at compile time to the logging targets. If a message is not routed, it will be
- discarded. This allows to additionally disable messages at run-time.
-
- \fixme Restructure compile-time configuration: Only allow stream based configuration using
- SENF_LOG_CONF. For more complex configuration, accept a macro SENF_LOG_CONFFILE which, if
- defined, must define the path of a file to be included. Area specific configuration must be
- done in this include file. The area must be complete (not only a predeclaration) at that
- point.
+ discarded. This allows to additionally disable messages at run-time. Message routing is managed
+ via the ::Target interface.
+
+ \section config_compile Compile time configuration
+
+ Compile time configuration is set on the compiler command line:
+ <pre>
+ g++ ... -DSENF_LOG_CONF="(( (senf)(log)(Debug),(_),DISABLED ))
+ (( (senf)(log)(Debug),(foo)(SomeClass),VERBOSE ))
+ (( (_),(_),NOTICE ))" ...
+ </pre>
+ The value is relatively complex; It's a Boost.Preprocessor style sequence of tuples, of which
+ the first and second elements are again sequences. What this boils down to, is that it allows to
+ configure compile time logging limits based on stream and optional area.
+
+ The above example disables all debug logging by setting the default log limit for all areas on
+ the \c senf::log::Debug stream to \c DISABLED. It enables debug logging only within the \c
+ foo::SomeClass area, where it is set to \c VERBOSE. Lastly, the global compile time limit is set
+ to \c NOTICE.
+
+ There are two standard uses for this configuration: Either to disable most logging in final
+ builds by changing the compile time limit to something like senf::log::IMPORTANT or to enable
+ senf::log::VERBOSE messages for some area:
+ <pre>
+ # Disable debug logging below 'IMPORTANT' level
+ g++ ... -DSENF_LOG_CONF="(( (senf)(log)(Debug), (_), IMPORTANT ))"
+
+ # Or enable verbose messages for the 'some::Area' area
+ g++ ... -DSENF_LOG_CONF="(( (senf)(log)(Verbose), (some)(Area), VERBOSE ))"
+ </pre>
+
+ All the entries specified via \c SENF_LOG_CONF are applied in a fixed order:
+
+ \li First the entries which have both a stream and an area specified are checked
+ \li next all entries with area but no stream given are checked
+ \li followed by all entries with a given stream but no area
+ \li and lastly if no match was found until now, a generic entry without stream and area is
+ checked
+ \li if no matching entry is found, the default compile time limit of the stream is used
+
+ So an area specification has precedence over a stream specification.
+
+ \warning Enabling a message at compile time does \e not ensure, the message is shown. You
+ additionally need to \e route the message (see next chapter). This is especially true for \c
+ VERBOSE messages, which are default disabled at runtime.
+
+ \see \ref SENF_LOG_CONF
+
+ \section config_runtime Runtime configuration
+
+ The runtime configuration is performed by routing messages to one or more logging targets:
+ \code
+ senf::log::ConsoleTarget & consoleLog (senf::log::ConsoleTarget::instance());
+ senf::log::FileTarget fileLog ("my.log");
+
+ consoleLog.route<senf::log::Debug>();
+ consoleLog.route<foo::Transactions, foo::SomeClass>(senf::log::Target::REJECT);
+ consoleLog.route<foo::Transactions, senf::log::IMPORTANT>();
+
+ fileLog.route<foo::Transactions>();
+ \endcode
+ Here we see an already relatively complex setup: All debug messages (that is, those, which are
+ not disabled at compile time) are routed to the console. We also route important transactions to
+ the console \e except transactions from the \c foo::SomeClass area. The \c fileLog simply
+ receives all transaction log messages.
+
+ The routing statements are processed by the targets in order, the first matching rule will
+ decide a log messages fate for that target.
+
+ \warning You can \e only route those messages at runtime which have been compile-time
+ enabled. By default, \c VERBOSE messages are \e disabled at compile time. They must be
+ enabled \e explicitly by setting \c SENF_LOG_CONF so they can be routed.
+
+ \section config_fallback Fallback routing
+
+ There are two cases, where this setup may lead to inadvertently lost log messages:
+ \li When using a library which does internally use the Logger but not initializing the logger in
+ your application.
+ \li When log messages are created during initialization of static objects.
+ Since no route is set up in these cases, the messages will be dropped.
+
+ To counter this problem, the logger is initially in <em>fallback routing</em> state. If any log
+ message arrives in this state, the message will be logged to the console if it is above the
+ default runtime limit of it's stream. The first routing statement on any target will take the
+ logger out of this state and normal routing will take place.
+
+ \see \ref senf::log::Target
+
+ \section config_timesource Log message timing
+
+ One auxiliary aspect of logging is message timing. Each message is stamped with a time-stamp
+ giving the exact time the message was created. How the current date/time value is created may be
+ changed by setting a \e TimeSource. A TimeSource is an instance derived from
+ senf::log::TimeSource which will return the current universal time (UTC) when called.
+
+ By default, the logging library will call gettimeofday() for each log message. To change the
+ time source, just pass the new class or instance to senf::log::timeSource:
+ \code
+ // Use senf::scheduler::eventTime() to time log messages
+ senf::log::timeSource<senf::scheduler::LogTimeSource>();
+ \endcode
*/
namespace senf {
/** \brief Compile time configuration
This define symbol sets the compile time logger configuration. This symbol should normally
- be set on the compiler command line:
- <pre>
- g++ ... -DSENF_LOG_CONF="(( (senf)(log)(Debug),(_),DISABLED ))
- (( (senf)(log)(Debug),(foo)(SomeClass),(VERBOSE) ))
- (( (foo)(Transactions),(_),NOTICE ))" ...
- </pre>
- (As this option can get quite long, you might want to use the '-imacros' option instead)
+ be set on the compiler command line.
The formal syntax of this option is:
\par ""
<table class="ebnf">
- <tr><td>conf</td> <td>::= \e element \e element* \n</td></tr>
- <tr><td>element</td> <td>::= <tt>((</tt> \e stream <tt>,</tt> \e optional_area <tt>,</tt> \e level <tt>))</tt> \n</td></tr>
- <tr><td>stream</td> <td>::= \e scope_seq \n</td></tr>
- <tr><td>optional_area</td><td>::= <tt>(_)</tt> | \e scope_seq \n</td></tr>
- <tr><td>level</td> <td>::= \c VERBOSE | \c NOTICE | \c MESSAGE | \c IMPORTANT | \c CRITICAL | \c DISABLED \n</td></tr>
- <tr><td>scope_seq</td> <td>::= \e scope \e scope \e scope* \n</td></tr>
- <tr><td>scope</td> <td>::= <tt>(</tt> \e name <tt>)</tt> \n</td></tr>
- <tr><td>name</td> <td>::= arbitrary C++ identifier</td></tr>
+ <tr><td>conf</td> <td>::= \e element \e element* \n</td></tr>
+ <tr><td>element</td> <td>::= <tt>((</tt> \e optional_stream <tt>,</tt> \e optional_area <tt>,</tt> \e level <tt>))</tt> \n</td></tr>
+ <tr><td>optional_stream</td> <td>::= <tt>(_)</tt> | \e scope_seq \n</td></tr>
+ <tr><td>optional_area</td> <td>::= <tt>(_)</tt> | \e scope_seq \n</td></tr>
+ <tr><td>level</td> <td>::= \c VERBOSE | \c NOTICE | \c MESSAGE | \c IMPORTANT | \c CRITICAL | \c DISABLED \n</td></tr>
+ <tr><td>scope_seq</td> <td>::= \e scope \e scope* \n</td></tr>
+ <tr><td>scope</td> <td>::= <tt>(</tt> \e name <tt>)</tt> \n</td></tr>
+ <tr><td>name</td> <td>::= arbitrary C++ identifier</td></tr>
</table>
- \ref SENF_LOG_CONF is a Boost.Preprocessor style sequence of 3-tuples. Each tuple applies to
- a specific stream which is defined by the first tuple element \e stream.
+ \ref SENF_LOG_CONF is a Boost.Preprocessor style sequence of 3-tuples.
+
+ The first tuple element \e optional_stream specifies the stream to match. If this is
+ <tt>(_)</tt>, the entry will match any stream.
The next tuple element, \e optional_area optionally restricts the entry to match only the
- given area.
+ given area. If set to <tt>(_)</tt>, the area is left unrestricted.
The last tuple element \e level defines the compile time log level. Messages with a level
below this are discarded at compile time.
- Both \e stream and \e optional_area are given as a \e scope_seq. A scope sequence is a fully
- qualified C++ identifier placed into a sequence: <tt>foo::bar::baz</tt> is represented by
- <tt>(foo)(bar)(baz)</tt>.
+ Both \e optional_stream and \e optional_area are given as a \e scope_seq. A scope sequence
+ is a fully qualified C++ identifier placed into a sequence: <tt>foo::bar::baz</tt> is
+ represented by <tt>(foo)(bar)(baz)</tt>.
*/
# define SENF_LOG_CONF
projects / senf.git / blobdiff