Some hints about Ceylan

• The test cases do not use the Log system in general (except of course for the Log system test!), in order to be as autonomous as possible, according to Kent Beck's principles.

• Source files are best viewed thanks to any text editor which is at least 100 character wide, set so that a tabulation is four character long. Ex: nedit, which can be launched thanks to nc (or neditc, or nedit-nc depending on the distribution), to avoid wasting too much memory.

  See our nedit configuration file, which we recommend to view our source code (one can put its content into ~/.nedit/nedit.rc to have it taken into account). One can use too our script called n, that allows to edit files with great simplicity.

• To make good use of the Log system within a main() function, one shall mimic the standard way of using it:

  
  
  	
  
  		
  
  		  
  
  			
  
  
  #include <iostream>    // for cerr, endl
  #include <exception>
  
  #include <string>
  using std::string ;
  
  
  #include "Ceylan.h"
  
  using namespace Ceylan ;
  using namespace Ceylan::Log ;
  
  
  /**
   * What my program does.
   *
   */
  int main( int argc, char * argv[] )
  {
  
  	LogHolder logger( argc, argv ) ;
  
  	try
  	{
  
  		LogPlug::info( "Starting my program" ) ;
  
  
  		// Blah blah blah
  
  
  		LogPlug::info( "End of my program." ) ;
  
  	}
  
  	catch ( const Ceylan::Exception & e )
  	{
  		std::cerr << "Ceylan exception caught: "
  			<< e.toString( Ceylan::high ) << std::endl ;
  		return Ceylan::ExitFailure ;
  
  	}
  
  	catch ( const std::exception & e )
  	{
  		std::cerr << "Standard exception caught: "
  			 << e.what() << std::endl ;
  		return Ceylan::ExitFailure ;
  
  	}
  
  	catch ( ... )
  	{
  		std::cerr << "Unknown exception caught" << std::endl ;
  		return Ceylan::ExitFailure ;
  
  	}
  
  	return Ceylan::ExitSuccess ;
  
  }
  
  
  			
  		  
  		
  	
  
  
  

• If an object inheriting from Ceylan::Object has chosen to track instances (trackInstance is true or not specified in Ceylan::Object's constructor call), and if it does not send explicitly messages in its life (using the send method), its name will be mangled in log aggregations. The reason for the mangling is that trackInstance will cause two log messages to be sent (thus the channel will be created), but one will be sent from the object's constructor, the other from the destructor. In these two cases, compilers like gcc will return mangled names for one's instance class name (such as sonata/PID-12363/N6Ceylan6ObjectE/0xbffff020) , which would have been automagically corrected by a Ceylan smart aggregator if a classical message had be sent.
• When running executables using LogHolder (which is the recommended way of using the log system), one has often a favorite log plug (generally, the HTML one) which is a bit painful to specify each time (for example by adding --HTMLPlug on the command line), so that the default plug, which is the classical (raw) one, is overriden. A convenient trick is to set an environment variable in your shell settings to define your plug (bash example: export PLUG="--HTMLPlug"), and use it accordingly: ./testObject.exe $PLUG.