Logging

The Acts logging facility supports several severity levels which allow you to control the amount of information displayed at run-time. Logger objects can easily be created using the Acts::getDefaultLogger() function which should be sufficient to get you started. In case you need more customized debug output, you can make use of the output decorators defined in Acts::Logging or even write your own implementation of Acts::Logging::OutputDecorator. In order to add debug messages to your program, you should use the provided macros for the different severity levels:

ACTS_VERBOSE(...);
ACTS_DEBUG(...);
ACTS_INFO(...);
ACTS_WARNING(...);
ACTS_ERROR(...);
ACTS_FATAL(...);

These macros correspond to the available log levels:

enum Acts::Logging::Level

Values:

enumerator VERBOSE
enumerator DEBUG
enumerator INFO
enumerator WARNING
enumerator ERROR
enumerator FATAL
enumerator MAX

The macros require that a function logger() returning a Acts::Logger object is available in the scope in which the macros are used. Inside classes containing an Acts::Logger object as member variable, this could be achieved by providing a private class method called logger() (for an example see e.g. Acts::CylinderVolumeBuilder::logger()). Inside free functions or member methods with local logger objects, the same effect can be achieved by using the macro ACTS_LOCAL_LOGGER(...) which is provided for your convenience.

Code example illustrating the usage:

#include <fstream>
#include <memory>

#include "Acts/Utilities/Logger.hpp"

void myFunction() {
  // open the logfile
  std::ofstream logfile("log.txt");
  // setup a logger instance for >= INFO messages, streaming into the log file
  // make sure you do NOT call the variable 'logger'
  std::unique_ptr<const Acts::Logger> myLogger
      = Acts::getDefaultLogger("MyLogger", Acts::Logging::INFO, &logfile);
  // make sure the Acts debug macros can work with your logger
  ACTS_LOCAL_LOGGER(myLogger);
  ACTS_VERBOSE("This message will not appear in the logfile.");
  ACTS_INFO("But this one will: Hello World!");
  // do not forget to close the logfile
  logfile.close();
}

Logger integration

In case you are using Acts in another framework which comes with its own logging facility (e.g. Gaudi) you can pipe the logging output from Acts tools and algorithms to your framework’s logging system by supplying different implementations of:

There are two approaches to logger integration:

  1. Overriding Acts::getDefaultLogger(). This has the downside that log levels cannot be controlled from top-level experiment specific code. This means that it is non-trivial to steer the log level of an e.g. Gaudi algorithm via the OutputLevel property, and have the ACTS code respect this log level. It is therefore now discouraged to use this approach.

    Note

    ACTS code has iteratively moved to not construct loggers via getDefaultLogger as much as possible, in favor of using Acts::LoggerWrapper. The latter is a lightweight non-owning wrapper around a logger instance, that can default to not containing a logger. It is more suitable to pass into functions that might be called from other ACTS functions (rather than construction a local logger via getDefaultLogger).

  2. Passing logger instances to high level components, and rely on ACTS code to pass them into lower level classes / functions.

Overriding Acts::getDefaultLogger()

Attention

Using this mechanism is now discouraged for integration with an experiment framework.

Since Acts makes extensive use of Acts::getDefaultLogger() to provide sufficient information for debugging, you might want to provide a modified implementation of this function (using your output filter and printing policies) to also pipe this output to your framework. You can use the following approach using the possibility to inject custom code by pre-loading shared libraries with LD_PRELOAD. You need to provide an appropriate implementation for a function of the following signature into a separate source file and compile it in a shared library

namespace Acts {
std::unique_ptr<const Logger> getDefaultLogger(const std::string&,
                                               const Logging::Level&,
                                               std::ostream*);
}

Then you can run your executable, which uses Acts tools and algorithms, in the following way (tested under Unix)

$ LD_PRELOAD=<YOUR_SHARED_LIBRARY> path/to/your/exectuable

For an example have a look at CustomDefaultLogger.cpp which you can use as follows:

$ cd <ACTS/INSTALL/DIRECTORY>
$ source bin/setup.sh
$ LD_PRELOAD=lib/libActsCustomLogger.so bin/Examples/ActsGenericDetector

Logging thresholds

Generally, log levels in ACTS are only of informative value: even Acts::Logging::ERROR and Acts::Logging::FATAL will only print a messages, and not terminate execution.

This is desirable in an experiment context, where jobs should not immediately terminate when ACTS encounters something that is logged as an error. In a test context, however, this behavior is not optimal: the tests should ensure in known configurations errors do not occur, or only in specific circumstances. To solve this, ACTS implements an optional log threshold mechanism.

The threshold mechanism is steered via two CMake options: ACTS_ENABLE_LOG_FAILURE_THRESHOLD and ACTS_LOG_FAILURE_THRESHOLD. Depending on their configuration, the logging can operate in three modes:

  1. No log failure threshold exists, log levels are informative only. This is the default behavior.

  2. A compile-time log failure threshold is set. If ACTS_ENABLE_LOG_FAILURE_THRESHOLD=ON and ACTS_LOG_FAILURE_THRESHOLD=<LEVEL> are set, the logger code will compile in a fixed check if the log level of a particular message exceeds <LEVEL>. If that is the case, an exception of type Acts::ThresholdFailure is thrown.

  3. A runtime log failure threshold is set. If only ACTS_ENABLE_LOG_FAILURE_THRESHOLD=ON and no fixed threshold level is set, the logger code will compile in a check of a global runtime threshold variable.

Note

If only ACTS_LOG_FAILURE_THRESHOLD is set, ACTS_ENABLE_LOG_FAILURE_THRESHOLD will be set automatically, i.e. a compile-time threshold will be set

Two functions exist to interact with the failure threshold:

Level Acts::Logging::getFailureThreshold()

Get debug level above which an exception will be thrown after logging.

All messages with a debug level equal or higher than the return value of this function will cause an exception to be thrown after log emission.

Note

Depending on preprocessor settings ACTS_ENABLE_LOG_FAILURE_THRESHOLD and ACTS_LOG_FAILURE_THRESHOLD, this operations is either constexpr or a runtime operation.

void Acts::Logging::setFailureThreshold(Level level)

Set debug level above which an exception will be thrown after logging.

All messages with a debug level equal or higher than level will cause an exception to be thrown after log emission.

Note

This function is only available if ACTS_LOG_FAILURE_THRESHOLD is unset, i.e. no compile-time threshold is used. Otherwise an exception is thrown.

Warning

The runtime log failure threshold is global state, therefore this function is not threadsafe. The intention is that this level is set once, before multi-threaded execution begins, and then not modified before the end of the job.