Skip to content

DEVELOPER.md

Latest commit

 

History

History
39 lines (31 loc) · 2.08 KB

DEVELOPER.md

File metadata and controls

39 lines (31 loc) · 2.08 KB

Developer Guide

Tests

The directory Exmpl_Bench contains a bunch of test and development. They are typically work in progress and not necessarily in a state of being finished.

The directory CI_tests contains some tests that are run in CI on GitHub. Their runtime should be short, from 5 minutes to 1 hour. Together they should cover as much as possible of the functionality of the code. If the test is working like normally, then it should be scheduled in the cron to run once per month. We do not want to overflow the credit that we have.

Compilation options related to debug

There are several environment variables that can be used during runtime.

  • DEBUG for making some print statements and making some checks. Those checks have to be fast.
  • KEY_VALUE for printing some KEY(....) VALUE=(....) that can be used for postprocessing of the options and heuristic optimization.
  • TIMINGS for printing some runtime information.
  • SANITY_CHECK for doing some checks and stopping if incoherence are detected.
  • SANITY_CHECK_EXTENSIVE for doing some long running checks and stopping if incoherence are detected. Since those are long running, a print statement must mark their beginning and another one their end.
  • TRACK_INFO for printing stuff for further work down the line.
  • METHOD_COMPARISON for considering two (or more) different methods and comparing their performance (mainly runtime but not only).

The options TIMINGS and DEBUG enable all the timings and debugging statements. For a more granular debugging, stuff like DEBUG_LINEAR_PROGRAM can be used. See the top of the header files.

A printout to std::cerr should occur if an error has been identified and the program will terminate with a call to TerminalException. Other print statement should be encapsulated in std::ostream & os that should be passed by reference from the initial call. So typically for serial output we pass the std::cerr while for the parallel runs we pass a stream to the output of that process. That way we avoid mixing between different sources.