Running the DALTON test suite

To check that DALTON has been successfully installed, a fairly elaborate automatic test suite is provided in the distribution. A test script, all the test jobs and reference output files can be found in the dalton/test directory. It is highly recommended that all these tests be run once the program has been compiled. Depending on your hardware, this usually takes 2--12 hours.

The tests can be run one by one or in groups, by using the test script TEST. Try TEST -h to see the different options this script takes. Also, have a look at CONTENTS for short descriptions of the various tests. To run the complete test suite, simply go to the dalton/test directory and type:

> ./TEST all

You can follow the progress of the tests directly, but all messages are also printed to a log (TESTLOG by default). After all the tests have completed you should hopefully be presented with the message ``ALL TESTS ENDED PROPERLY!''. If not, you will be given a list of the tests that failed to run correctly. Please consult the file KNOWN_PROBLEMS too see if these tests have documented problems on your particular platform.

Any tests that fail will leave behind the .mol and .dal input-files (these are described in more detail in Chapter 5), and the output-file from the test calculation which will have the extension .log (DALTON output-files usually have the extension .out). For all successful tests these files, as well as some other auxiliary files, will be deleted as soon as the output has been checked, unless TEST is being run with the option -keep.

If most of the tests fail, it is quite likely that there's something wrong with the installation. Look carefully through Makefile.config, and consider turning down or even off optimization.

If there's only a few tests that fail, and DALTON seems to exit normally in each case, there may just be some issues with numerical accuracy. Different machines give slightly different results, and while we've tried to allow for some slack in the tests, it may be that your machine yields numbers just outside the intervals we've specified as acceptable. A closer comparison of the results with numbers in the test script and/or the reference output files should reveal whether this is actually the case. If numerical (in)accuracy is the culprit, feel free to send your output-file(s) to dalton-admin@kjemi.uio.no so that we can adjust the numerical intervals accordingly.