Unify the output of the test programs

[Alpar Juttner]

The rules on thumb:

• Print something to the output only if it really seems necessary. Check the validity of the results of the algorithms instead (using the check() macro).
• Do not print a message saying all test passed at the end of the output.

[Peter Kovacs]

It should be done all at once both in HG and SVN repositories.

[Akos Ladanyi]

What about introducing an environment variable (e.g. LEMON_TESTS_VERBOSE) and only print anything when it is set? So by default (when it is unset) the test programs would report their result only by their exit status. If somebody is interested in the details, then he could rerun the failed tests with LEMON_TESTS_VERBOSE set.

[Alpar Juttner]

Replying to kpeter:

It should be done all at once both in HG and SVN repositories.

Probably not. We should focus on the HG version as much as we can. Only critical bugs are to be fixed in the SVN version.

[Alpar Juttner]

Replying to ladanyi:

What about introducing an environment variable (e.g. LEMON_TESTS_VERBOSE) and only print anything when it is set?

It is quite unnecessary. If everything is OK, we don't need any output. When something goes wrong, the chech() macro reports the corresponding line (with emacs you can directly jump to that line by hitting C-x `), and also an error message. These two together should be enough.

[Peter Kovacs]

Replying to ladanyi:

What about introducing an environment variable (e.g. LEMON_TESTS_VERBOSE) and only print anything when it is set?

I think it is not important.

• If a test fails, one of the check() functions prints the proper message. That's enough.
• If a test passes, there is usually nothing intresting to be print. Maybe there is only one exception: when benchmark tests are also involved in the file and the results are printed.

By the way validity checks and benchmark tests should be separated and "make check" should not run benchmark tests.

[Peter Kovacs]

The changeset [02f4d5d9bfd7] solves this issue.