Note: I abandoned this project ages ago when it became clear that no one besides myself here at work was interested in using this, and I haven't been writing any new Fortran lately. It is incomplete; use it at your own risk.

This unit testing framework aims to be nicer and more flexible than existing unit testing frameworks for Fortran.

Edit the Makefile for your system. Type +make+. If this succeeds it will produce a command +funit+. Place this somewhere in your path.

I tried to learn autotools once and gave up. I may write a configure script at some point, but if you know how, I would accept a contribution to the project.

Write your tests (described below) in a file ending in '.fun'. For now, you will have to generate the Fortran test code file with funit, then compile it and run the test program yourself, e.g.:

$ funit -o test.f90 test.fun
$ gfortran -o test test.f90
$ ./test
 Running all macros
  test all macros   FAILED
  'an .and. expression' is false

Finished in 0.00 seconds
1 tests in 1 sets, 1 failures

In the future, this will be handled more automatically.

Tests are written with special syntax around Fortran code implementing each test. The special syntax also specifies dependencies which need to be built, setup and teardown routines, etc. By convention, these are saved in files ending in ".fun".

set set-name
  dep "../file1.F90"
  dep "../file2.F90"
  use a_module

  tolerance 0.00001

  setup
    ! fortran code to run before each test
  end setup

  teardown
    ! fortran code to run after each test
  end teardown

  test case1
    ...
  end test case1

  test case2
    ...
  end test case2
end set

Note: dependencies must be quoted and unlike Fortran, the strings must not be continued with an ampersand (&).

• assert_true(expression[, msg])
• assert_false(expression[, msg])
• assert_equal(a, b[, msg])
• assert_not_equal(a, b[, msg])
• assert_equal_with(a, b[, tol][, msg])
• assert_array_equal(a, b[, msg])
• assert_array_equal_with(a, b[, tol][, msg])
• flunk(msg)

The array assertions should be given array variable names rather than complex expressions because their arguments are evaluated several times.

FUnit looks for config files in three locations:

1. +.funit+ in the current directory,
2. +.funit+ in the user's home directory (via the +HOME+ environment variable),
3. and finally +/etc/funitrc+.

The first time a config value is seen, that is its value. This way, the current directory's +.funit+ will override values in +$HOME/.funit+ which in turn will overrid values in /etc/funitrc.

The syntax is like many Unix-style rc files: comments starts with '#' and continue to the end of the line; configuration settings are made with unquoted values like

key = value

or quoted values like

key = "value"

or

key = 'value'

Unquoted values cannot contain spaces, tabs or the comment character ('#'). Quoted values may contain spaces, tabs or '#', but they have no way of escaping the character that opened the quotation (a quotation mark or apostrophe). They may contain the other quotation character however.

The following configuration keys are recognized:

build = "BUILD COMMAND"

default: TDB example:

fortran_ext = .EXT

default: .f90 example:

template_ext = .EXT

default: .fun example:

Note: this is not working yet!

$ funit test/test_XXX.fun

Building set-name...
Running set-name
  test test-name        PASSED
  test name2            PASSED
  test third            FAILED

Finished in 2.3 seconds
3 tests in 1 set, 1 failures