Skip to content
/ atto Public

Atto is the simplest-to-use C unit test framework, just one header file, no malloc, no fork, no dependencies

License

Notifications You must be signed in to change notification settings

TheMatjaz/atto

Repository files navigation

Atto - the microscopic C unit test framework

Build Status Master Build Status Develop GitHub release (latest by date) GitHub

Atto is the simplest-to-use C unit test framework, in just one header file, without malloc(), without fork(), without dependencies, ready for embedded systems that can at least call printf() - and even those who cannot, can easily adapt it!

Most probably you will understand most about Atto if you just read its header: atto.h. I promise it's not that long, if you exclude the documentation.

Why should I use Atto instead of another framework?

The goal of a unit test is generally very simple: verify that the obtained value equals to the expected one. More generally if a boolean expression is true (e.g. a value is in the allowed range).

There are so many complex unit test frameworks to do this in C and C++ which just increases complexity of the whole project. Suppose you are just writing a simple program and want to test it. Oh, boy! Start managing the dependencies for the framework, check if you can install it, maybe you need Docker etc. Some frameworks even require fork() - how could that work on an embedded system?

Atto is born to remove all of that complexity. Heavily inspired by MinUnit, which has just 3 lines of code, Atto is just a header file you can include statically (copy-paste) in your project and start writing your tests. Atto is so tiny that even its name means "tiny" ;)

The output is basic, but enough: indicates just where the test case failed, by filename and line number. Open that file, go to that line. There is the error. Start using the debugger around that point.

Passing tests are not printed, to avoid cluttering the output. You can still explicitly request a status report at any point in the test suite codebase (e.g. after a cluster of testcases of similar nature) with atto_report().

Dependencies

Only the C standard library!

  • stdio.h, for printf() - if your system does not have it, replace the one call of printf() in atto.h with something else!
  • math.h, for fabs(), fabsf(), isnan(), isinf(), isfinite()
  • string.h, for strncmp(), memcmp()
  • stddef.h for size_t

No malloc() or fork() required

How to use

Basic idea and example

Use the atto_* macros in your test cases.

Here is an example where we test the sqrt() function from math.h:

#include <math.h>
#include "atto.h"

static void test_sqrt_valid_values(void)
{
    atto_eq(0.0, sqrt(0.0));
    atto_eq(1.0, sqrt(1.0));
    atto_eq(2.0, sqrt(4.0));
    atto_ddelta(1.4142, sqrt(2.0), 0.001);
    atto_ddelta(1.7321, sqrt(3.0), 1e-8); // This fails!
                                          // 1.7321 is not accurate enough
                                          // The test case stops here.
    atto_ddelta(2.2361, sqrt(5.0), 1e-8); // This is NOT executed!
}

static void test_sqrt_negative_values(void)
{
    atto_nan(sqrt(-1.0));
}

int main(void)
{
    test_sqrt_valid_values();
    test_sqrt_negative_values();
    atto_report();  // Print a small one-line status report
    // Looks approximately like this
    // REPORT | File: /path/to/my/file.c:27 | Test case: main | Passes: 5 | Failures: 1
    return atto_at_least_one_fail; // Non-zero in case of error to provide a proper exit-code
}

Real-world examples

Check some of my other personal projects, where I use Atto for unit testing!

Complete recipe for the creation of a test suite with Atto

  1. Add the files atto.h, atto.c to your project:
    • You can copy the whole src folder into your project. This is probably the fastest and simplest solution.
    • Alternatively, use git subtree to include this repo into your own
  2. Create a file with your tests, say test.c.
  3. Add test case functions returning void to test.c, as the two test_sqrt_* functions in the example above.
  4. In each test case call atto_assert(), atto_eq(), atto_flag() etc. to verify the values you are testing. On a fail, the test case is terminated early and the lines of code after that are not executed.
  5. Add a main() function to test.c which returns atto_at_least_one_fail. By doing so, the exit code of the test executable will be 1 in case at least one test failed. Particularly useful when running the test executables in a pipeline that should stop when something is broken. Of course on embedded systems one can use atto_at_least_one_fail in different ways, as there is no returning from main; for example by transmitting it via UART.
  6. Call the test case functions from the main().
  7. Compile test.c into an executable and run it.
  8. Check its standard output and the process exit code for failed tests.
  9. Bonus: add some calls to atto_report() wherever you want in the test suite code. At least one call at the very end (before the main returns) is suggested, but also after a set of similar test cases is a good choice to see where something is making the test suite crash, in case so happens.

A test case is failing. Now what?

The output will contain one or more lines like:

FAIL | File: /path/to/some_project/test.c:182 | Test case: test_valid_input_length
  1. Open the file /path/to/some_project/test.c

  2. Go to line 182 (use some keyboard shortcut), which is in the function test_valid_input_length()

    Note: sometimes the path:linenumber pattern is identified as clickable by your IDE. Maybe that speeds up your search.

  3. The assertion on that specific line failed. Now up to you to debug why. The easiest way is to place a debugger breakpoint earlier in the test case.

Once the standard output does not contain any FAIL lines and the exit code of the process is 0, you are good to go!

But this framework does not fit my needs!

It may not be the best solution for your scenario - it was born for my personal projects where I needed something simple. You are more than welcome to customize it to your needs within your project or simply to use other frameworks.