title: Unit testing framework testc
author: Sergey Kalichev <serj.kalichev(at)gmail.com>
date: 2020
...
The testc
is a unit testing framework for code written in C. The testc
is a part of faux
project (library of aux functions). The utility executes a set of unit tests, gets a results (success/fail) and generates a report. Each test is a C-function. The source of tests is a shared object - executable binary or shared library. The tested shared object must contain the symbol with prefixed name. This symbol references the list of test functions. So the shared object can contain production code and testing code simultaneously. Also testing code can has its own shared object that linked to tested libraries.
The goal of testc
framework is simplicity and integration of testing to the development process.
testc
utilityThe testc
utility gets a list of need to be tested shared objects from command line, executes all the test functions from them. It generates a report while tests execution.
The utility executes fork() for each test. So broken test interrupted by signal can't break testc
itself. The utility prints a return value for each test. It can be success
, failed
, interrupted by signal
. For failed tests their output (stdout, stderr) is printed to report for debugging purposes. The successfull tests are silent.
The utility's return value will be 0
if all the tests are successful. Single failed test will lead to non-null return value.
The testc
utility has no dependencies excluding standard libc
library. This fact allows to use LD_LIBRARY_PATH
environment variable to define path to tested shared objects and tests the software in place without installing it.
User can define tested files in command line using absolute path, relative path or library name only. If path is not specified then utility will search for the library within LD_LIBRARY_PATH
and standard system paths.
There are examples of testc
utility execution with defining relative path, absolute path, search within LD_LIBRARY_PATH
, search within system paths accordingly.
$ testc ./.libs/libfaux.so.1.0.0
$ testc /home/pkun/faux/.libs/libfaux.so.1.0.0
$ testc libfaux.so
$ LD_LIBRARY_PATH=/home/pkun/faux/.libs testc libfaux.so
-v
, --version
- Show utility version.-h
, --help
- Show help.$ LD_LIBRARY_PATH=.libs/ testc libfaux.so absent.so libsecond.so
--------------------------------------------------------------------------------
Processing module "libfaux.so" v1.0 ...
Test #001 testc_faux_ini_good() INI subsystem good: success
(!) Test #002 testc_faux_ini_bad() INI bad: failed (-1)
Some debug information here
[!] Test #003 testc_faux_ini_signal() Interrupted by signal: terminated (11)
Module tests: 3
Module errors: 2
--------------------------------------------------------------------------------
Error: Can't open module "absent.so"... Skipped
--------------------------------------------------------------------------------
Processing module "libsecond.so" v1.0 ...
Test #001 testc_faux_ini_good() INI subsystem good: success
(!) Test #002 testc_faux_ini_bad() INI bad: failed (-1)
Some debug information here
[!] Test #003 testc_faux_ini_signal() Interrupted by signal: terminated (11)
Module tests: 3
Module errors: 2
================================================================================
Total modules: 2
Total tests: 6
Total errors: 5
The tests source code doesn't use any special headers and test binaries don't need to be linked to any special testing library. The only necessary thing is to define three special symbols within shared object and define testing functions accordingly to predefined prototype. These three special symbols are API version (major version byte, minor testing byte) and list of testing functions.
The testing function prototype is:
int testc_my_func(void) {
...
}
The function name is arbitrary. It's recommended to start name with testc_
to differ testing functions from the other ones.
The testing function returns 0
on success or any other value on error. The function can output any debug information to stdout or stderr. Only failed function's output will appear in report. The successful function is silent.
The following functions are simple examples of tests.
The following function are always successful. If test needs any external data for processing the environment variables can help. This example prints nothing because it always returns 0
i.e. function is successful.
int testc_faux_ini_good(void) {
char *path = NULL;
path = getenv("FAUX_INI_PATH");
if (path)
printf("Env var is [%s]\n", path);
return 0;
}
The following function is always failed. The report will contain the printed text string.
int testc_faux_ini_bad(void) {
printf("Some debug information here\n");
return -1;
}
The following function leads to Segmentation fault
and test will be interrupted by signal.
int testc_faux_ini_signal(void) {
char *p = NULL;
printf("%s\n", p);
return -1;
}
The corresponding report (execution of these three example functions) you can find in Report example
section.
The testing function prototype can be changed in future and format of functions list can be changed too. The testc
utility must know about it. For this purpose the tested object must contain the following special symbols:
const unsigned char testc_version_major = 1;
const unsigned char testc_version_minor = 0;
It's API version (major and minor parts). The module declares API version that it uses. The names of symbols is fixed. Now the single API version 1.0
exists but it can be changed in future. Frankly the API version declaration is not mandatory but it's strongly recommended. If version is not specified the testc
utility will suggest the most modern API version.
Each testing function must be referenced from testing functions list. It allows testc
utility to find all the tests. The symbol that contain a list of testing functions has a special name and type. It's an array of pairs of text strings.
const char *testc_module[][2] = {
{"testc_faux_ini_good", "INI subsystem good"},
{"testc_faux_ini_bad", "INI bad"},
{"testc_faux_ini_signal", "Interrupted by signal"},
{NULL, NULL}
};
Each text string pair describes one testing function. The first text string is a name testing function. The testc
utility will use this name to find corresponding symbol within shared object. The second text string is a description of the test. It will be used within report to identify tests.
The testing functions list must be terminated by mandatory NULL-pair {NULL, NULL}
. Without it the testc
utility doesn't know where the list ends.
All the testing functions can be contained by separated shared library. This library can be separated from tested project.
Testing function can be part of tested project but be isolated in testing-purpose files. These files can be compiled or not compiled depending on the build flags.
# Makefile.am
...
if TESTC
include $(top_srcdir)/testc_module/Makefile.am
endif
...
The production source code files can contain testing functions. These functions can be compiled or not compiled depending on the build flags.
# Makefile.am
if TESTC
libfaux_la_CFLAGS = -DTESTC
endif
int foo(int y) {
...
}
#ifdef TESTC
int testc_foo(void) {
...
if (foo(7)) ...
}
#endif
In this case testing function can test local static functions but not library public interface only.