The new implementation of UnityPrintFloat was mostly worse than the existing one. This branch reverts the changes and then adds back the key features of that branch.

The new attempt had a few drawbacks. It only printed 6 digit of precision, whereas the existing code printed 9+ correctly. You need 9 digits to differentiate single-precision floating point numbers. It had mistakes, rounding errors in the 6th digit on millions of float values. Finally, it covered up the correctness problems by using tests that ignored the last digit.

The existing implementation didn't attempt to print very small values (< 0.000001). Don't throw the baby out with the bathwater! It also didn't print negative zero as -0.0. Let's keep these two ideas: printing -0.0, and print 6 digits (for now) for very small float values.

• Revert "Merge pull request #291"
• Cherry pick code attributed to author of commit 0e7eb54
• Clean up and match printing style

Hi,

We are using Unity with output-char putting a character directly to the console. But we can't omit the new Flush function added in commit c5c392b18a21b048ceb6f728bb7dea84c6e20db5.

When I define UNITY_OUTPUT_FLUSH or UNITY_OUTPUT_FLUSH() it fails at the https://github.com/ThrowTheSwitch/Unity/blob/master/src/unity_internals.h#L308 declaration.

For -D'UNITY_OUTPUT_FLUSH()': error: macro "UNITY_OUTPUT_FLUSH" passed 1 arguments, but takes just 0

for -DUNITY_OUTPUT_FLUSH:

error: expected identifier or '(' before numeric constant
../Unity/src/unity_internals.h:308:13: note: in expansion of macro 'UNITY_OUTPUT_FLUSH'
 extern void UNITY_OUTPUT_FLUSH(void);
             ^

for -DUNITY_OMIT_OUTPUT_FLUSH_HEADER_DECLARATION -DUNITY_OUTPUT_FLUSH:

<command-line>:0:20: error: expected identifier or '(' before numeric constant
../Unity/src/unity.c:15:5: note: in expansion of macro 'UNITY_OUTPUT_FLUSH'
 int UNITY_OUTPUT_FLUSH(void);
     ^

for -DUNITY_OMIT_OUTPUT_FLUSH_HEADER_DECLARATION -D'UNITY_OUTPUT_FLUSH()':

../Unity/src/unity.c:15:28: error: macro "UNITY_OUTPUT_FLUSH" passed 1 arguments, but takes just 0
 int UNITY_OUTPUT_FLUSH(void);
                            ^

I'm out of combinations. I also tried thinks like: -D'UNITY_OUTPUT_FLUSH()'='(void)' I just want to omit this feature which is unneeded for our case.

Having odd forward declarations all over the shop seems a bit odd. Perhaps a UNITY_HAS_FLUSH and UNITY_FLUSH_FORWARD_DECL config options are more desirable than the current situation.

Kind regards,

Rik

Please revert commit 45020b0d3b03; it causes the following build errors on MinGW when setUp() and tearDown() are not overridden (i.e. when relying on the weak symbols):

ld.exe: unity.o: in function `UnityDefaultTestRun': unity.c:1748: undefined reference to `setUp'
ld.exe: unity.c:1753: undefined reference to `tearDown'
collect2.exe: error: ld returned 1 exit status

Note that under MinGW, weak symbols need to be defined in the same module that references them. I made a note of this in https://github.com/ThrowTheSwitch/Unity/blob/2939c420ed9f2697bcb82c1b193bb48c4a8ad7a9/src/unity.h#L33 but it appears to have been overlooked.

The correct fix for #417 would be to make sure that neither UNITY_WEAK_ATTRIBUTE nor UNITY_WEAK_PRAGMA are defined when using the TI C55x compiler. That's how you're supposed to tell Unity that the compiler doesn't support weak symbols, and in that case, UNITY_INCLUDE_SETUP_STUBS will do nothing.

• Implementing feature requests #630 & #574
• Fixing copy-paste bug with UnityFloatsWithin usage instead of UnityDoublesWithin
• Adding lost entries for exclusion (float & double)

If i understand correctly this is the script for the raw unity output to JUnit style xml and overall test summary. I modified the script in four different ways to get it working:

• The String split now works correctly didn't work correct on windows ("") platforms and unix based ("/") platforms.

• Removed the unnecessary whitespaces in the xml output, i needed this for the interpretation of the tools. It was really ugly and sometimes there were 2 whitespaces at front of a message.

• Added support for the TEST_IGNORE() macro which outputs no message. This wasn't considered in the summary at all.

• Added the name property to the testsuite tag in the xml output for completeness reasons.

From a prompt I can run this: xterm -T ./SomeProgram -e /usr/bin/cb_console_runner LD_LIBRARY_PATH=$LD_LIBRARY_PATH:. /home/pi/Documents/Codeblocks_Projects/SomeProgram_WithUnitTests/bin/UnitTest/SomeProgram

...and it works just fine. From a terminal window if I run ./SomeProgram it stops after the first test. I did try to take away the xterm -T from the above but it does the same thing. What am I doing wrong?

Hello kind folks, I come seeking wisdom!! :-)

There is a specific issue with Unity that I have come across yesterday and despite my best efforts I am unable to fix it. I have a header file (DriverDS.h) that contains the structure definitions for some custom datastructures I would like to use. The I have DS.c/.h and test_DS.c files which I have created using ceedling module generator. I would like to test a function DSInit.

DSInit takes two parameters - an enum which indicates the type of datastructure to initialize and a void pointer which has to be initialized based on the enum.

I have a test in test_DS.c which basically creates a void pointer (pPtrUnderTest), a pointer to the custom datastructure (pExpectedValues). The test first initializes the pExpectedValues pointer to the default zero state, then passes in the pPtrUnderTest to the DSInit function.

The function itself is supposed to malloc memory for the pointer, depending on the enum, cast the pointer and set the values to null.

I check if both the pPtrUnderTest and pExpectedValues are equal after the function invocation and would expect them to be equal. However, this test fails. I am not sure where I am making a mistake. I would appreciate any help in fixing this error.

DriverDS.h

#ifndef DRIVER_DS
#define DRIVER_DS
#include "stdint.h"
struct POSTResults
{
	// Good is 1, Bad is 0
	bool Battery;
	bool Authentication;
	bool Transport;
};

typedef struct RuntimeSettings
{
	char TimeStamp[20];
	uint16_t POSTInterval; // Interval is specified in seconds
	struct POSTResults POST;
} Settings;

#endif // DRIVER_DS

DS.h

#ifndef DS_H
	#define DS_H
	#include "BSP.h"
	#include "DriverDS.h"

	enum Error_codes DSInit(enum DataStructures, void *pStructure);

#endif // DS_H

DS.c

#include "DS.h"
#include "stdlib.h"
#include "stdio.h"
#include "string.h"
//Pass in a void pointer for the memory block
// cast the void pointer to a the appropriate structure type and malloc memory for it.
// Set each member to null
//
enum Error_codes DSInit(enum DataStructures dstype, void *pStructure)
{
	switch (dstype)
		{
		case SETTINGS:
				pStructure = malloc(sizeof(Settings));
				memset(((Settings *)pStructure)->TimeStamp, 0, sizeof(((Settings *)pStructure)->TimeStamp));
				((Settings *)pStructure)->POSTInterval = 0;
				((Settings *)pStructure)->POST.Authentication = true;
				((Settings *)pStructure)->POST.Battery = true;
				((Settings *)pStructure)->POST.Transport = true;

				break;
			case HWINFO:
				break;
			case DATASTRUCTURES_MAX:
				return ERROR;
			}
		return SUCCESS;
	}

test_DS.c

#ifdef TEST

#include "unity.h"

#include "DS.h"
#include "DriverDS.h"
#include "string.h"
#include "stdlib.h"
void setUp(void)
{
}

void tearDown(void)
{
}

void test_DS_DSInit_Settings_initializes_values_to_0(void)
{
	void *pPtrUnderTest;
	Settings *pExpectedValues = malloc(sizeof(Settings));
	memset(pExpectedValues->TimeStamp, 0, sizeof(pExpectedValues->TimeStamp));
	pExpectedValues->POSTInterval = 0;
	pExpectedValues->POST.Authentication = true;
	pExpectedValues->POST.Battery = true;
	pExpectedValues->POST.Transport = true;
	DSInit(SETTINGS, pPtrUnderTest);
	// TEST_ASSERT_EQUAL_PTR(pExpectedValues, (Settings *)pPtrUnderTest);
	TEST_ASSERT_EQUAL_MEMORY(pExpectedValues, (Settings *)pPtrUnderTest, sizeof(Settings));
}

#endif // TEST

Now the test fails whether I compare PTR or MEMORY.

I see this error on the console.

 ceedling test:DS


Test 'test_DS.c'
----------------
Generating runner for test_DS.c...
Compiling test_DS_runner.c...
Compiling test_DS.c...
Compiling unity.c...
Compiling DS.c...
Compiling cmock.c...
Linking test_DS.out...
Running test_DS.out...

-------------------
FAILED TEST SUMMARY
-------------------
[test_DS.c]
  Test: test_DS_DSInit_Settings_initializes_values_to_0
  At line (30): "Memory Mismatch. Byte 0 Expected 0x00 Was 0x40"

--------------------
OVERALL TEST SUMMARY
--------------------
TESTED:  1
PASSED:  0
FAILED:  1
IGNORED: 0

---------------------
BUILD FAILURE SUMMARY
---------------------
Unit test failures.

The file BSP.h contains the enum Error_codes definition which for now is just ERROR or SUCCESS.

From what I see here, the pExpectedValues are being set to 0 but the function under test is not setting the memory contents to 0. However, it is the same sequence of operation in the test module and the actual function under test. I have literally copy pasted the code from the test file to the function and renamed it accordingly. I m not sure what is the error I m commiting here.

I appreciate any insight and help into where I am goofing this up.

Thanks in advance again folks!! 😄

PS: To help troubleshoot, I have inserted printf statements inside the function DSInit to print the values after setting them to 0. They are set correctly.

I'm submitting this pull request as a "Request For Comment" since it's a little bit of a design change to the test-runner-generator, so I'd like to hear any concerns or ideas as to how to do this better.

(Summary from the commit message follows:)

This standardizes the form of many of the functions in the auto-generated test runners, using the UnityRunner_ prefix. The initial goal was to make UnityRunner_Verify() accessible from within tests -- this is useful for verifying the call chain halfway through a longer, integration-style test.

A new UnityRunner_RunTest function replaces the generated RUN_TEST macro. It requires test functions to be of the form "void func(void)", so parameterized tests are now handled by creating a tiny wrapper that passes the desired arguments to the actual test.

Commonizing UnityRunner_RunTest significantly reduces the size of the compiled binary. On amd64, the largest test runner in the test suite (testsample_DefaultsThroughCommandLine_runner.o) was reduced from 3.3 kB to 2.4 kB (stripped).

A new header, unity_runner_common.h, contains the implementation of UnityRunner_RunTest. This header is included in each generated test runner, so users shouldn't need to do anything with the new header directly.

The :framework option is no longer supported. I don't think that setting this to anything besides "unity" would have worked anyway. The generated test runners use functions in unity.h, so they need to include unity.h. All other changes should not affect existing usage.

Currently TEST_ASSERT_EQUAL relays to TEST_ASSERT_EQUAL_INT and casts the variables to compare to INT. This can lead to unexpected behavior whereas it should really just be the opposite of TEST_ASSERT_NOT_EQUAL

Concept

I would like to use printf like formatting to add additional messages to my test output. To do this I propose a function which builds a string using snprintf, and passes it to UnityPrint. The proposed function name is UnityPrintf.

Possible Implementation

UnityPrintf(. . .)
{
    char * logLine;
    logLine = calloc(UNITY_PRINTF_MAX_LEN * sizeof(char));
    snprintf(logLine, UNITY_PRINTF_MAX_LEN, . . .);
    UnityPrint(logLine);
    free(logLine);
}

snprintf is chosen for avoiding buffer overflow issues. A define will be made with a default of maximum 160 characters, and defined in unity_internals.h inside of a check to see if the define already has been made. If a user wants more or less characters as their max log output, they should override the define by specifying it earlier.

• [ ] Create UnityPrintf function.
• [ ] Add define for the maximum line length.

Does anyone else think this would be useful? I am just using a macro for now.

This is the test:

void testHexPrintsUpToMaxNumberOfNibbles(void)
{
#ifndef USING_OUTPUT_SPY
    TEST_IGNORE();
#else
    startPutcharSpy();
    UnityPrintNumberHex(0xBEE, 21);
    endPutcharSpy();
    TEST_ASSERT_EQUAL_INT(sizeof(UNITY_INT)*2, strlen(getBufferPutcharSpy()));
#endif
}

On TI, we have:

• UNITY_MAX_NIBBLES == 8
• UNITY_INT_WIDTH == 16
• sizeof(UNITY_INT) == 1
• actual strlen: 8
• expected: 4

So, fail.

For comparison, on an amd64 PC we have:

• UNITY_MAX_NIBBLES == 16
• UNITY_INT_WIDTH == 64
• sizeof(UNITY_INT) == 8
• actual strlen: 16
• expected: 16

Pass.

I cannot figure out how to make this test pass consistently on all platforms without introducing a dependency on CHAR_BIT ie. limits.h.

I am getting this warning when trying to generate test runners:

...unity-src/auto/generate_test_runner.rb:344: warning: Passing safe_level with the 2nd argument of ERB.new is deprecated. Do not use it, and specify other arguments as keyword arguments.

Ruby version:

ruby --version
ruby 3.1.3p185 (2022-11-24 revision 1a6b16756e) [x64-mingw-ucrt]

Is there a plan to update generate_test_runner.rb and get rid of the deprecated function/usage?

/* Test.cpp */
#include <unity.h>
#include "functions.h"

void test_something_specific_to_this_test_file_ok() {
    TEST_ASSERT_TRUE(true);
}

void test_something_specific_to_this_test_file_fail() {
    TEST_ASSERT_TRUE(false); // line 10
}

int main() {
    UNITY_BEGIN();

    RUN_TEST(test_something_specific_to_this_test_file_ok); // line 16
    RUN_TEST(test_something_specific_to_this_test_file_fail); // line 17

    RUN_TEST(test_something_more_universal_from_the_other_file); // line 19

    return UNITY_END();
}

/* functions.h */
#include <unity.h>

void test_something_more_universal_from_the_other_file() {
    TEST_ASSERT_TRUE(false); // line 5
}

The result I get:

Building...
Testing...
test/native/test_including/Test.cpp:16: test_something_specific_to_this_test_file_ok	[PASSED]
test/native/test_including/Test.cpp:10: test_something_specific_to_this_test_file_fail: Expected TRUE Was FALSE	[FAILED]
test/native/test_including/Test.cpp:5: test_something_more_universal_from_the_other_file: Expected TRUE Was FALSE	[FAILED]

Program received signal SIGINT (Interrupt: 2)
----------- native:native/test_including [ERRORED] Took 2.93 seconds -----------

For the last test, the line given 5 is OK, but the filename is wrong: Test.cpp – in reality, the assert that failed is in the functions.h file.

And that makes debugging confusing and hard.

Adding possibility for passing params to TEST_RANGE without any brackets, that will operate as usual TEST_CASE params.

It will be helpful for implementing param tests with partial params as ranges, partial params as usual values.

For example:

TEST_RANGE("example string value", [20, 40, 4])
void test_unifiedArgs(const char* s, int i)
{
  // test logic
}

expected to be equal to usual TEST_CASE cases:

TEST_CASE("example string value", 20)
TEST_CASE("example string value", 24)
TEST_CASE("example string value", 28)
TEST_CASE("example string value", 32)
TEST_CASE("example string value", 36)
TEST_CASE("example string value", 40)
void test_unifiedArgs(const char* s, int i)
{
  // test logic
}

Sometimes it would be useful to indicate that a test failed before even reaching its test focus. JUnit-like test framework have a dedicated test result for this: ERROR.

Would it be possible to add this to Unity? Something like "TEST_ERROR()" and "TEST_ERROR_MESSAGE()".

Currently unity supports, TEST_ASSERT_EQUAL_DOUBLE_ARRAY. In most cases, it's not helpful to do equality checks on floating point numbers, since a very small rounding error could result in a failed test. It would be much more helpful to implement TEST_ASSERT_EQUAL_DOUBLE_ARRAY, which would allow the check to be done within a certain range of the expected value or even, better yet, to have a way to do the check within a certain percentage of the expected value. Please consider adding this as a feature for double-precision and/or single-precision values.