The internal malloc must free in LIFO order

 This reverts commit 783fcaea
 The guard memory bytes should never be freed inside unity_malloc()

Update unity_fixture.c (Thanks, Bryon!)

Update unity_fixture_Test.c (Thanks Bryon!)

I believe if realloc() returns a pointer to a different location, the old location is freed. However, the pointer 'n1' is not freed if realloc fails to obtain a large enough block of memory and returns NULL. (more details on [StackOverflow](http://stackoverflow.com/a/16676964)).

Sorry @mvandervoord -- not your code, your repository 😁

Regarding the double free, cppcheck has a check for that too (just saying) 👍

Free guard if UNITY_EXCLUDE_STDLIB_MALLOC is not defined. 👍

Thanks @ for the hint.

[../Unity-master/extras/fixture/test/unity_fixture_Test.c:530]: (error) Deallocating a deallocated pointer: n1

This solution from Daniel Fischer was helpful in fixing the error because if realloc returns a pointer to a different location, the old location is freed.

Found by https://github.com/bryongloden/cppcheck

[../Unity-master/extras/fixture/src/unity_fixture.c:210]: (error) Memory leak: guard

Found by https://github.com/bryongloden/cppcheck

Weak attribute fix for TMS470 Compilers.

Escape backslashes for windows paths, fixes #137.

Change-Id: I901bf192d3c06457d67c6d3a64bdb9aea22c53c2

Change-Id: I2dbae30e21ab31921817f68fd7a5e08d21c76626

add unity_to_junit.py (Thanks!)

Change comments style to compile with std=c89 (Thanks Andy... and everyone else who helped make this decision)

Add template example unity_config.h file. (Thanks wolf99!)

As recommended in PR #193

The class gives back to the community: better documentation produced for the class and now part of the open source project

add UNITY_OUTPUT_FLUSH to make sure we get the output we need on aborted tests and whatnot.

100% code coverage for Unity project (Thanks jsalling!)

 Add test for failure count and UnityEnd return value
 Cover printing escape codes with length parameter
 Full statement coverage

 More statement coverage on NULL input feature

 Get test coverage on this feature for int, float, & double

 Add config flag for 64-bit, max covered lines even with 32-bit compilers
 Always print uncovered lines, delete separate make target

 Remove what looks like an overflow check. 'divisor' cannot overflow
  since it is the same type as 'number' and cannot be greater.

Add option to print FAILED instead of FAIL during the summary

 For easier automation and searching for a test suite failure

Thanks @rikvdh for catching some instances where we should have been using isnan and such that were missing!