I do. Do you want to try to find it first? If not I'll tell you now...

Few people are refactoring; Good Job

Isn't programing for speed (O(n) time vs O(n Squared) time) a best practice???

Your solution has a potential crash bug in it; the same one that mine has.
Do you see it? (The Kata Test can't catch it that's why ours passed)

No falut in the Kata; it could not have caught it...

My "solution" has a severe bug, do you see it?

Assume! That sounds reall good.

I'm very impressed with your turn around time and dedication!

Minor issue since as you pointed out you can still do the Kata...

If you look at the test code:

#define ASSERT_ARR_EQ(expected, submitted, sz)
if(sz == 0 && submitted != NULL)
{ free(submitted); submitted = NULL;
cr_assert(0, "n == 0, should return NULL\n"); }
for(size_t i=0; i<sz; i++)
if(submitted[i] != expected[i])
cr_assert(0,
"Incorrect value at index %i: %lld should equal %lld\n",
i, submitted[i], expected[i]);
cr_assert(1);

It does not NULL check submitted if sz != 0...

Actually when I simply return NULL there does appear to be a testing error:

Test Results:
Sample_Tests
should_pass_all_the_tests_provided
Test Crashed
Caught unexpected signal: SIGSEGV (11). Invalid memory access.

Should the Test crash?

NULL is supposed to mean 'empty array' in this test.
Shouldn't it be reported as a wrong return value as opposed to the test crashing?

Ah yes, Correct.

I guess my mind set is warings == errors

I don't let warnings stand because they either contain important info or obscure other important info when out of hand...

I believe this kata has bugs in C. If I do nothing but return null I get 8 pages of errors starting with (note error is in fixture.c):

fixture.c:24:29: warning: passing 'const long long *' to parameter of type 'void *' discards qualifiers [-Wincompatible-pointer-types-discards-qualifiers]
ASSERT_ARR_EQ(expected, submitted, n);
^~~~~~~~~
fixture.c:7:14: note: expanded from macro 'ASSERT_ARR_EQ'
{ free(submitted); submitted = NULL;
^~~~~~~~~
/usr/include/stdlib.h:563:25: note: passing argument to parameter '__ptr' here
extern void free (void *__ptr) THROW;
^
fixture.c:24:5: warning: format specifies type 'int' but the argument has type 'size_t' (aka 'unsigned long') [-Wformat]
ASSERT_ARR_EQ(expected, submitted, n);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
fixture.c:13:17: note: expanded from macro 'ASSERT_ARR_EQ'
i, submitted[i], expected[i]);
^
/usr/include/criterion/internal/assert.h:529:62: note: expanded from macro 'cr_assert'
#define cr_assert(...) CR_EXPAND(cr_assert(VA_ARGS))
^~~~~~~~~~~
/usr/include/criterion/internal/assert.h:159:28: note: expanded from macro 'cr_assert'
CR_VA_TAIL(VA_ARGS)
^~~~~~~~~~~
note: (skipping 31 expansions in backtrace; use -fmacro-backtrace-limit=0 to see all)
/usr/include/criterion/internal/preprocess.h:35:43: note: expanded from macro 'CR_EXPAND'
#define CR_EXPAND(x) x
^

I would have voted for best practices if 'a' was used instead of 97 etc...

The parameter 'a' of twos_difference is supposed to be (const int *) according to the kata