oh, indeed.... haha

I always overlook things the first time. haha

Oh, I just realized you're the guy who approved it. Haha. Thanks for all the help, again!

Cool! I learned more about cmov after I wrote this like a week ago.

Nice to see we got similar answers, and how I could have used cmov here!

Changed ssize_t to int everywhere.

Why ssize_t if 0 < n < 1e6 and it has nothing to do with size?

Approved

Macro parameters are not real variables. They are just textual values which will be inserted into the macro body during the macro expansion phase. So we need to create a proper variable to ensure that the type of the input parameter is correct.

Updated!

So basically reassigning the macro'd variables to their proper types?

The tester macro should be updated:

#define tester(n, exp) {\
    ssize_t submitted = prev_mult_of_three(n);\
    ssize_t expected = exp;\
    cr_assert_eq(    submitted,      expected,\
        "n = %zd\n \nSubmitted: %zd\nExpected:  %zd\n \n",\
         n,          submitted,      expected\
    );\
}

The current version has two issues (which I didn't notice before):

1. It does not assign a correct type to literals in sample tests and -1 is printed as 4294967295.
2. It calls the reference solution twice in random tests.

This was an accident.

Alright. I think I did it. Sorry, I messed up and I think I published a c fork on accident.

Edit: okay I fixed the accidental publish, too, I think.

edit2: this might help: New NASM Translation

Sorry, I didn't see this! I will give that a shot!

Duplicate issue below

I realized if I dec rbp when it's an empty string, I write to memory that's before the string, which is bad.

I could have solved this with starting the string one byte forward, since I have 4x the string length to work with, but figured it would be best to just do another check and return an empty string without modifying memory outside the string.