Also in C++, Python and original JS (probably some others)

So we have to two options here:

1. update all translations to through away empty strings
2. edit the description to include a case with empty stings

And because this kata is 4kyu, then the second solution seems more reasonable!

int len1 = strlen(a)-1, len2 = strlen(b)-1, add = 0;
char* ans = malloc(sizeof(char)*(len1+len2));

for "8" + "9", this will allocate 0 bytes of memory, while 3 are needed.

what makes your code crash however it the fixed test with an empty string (which should not exist, but that's a separate issue already reported here

Your solution has a bug. For some inputs it returns a non-freeable buffer, and tests crash on attempt to free it.

C
testcase: strsum("99999999999999999999999999999", "9999999999999999999"), "100000000009999999999999999998"
My code passes when I use the same testcase on the sample tests, yet I receive a "Caught unexpected signal: SIGSEGV (11). Invalid memory access." when attempting the generic tests.
What could be the problem?

My dumbass tought he could just convert to long long, make the sum and convert back to string.
Amazing kata!

The two functions are not the same; the second code nests the y loop inside the x loop, causing different behavior, while the first runs the loops separately.

langs do not work like that, at least 4 this level
try another algorithm

They are totally not identical, if you open your eyes you will see a difference between them :D

Leading 0's are not perrmited, as no "real" numbers have them (except 0 itself).

Your solution does have leading 0's, "03" != "3"

This kata is great !

This kata is great ! i scratched my head so hard trying to code a simple addition xD

You should treat each string as a representation of a base-10 integer. So, "0" -> 0 and "5" -> 5, so the expected answer is 5 and not the addition of 2 strings which yield NaN.

I think if leading zeros are allowed, it makes sense to say so in the condition or show it in the example for the task.