The fixed tests are only the most basic "happy flow" tests that don't check for error cases such as:

• / and % with 0 as second argument (division/modulo by 0)
• non-integer character during integer mode
• attempt to convert an unmappable number to char code like -1 and 123456
• = with 0/1 stack items, or : with an empty stack
• ~ when input is already exhausted (0 is not mappable to a string in languages where strings are 0-terminated).

They're 80% similar but the command set has a few differences.

(But then, lots of esolang interpreter katas are really very similar to each other. A good spot to farm kata authoring points...)

There are no random tests.

It is not specified what value should be fall back to when reading past all inputs/popping from empty stack. (it's 0)

It is also not specified how "integer" and "character" are converted between each other. (it's character code)

Here's a fork that fixes both of those issues, plus another one where the random tests use the result of the user's solution for the input

Ruby should be upgraded to 3.0

input is used after being potentially mutated by a soluiton. One of consequences is confusing assertion messages.

approved, but note that the assertion message crashes if the user returns null (super minor issue)

System.ArgumentNullException : Value cannot be null. (Parameter 'values')

C# translation (author inactive)

Ruby: No random tests.

Don't forget .sort on your result

duplicate of https://www.codewars.com/kata/befunge-interpreter

Use [text](url)

What if there's more than three of a given number?...

sorry, my bad.
Everything is ok with that!