How did Fortran prevent recursion prior to Fortran 90?

Question

I have been reading a bit about programming languages and saw that prior to Fortran 90 recursion was not permitted. How was indirect recursion prevented?

While I see how direct recursion (inside the body of the function itself) can be trivially detected, the problem seems much harder in the general case.

It just means it doesn't have to work. It doesn't have to be detected. — user207421, Jun 27 '20 at 01:13
It is likely more technically correct to say the program was non-conforming. — evets, Jun 27 '20 at 01:55
Some of the pre-F77 vendor specific compilers (like PRIME) allowed recursion. The depth of recursion had to be specified. — cup, Jul 04 '20 at 19:42

score 3 · Answer 1 · answered Jun 27 '20 at 09:26

As the comments on this question note, recursion was prohibited simply by saying that recursion was not allowed. Fortran 77 for example has the statement:

A subprogram must not reference itself, either directly or indirectly.

The consequence of this is that it is a restriction on the programmer to ensure that recursion doesn't happen, and a compiler can assume that this condition is met.¹

Fortran 90 permitted recursion, and from then up to Fortran 2008 a subprogram which is to be potentially used recursively must have the recursive prefix.

Even in Fortran 2018 (where allowing recursion is the default), correct use of recursion is a restriction on the program:

The NON_RECURSIVE prefix-spec shall not appear if any procedure defined by the subprogram directly or indirectly invokes itself

This remains even now something that the compiler is not required to validate.

¹ The discussion in the answer about detection of violations doesn't truly apply to Fortran 77. Back then, a compiler could be much more trusting of the programmer to give a correct program.

How did Fortran prevent recursion prior to Fortran 90?

1 Answers1