The evils of QUIT.

[SirWumpus]

I still have a major concern with QUIT not being catch-able (given past discussions). Not being able to CATCH QUIT means that its not possible to clean-up nested input file-id , FREE any allocated memory, and any other resources, which can leave the system in a potentially unstable state with orphaned (leaked) resources.

[ruv]

I still have a major concern with QUIT not being catch-able

It is possible to define a similar word in a standard program:

: quit ( i*x -- never )
  begin
    [: ." #> " pad 80 accept pad swap ;] catch if bye then
    ['] evaluate catch
    [: ?dup if cr ." ### Error: " . depth ndrop else ." Ok" then cr ;] catch if bye then
  again
;

This word is not catch-able to the same extent as the standard word QUIT. And it also does not allow to free any resources. Why is this a problem?

The behavior of the word is documented — if the user uses this word, he is aware 🙃

which can leave the system in a potentially unstable state with orphaned (leaked) resources.

There are millions of ways to achieve this. The Forth system cannot prevent a Forth program from doing this.

[SirWumpus]

which can leave the system in a potentially unstable state with orphaned (leaked) resources.

There are millions of ways to achieve this. The Forth system cannot prevent a Forth program from doing this.

Well a Forth system could help a lot more by itself avoiding as much as possible these issues within its definition and implementation of standard words. It should not promote bad behaviour by being an example of such. You don't see C or POSIX standards defining functions that perform badly; they simply leave that to the developer to do. C provides the developer enough rope to hang themselves. -- just doesn't provide the gallows too.

[ruv]

And what we can do with QUIT? Do you suggest to obsolete it?

On the other hand, I have never heard of any problems due to using QUIT. And I don't know any other use of this word except for debugging.

[SirWumpus]

And what we can do with QUIT? Do you suggest to obsolete it?

• That would be one clear option that would remove IMO one of the more confusing parts of Forth in favour of just using ABORT which is clearer and more consistent in behaviour IMO.

And all the references to QUIT being the interpreter loop maybe, create a better defined word (repl) ( -- ) for the setup and input loop, and then (interpret) ( -- ) as the (sole?) part of the loop body.

SideSwipe: I've seen examples reference an INTERPRET word with no clear stack diagram; some example it behaves like ( -- ior ) while others ( -- ).

• The other would be let it be caught. We've discussed before else where. However this would probably upset everyone (I can imagine lots of shouts, possibly thrown coffer mugs that hit or CATCH, maybe a pencil duel at dawn on some foggy marsh).

• Replace it with something that had portions of QUIT like emptying only the return stack, can be caught, and returns to the REPL. DONE, FINISH, STOP, (repl).

On the other hand, I have never heard of any problems due to using QUIT. And I don't know any other use of this word except for debugging.

I can see it used as a deep return, but IMO ABORT is better suited to that since it can be caught and do clean-up as it bubbles up the exception stack.

If you need a debug word in order to debug and examine stacks and date, then build those in: PAUSE and RESUME / PLAY. But like many things it seems this would be resisted as creating a new word, that has no historical context, to replace something that was meh ok in their minds like QUIT.

[ruv]

And all the references to QUIT being the interpreter loop

It's possible in the future. But it's a big work, and we have a long backlog of things to be done.

And, anyway, a slightly simplified version of QUIT can be implemented in a standard program. This means that this word is not as difficult to implement as it may seem.

I can see it used as a deep return, but IMO ABORT is better suited to that since it can be caught and do clean-up as it bubbles up the exception stack.

Therefore, nobody uses QUIT as a deep return. If you need a deep return, you use ABORT or THROW.

If you need a debug word in order to debug and examine stacks and date, then build those in: PAUSE and RESUME / PLAY.

Probably, it is unnecessary to standardize such words because they can be implemented in a standard program (and be provided as a library), and they are not frequently used.