• bss03@infosec.pubEnglish
      4·
      6 days ago

      Java doesn’t allow goto, but specifically does have labels for labeled break/continue to support the multi-loop exiting case.

      I imagine these two “structures” will always be implemented in C source through disciplined use of goto.

      • Gladaed@feddit.org
        51·
        6 days ago

        It’s literally the only way to do this. Other ways include checking of loads of bools. That’s slow.

        • bss03@infosec.pubEnglish
          3·
          6 days ago

          In C maybe. In language that support proper recursion schemes, the apomorphism models the early-exit loop.

    • ZILtoid1991@lemmy.world
      3·
      6 days ago

      I also use it for avoiding recursive function calls. In theory, this will tell the compiler to just ditch the current stack and go back to the beginning of the function.

  • Frezik@lemmy.blahaj.zone
    10·
    6 days ago

    IIRC, this is because gcc optimizes goto very well, or at least it did back in the day. It also is a genuinely workable solution for error handling in C.

    Consider if you need to setup three things, do something with them, and then tear them down in reverse order. If there’s an error on the second thing, you want to jump right to the part where you tear down the first thing. Using goto tends to make cleaner code for that in C compared to, say, nested conditionals.

  • stormeuh@lemmy.world
    4·
    6 days ago

    Seems like it’s mostly error handling, which makes total sense to me. In a function with a lot of error conditions, where it also takes more than return <nonzero value> to report that error, the code would get very cluttered if you handle the errors inline. Using goto in that case makes the normal case shorter and more readable, and if proper labels are used, it also becomes clear what happens in each error case.

    Sure, you can do that with functions too, but it’s much nicer staying in the same scope where the error occurred when reporting on it. Putting things in a function means thinking about what to pass, and presents extra resistance when you want to report extra info, because you have to change the function signature, etc.

    • squaresinger@lemmy.world
      6·
      6 days ago

      Not really. There are quite a few of structures you can make with gotos that can’t directly be translated into functions. Sure, you can implement any functionality in any programming paradigm, but it might require much more work than to just replace goto with a function call.

    • calcopiritus@lemmy.world
      3·
      5 days ago

      In C, goto is basically a necessity though. There is really no good way of error handling.

      Options:

      1. Using goto
      void func(void *var) {
    void * var2 = malloc();
    if var == Null {
        goto err;
    }

    do_something();

err:
    free(var2);
}
      1. Early returns:
      void func(void *var) {
    void * var2 = malloc();
    if var == Null {
        free(var2);
        return;
    }

    do_something();

    free(var2);
}
      1. Skipping with conditionals:
      void func(void *var) {
    bool error = false;
    void * var2 = malloc();
    if var == Null {
        error = true;
    }

    if !error {
        do_domething()
    }

    free(var2);
}
      1. Early return + cleanup function.
      void cleanup(void *var2) {
    free(var2);
}

void func(void *var) {
    void * var2 = malloc();
    if var == Null {
        cleanup(var2);
        return;
    }

    cleanup(var2);
}

      Option 1 is really the only reasonable option for large enough codebases.

      Option 2 is bad because duplicate code means you might change the cleanup in one code path but not some other. Also duplicate code takes up too much valuable screen space.

      Option 3 has a runtime cost. It has double the amount of conditionals per error point. It also adds one level of indentation per error point.

      Option 4 is same as option 2 but you edit all error paths in one single place. However, this comes at the cost of having to write 2 functions instead of 1 for every function that can error. And you can still mess up and return while forgetting to call the cleanup function.

      You must also consider that erroring functions are contagious, just like async ones. I’d say most of the time a function is propagated upwards, with very few being handled just as it ocurrs. This means that whichever downside your option has, you’ll have to deal with it in the whole call stack.