in reply to Data copied with fork() -> how to access true data?

So I grabbed the fork documentation and I found out that variables and such are copied. How can I, from within the fork, access the data of the main process?

You can't!.

That's the main difference between forks and threads. Threads allow shared data access, forks do not. You can use various mechanisms to indirectly share inormation between forked processes, sockets, shared ram etc. but you cannot share direct access to the same data.

I'm confused. You say that you looked at threads, but they aren't supported on your OS/perl combination, but that fork works. What confuses me is, on Windows, fork is emulated using threads. So if fork works, then I thought that Thread should also work, but maybe not.

Update: How quickly we (I) forget. I found this whilst attempting to refresh my memory.

I've never used Perl on '98. It is possible (though it seems unlikely) that the fork emulation is enabled but not the threading?


Examine what is said, not who speaks.
"Efficiency is intelligent laziness." -David Dunham
"Think for yourself!" - Abigail
  • Comment on Re: Data copied with fork() -> how to access true data?

Replies are listed 'Best First'.
Re: Re: Data copied with fork() -> how to access true data?
by muba (Priest) on Jun 01, 2004 at 19:14 UTC
    So, it seems that the most easiest way is to install 5.8.x?
    That's ok, then.

    I know this is some kind of a FAQ, but will my code (not only for this game, but all of my Perl code) need to be rewritten?

      That's probably a good move anyway, though if there is any chance that you could upgrade your OS to something less than 10 years old, that would probably help your cause enormously also :)

      With respect to rewriting your perl code, probably not.

      For the most part, anything that runs on 5.6.1 should run on 5.8.x with minimal changes. That's a risky generalisation to make without having seen the code in question, but based on my own usage of both, it's not too wild a guess. Of course, my memory could be playing tricks on me again.

      The harder question is how much effort will be required to make your current single threaded code run multi-threaded. On that, I will not even hazard a guess.

      From the little information I have, your app is basically a big loop that get user input, uses it to modify the state of the current player, takes the opportunity to update the state of your NPC's and loops back for input.

      Your desire is for the NPC's to run independantly of the user input. In theory, sharing the variables that contain their state with a background thread that loops over a sleep and modifies their behaviour is pretty trivial, but the devil is in the detail.

      If your NPC's are OO-modelled, or if their state is maintained in nested hashes, or any number of other standard perl techniques, then the modifications could be awkward.

      It may make more sense to simple use a timeout on your user input (using Term::ReadKey or similar), and just allow the NPC's to continue their state changes whilst the user thinks about his input.

      How much effort would be involved, and whether multithreading would be beneficial relative to the other options available, will depend highly upon how your code is currently structured and how you want to modify it.


      Examine what is said, not who speaks.
      "Efficiency is intelligent laziness." -David Dunham
      "Think for yourself!" - Abigail
        Yes, the NPC's are OO-modelled and their state is indeed maintained in nested hashes (if that means the $dude->{hitpoints} thing). Altough the Object Model will have to be revised (don't know if you care, but whatever)

        The current system indeed gets user input, executes the user command, lets the NPC do their moves (some kind of AS - Artifial Stupidity - they're not quite intelligent yet :) ) and then gets user input again. Term::ReadKey seems to be an attractive solution, thanks for the hint. But that also seems not to work on Windows systems. That is, it never worked for me.

        But I'll have a look at it!

        You say it'll depend on how my code is currently structured, but that's no important issue. Some procedures that should/could be part of the Classes, are still stored in the main part of the code (non-OO style), while others are stored in the class files. Some procedures that should be a method of class X that inherets from class Y, are stored in the file of class Y. So the current structure is Bad and really needs to be revised, as stated before.