Login
You're viewing the functional.cafe public feed.

Replies

  • Nov 19, 2026, 5:27 AM

    @a2_4am @a2_qkumba Well, it should be simple enough to fix. The game would still have to be thoroughly tested though, because sometimes the leaked source code deviates a bit from what was used for the relreased game, and Zilf isn't 100% compatible with every quirk of the original compiler.

    Which is not to say Zilf is doing something wrong, just that the behavior may have been poorly defined to begin with.

    But I was thinking of the original release of the game.

    💬 1🔄 0⭐ 0
  • Nov 19, 2026, 5:40 PM

    @et_andersson @a2_qkumba I understand the Infocom release is what is is, but I would love to see this and other known bugs quashed and fixed versions distributed. (If licensing is the issue, perhaps the community could start with Zork 1-3, which Microsoft has already relicensed.) Is anyone working on that?

    💬 2🔄 0⭐ 0
  • Nov 19, 2026, 6:00 PM

    @a2_4am @a2_qkumba I started playing around with the Zork I-III source code a while back, making sure to start with the exact repositories in the Microsoft announcement. But I ran out of steam, and haven't even looked at it for months.

    I remember I was still in the stage where I was doing force pushes to clean up the version history, and was trying to decide exactly how the fixes should be documented.

    I may have gone a bit overboard with some of them.

    github.com/eriktorbjorn/zork-t

    💬 2🔄 2⭐ 0
  • Nov 19, 2026, 6:20 PM

    @et_andersson @a2_4am @a2_qkumba Your bug lists have been very useful to me on their own!

    But I agree that it's hard to feel motivated about a bug fixing spree. There's simple "variable used wrong" bugs, but then there's bugs which come down to “the designers didn't think of this corner case." Or “the designers *did* think of this corner case and put in an imperfect hack to avoid it.”

    💬 1🔄 1⭐ 0
  • Nov 19, 2026, 6:23 PM

    @et_andersson @a2_4am @a2_qkumba Now you’re fixing design bugs.

    You wind up trying to design the game that someone would have made 40 years ago if they’d thought of it, and then it's not a historical exercise any more. If I were writing Zork 1 today, I wouldn’t.

    💬 1🔄 1⭐ 0
  • Nov 19, 2026, 6:27 PM

    @et_andersson @a2_4am @a2_qkumba Also, on the tediously practical level, you'd have to write a very large test suite first, with 100% coverage of all printable game messages. Otherwise you run the risk that a bug fix breaks another interaction elsewhere in the game. (This happened *all the time* in Infocom's post-launch updates.)

    💬 2🔄 1⭐ 0
  • 💬 1🔄 0⭐ 0
  • Nov 19, 2026, 8:24 PM

    @zarfeblong @et_andersson @a2_4am @a2_qkumba a few years ago I wrote a fuzzer for Z-machine (only Mini-Zork so far) and kinda just looked for novel text output, and tried to do some strategic pruning. It did discover a bug where it printed "Clrthatrqdc"! But it was a pretty large space to explore, I always wondered if there was a smarter theorem-solver based approach or something.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Nov 19, 2026, 8:29 PM

    @zarfeblong @a2_4am @a2_qkumba

    One good example of things that used to work but broke at some point is the buoy in Zork I. You're supposed to get a clue that there is something inside when you pick it up, but that doesn't happen in revision 88 at least.

    It's supposed to be handled by RIVR4-ROOM, but if memory serves it's the vehicle, not the room, that gets the chance to handle the action instead.

    💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Nov 21, 2026, 9:12 AM

    @a2_4am @a2_qkumba I don't remember. I thought I had a checklist of things I wanted to test but I can't find it now so maybe I got it mixed up with something else.

    I think I felt pretty much done with Zork I by the time I dropped it, though. So yes, probably a lot of testing. And the BUGS.md files contain things I haven't fixed.

    💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Nov 21, 2026, 4:03 PM

    @a2_4am @a2_qkumba I guess if someone filed bug reports I'd feel obligated to at least look at them.

    And of course, you or anyone else are free to fork the repository or cherry-pick anything you like from it. I wouldn't mind.

    💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Nov 21, 2026, 6:17 PM

    @et_andersson @a2_qkumba it seems that Quetzal save files are less helpful for this than I would have hoped, since they are invalidated every time you rebuild (if the checksum changes). But at least Voxam (used here) has an easy way to set the initial random seed so theoretically each test could just be a complete set of commands up to the point of test, and the tests would be reproducible since the randomness would be controlled.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0