Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - Simon

#1
Yes, now I remember how I had this idea during the 5 streams in summer: Show teleporter association in unclear normal view mode. It's important, and it's an excellent design question how to do it.

I've never made a topic because I couldn't decide exactly what to suggest: Show letters and you must find the second half yourself, or draw lines/arrows across the map to the second half, show trigger areas or not, ...

-- Simon
#2
100 % agree that the game should not crash. It's a good immediate first fix.

So far, we know:

0-26 pairs work in play, and clear-physics tells pairings with A-Z.
27-55 pairs work in play, but clear-physics used buggy symbols or doesn't tell the pairing.
56-126 pairs work in play, but either crashed clear-physics or don't tell the pairing.
129 pairs don't fit into play (see #2 below).

At this point, let's look at the remaining ecosystem for guidance:

  • What does the existing NL editor offer? Does it tell you a limit? At what point do you have to work around the obvious tooling to put more pairs?
  • How does the game react to 129 pairs or higher? In the quote: "256 slots which allows for a max of 128 teleporter pairs", how does the restriction of the "allows for a max" manifest in practice? Game presents an error? Ignores silently? If there is some code to explicitly catch list overflow, we have a soft argument that 128 pairs is indeed to be supported.
  • I too remember lots of lag by interactive objects, e.g., on the huge map 10,000 AD in Egypt by Turrican from 2 years ago. WillLem has already looked into the code and didn't see any easy performance fixes. All this lag is a soft argument for a lower cap.
  • What does namida think? Reason: namida decided that what has NL in its name, must remain well-compatible with vanilla NL in certain ways. What is this exactly? Is putting 27 pairs already relying on a bug that CE should fix? Putting 56? Or does NL call for supporting 126/128 pairs?

-- Simon
#3
Quote from: WillLem on November 15, 2026, 08:50:37 PMinstinct is to simply disallow more than 26 teleporter/receiver pairs

Then CE won't be a drop-in replacement for NL anymore.

To reject >= 27 pairs, you'd need to get wide consensus that putting more than 26 pais is a level design error even in namida's NL. The exact rule for the future (explicitly limit to 26, or allow >= 27) doesn't matter much to me, but I recommend against a split of the ecosystem.

-- Simon
#4
We're still 5 people confirmed: Flopsy, Kingshadow, Will, Icho, me.

German forumers, feel free to let me know on short notice. But Icho is currently figuring out group accomodation. If you commit early, you'll be able to book cheap lodging. Icho knows a holiday apartment and you don't have to take a hotel.

Kingshadow: By coincidence, I've received Magical Athlete a.k.a. Turbo-Flitzpiepen 2000 for birthday. No need to bring that game from England anymore. :-)

And there are enough Zendo sets here to run two or more groups. No need for Will to bring the set that I gave to him in October 2024.

-- Simon
#5
Other Projects / Re: Rustlings: a WIP lemmings engine
November 02, 2026, 09:06:31 PM
Rustlings is exciting to follow.

The closest engines in spirit are Lemmix and Golems, for bug-compatible DOS Lemmings 1 physics. But Lemmix is written in Delphi for Windows. Golems runs in the browser and has a Windows release, but it's not open-source yet. Rustlings is both cross-platform and open source, building it was easy (get the source, then cargo build) and it all works offline after cargo has fetched the dependencies.

And you've reverse-engineered parts of DOS L1 yourself. We believe that Lemmix and ccexplore's descriptions replicate DOS L1 exactly, but it's always possible that there are bugs that we haven't found even after 20 years of Lemmix.

I don't think I'm going to play a lot of Lemmings under exact DOS Lemmings 1 physics. I play DOS L1 itself occasionally in Dosbox for nostalgia. Still, I'll see how much modern convenience, replay fine-control, ... you'll add.

You show VGA level graphics with the EGA skill panel and information bar. Is that intentional?

-- Simon
#6
General Discussion / Re: Simon blogs
October 26, 2026, 10:49:55 PM
exp(−(x − μ)² / 2σ²)

There is a curious pattern with getting good at a skill.

The skill can be anything: Playing strength in a certain game, playing an instrument, typing fast, getting good at drawing, playing a sport, finding good designs to engineering problems, ...

When you start learning, you know that you suck, and your output varies between absolutely terrible and pretty lousy. Maybe it's good for small examples that you've practiced in detail, but for a real-world application of the skill to a new random real-world case, your output will be be terrible to lousy.

You get better, and you increase both your mean and your standard deviation. Now, your output is mediocre, sometimes bad, sometimes all right.

Eventually, you'll get even better, and you nail an outstanding piece of workmanship. Congratulations! You're in the spotlight! Maybe you'll appear on the front page of some website, maybe you'll get a strong rank for your game account.

The problem is: The very first time you're in the spotlight, you've probably increased your mean during your recent practice, but you haven't shrunken your standard deviation yet. That means that your outstanding piece of workmanship is exactly that: An outlier in the positive direction. You'll have a hard time reproducing this quality. It takes months of hard work to prune old bad habits and stabilize the skill (= shrink the standard deviation) all while maintaining the same or an even higher mean.

Do not fear the subsequent garbage output. It's sad, but it's natural. It may feel like getting bad again, but it's regression to the mean after the lucky shot.

The spotlight can only show single pieces of craftsmanship output because it can't show the craftsmanship itself.

-- Simon
#7
Lix Main / Re: MIDI music file support
October 23, 2026, 02:54:34 AM
I looked through the Allegro sources. Still no MIDI support in Allegro 5.

Allegro 3 and 4 supported it in the 90's and early 2000s. It looks like it was dropped for A5 around 2010-2015 and hasn't ever come back.

-- Simon
#8
Quote from: Forestidia86 on October 22, 2026, 06:16:08 PMAny Way You Want

Hmm, Lovely/anyway.txt ships with LF, and Lovely/escapethepit.txt ships with CRLF. This kind of mish-mash has been in all Lix releases (current is 0.10.36) for all operating systems. That's annoying. I'm going to normalize this in git independently of what we're eventually going to ship in the releases.

Quote from: Silken Healer on October 22, 2026, 05:01:49 PMjust work too on the default text editor
Quote from: Forestidia86 on October 22, 2026, 06:16:08 PMNotepad on systems older than Win10 usually doesn't understand the LF line ending. This means that the text in the file appears to be in one line without line breaks

Right, Windows Notepad in Windows 8/7/earlier ignores lone LF characters, and instead shows all LF-separated lines text together as a single long line. I consider this an outright bug in Notepad.

One idea is to automatically convert all levels/docs/... to CRLF when I create the Windows release. I.e.: Keep LF in git, ship LF for Linux, and ship CRLF for Windows. Then the files will look different on different systems, have different sizes, different checksums, ... but it's probably the strongest contender to shipping LF everywhere.

Still I'd prefer to ship LF everywhere. Everybody will get the same files, everybody will see shared semantically-identical levels as file-identical, all regardless of tooling. Among only the Windows users (global OS share, not specific to Lix users), the ratio of Win 8/7/earlier dropped from 3 % in January 2025 to 1 % in October 2026, source. I expect this downward trend to continue. And in that small ratio, you're only hit if you use the broken default text editor. I am willing to work around some bugs, that's why I've shipped the documentation in CRLF for years, but the benefit is getting too low.

Quote from: Silken Healer on October 22, 2026, 05:01:49 PMWould there be any way for Windows 7 users still edit Lix levels on Notepad and it would just be trickier

On Windows 7, mass-convert everything manually to CRLF, then edit in Notepad. A common way to convert on Linux is unix2dos. I don't know about the easiest way in Windows. Stack overflow question: Anything like dos2unix for Windows? says: Install Notepad++ and use its mass conversion feature. Or use a Powershell one-liner.

-- Simon
#9
Lix Levels / Re: Chicago-Themed Terrain
October 20, 2026, 12:14:45 AM
Quote from: Simon on April 17, 2023, 11:58:34 PMMore feedback on the entire set to come, again with a deadline: Tuseday, April 25th

I'm over 3 years late, or really 6 years because Liebatron's first post is from November 2020.

I will not merge the Chicago tileset into the main Lix download in the upcoming years.

The Chicago set agrees with itself, and it has a lot of tiles/steel/idioms, but they don't mix well with the remainder of Lix. The lix get stuck in the fine details to the point where the tiles feel like they should be upscaled by 2x or 4x. But the set would lose a lot of character after the upscale. The ideas for traps, e.g., moving train is water, work in the Chicago set, but really need better support from the engine to shine. I'm not ready to implement that needed support. And the sheer number of steel tiles, terrain, ..., is required to make what you envision. but yet again, Lix's list-it-all steel browser is made for smaller amounts of steel. Levels want to be either all-Chicago, or no-Chicago. Mixing is hard. I believe that, for Lix as a project, it's more correct to reject the Chicago set than to merge it.

But because Chicago mixes well with only Chicago, it's still a lovely and consistent add-on. I don't want to erase it from anyone's memory. You have your fans in this thread, and the set is available for everyone here, and maybe we'll see more Chicago levels one day.

Liebatron, I must, in light of the decision not to merge Chicago into the main download, apologize for my early enthusiastic feedback. The early feedback came complete with easy-to-implement suggestions, and you followed through to improve it. This is the biggest sin that a maintainer can do to a contributor: Make you do extra work, and only then drop the news that the improved contribution won't be a good fit regardless. I didn't realize it in the first few months, and the set has found its fans here, that's what has made it so hard to say no outright for 6 years. The indecision has bugged me throughout this entire time.

Now there is closure.

It's a cute set and a labor of love. It's even, as you say, silly in a good way. Looking at the Chicago screenshots here still makes as happy as it has made me back in the day.

-- Simon
#10
Yeah, no need to upgrade. Anything from 2024, 2025, 2026 will be fine. It's even good if a few users play with older Lix. This way, I'll be able to see that I didn't break anything.

-- Simon
#11
Right, the level filename decompression.txt, but the replays start with decompressionmeth1- and should instead start with decompression-. It has 8 replays, many are practically identical. I'll toss all but 1 or 2 and rename the surviving ones to start with decompression-.

Pointing from within the replay is fine, they all point to decompression.txt. The mass replay verifier ignores the replay filename and looks only at the pointing. The replay file name is relevant for Rep4Lev, i.e., for the Load Replay button from the singleplayer browser.

Yesterday, I wrote a D script matches-pointer.d to find these misnamed replays in the collection. Only the 8 replays for Decompression Method had this issue. All other replays in the collection are fine.

More to reply these days.

-- Simon
#12
Renaming replays to include a judgment is good. There is no support in the replay format to mark this information, it's been a loose feature idea but I've never done anything toward it. Therefore we must resort to the filename, which is good anyway.

There is no precedence. We can make something up:
levelname-player-judgment-timestamp.txt
levelname-judgment-player-timestamp.txt
levelname-player-judgment.txt
levelname-judgment-player.txt

And the judgments might be strings over a-z0-9, e.g., intended, intended1, intended2, accepted, backroute, bug, ..., whatever proves useful. There is no tooling that looks for these strings yet.

Identical replays: Feel free to point them out. Back in the day, we tossed everything in and hoped that it become useful. It's still possible that future physics bugfixes will invalidate some replays but not others, and then levels will stay covered without extra work ... but that's not a strong argument. The levels with many replays tend to be easier levels that are quick to re-cover. And most levels with multiple replays will still have only one intended replay, which should pass (regardless of whether other replays pass) under this new system where we use replays to document intended routes. It makes sense to delete some dupes.

Failing replays: No need to point them out, I've found them with full replay verification. 22 replays fail in Lix 0.10.36, and 14 of the fails are in lemforum:

14 failing replays in lemforum
Daunting/cornerstone-Michael-2016-07-04-223017.txt
Hopeless/eyeoftheneedle-Michael-2017-11-30-022958.txt
Hopeless/fromtheotherside-geoo-2017-08-02-103750.txt
Hopeless/fromtheotherside-Michael-2018-04-29-162614.txt
Lovely/putyourlix-Michael-2016-06-27-144736.txt
Vicious/downamong-Michael-2018-04-14-170928.txt
Vicious/thelastlaugh-Michael-2016-07-05-120031.txt
Vicious/tinkertailorsoldier-mobius-2022-12-09-211843.txt
Vicious/tinkertailorsoldier-mobius-2023-01-16-194919.txt
Lovely/climbtofreedom-sekti.txt
Lovely/pipedream-Simon-2018-01-08-005723.txt
Vicious/fiveforfighting-geoo-2017-08-02-120536.txt
Vicious/theresa-Nepster-2017-08-09-131654.txt
Vicious/waitwhyisthereatree-geoo-2017-06-21-132852.txt

I'd say we delete them unless you pick some for explicit preservation. You can look at them to see if a replay documents an intended route found nowhere else, but we should re-cover that level then anyway.

Thanks for the new replays!

-- Simon
#13
Yes, this is an excellent undertaking. The original point of the collection was automated proof of solvability, and we submitted whatever passed the level, but it's equally important to collect the intended solutions.

I have these bugs left open from 2016:
#109 Rename replay directly in the replay browser
#102 Replay <-> level: run against arbitrary level

Do you still need support for rename-replay from the replay browser? Or do you use the show-replays-for-this-level dialog ("Rep4Lev") of the singleplayer browser instead? Improving rep4lev is harder than improving the replay browser, but I don't mind looking at what helps you with your archiving project.

-- Simon
#14
Lix Main / Re: Lix 0.10.36 released
October 11, 2026, 05:28:45 PM
Lix 0.10.36 released.

:lix-cool: Download for Windows 64-bit -- recommended
:lix: Download for Windows 32-bit -- fallback for ancient machines
:lix: Download for Linux 64-bit
:lix-evil: Source code
:8(): Changelog
:8:()[: Issue tracker

How to update
Quick and dirty: Extract over your old installation. Level moves will then result in duplicate levels that remain in your level tree.

Clean method: Extract to new directory, then copy these directories from old Lix into new Lix:
user/
replays/
...and any levels that you've built yourself or added manually.

  • NepsterLix: House of the Mad Trickster (Black Hole 20): Cover therectangular hole's entire ceiling with buzzsaws. Climbers will now die in this hole. That makes it more honest, the deadly-looking nook is indeed deadly now.
  • NepsterLix: Write the unenforced time limit in ticks, not seconds, into the level titles of the 3 Outtake levels that had time limits: It's Hero Time Again, Just a Quickie, Time Out. Reason: Lix counts in ticks only and doesn't convert 15 ticks to 1 second anywhere.
  • Fix #502 and #304: Detect hotkey duplication across tabs in the options menu between functions that can be available simultaneously. E.g., if you assign the same key to hold-to-scroll (on the general controls tab) and to a game feature (on the game hotkeys tab), the duplication gets highlit in red. You can still duplicate keys as much as you like; it's merely shown.
  • Fix #501: When you drag a selection frame in the editor, the info bar at the bottom will tell you the size of the dragged frame.
  • Fix #500: The dialog that loads replays for the current level now tells you where it looks for replays, typically in ./replays/, and what it lists, filenames starting with the level basename without extension. Before, the dialog showed only a list of files, possibly empty, without explantion.



-- Simon
#15
Hmm, in that case, your next prompt should be: Please run over all the singleplayer levels from the upstream Lix download, look for levels that match the pattern n-each-of-m-different-skills for n > 0, m > 7, remove from those files all skills that Walkers doesn't support, and print me the list of these so-modified levels so I can manually test them.

From that point, you'll have to play some of it yourself, but I'm sure you can do it. :-) We're happy to help with concrete suggestions if you get stuck. Examples:

  • Any Way You Want will certainly pass, and it's the iconic opener that would be nice to preserve.
  • Solar Eclipse will certainly fail even though it's easy with the full Lix skillset.
  • Put Your Lix on Ice is solvable with 5 climbers, 5 miners, but it will become harder than Lovely (more timing, no jumpers to help) and should rank higher or stay out.
  • Adventure Playground should pass.
  • The Hopeless n-of-everything will likely all fail after cutting skills. Labyrinth of Despair may be open-ended, but it's hard, it needs the many skills and it needs the flexibility.

Ah, and a curiosity from looking more into your linked list of levels: Was it really a prune according to skills only, or did you prune for other criteria, too? I'm wondering because Little Miner Puzzle (lemforum/Lovely) made it and Advent Calendar (simon/xmas20218) didn't, and both levels' skillset is exactly 5 miners and nothing else. Ah, traps aren't supported in Walkers, that explains it.

-- Simon