Experimental Releases for Physics Bugs

[Nepster]

Edit by namida: Check the latest posts in this topic for the most up-to-date experimentals. These are released by me rather than Nepster so I haven't edited them into his original post here.

DOWNLOAD THE MOST RECENT VERSION Exp V1.44-D HERE:
https://www.dropbox.com/s/29zxv1iis02n358/NeoLemmix_Exp%201.44-D.exe?dl=1

Here is an experimental release fixing some of the physics bugs. This version is even less official than namida's experimental releases and it might very well be that some bugs get removed in a totally different fashion in the official release. You can get the experimental version here:

https://www.dropbox.com/s/pkee100jdjlmlcc/NeoLemmix_Nepster_Experimental.exe?dl=0

The following glitches were fixed:

• Cloning platformer leaves gap: The latest brick is now cloned as well.
• Less strict terrain checks for platformers at left edge: Same terrain checks now everywhere.
• Builder-Walker-combo to ghost through 1-pixel walls: If builders now find solid terrain at the pixel they want to go to, they directly turn around and not only after moving into this position. At sufficiently high walls an equivalent check was already in the original code, so nothing should have changed there.
• Building horizontally through walls: Automatically fixed by the earlier turning position.
• Miner falls through 1-pixel high OWW-ground: Lem turns around if after mining, he cannot move down due to steel or wrong OWW. He will not stop if there is normal terrain (e.g. by placing bricks).
• Updrafts don't effect OhNoers: They do now.
• How bashers turn at steel: The behavior is still the same as for the experimental release attached to the posts in the basher topic.
• Triggered traps are only checked during some frames: Now they are checked on all frames (as long as the lemming is no blocker)
• Assigning miners on OWW: The OWW-check for assigning miners was moved from somewhere in front of the lemming to directly below the lemming, just where the steel check is already.
• Skill-count for cloned lemmings: Now only the original lemming is remembered to have used a skill. Warning: I was not able to really check whether the behavior is now correct!
• Swimmer bugs: Here lots of terrain checks were moved one pixel. Probably the biggest change occurs when a swimmer moves upwards in water: He will usually be one pixel below the position in V1.43. In a few other cases, the lemming position will (at least for a few frames) differ by one pixel when compared to V1.43.
• Platformer moves into isolated pixels: This was already discussed with namida on IRC yesterday, so I just went ahead and let the platformer stall a bit. If you don't like this change, it can easily be reverted to the original behavior.
• Non-deleted blocker fields: When a lemming gets killed by a trap (or is teleported away), the blocker field is deleted now.

I tested this version using 455 replays of NepsterLems and various LP level packs. Result:

• 367 replays succeeded.
• 85 replays failed due to builder bug fix
• 2 replays failed due to assignment of miners on OWW
• 1 replay failed due to copied brick of cloned platformer

Because of the high number of the failures due to the builder bug fix, I made a second experimental version with all fixes, except for the two builder bugs.

[IchoTolot]

Results:

Reunion: 28 Failed   15 Undetermined   107 Passed -----  43 broken  ( nearly 1/3 )
Pimo:      21 Failed   12 Undetermined   112 Passed -----  33 broken  ( a bit more than 1/4 )

I did not go through my single levels, but I expect quite a few broken replays there, too.

So in total I would say ~ 100 broken replays overall for me.

As things like the miner/cloner behavior that glitches through terrain were not yet included, prob a few more will come to this.
If there is room for backwards-friendly improvement it would be in fixing that builder behavior.

Also what about the deadly left/right sides, or the possibility of assiging miners on wrong facing OWW's rather than treat it as steel?  Nobody posts in the dedicated discussions anymore and I would like to know what the final result is, so I/We can prepare.

[namida]

Chances are that (pending feedback, and my own actual testing of them) I will use these fixes as much as possible. It'd be a shame to waste the effort that went into making them, after all. That being said, it's never a good idea to rely on the fiddlier details of an experimental release, even if it was one that came from me.

There's no final decision in regards to those things yet. If you feel it's important to prepare your levels now, I would say the best option would be to make them independant of the outcome on those. Be aware that the next update will end up being a pretty major one due to the large amount of physics changes / fixes, so it's to be expected that not all content will be upgraded to it right away. Due to that, and the removal of gimmicks etc, V1.43n-F will remain easily available, so there won't be so much pressure to upgrade as quickly as possible.

[Nepster]

Here is a new experimental release, which includes some changes mentioned in topics that have not yet reached a final decision. In addition to the changes made for the first experimental version, the following changes were added:

• Deadly sides of levels: The left and right edge of the level are now void and behave like the bottom. Please try the experimental release and post your feedback in the deadly sides topic so that we may reach a decision soon.
• Stackers turn too soon: The turn-around rule was changed to: One frame after placing a brick, that does not add any terrain pixels, the stacker turns around.
• Climber may bomb deeper holes: A special rule was added: If a climber reaches a 7-pixel wall, then the hoister starts at ground level, not one pixel below.
• Intermediate checks for trigger areas: Intermediate checks were added so that the glitches described by möbius, ccexplore and others should not be possible any more. However to avoid adding new glitches (seperation of builder from crowd), checks for blocker fields are made only at the final lemming position (same as in the current NeoLemmix version). Whether (or rather which) new glitches appeared by adding intermediate checks has to be seen. This needs a lot more testing than what little I just did. 

Moreover the floater, glider and stacker sprites were subtly changed to remove exceptions in the code. It is a known graphical glitch, that neither stacker bricks nor parachutes get recolored. Please ignore this for now.

Download the experimental release here:
  https://www.dropbox.com/s/00wa2i8sfr5qspf/NeoLemmix_Nepster_Experimental_2016-04-02.exe?dl=1

[IchoTolot]

Test Results:

Reunion: Same as before.

PimoLems: +1 undetermined. Hurricane 12 "Rod Island" Reason: A basher stops now instead of continuing. --- start a pixel lower and it's fixed, so no break here

If the miner/cloner thing is fixed + a decision in the miner/wrong OWW has been made, I might go through my single stuff where I also use all the new skills.

[Nepster]

Oops, the cloner-miner-bug is already fixed in the new experimental version. I just forgot to mention this.

[namida]

Danger 19 from LPI's intended solution replay raises a possible issue (level and replay attached). Not entirely sure what the ideal result of this situation is.

Note that by slightly changing the timer of the blocker assignment, you can avoid the "stuck in one position, facing same direction" matter. However, an extra digger is still needed compared to previously in order to free the blocker (note that this digger can merely substitute for the miner, rather than being in addition to it). Alternatively, by waiting longer to assign the blocker, one can mine at such a time where the miner will no longer be affected by the blocker. Also note that this is probably not the only way to solve this level, so it isn't even close to level-breaking as far as this level is concerned; just breaks replays and increases the execution difficulty of one possible approach.

[Nepster]

The previous experimental version turned out to be pretty buggy (but what do you expect with an experimental release ): One builder check was missing (which should have been kept from V1.43) and the intermediate trigger area checks introduced quite a few bugs/gltiches (as predicted by Simon).
So here is the next try:
  https://www.dropbox.com/s/umwt758tnqczefj/NeoLemmix_Nepster_Experimental_2016-04-03.exe?dl=1

[IchoTolot]

Ok, here are my results:

Reunion:     
Failed: 18 (-10)  Undertermined: 10 (-5) Passed: 122 (+15)    --->  28 replays broken   ~ 1/5

PimoLems:
Failed: 14 (-7)    Undetermined: 10 (-2)  Passed: 121 (+9)      --->  24 replays broken   ~ 1/6

A big improvement in terms of backwards-compability here

Take your time with the final version! Rather fix and check the exp. version a few weeks more, than have a bugged official release which would lead to multiple hotfixes that also break newly fixed replays again along the way.

[IchoTolot]

Ok tested my single levels with the newest exp version.

Results:

Total levels: 35    Passed: 26   Failed: 9     (~ 1/4 failure rate)

The failes were all just broken replays, because of following reasons:

Builder fix: 4 (Among it "T=0K")

Stacker change: 2

Assignment is off due to cloner platformer/builder fix: 1

Basher against steel: 1 ("Lemmings in Unison")

Walker assignment off. 1 ("Spider Dance")

[Nepster]

Here a new experimental release: Lots of small bugs were fixed, like lemmings stopping when meeting (on intermediate checks) some trigger areas of an already triggered traps, and many more...
There are three big changes:
1) Trigger area checks after stepping out of receiver: If there is no trigger area, there should be no difference at all to V1.43. It just does some more checks now. I haven't discovered any irregular behavior, but please check again, especially given namida's worries that this introduces bugs.
2) Disarmers and harmless one-kill traps: Disarmers should no longer disarm harmless traps.
3) Blocker areas overlapping other trigger areas are again working properly. This was a bug introduced in the previous experimental release.

I haven't yet fixed the bug "Walkers allow to pass OW-Fields in wrong direction", because there is no preference yet visible how it should be fixed and I am waiting for more input.

Download link:
   https://www.dropbox.com/s/zd1xyxr3qnfplwo/NeoLemmix_Nepster_Experimental_2016-04-10.exe?dl=1

[IchoTolot]

Nothing new has broken, everything is as before for me.

Tested: Reunion + PimoLems + Every single level from me that uses teleporters

[Nepster]

Time for a new experimental release: Apart from a few bug fixes, lots of behind-the-scenes stuff regarding objects was done. This means on the one hand side, that object rendering is now much faster and even lemming rendering should be improved now. On the other hand, chances are that a few new bugs got introduced by this. So please test it and notify us if you see any graphical (or other) bugs.
The main bug-fixes concern:
1) Walkers passing through One-way-fields: The current fix is kind of a hack-job. Even after dealing with the normal set-up, I needed three more tries before it too worked as expected, if the terrain below the lemming in the one-way-field got removed. In particular this means, that it is kind of doubtful, that I caught all exceptional cases. So please stress-test this bug-fix.
2) Exploder time of 9 seconds: I changed the exploder time to full 9 seconds = 153 frames. Please check, whether you like the current behavior better than the old one. It is a matter of 5 seconds to change the exploder count-down again.

Download link:
   https://www.dropbox.com/s/1qnvrx8zyk6h157/NeoLemmix_2016-04-20.exe?dl=1

[Nepster]

Time for a new experimental release:
1) No leading zeros and black-out instead of white-out in skill counter: Thanks to a change made by namida, leading zeros are now always disabled and instead of whiting out, the option is now to black out the skill count once it reaches zero. These changes were made to give you the possiblity to test them in actual game-play (compared to just arguing theoretically about this). So please give us feedback, whether you like the new way better than the old one, or not.
Important: Old settings for using "White-out" seem to be ignored and so "Black-Out" will be disabled if you download the experimental player. So be sure to check the options menu!
2) New priority algorithm for skill assignments: Due to yesterday's discussion on IRC and some ugly looking parts in the corresponding code, I decided to replace the prioritizing algorithm with a new overly complicated very sophisticated one, which should be much more flexible as well (e.g. easily permits full priority invert is this is desired). In most cases, you will probably not feel any difference, but hopefully the changes remove a tiny bit of frustration once in a while .
So please give it a try and tell us, whenever NeoLemmix assigns a skill to the wrong lemming. The new algorithm can certainly be improved in many ways!

Download link:
  https://www.dropbox.com/s/f1ykzxx9n491mnu/NeoLemmix_Exp%201.44-A.exe?dl=1

[Wafflem]

I get an Access Violation error when I assign a bomber to a lemming that is not falling. The lemming still explodes after clicking "OK", so the error is just an annoyance but still very important to fix.

Also, the green/blue color inversion when assigning a lemming a permanent skill is a very awesome touch!