The reason Minecraft bug reports can be so datailed is because the Minecraft EULA allows end-user to decompile the game. So technically Minecraft is a source-available software.

finally, something more than "the game stopped working, please fix it :("

I agree its great but honestly not sure how i would fix that bug, without breaking anything else or have more computation and checks.

@@aladanor personally I would give the same tolerance for the onGround check, as the status has, so there wouldn't be a discrepancy between the two. But yeah... I'm not familiar with the rest of the code, so I don't know if this wouldn't break something else xP

​@@aladanorI mean, the system (at least as explained here) is definitely a little clunky. Probably because Minecraft is approaching 15 years old after all. I'd have to look at the actual code, but I imagine some cleaner logic for handling the fall would either eliminate this bug all together or at least make it easier to fix.

11/10 comment

Extremely underrated comment. They should pin this. 😆

Gosh this is genius

@@abebuckingham8198 no worries; it has floated to the top

@@landru27 just because is a good point

Lovely fact!

❤

❤

That is far more than the maximum height....

@@reIONEre yeah but you can get mods that remove the build height limit. You can also just teleport the boat to that height.

haha

This video simplified: Computers do not like numbers that are not multiples of 2^integer.

Oh, hello there)

and DID you need to know about maths to enjoy the video?

hi moggy swampy

And, specifically relevant to this, being in water negates all fall damage.

that could also possibly be the status of being on a ladder

in lava, yes. :)

Or cobweb, right?

I have a feeling that a lot of things that negate or reduce fall damage would be relevant here. I'm not in the air, I am actually: in water, on a ladder, in a cobweb, in lava (it doesn't negate it then for some reason), in scaffolding, in powdered snow...

But only if they're distinct!

Yes we all saw the video, mate

@@aurelia8028 Not necessarily. Every once in a while I will space out and miss a line like that. Regardless, this is a comment conveying appreciation for a good joke; it's positive feedback, and good for the algorithm. It has value, even if it has none directly for you.

To be fair, this makes more sense when you know how floating point numbers work. It's a reference to rounding errors that crop up all the time.

That was profound.

"Oh, I actually know why the problem is here! thank you, thank you THANK YOU!!! Now I will break 60% less when trying to fix it!" That's what I imagine they would say.

He’s in the loony bin now.

Also probably sinking through powdered snow or scaffolding. Maybe sliding on the side of a honey block. And potentially gliding with an elytra because the damage taken from crashing with an elytra is calculated separately from fall damage, so there's probably a different state specifically for that.

This makes me wonder if Matt has his explanation slightly off and it is the case that the boat’s in-air value is true but your in_air value is false when riding in the boat and usually means the boat bypasses its damage calculation except for the floating point cases and separately you bypass the damage calculation except for these floating point errors where maybe the boat has reached the ground; been destroyed; your in-air gets set to true; so in the next tic you don’t bypass the checkFallDamage method and instead die.

Actually, horses don’t count (unless it’s Bedrock version) if I recall correctly Edit: I made a mistake. This is for fall damage when riding horses.

climbing was the first thing i thought of

@@ejvalpey good point. If the boat counts if could mean the player isn't in_air, since they're on a boat.

This comment needs to be pinned! 😀

This suggests that if you use them while in the air you would be immune to that damage because your status would be IN_AIR. Is that correct?

​@@petergerdes1094you take damage regardless 😅

​​@@petergerdes1094Ender pearls teleport the player, and I'm assuming that the position (and therefore the ON_GROUND status) happens before the damage check is done.

@@ShadowlessDeath47 ender pearls do not have this check, it simply deals 5.0 fall damage to the player

Mario?

@@IcecalGamer M(a/2)rio

Matpat

i wonder how many half a presses was needed for this.

Super Mario 64 TAS Speedrun for anyone wondering.

OH NO THIS PHRASE IS GONNA HAUNT ME EVERYWHERE NOW THAT I'VE WATCHED THE HOPPER SAGA NOOOOOOO

I vaguely remember losing my boat when driving into land too fast. Was that real....?

@@HappyBeezerStudios yup. Used to happen a lot

swimming maybe too? (if you can swim in that game)

@@uu8217 You can swim in the game (properly only as of only a few years ago), and it has its own state.

doesn't*

You can also climb vines. And scaffolding, though that may be considered on the ground. You may also be in lava or falling though a cobweb. Or lying in a bed, or riding a horse. I'm not sure what any of those mean. .

@@uu8217 There's boats, so obviously you can swim

Great tip!

I did not expect to get this kind of wisdom from a youtube comment. Thanks.

This may solve the problem of repeated adds, however because of the use of the mantissa and exponent in IEEE floating point representations that is not guaranteed. This would produce deterministic results in fixed point arithmetic with 8 bits or more reserved for the fractional component. However, this bug is actually due to floating point/integer rounding errors and therefore your suggested improvement is not a solution.

My programming tip is only use floating points when absolutely unavoidable. If you are only using numbers with a fixed number of decimals, prefer switching the decimal point and using whole numbers for all the calculation. For example properly coded accounting applications use whole numbers to represent the value in cents, not floats to represent the value in dollars with a decimal part. And in modern languages, there is even a "decimal" type that is specifically designed to do that for you.

@@christianbarnay2499 Yeah no. Many times, fixed-point math can actually be *slower* than floating-point counterpart. Like multiply/divide across multiple orders of magnitude, trigonometry, exp/log, etc. Point being (pun intended), modern computer are absolute beast when numbercrunching IEEE 754 floating-point, to it being a measure of power (FLOPS). I would dare to say the inverse, *don't* use fixed-point unless you absolutely sure: 1. All numbers fits, including intermediates. 2. Has sufficient precision that is more than standard floating-point. 3. You *really* need that extra precision, with trade-off of more memory/register/cache/etc. 4. Use good generic fixed-point library (don't try to manually implement, it's 100% not worth it).

This is such a high quality pun it hurts. +1

genius

+2 moment

You win the internet for today.

pff, get real

The case where the outcome is contrary to the intended logic is the bug. In this case, the intended logic is that being in a boat negates fall damage.

​@@SnakebitSTI it clearly started life out as a bug which they've kept around because people like it's quirky/exploit nature.

@@TimSheehan > a bug which they've kept around because people like it's quirky/exploit nature. it is a feature then, not a bug

@@zocker1600 ... you know that "not a bug its a feature" phrase is from people making fun of buggy/ faulty things right? E.g. Oh the door handle has fallen off... Not a bug its a feature car is very secure against breakin now!

​​@@TimSheehan It becomes a feature once its presence becomes intentional. Your examples are different.

People who code in Java must love this!

@@Vifnis I do 😅

Now go compute some inverse square roots :p

Yeah the missing 198 is at least part of what kept me from seeing a pattern.

Yeah, I'm bad at maths but decent at pattern recognition and the pattern didn't take me long to see: Along with what you said, the distance between each pair of numbers is: - 36 = 12×3, - 60 = 12x5, - 84 = 12x7, - 108 = 12x9, Following the pattern, the next 10 numbers in the sequence would be: 447, 453, 609, 616, 796, 804, 1008, 1017, 1245, 1255...

And why I avoid floats unless I'm _really_ certain I need them

​​@@adfaklsdjfI remember that once I actually created a class 'fraction' that had integer numerator and denominator to be able to compute some numbers exactly, I remember I needed it for something haha

I'm not a programmer, but I did some gauge stuff for Flight Sim back in the day. One involved some INT(float) operation that wouldn't work (iirc I was converting to octal for transponder), and I was completely unaware of the float rounding error. Obviously, my "programming" failed and I couldn't figure out why. So I set it up in Excel to see what was going on, and to my amusement even excel did it wrong. Only when I expanded number of decimals did it show i.e. 4.9999999543xxx instead of 5. I've been aware of it since, even if I've never had to deal with it again.

it helps when you understand that they are in fact precise, just inaccurate

​@@adfaklsdjf Why would you be using any data structure that you aren't sure you need?

I just wish he tried to find a pattern on his own for a bit longer, both before OEIS and especially after OEIS failed. That feels like a more rewarding way of doing things.

I am disappointed that as of the start of April 2024, it is still not in OEIS.

Now the "Minecraft Boat Fall Damage Sequence" will be introduced to OEIS.

I asked ChatGPT (not the best option, I know, but it was worth a shot) It started strong (-ish?) and then started (convincingly/confidently) pulling stuff out of it's virtual arse. :)

Same.

Thank you for reminding me of this! I felt like i was missing one piece of information and i think this was it. Boats used to break and now they don't! Matt keeps talking about a fall damage calculation but I think a better name would be *collision* calculation. And to protect them against damage this is the code they added which in turn inadvertently also keeps the player from damage as well!

So the bug isn't really that the boat breaks at certain heights, but that the "avoid having the boat break all the time" workaround causes a bug where you _incorrectly_ don't receive fall damage while sitting in a boat _except_ with these rounding errors.

@@SteinGauslaaStrindhaug this is a good way to put it!

I remember boats breaking... I always hated that but maybe I should go back and play old minecraft

​@@SteinGauslaaStrindhaugit used to be a bug, now it's a feature!

I mean, unless you're on the bottom

if you jumped into a cobweb or are sliding down honey as well

My first thought was that it might apply to ladders.

My first thought was "when you're swimming in lava". I guess that resembles my fortune, errr, skills, when playing Minecraft. :)

You watch at Matt Parker not knowing numberphile's most featured website? Quite suspect

Neil Sloane

They have the founder of it -Neil Sloane on quite a few numberphile videos. He's pretty cool

I think Matt has a copy of the EIS (i.e. the printed version, before it was the OEIS).

@@nocturnomedieval They're probably new here - be gentle.

it rounds to the nearest number of blocks fallen minus 3 (meaning first 3 blocks don’t deal damage, every block afterwards deals 1 point of damage). so if you fall 3.49 blocks, you take 0 damage. if you fall 3.5 blocks, you take 1 point, half a heart of damage.

@@davidbarroso1960 but would you take damage in a boat if you fall 12.5 blocks, for example?

@@davidbarroso1960That doesn't matter, because floating point inaccuracy could still occur and place you very slightly above the ground

@@yonaoismeI'm assuming it hasn't been tested yet because it isn't really necessary to fix the bug, but it could be really easily tested by spawning boats with command blocks, slowly travelling upwards by using an invisible armour stand

@@yonaoisme no. for you to take damage, the resulting position has to be “equal” to the floor position. but the resulting position never ends in .5, so it’ll never equal the floor position if the fall height ends in .5.

Bells, enchantment tables, cakes, are some more specific ones, but (and this is slightly cheating) there's also falling ANY distance up to 0.01 precision by falling sideways out of a water stream or honey blocks (or even technically off of vines/ladders/scaffolding but that's way harder to be precise ironically)

Are you actually lower than a block on soul sand or is that implemented differently? It's unlike grass paths because you can place rails and redstone.

@@fwiffo Yes, you are lower, which is why you can get effects from blocks below the soul sand, if ice for example is below the soul sand you will find it gets slippery, and this isn't hard coded, it's just the fact that the block your feet are in is the block above the ice. You can also see the Y coordinate in F3 go down when you step on soul sand.

@@fwiffo I'd guess that Soul Sand simply has a lower collision for the top part, so it isn't actually a full block, only visually represented as one.

​@TheRenaSystem and different stages of amethyst shard growth, pitcher plants, and...

Stan-up maths

Stand-Up Crafts

And the bad thing about this...

@@babilon6097 Is that nothing ever goes wrong with a bit of Math, right...

At this rate Matt will have to actually play it...

_[Agile/Scrum PTSD intensifies]_

oh god please no

I'm just surprised that SciCraft hasn't figured out how to hack Minecraft physics using this.

I was looking into horse jump heights a while back and found out that if you're riding a horse in moderately-recent versions, it experiences .02 vertical drag twice per tick, or around .0396 per tick. Do you know if other entities experience this sort of effect in any direction?

I was about to comment that I remember seeing somewhere some acceleration graphs for MC and they weren't constant and then the higher numbers in Matt's sequence prolly won't work but you pointed it out way more detailed *kudos*

@@Kirby703 As far as I'm aware no, but I haven't looked much into mobs riding other mobs like in this case. Maybe that could be something interesting to look at...

One time I was modding a monster that jumps in the air to land on the player and crush them, and I was having *so* much trouble until I discovered Minecraft has **different air resistance** for *vertical* and *horizontal* motion 🤨

Like falling onto ground layered with snow from a snow golem's path?

Rounding floating point numbers is a legitimate nightmare that you should only do if you don't actually care about the result being consistent with anything else in your system.

Thankfully 0.5 is exactly representable in binary

@@jesselapides4390 yes, but should 0.1+0.4 round up or down? Should 0.1x5 round up or down? Should 0.05x10 round up or down? If you're dealing with precise numbers you usually don't need to round

@@jesselapides4390 Sure, but not all numbers with a .5 at the end are representable. For example, the number 65166464.5 is stored as 65166464, which means if you round 65166464.5, it will round down instead of up, like it should.

@@xanderh2404 You cannot store 65166464.5 in a 32-bit float. You can't store even 65166464 vs 65166463 in a float (ie, you'll get the same value (65166462 to 65166466 all "round" to 65166464))

While true, I feel like I should point out that data packs can change worldgen to go from -2032 to 2032 minecraft.wiki/w/Custom_noise_settings

Yeah you're right

Wait until a work colleague wants to use floating point numbers with currency 😨

@@SushiElementalDo it and pull an Office Space 😂 (but unlike the movie, do your arithmetic correctly)

@@natsuzkan Haha I'll think about it 🤓

@@SushiElemental There's an open source ERP (not super well known but actually used by some companies) that uses floating points for currency. ERP does inventory, accounting, etc. One of the many reasons that made me want to tear my hair out when working with it.

@@bobson_dugnutt It seems some people just want to lose money.

🙃 Minecraft physics

I was trying to figure out why no one was pointing out that there was a missing value in his list... several times.

12 and 13 also don't cause player damage, though, (and also don't break the boat when a player is in it), so I don't think that's the reason. I think people just repeat the "theoretical" values without testing them very thoroughly. Something about "player in boat" is different from other cases, for both the rider and the boat itself.

I was thinking of that, though that was included in the bug report. The video that was documenting this though was specifically about the java edition, which I don't know if that's either significant, or that wouldn't make a difference and it was just overlooked when the short was published. Either way, the important issue is that this phenomenon was recorded through the bug report, which I'm more inclined to monitor.

@@BlueCyann There's about 2 kinds of negligence upon the short's inclusion of 12 and 13, as it's omission of 198. However, nothing about this is theoretical. It's in the bug report, and the report does mention that there are different conditions which the player does/doesn't survive (but never specified which). They simply used examples of the affected values to illustrate the point about the bug and how that's handled. I would think that would've been beyond the point to include the player in the test used in this video when it's illustrating the fall damage to the boat with those values. When the boat is "destroyed" is a whole matter unto itself with the game's code in relation to the distance. That's already covered.

So does that mean that when you use the standard == comparison operator in your code that under the hood the CPU is actually subtracting one from the other and checking that the result is within plus or minus epsilon of zero? So that in something like Java, that's a thing that's being done without the programmer necessarily knowing it?

@@AlRoderick I think it's something the programmers integrated (and OP just means it's common)? Because I definitely have been screwed over by this before

​@@AlRoderick actually no! "==" does a direct comparison, which is why 0.1+0.2==0.3 (as floats, idk about doubles) returns false. There is probably some package for comparing floats somewhere :)

@@catmacopter8545 Yup, exactly this. The language can't make the == operator do this by default because somebody someday might actually want to do a true strict comparison of floats. Math packages will usually have a way to do this, or you can just do it yourself. It's not exactly a common problem, most of the time you're working with integers (not necessarily the _datatype_ integer, the mathematical integer).

@@AlRoderick No, == does an exact comparison, which would cause you to fall through the ground unless you were clamping the value, because you would never exactly be at the same level as the ground. Clamping instead of comparing an approximate value may be a way to fix this bug though, like only comparing if you are at or below the ground and correcting the position to be exactly equal to ground level to avoid ever rounding in a way that causes fall damage.

Which can be covered by the status "is_swimming" or something :P

yes. there can often be good reasons to have an enum status variable but also a bool that covers a specific case of that enum

​@@Khantiathat would bloat the code and if you add enough stuff like this your performance will suffer It's more efficient to cover all 3 options with just the two states

@@bilboswaggings first time I'm hearing of such practices... since when is it more optimal to check if (this.status == ON_GROUND && !isGrounded) instead of just (this.status = IN_WATER)? Computation-wise the latter option is more efficient. It makes little sense for this to be the reason for the existence of isGrounded :P

@@Khantia as shown in the example both are needed for the game to function as it's built Adding a status for in water would also require you to add the code necessary for that check (how else would you know what the status is)

When would it though? Considering the video deals with all the height values in MC

@@Baddabythere are ways to launch a boat upwards so it goes above build height, like hooking a bunch of fishing rods to it and pulling them up at the same time

@Baddaby Okay, this is an excellent question because my initial answer would have been: "Because of the build height limit in Survival worlds there is only enough height in Sandbox worlds to achieve terminal velocity so Matt must be testing in a Survival world... etc" BUT after reviewing the video again it does appear that Matt WAS indeed testing at heights above the Survival height limit (it makes sense that one would do certain tests in a Survival world) so in theory some of the higher tests should be reaching terminal velocity and no longer be subject to acceleration, so what's going on here? 🤔 I did a little more digging and my current theory is that he is testing on a Super Flat world that starts at Y=60 and bedrock is directly beneath one layer of Grass blocks, which would potentially reduce the maximum falling distance by 60 blocks e.g. a result from Y=300 is actually only falling 240 blocks. Again I could be wrong, perhaps Matt is the only person who can answer this, I wonder if he considered terminal velocity at all or if it would even effect the results but I feel like it should, right?

​@@Baddaby *deep breath* And just to complicate this further, terminal velocity in Minecraft is different for different entities! For a; Player=78.4m/s, Item or Block=40m/s, Arrow=100m/s I don't know which of these is applicable for a player in a boat, your guess is as good as mine 😂

what's the point of the emoji?

Stuck in a spider web…

perhaps falling inside of a non solid block like fire might not count as in the air? just a thought

Climbing a ladder

@@Halinn would you be in the air while climbing a ladder? or are you just "in" the ladder?

This is entirely dependent on the 0.04 acceleration constant, so it doesn't really seem like something they would add. "Integers 1/25th of a triangle number" is pretty arbitrary.

@@cadekachelmeier7251 Methinks we need a way to search the OEIS for sequences that are offset or a multiple of some sequence then. Because triangle numbers are probably in there and would come up.

"Block heights in Minecraft, from which being in a boat will not save you from fall damage"

​@@cadekachelmeier7251as if they don't have completely arbitrary sequences

@@LethalChicken77 they do but they try to keep them to a minimum . . . however I think an exception will probably eventually be made for this sequence, just because it's in this video.

They never removed it that you can break the boat. They made it way sturdier and able to pick up.

@@Systox25 but the code is vestigial because these bugged block heights are the only time breaking a boat drops sticks and planks now

@@ordanarymods4990 thought explosions would also destroy them into pieces.

@@Systox25 fun thing is, it always gives you oak planks no matter the wood type used, meaning you can use boats to convert any wood type into oak

@@Systox25 they don’t

I would guess that some of Matt's viewers play minecraft as it is probably popular in programers too. But most of the minecrafters are probably children who do not know Matt and math.

​@@Neo-vz8nhnerdy 20 year olds likely make up a sizable chunk of his audience and I'd bet those people grew up on Minecraft (or were at least around it) and would have played Minecraft. Though I agree that obviously the majority of Minecraft players have never seen one of his videos

i watch(ed) ilmango on the scicraft server and i watch people like marcel vos in openrct2 and im also here, so theres at least one XD

Someone made a CPU in Minecraft. Like a .0000001 MHz 8086 or something.

@@BobCat0someone also made a 1 hz cpu in Minecraft. Most other ingame redstone computers run way slower though

Pretty sure that IS "doing it properly", given the context of the problem.

​@@chrishillery I would argue it's not, because you're not getting the analytically correct answer for any given input. Floating point was designed to be a jack of all trades format; it's good enough at representing basically any kind of number, but not the proper way for anything specific. In this case, a proper way to do it could have been to store postion as an integer multiple of 0.04 (essentially, a fixed-point format rather than floating point). This way you can get a proveable, exact algorithm.

@JochCool Sure, that's why financial apps don't use floating point. But that's what I meant by "the context of the problem". *Given that* you're working with floating point, the only valid implementation of "equals" is that kind of epsilon comparison.

​@@chrishillery I still disagree: there already exists code that makes sure you don't fall through the ground - the code that checks for collisions. If you can detect for collisions, you can certainly check whether you have "finished falling" and are, in fact, on the ground, in the exact same way as the collision - this should hopefully eliminate the distinction of becoming "on ground" and touching the ground. But hey, the margin is a neat trick that probably made for less code and fewer bugs than my idea, so I don't mind it.

​@@JochCoolActually, your position isn't limited to 0.04 increments if there are forces other than gravity

hi misspotato

wait, scaffolding stops fall damage?

@@endernightblade1958 I'm not certain actually. I just assumed so because they're just cube shaped ladders, but I haven't actually tested it.

@@misspotato813 i think i vaguely recall them stopping me when crouched at least, but as far as i know without that they’re practically a solid block

@@endernightblade1958 if you are holding shift when falling onto scaffolding you will not take any. If you're not holding shift you're taking full dmg

I'm really glad I wasn't the only one

or that they are not updated at the same time. either way, the real bug isn't floating point rounding, but de-sync of state.

Moreover, blocks in unloaded chunks don't have collisions. If an entity is so fast, it moves into an unloaded chunk in one tick, it will go through all blocks it would otherwise collide with in that one tick (before also becoming unloaded)

Thank you! I was confused his explanation here too

So, if they used the same method/formula to calculate when (this.status = ON_LAND) and when (onGround = TRUE), then this bug would be fixed, right? Or are they somewhat different things?

So this is a case where they decided that a bug was a feature, then got a "bug report" for cases where the game was working the way it was originally supposed to? I haven't the words to tell you how much I love this. ETA: Apparently, the "no falling damage while on a boat" thingy was a by-product of fixing a different bug (boats broke when rammed into land or something). Dear gods, how I love it all.

Isn't the issue that the boat is breaking, and then the player falls through and hits the ground? The boat, without even having a player in it, breaks from certain fall heights.

This makes much more sense! I do hope this gets pinned!

how is waterlogged counted? because i think it is, when you hit a solid piece of a block, that is also waterlogged

I do not know, but it seems possible that it's the same issue between the math and the conflicting status.

@@VoltisArt Maybe it's when you get a fall damage death (some kind of floating point issue) while the game is telling you you're in water.

I would bet nearly any amount of money that the death happens when your falling speed is such that after one "tick" you are exactly aligned with the top of the slab, having been outside of the water the previous tick. I can think of several different code weirdnesses that could actually be responsible for the death, but I bet that's the edge case regardless of the specifics.

I think that message occurs because the developers didn't account for you dying in that specific way, and didn't assign a proper message to play when it happens.

If you're in the ground, you count as on the block beneath your feet, I think If you're in the void, you count as in the air but just don't have anything to hit Liquids shouldn't count as in the air or on the ground, and probably don't Cobwebs shouldn't change if you're in the air or on the ground, and they'd just affect your falling speed/acceleration

You can also just exit the boat right before hitting the ground

@@thesurlywombat Yes, this is far easier. Getting back into a falling boat reliably is a little bit tricky, mechanically.

There were a _lot_ of caption errors. One of the few times I've turned off non-automatically generated captions other than because it was covering up parts of the video.

There's been plenty of caption errors in recent videos. I'm almost inclined to volunteer to contribute to the subtitles myself.

Hey man, don't be boolean the captions.

There's also "12th" instead of "20th" when talking about ticks.

Voice-to-text seems to have quite a problem with uncommon words, nor does it judge well by context. That is also similar to why I do not trust auto-correct. I prefer it mark suspected misspelling and let me correct it myself. Those random word switcheroos are really annoying.

Someone mentioned here that the boat has a maximum velocity but it needs to fall from higher than here shown

actually, when a player is sitting in a boat, air resistance is added. However this does not appear to be the case for mobs, so when you put a pig with a saddle in a boat it doesn't have air resistance. You can then hop on the pig and enjoy gravity freely

@@breznknedlwith enough Hermits with some fishing rods and you got yourself a space program 😂

I notice this when speedrunners launch a shulker riding boat high in the sky for transportation. The boat comes down at such speed it looks like teleporting to the ground

It could be read for players or changed for other entities, but not changed for players, right?

@@thesurlywombat correct. you can't actually modify player data.

Ladders

I would guess OnGround is used to reduce your mining speed while it's false (ever tried mining something while on a ladder? It takes like twice as long)

No, because 1/25 has no common divisors with 1/8. The closest you get is .125: 2 ticks in you're at y-0.12 leaving you 0.005 above the ground. You'd need an accumulated error of 0.004 to pass the check. The precision error is 0.0000000009, so you'd have to fall for 4.4 million ticks to accumulate enough errors to trigger the bug.

It rounds