mjmcacctjira
Assigned
Reported
Comments
I have not done switch development, but, if the switch allows attachment to arbitrary processes by a debugger, and you let me know, happy to do so or setup a VM on a local machine you can remote desktop into and control. If anyone on the MC team is in the Microsoft building on 17th street in Atlanta by Atlantic Station happy to bring down the switches in person to demo it for them (or hell, have a visit at my home if they care, it’s not far).
If I seem overly interested in getting y’all a reproduction: It’s clearly the hard part, both technically and “business decision” wise. I believe this is a QOL issue for switch players which effects “low to medium” revenue customers, the exact ones who are most experimental with “low cost and microtransaction” games. I don’t want competitors to become the default everywhere, etc, because y’all start missing the low spec but still current generation consoles and the like, but I bet whatever changes causes this issue likely advanced some other business or development goal, but not with the intent to disable this capability.
And as to eliminating the chance of bad RAM/radiation/high elevation on one of our three switches: We are only about 1000ft above sea level.
We see crashes only in places in other games other people report crashes. We see them all at the same places in games, including when not playing networked games. (Like Dead Cells only crashes at load screens that are widely reported for crashes, or upon specific unlocks which crash other people). Most games never crash. (We definitely understand how unreasonable of an expectation that is for a game the age of minecraft)
We live in a place with low radiation brick, in a brick house (which lowers cosmic ray interference from the side) and almost never have radon in basement’s in our area. We are friends with professional radiation maintenance workers and the last time a geiger counter was in our house there was not incidental radiation (for fun, not because of contamination).
Our Switches are powered through surge protection if not surge protection and battery backup. Power interruptions are rare and uncorrelated with the issue.
Do you need me to upload worlds again? I think the minecraft scenario “The Mountain” is a much faster repro than the vanilla world. It truly is very quick to show the error. (see other comment). If it’s a matter of “our testing lab for switches isn’t on the eshop so we can’t download that”, I can attempt to generate a more trivial procedure to reproduce the error.
If you believe this error was specifically, intentionally remedied in a version (which I can download on a switch), I will test it.
If you believe this may be related to mesh networking/eero (even though we saw it on simple wifi too), I likely can find three iOS or desktop OS or a chrome OS version and try to reproduce it, but I’d need a little guidance to make sure I had an equivalent version.
*strong text*Internal and SD card storage both show this issue. Multiple different switches as host show this issue. Using mesh wifi (eero) or direct wifi (google fiber, dlink, att) show this issue. Using resource packs or no shows this issue. Using no packs, simplest fast wifi and “not janky machines” eventually still shows this issue.
I only mentioned scenarios and such as it was fast to reproduce. There was not a workaround, very basic small farms or any standard exploration would eventually show the same issue.
You can repeatedly stimulate this issue with small farms (a few foods grown, then a pen for animals which you can walk around and toss lots of food), then send one of the other two players out of chunk, perhaps through a portal to another area. The whole goal is getting the sync effort high. The configuration of three players on local network sync on three switches with a ~50 mb save file alone will repro the error when done naturalistically. I do not know how big a “trivially” generated large save file (like overly exploring with a boat or something) would have to be to generate the error but basic minecraft “build a base” stuff fails after a couple of days of play with this limitation.
We have tried a few nintendo switch versions to see if the error is corrected. We can do so again periodically. Right now, minecraft was MIA in our gaming rotation though due to this bug.
When you're trying to restart the Switch to see if that helps fix errors: Make sure you're not just sleeping the switch. That's like when you close a laptop, it doesn't really restart it and make all the programs load anew. You usually are just "sleeping" the switch when you "turn off" the Switch, so it's easy to not know how to really reboot it.
Here are Nintendo's directions on how to fully restart the switch:
Hold down the POWER Button on the top of the console for 3 seconds.
Select "Power Options" from the Power Menu that's displayed.
Select "Restart".
Nintendo says if that doesn't work:
If holding the POWER Button for 3 seconds fails to bring up the Power Menu, you can press and hold the POWER Button for at least 12 seconds to force the console to shut down.
One relevant point I'm not sure I said: The host for this game does not have a Microsoft account (yet). The secondary and tertiary Switches were played by players who do have one, but the errors occur whether or not those players were logged into the MS account or not.
First of all: My main motivation for providing the bug report at all. Bold is just emphasis, as this is a long post, not yelling, btw 😃:
I feel recent {}software changes, likely made for real reasons, have likely significantly degraded the for-free multiplayer experience which Minecraft has had up until very very recently. It is incredibly unlikely this was intentional, but as there are servers you sell to run the game, and you have new paid Minecraft IPs (Minecraft Legends), etc, the point should be made that it's "not a good look" for people looking to cast that aspersion (which I am definitely not).
I personally don't think these are at all linked. I think the internet, generally speaking, has a pretty dodgy grasp of correlation vs causation. Summer vacation and it's attended travel just started in many parts of the US and lots of families are going to be experiencing this bug (as realms and wifi are going to be less certain, and more local play will occur). I assume, if this bug is as widespread as I suspect it likely is, fixing this may become surprisingly high priority if that happens and wish to equip the team to do so if they choose to do so. **
As to why I pointed at the scenarios:
When we played Arctic Phoenix after getting it for free a couple years ago originally, it had basically no real problems loading for us on a 2017 switch (plugged in and shown on a TV) and 2 clients (2020 normal switch in handheld mode and a 2020 switch lite also in handheld mode). We are fairly religious about moving games to the system memory vs (the fairly high speed) SD card if there are performance issues, so that may have been the performance difference for us?
Apologies if I was unclear about why I was mentioning the scenarios: I'm not specifically charging the core Minecraft team with responsibility to make sure all these scenarios work in all situations. I felt pointing out scenarios available on the Minecraft store which quickly demonstrated this bug when 1 switch host and 2 switch clients were being used (or even 1 and 1 often), would be an easier way to reproduce than just giving a Minecraft world which is not a predictable result to get to. 1 server 1 client seems basic "gotta have that working to advertise the game has local multiplayer" level of functionality.
I'm reporting failures in very typical vanilla worlds which have very reasonable sizes recently becoming unplayably bad in local multiplayer. I bring up these store scenarios as clear exemplars of things which stimulate essentially the same bug for us. Kids playing in vanilla survival would easily get to this complexity of a world in a couple days of plays.
We have experienced this on multiple vanilla worlds as well. The main test world is a vanilla world and we actually only checked the Legendary Phoenix and Minecraft: The Mountain to try to find an easy reproduction to show what would cause this bug as it really did feel like a general performance degradation of the core stack, and not a "world based problem".
As in, I thought the test team could restart the Mountain scenario repeatedly to see what the threshold is for failure as it quickly gets to these sizes, but the vanilla world we most recently worked on, for instances, would take many hours of replay to get it to a threshold place which definitely causes the behavior and it might not if you don't do something specific with animals or the nether or a village etc. It only takes about 15-25 minutes of play for even a 7 year old, for instance, to get Minecraft: The Mountain far enough along to cause this issue. We saw the issue by the time we got to the "Collect these 6 items quest". I assume a professional tester would take about 7 minutes to repeatedly get to that point.
I say this because the fault (saving becoming so overly prioritized it boots clients after causing lots of lag), is very likely (from a software perspective) not simply "exactly 50MB" and you'll definitely see it. Serialization of a Minecraft world to a file possibly takes longer based on specific complexity and specific happenings in world, and I knew both of these scenarios had a lot going on there and would be likely to demonstrate it. It may not even be serialization, but the snapshotting or something like that. I just wanted a state easy to get to and to get to similar states repeatedly which do not exhibit the error as well. For instance, at the beginning of that scenario, for instance, all 3 people can play and the saves do not lag us out, take an undue amount of time, nor anything like that. The scenario itself (if you follow it's directions) also quickly tells you what to do to reproduce the bugs (as you just play with 3 switches enabled and you'll see it happen after not that long).
In our recent vanilla world, there are a few animal pens, some growing crops, and a things like that near our home base. Nothing YouTube worthy, but nothing "this should clearly break the game". We probably haven't built a mine cart yet and the most red stone we have is 5 observers on a tiny automated sugar cane greenhouse which is tiny (like 5-6 stalks wide, one line) and textbook. But the fact there are X sheep or Y cows could be the issue in that world, rather than the exact number, and we'd never know. (Note: being far away/in the nether while the animals and crops exist does nothing to help the bug). We moved some work tables around in a village fairly far away, and discovered a Forest Mansion and built a long cobblestone path through the nether. We built a couple mines (not automated). We have a box which automatically composts stuff you put in a hopper, and a similar box which automatically cooks food you put in a different hopper. Essentially very basic, not very demanding (from what I understand) of a world.
Many of our other former worlds have much more in the way of minecarts and things like that I feel would likely exacerbate the issue/make people go "it's clearly the whatever which is too much", but I'm willing to investigate if that would be helpful for some reason. We built this world "light on the machine" per se, from what we understood.
If uploading our particular file of this particular vanilla world would be that helpful, I'm happy to do it; I just need to link my daughter's account, upload it to a realm, then download it on a desktop computer. Hell, I'm happy to connect up a debug mode client or something locally on a Mac or windows computer and export logs. I am a professional programer (mostly iPhone) and am willing to climb down the "enable all the log files" rabbit hole if it's of use next to a system reliably exhibiting the fault.
I'd honestly expected this bug report to be "oh, that's a better reproduction for [Known bug so and so] than the one we have" and just be hot linked to a dupe in the JIRA. I really don't expect this to be a unique bug at all: I think this is a performance regression, probably deeper than was intended, and the team is likely partially aware of the chance of this happening, just not how quickly it does. Just tried to bring value to show "Look, all this stuff that used to work now doesn't, things really are getting worse, and likely past the commercially acceptable point all things considered".
For instance: Despite repeated 10x - 20x times we logged into the game from the switch lite above on 6/4/23, the Nintendo Error log only shows 2 error 2306-0332 that day (that's the black screen switch error you get when the network disconnects). Those two errors are when hard shutting down the Minecraft software (home button then X), not the game. When in the past we've ever seen a game repeatedly have network issues, we've seen those happen far more in these kinds of error logs.
I'm happy to provide account IDs and switch IDs the appropriate way and upload it if there are error reporting analytics to correlate to. If you have a document about "How you don't get pwned when helping us solve bugs and providing PII!" document to make sure I do that the "appropriately secret way", let me know. Re bog standard JIRA, "Viewable by All Users" toggle will not let me select any group other than "users" when changing who can see comments. so I bet "just drop that info in a comment" is not the right way to provide that data with this particular site. Happy to become more well informed as well on "how Minecraft works" to try to repro better if needed.
Can I get other players who are on three switches but have problems to try to do the mountain scenario to confirm for the devs how quickly they can generate the same error messages I showed? It’s a truly fast reproduction of the error. Less than an hour.