Time Nick Message 00:00 IhrFussel Then it will be (ab)used by server owners for non-testing 00:00 sofar nobody will play on those servers 00:00 sofar or, people will play but without the csms actually running on the players clients 00:01 sofar and if the server kicks clients if they detect that their csms aren't running, players will leave 00:01 sofar I think that's a good tradeoff 00:01 sofar excellent, even 00:01 sofar developers don't need to wait for cdb to test code 00:01 sofar experienced players can 'trust' if they want 00:01 sofar most players will likely never and are secured 00:02 IhrFussel You need to explain to me again hwo I can, as server owner, send my own non-distributed, server-specific customized mod code to clients 00:02 Shara A server would need a strong player base with a high trust level for the owner to be able to push such CSm to any great extent, and the more it was required as opposed to optional extras, the more difficulty there would be getting more players. 00:02 sofar IhrFussel: same way that signed mod code goes to the client, direclty from server to client 00:02 Shara The strength would really be in optional extras 00:02 sofar only dfference is the embedded signature 00:03 sofar so what me and rubenwardy envision is something like this: 00:03 sofar mod writer makes a mod with an embedded csm 00:03 sofar posts the code to cdb 00:03 sofar cdb reviews and signs the csm part, 00:03 sofar cdb then publishes a zip file with the embedded signatures for the csm parts 00:04 sofar the server owner puts THAT zip file contents on his server 00:04 sofar the client connecting then downloads the signed csm and signatures directly from the server 00:04 sofar the client verifies the signature with the cdb pubkey it has 00:04 sofar the client sees they match and runs the csm 00:05 IhrFussel Shara, think about the possibilities one day...when you can tell the CLIENT to update the HUD every sec instead of having to loop through every client server side to do that... and particles, physics etc... it can make gameplay a LOT smoother 00:05 sofar I think in the end players will want CSM more than server owners 00:06 sofar since it will allow resources on the client to be used that are otherwise idle 00:06 sofar and on the server side resources are likely at a premium 00:06 sofar because of scale 00:06 Shara Fussel: yes, that's why I'd like to see it happen like this. 00:07 Shara But giving clients the ability to decide what to accept is pretty important 00:08 sofar I'm myself a fan of the "mozilla certs" type where you can say "I trust these people to decide for me" 00:09 sofar which is why I think that CDB review model will work very well for everyone 00:09 sofar both players and server owners 00:09 sofar as for server tweaks 00:09 sofar IhrFussel: I think that it will force mod writers to properly use settings and non-hardcoded defaults and behavior 00:09 sofar which is going to be better code 00:10 sofar e.g. better use of mod channels and local storage 00:11 sofar here, imagine this scenario: 00:11 IhrFussel Allowing clients to decide which code runs and which not will cause more work for server owners still ... and it means servers still have to run the same code server side for clients that refuse it to run locally ... was it planned like this? To have duplicated code server and client side? 00:12 sofar forget about that 00:12 sofar there's no reason to worry about that 00:12 sofar if you would, you're going to write terrible code that will make your server crap 00:12 sofar you can never offer the same experience to players who refuse to run any csm 00:12 sofar so don't 00:13 sofar and the whole idea that the code is "duplicated" is a bit of a fallacy 00:13 IhrFussel But Shara, nerzhul and possibly others want it like that ... to give the client the freedom to run only the functions/callbacks the user allows 00:14 sofar this whole "optional" / "required" / "soime third level" is bullshit imho 00:14 sofar I've only heard nerzhul talk about it 00:14 sofar is anyone even in favor of that? 00:14 paramat (ok i see the issue, a server could alter sscsm code from what is published and a client wouldn't know and can't check the code) 00:15 IhrFussel Shara supports it on GH and another user 00:15 sofar honestly I think it's just a waste of time to even implement it 00:15 Shara Fussel, quite honestly I want CSM in any form to go bye bye, but we know that's unlikely 00:15 IhrFussel Let me check the post 00:15 sofar paramat: no, it can't 00:15 Shara Fussel, I'm in favour of the server getting more control. 00:15 Shara If there is a better option than one specific approach, by all means take the better option 00:16 Shara But I mostly think you are worrying way too much 00:16 sofar signed csm's would give the server control with the exception of clients not running any 00:16 paramat "to give the client the freedom to run only the functions/callbacks the user allows" i oppose that too 00:16 Shara I see most things not being required 00:16 sofar to me, it's all or nothing, and signed csm implements just that 00:17 IhrFussel 1. https://github.com/minetest/minetest/pull/8000#issuecomment-448514658 2. https://github.com/minetest/minetest/pull/8000#issuecomment-448941129 00:17 paramat yes, accept sscsms or nothing, that's my preference 00:18 sofar there is nothing wrong with people disabling javascript in browsers 00:18 paramat my comment https://github.com/minetest/minetest/issues/8002#issuecomment-448800634 00:18 sofar there is nothing wrong with players disabling sscsm in minetest 00:18 Shara Fussel, if you think I mean clients should go through and accept each individual thing, then no, not really what I mean at all 00:19 sofar if the server owner doesn't like it, they can kick the player ... 00:19 sofar as I said, I will even write a CSM for server owners to detect if players don't enable csms for them 00:19 Shara But if I join some new server, know nothing about the owner, and have no way to check what the code is or does.... then I want to be able to say no to running it 00:20 sofar right 00:20 Shara If it's possible to confirm it's on CDB and has been reviewed by someone there, then I will feel I should be able to trust it... If someone I know has written it, again, I may feel I can trust it... otherwise? no thanks. 00:20 IhrFussel Shara, I think you don't see a big issue: What if my mod now includes important SSCM code that NEEDS to run on the client or else the client will have a broken gameplay...now that client refuses to run that code because of "restrictions" ... it will make everything extremely complex 00:20 Shara It's nto complex at all. 00:20 Shara not* 00:20 sofar as I said, I will even write a CSM for server owners to detect if players don't enable csms for them 00:21 sofar in other words, to kick them 00:21 Shara It would just mean you have to either release that on CDB or somewhere else that will be trusted 00:21 Shara Or you need to accept you may not be able to offer a good experience to some players as a result of your own choices 00:21 IhrFussel We are not talking about that... we currently talk about clients deciding which features of the API can run and which can't 00:21 paramat yeah not complex. servers can refuse clients tha refuse sscsm 00:22 sofar IhrFussel: remember one thing: it's trivially simple to modify CSM and make everything a stub "faking" csm to the server 00:22 paramat nevermind nerzhul's silly idea 00:22 Shara Fussel, as a server owner YOU decide if you will demand clients to accept a certain thing in order for the experience to not break 00:22 sofar IhrFussel: if players want to do this, why would you bother defending against it? 00:22 paramat (*that) 00:23 IhrFussel You people confuse me: So do we all agree that nrz's proposal (let the client have different allowed "API levels") is bad? 00:23 sofar honestly, I think there shouldn't even be a way for the client to notify the server that sscsm is disabled or the csm mod code wasn't trusted 00:24 sofar different levels is bad, and not needed 00:24 paramat Ihr yes i agree 00:24 Shara I'd favout a system where you trust the source of the mods instead 00:24 Shara favour* 00:24 Shara With CDB then being the main official source 00:25 IhrFussel Okay so the thing you want is the required signing of the SSCM mod which forces server owners to distribute their modified mod "somewhere" 00:25 sofar "forces" only by popular vote, essentially. They can sign code themselves. 00:26 sofar signing by cdb allows oversight to happen that players need to be able to trust it 00:26 IhrFussel But who then checks if the code is dangerous? 00:26 sofar cdb reviewers 00:26 sofar rubenwardy, me, those type of people 00:26 IhrFussel Yes so server owners have to distribute their custom versions to the cdb in order to get it reviewed and let clients dwnload it 00:27 IhrFussel download* 00:27 sofar if, if, they want cdb to sign it 00:27 sofar yes they need to show it to cdb 00:27 sofar cdb will not sign unseen code 00:27 IhrFussel Else the clients will be warned that the mod is unsigned and might be dangerous? 00:27 Shara Not unsigned... but not signed by CDB 00:28 sofar detail problem 00:28 Shara You could sign it, and players could choose to trust you :) 00:28 sofar I'm not entirely sure about how to handle it 00:28 sofar most likely silently not trust it 00:28 IhrFussel Okay but you don't plan to outright CANCEL the option to load "non-cdb signed" mods from the server? You still want to let users decide? 00:30 sofar hold on, complex sentence, parsing 00:30 sofar well, so, I haven't made up my mind as to how the client should behave exactly 00:31 IhrFussel I mean you want to give users the option to still load the SSCM without it being signed on cdb 00:31 sofar however, I do believe that "warnings" to the player are bad 00:31 sofar oh wait 00:31 sofar so what I actually want is that the client can trust MORE reviewers than just cdb 00:31 sofar but, adding trust should be "explicit" 00:31 Shara Warnings can be bad, because they can scare people away big time 00:32 sofar as in, you go and download a file and put it in ~/.minetest/certs/ 00:32 IhrFussel rubenwardy said that this was his original goal but is more work before he left 00:32 paramat yeah, cdb shouldn't be the only trustworthy reviewer 00:32 sofar and then minetest will allow a non-cdb signature 00:32 sofar we need this, otherwise 00:32 sofar developers could never test sscsms 00:33 sofar and even 00:33 sofar I would allow players to remove 'cdb trust' 00:34 sofar which is effectively the same as "don't run any sscsm" 00:34 sofar I mean, if there's no other cert of another reviewer 00:34 sofar maybe we can even chain signatures so you could allow mods to be signed by multiple reviewers 00:36 IhrFussel SSCM sounds more and more complicated to implement every time people talk about its implementation 00:36 sofar anyway, I have to skedaddledoo 00:36 sofar this isn't that much more complicated 00:37 sofar even review and signing are relateively simple 00:37 sofar most likely we can make MT even sign stuff and trust local signatures in singleplayer mode 00:37 * sofar & out 00:42 IhrFussel BTW I found it pretty rude to be called an "extremist" by nrz just because I don't support his ideology of SSCM 00:43 IhrFussel If we take the "client needs all the freedom it can have" too far we can just scrap the whole SSCM idea 00:45 paramat 'some server owners are all extremists' =D 00:45 IhrFussel SSCMs are run locally then by definition, so some people will always say "this is my computer and i decide what can be run by servers on it" ... server and client opinions clash here 00:46 Shara and those people should go play singleplayer :) 00:46 IhrFussel I mean the server wants the client to run x but the client says "no this is not the server's decision I only run y" 00:48 IhrFussel Until now it was pretty easy to distinguish between "server rights" and "client rights" cause all mods are 100% run server side and the client only receives commands from within the c++ code ... that will change with SSCM and I think there will be quite a few big arguments...controversies 00:49 Shara What do you think we already have with CSM? 00:50 paramat any questions about csm ask in #7995 i will attempt to explain 00:50 ShadowBot https://github.com/minetest/minetest/issues/7995 -- Attempt at a summary of CSM, as requested (plus possible issues) 00:51 paramat lol right 00:51 IhrFussel Yes that is onyl the beginning...cause right now many people who "need" their client provided mods get mad when they are told to disable them 00:51 Shara So what is your solution, Fussel? 00:52 IhrFussel The question will be "which side is more important to obey" 00:52 paramat the client will be able to refuse sscsms of course, and the server will be able to refuse clients that refuse sscsms 00:52 IhrFussel IMO the clients have to obey when they play on my server ... but others might see it differently 00:53 Shara Which is no different than any other aspect of runnign a server ever 00:53 Shara Or we would not get rule breakers or ban dodgers or peopel trying to enable privs by editing client code 00:54 IhrFussel Yes they can disable SSCM that is not the question...I mean some "freedom elitists" might argue that they want the "API level" crap added to control the server sent mods more 00:56 Shara Sometimes it's like we invent issues just to have issues to argue. 00:56 paramat lol 01:06 IhrFussel Well it#s not 100% clear where the line is drawn between server and client ... and SSCM will blur that line even more 01:12 Shara I've already asked you what your solution would be 01:14 Shara CSM is already a thing. As much as I'd love to wish it away, that's a bit difficult. 01:14 IhrFussel Not sure if there is one... server owners wants as much freedom as possible and client wants that too 01:14 IhrFussel I mean freedom in terms of features executed client side 01:19 paramat there's still a part of me that wants to pull it all out, may still happen, but looks like we're giving it a try first. i'll be keeping a close eye on it from now on 01:21 tumeninodes damned if you do / damned if you don't ... choose your flavour 01:21 * VanessaE peeks in 01:21 tumeninodes :) 01:22 * tumeninodes throws VanessaE a creepy smile 01:22 VanessaE >:-) muahhahah 01:23 VanessaE I'll see your creepy and raise you an evil laugh :D 01:23 VanessaE also hi 01:23 tumeninodes howdy 01:24 tumeninodes my version of home decor ; ) https://forum.minetest.net/viewtopic.php?p=338529#p338529 01:24 VanessaE seems I've walked into a ...ahem... discussion of great volume. :) 01:24 VanessaE nice stairway :) 01:24 tumeninodes yes... I came in advanced and "disposed" of all the evil wine and scotch before it started 01:25 tumeninodes thank ya 01:25 VanessaE odd. the forum hasn't been emailing me with updates on that thread. 01:26 * VanessaE checks spam 01:26 tumeninodes haha, fitting place for it :P 01:26 VanessaE hm, nothing relevant 01:27 tumeninodes exacticall.... 01:27 VanessaE *moves a couple unrelated things out of spam, deletes all* 01:27 tumeninodes hey?! 01:28 VanessaE ? 01:28 tumeninodes nvm I thought the nothing rel comment was regarding the thread, I was tying to be funny-ish 01:28 VanessaE heh 01:29 VanessaE nah, nothing relevant in spam I meant :) 01:30 tumeninodes spam folder holds the only emails I get :( 01:30 VanessaE :< 01:31 tumeninodes gonna tackle adding these this function to walls and stairs mods soon, and then another half ass attempt to work it into mapgen via engine 01:31 tumeninodes https://pastebin.com/NFX5ThJv 01:31 VanessaE good luck with it :) 01:31 tumeninodes definitely gonna get messy... I am sure I will cry too at some point 01:33 tumeninodes I find the best way to have fun, is not having any idea wtf you're doing 01:33 VanessaE heh 01:34 VanessaE isn't that how the best discoveries are made? :) 01:34 tumeninodes those... 01:34 tumeninodes and tragedies 01:35 VanessaE bigger risk, bigger reward, eh? 01:35 tumeninodes that's what I've heard 01:35 tumeninodes not sure anyone has tested it out thoroughly 01:36 tumeninodes unless my reward system is broked 02:28 rubenwardy another coursework submitted! 2 to go 02:36 VanessaE Achievement unlocked: homework finished! 02:36 VanessaE :D 11:00 ANAND Hello. Has anyone else come across a runtime error which does not specify a modname, the line in file that causes the error, or a stack traceback 11:00 ANAND ?? 11:00 VanessaE aside from just flat-out segfaults? 11:00 ANAND Doesn't mention segfaults 11:00 ANAND It just says this: 11:01 ANAND ERROR[Main]: ModError: Runtime error from mod '' in callback on_receiving_message(): attempt to call a nil value 11:01 ANAND ERROR[Main]: stack traceback: 11:01 ANAND ERROR[Main]: Check debug.txt for details. 11:01 VanessaE I've seen that before. I get lots of segfaults more than anything else 11:03 ANAND But doesn't a segfault mention "segmentation fault" in the logs? 11:03 VanessaE probably not. 11:03 ANAND Hmm... I see 11:03 VanessaE (but I capture the console output, rather than relying on debug.txt) 11:04 ANAND I've done that too 11:04 ANAND I've also tried --trace 11:05 ANAND I just get that single error message. Nothing else 11:05 ANAND This happens when I try to run capturetheflag (sub-)game on 5.0.0-dev 11:07 ANAND Until recently, it ran just fine on 5.0.0-dev, which is even stranger. 11:08 ANAND I think a recent commit probably broke compat between 0.4.x worlds and 5.0.0-dev 11:08 ANAND I'll try creating a new world 11:28 ANAND It doesn't throw the error in a new world. o_O 13:25 * rubenwardy sighs https://forum.minetest.net/viewtopic.php?f=9&t=21599 14:42 bobr fricken hackers man 14:42 bobr theyve got noclip and creative now apparently 15:05 Fixer fuck github/ms 15:05 Fixer do you remember layout for posting a code in github? 15:05 Fixer --- or smth 15:06 T4im ``` 15:06 Fixer thanks! 15:25 IhrFussel bobr, modified clients cannot cheat privs and it's up to the mods to prevent users with insufficient privs from accessing items ... there are also quite a few mods with item duplication bugs so that's another way to cheat free items you shouldn't have 15:28 IhrFussel A big security hole are storage nodes which don't correctly check the privs in the allow_take() callback ... for example if someone codes a CSM which is able to access metadata of inventories and that check is missing then you CAN take the stuff 15:29 IhrFussel Many inventory mods only check privs in on_rightclick() which is a BIG no-no 15:37 IhrFussel unified_inventory seems to be safe from such an exploit since you don't have actual item stacks in the menu list (right side) but only formspec buttons which then check your privs as soon as you click them... shouldn't be able to bypass that with CSM 15:41 Shara Fussel: they might not be able to cheat privs as such, but they can skip the requirement for having certain privs 15:43 IhrFussel Shara, only if the inventories/storage nodes are missing the checks I wrote above...there shouldn't be another way 15:43 Shara fly/fast/noclip have nothign to do with inventory checks 15:43 IhrFussel Before CSM the allow_take() could never be executed without on_rightclick() (without client hacks) 15:43 Shara nothing* 15:44 IhrFussel But bobr said they can have creative now 15:44 Shara He also mentioned noclip 15:44 Shara And your reply was general - implying clients cannot "cheat" any privs, which is not really true 15:44 IhrFussel Yes but I assumed it's well-known that noclip can be enablked with hacked clients...just like fly and fast 15:45 IhrFussel Pretty much every movement related stuff can be cheated 15:45 Shara So make sure you are specific when you dismiss a claim of cheating 15:46 IhrFussel But AT LEAST items are 100% safe from hackers if the definition of the inventories/storage nodes are correct and check privs on EVERY action not just rightclick/punch 15:47 IhrFussel Yeah I should've mentioned that...I forgot about the other part of his message when I wrote my reply 15:47 Shara Are they really? 15:48 IhrFussel There is no way to bypass the allow_take() check of inventories AFAIK 15:48 Shara As far as you know 15:48 Shara 100% is just a pretty bold claim 15:49 IhrFussel 100% safe up until this day 15:49 IhrFussel Nobody knows if people will find a way to bypass that check... but I doubt it's possible... we can just ask some core devs how secure allow_take() is 15:50 Shara Well, I can recall at least two ways I could have helped myself to items from other players locked chests on current stable servers (I reported the issue to the server owner in each case) 15:50 Shara And those were nothing to do with inventory checks 15:51 Shara Mods don't always play nicely together in the ways that people might assume, and there are very few things you can ever be 100% certain of when it comes to security 15:51 IhrFussel Are you talking about mesecons? 15:51 Shara No, and I'd rather not say more 15:52 IhrFussel You mean you don't access the contents directly but still receive them? That sounds impossible 15:52 Shara Nopt my fault if you can't think creatively 15:52 Shara Not* 15:52 Shara AFK 15:53 IhrFussel Does it require certain broken mods to do it? Or it works in vanilla MTG? 16:14 IhrFussel Okay so some of my previous messages might have sounded confusing/misleading ... what I ultimately mean is that, as long as the server has no mods that have broken code/missing checks, even the most modified client cannot just give itself any item it likes 17:05 rubenwardy ContentDB redesign: https://i.rubenwardy.com/95enj.png 17:08 rubenwardy when you're not logged in: https://i.rubenwardy.com/Rwlki.png 17:20 Andrey01 rubenwardy: i don`t see the Content DB is redesigned now 17:20 rubenwardy because it's not been deployed yet 17:21 Andrey01 i even went to Mesecons page, this is same as before 17:21 rubenwardy because it's not been deployed yet 17:21 Andrey01 that page that you showed 17:21 sfan5 rubenwardy: the download button could use a greener green if you ask me 17:21 rubenwardy yeah, I was actually thinking that 17:21 rubenwardy I'm also not happy with the button placement 17:21 rubenwardy the ones in the image should probably be icons instead 17:25 Andrey01 and next question: does "content" menu section in the game display dependencies, more full description and releases now from content.minetest.net? 17:25 Andrey01 of each mod 17:28 Andrey01 does it display now? 17:31 rubenwardy no, the displaying hasn't changed in a while 17:31 rubenwardy there's related issues for this 17:31 Andrey01 ohhh 17:31 rubenwardy the point of CDB is to have a well structured source of information 17:32 rubenwardy so that all the mods[tm] can be added before full support is even implemented 17:53 Krock hi Fixer 17:53 Fixer hi Krock 17:54 Krock you didn't see that one coming :D 18:21 IhrFussel shivajiva, after the import I no longer need 'auth.sql' correct? 18:22 shivajiva correct 18:23 IhrFussel Doesn't sauth delete it automatically or went something wrong? 18:25 shivajiva don't think it deletes it, the main mechanism for getting doing a transfer is manual import 18:25 shivajiva -gettting 18:26 shivajiva I can only see a cleanup for sauth.sqlite on re-run 19:46 Fixer minetest 5.0 runs on my kitkat, but not for long if going into contentdb, since I have at most 400mb of free RAM, if you don't go there and just create a world - works with 24+ fps 19:53 Fixer kinda sucks it can't dump unused parts of ram when low on ram, including on 32 bit systems 20:00 IhrFussel The "content" tab can even make my 1 GB RAM phone crash 20:00 IhrFussel Something doesn't seem right there 20:00 rubenwardy It should probably just be disabled on Android 20:01 rubenwardy and link to this instead: https://play.google.com/store/apps/details?id=com.rubenwardy.minetestmodmanager&hl=en_US 20:01 IhrFussel rubenwardy, any reason why it requires so much mamory? 20:01 IhrFussel memory* 20:02 rubenwardy your phone is probably already using half of that 20:02 IhrFussel It#s weird that MT SP runs fine but the cdb menu can crash 20:03 rubenwardy SP is C++, cdb is Lua 20:03 rubenwardy each value in lua is at least 64 bits 20:03 IhrFussel Oh WOW 20:03 rubenwardy also, the list is stored twice 20:03 IhrFussel It crashes after 30 secs 20:03 rubenwardy once is sorted 20:04 IhrFussel The problem is likely that it needs to load all results at once 20:04 rubenwardy opening the store increases usage by 66MB on a computer 20:05 IhrFussel Let me check a realtime ram usage widget 20:07 rubenwardy flicking through the pages also causes the memory to rise, probably because of the images 20:07 rubenwardy MMDB had this problem too though 20:07 rubenwardy just had a lot less mods 20:07 rubenwardy (the cdb code in the main menu is inspired by MMDB) 20:08 IhrFussel 70% RAM free when I start MT 20:08 rubenwardy going in-game causes it to use 300MB quickly 20:09 IhrFussel As soon as I open the cdb menu the RAM overlay disappears which means memory low and after 20-30 secs it crashes the whole app 20:09 IhrFussel That cannot be normal 20:10 rubenwardy what does "after 20-30s" mean 20:10 rubenwardy can you see the store? Do images appear? 20:10 IhrFussel 20-30 secs after I tap "Browse online content" the app crashes 20:10 IhrFussel Yes 20:10 IhrFussel It all appears but after that time it crashes 20:11 rubenwardy do you do anything, like change pages? 20:11 IhrFussel Basically it idles in the menu until the crash 20:12 IhrFussel I can change pages but it still crashes... no matter what I do it will crash after that time 20:13 IhrFussel I'll try to stay in the search field 20:14 IhrFussel No crash when searching it seems but I assume it will crash as soon as I leave the type mode 20:15 IhrFussel Nope, still works...now I wonder if it crashes again after leaving the search mode for 20-30 secs 20:15 IhrFussel Yep it does 20:15 rubenwardy what does search mod mean? 20:16 IhrFussel When your phone shows you a blank field to type 20:16 rubenwardy ah, that will be pausing the activity 20:16 rubenwardy so Minetest won't be running in the background 20:16 rubenwardy probably 20:16 IhrFussel Yep but it still takes 20-30 secs to crash in the normal cdb menu... 20:17 IhrFussel Even after leaving the "type mode" 20:17 IhrFussel Now I search "mobs" and maybe it will not crash anymore cause of only 3 pages 20:18 IhrFussel Still crashes...amount of pages doesn't matter 20:31 IhrFussel rubenwardy, the monitor app I used was brokjen it seems...now I found a working one and my free RAM was 340 MB and as soon as I opened the cdb menu it went down to 90...that's when the monitor service was closed by the OS 20:32 rubenwardy I may have reproduced the crash 20:32 rubenwardy not sure 20:33 IhrFussel I will now reboot the phone and see how much it has free...maybe it will survive longer 20:34 Fixer my tablet has about 300 mb free ram before minetest start 20:34 rubenwardy yeah, I reproduced it again 20:34 Fixer rubenwardy: i had weird problem with cdb on android that images on first page were ok, but others pages did not have images (once), then it crashed on page 4 or so 20:35 rubenwardy I have a suspicion that it's related to malformed images 20:36 IhrFussel I get my phone to 400 MB free now...let's see if it's still not enough 20:36 Fixer rubenwardy: yet it did happen like once or so, other times they were ok 20:37 IhrFussel RAM usage goes up at a rate of 10-20 MB per SECOND...I guess it really is lua 20:37 rubenwardy I think it's a screenshot issue relating to how the libraries work on Android 20:38 IhrFussel It still crashes...so 400 MB RAM is not enough on Android 20:39 rubenwardy oh, ffs 20:41 Fixer ofc it is not enough, jordach will say 640 gigabytes should be enough for you phone 20:41 Fixer on a serious note, I doubt you can squeze it into 400 mb 20:41 Fixer At start it is something like 300mb in windows 20:41 Fixer and then it grows 20:45 Fixer minetest on win7, main screen: 47 mb 20:45 Fixer rubenwardy: each page of CDB: +5 megabytes of memory (win7) 20:46 Fixer minetest singleplayer: 223 megabytes of RAM just after start 20:46 rubenwardy images 20:46 Fixer not sure if memory released after some time 20:47 Fixer if releases* 20:52 IhrFussel But we are not talking about singleplayer... I think it's a bit much to require 400+ MB RAM JUST for 1 single menu..let me check how much SP actually needs 20:53 rubenwardy CDB requires 60MB 20:53 rubenwardy on Linux 20:53 rubenwardy and SP is around 300 20:54 IhrFussel Starting MT eats 30 MB ... SP with just vanilla MTG needs ~ 100 MB 20:54 IhrFussel It only needs 100 on my phone 20:55 IhrFussel Of course it depends on the mapblock cache limit/timeout 20:56 rubenwardy ok, I'm going to try something 20:56 IhrFussel WOWOWOWOWOW.... 20:56 rubenwardy although, hmmmm 20:57 IhrFussel Just tapping "Content" starts to drain the RAM 20:57 rubenwardy it doesn't on PC, which is weird 20:57 IhrFussel And the pictures don't load 20:57 rubenwardy oh, what do you mean "Content"? 20:57 rubenwardy the tab? 20:57 IhrFussel Where you select the game 20:57 IhrFussel Packages* 20:58 rubenwardy ah yeah, there's a screenshot there 20:59 IhrFussel Not sure how but suddenly my phone has 520 MB free 20:59 rubenwardy a background instance of MT was killed, probably 21:00 IhrFussel From 520 allthe way down to.......1XX ... there must be a HUGE bug in the Android version 21:02 IhrFussel When I switch the tab WHILE it drains it doesn't go back up 21:03 IhrFussel I had 420 MB in "Join Game" then "Content" drained it down to 375 and I switched back to "Join Game" ... still 375 MB 21:04 IhrFussel And every time I switch back to "Content" it drains more without giving it back after 21:04 rubenwardy well, there's somethings I can do to improve this 21:04 rubenwardy namely reduce the image size server side 21:05 rubenwardy but I'd like to wait on that until we fix this issue, because it would make debugging harder 21:05 IhrFussel So you think it is just the download of all the images? 21:05 rubenwardy I think that images are filling up the menu for some weird reason 21:05 IhrFussel It makes sense if the download pauses when switching the tab 21:06 rubenwardy *memory 21:07 IhrFussel But only on Android? 21:07 rubenwardy apparently so 21:08 IhrFussel Let me compile a 5.0.0-dev real quick on my pC 21:11 Fixer IhrFussel: no-no, main menu just takes 45 mb or so 21:11 Fixer IhrFussel: singleplayer is 200+ mb 21:11 Fixer IhrFussel: content db page requires 5 mb per page flip 21:12 Fixer IhrFussel: with vrange 240 ofc, with android one can probably be 100mb + 21:13 IhrFussel Is it normal that it says "target version 0.4.17-dev-blahblah" ? 21:13 IhrFussel I didn't use a checkout command 21:14 IhrFussel I only used this https://forum.minetest.net/viewtopic.php?f=3&t=3837 21:14 * tumeninodes is starting to wonder if it might be best to build a version of MT/MTG native for Android 21:14 Fixer IhrFussel: how you monitor the memory use with adb? 21:14 Fixer with top? 21:14 rubenwardy he's not 21:14 IhrFussel No, there are monitoring apps for that 21:14 Fixer or is there something more user friendly 21:15 IhrFussel "Resource Monitor Mini" 21:15 IhrFussel Puts a nice screen overlay 21:16 IhrFussel Guys is this NOT compiling 5.0.0-dev? https://forum.minetest.net/viewtopic.php?f=3&t=3837 21:17 Fixer is it even up to date? 21:18 IhrFussel It should compile master...always 21:19 tumeninodes ughhhh... android is such a royal PITA 21:19 IhrFussel Wait...uhh one sec 21:19 Fixer nah, it has aids, will not install, also 3 megabyte download for """"mini"""" 21:20 tumeninodes is it quite possible, that Android has overtaken MS Windows in the #1 sucks spot? for adware, and just being down right specific? 21:20 ShadowBot https://github.com/minetest/minetest/issues/1 -- GlowStone code by anonymousAwesome 21:22 IhrFussel Fixer, you didn't say that you need some ad-free ... there might be another monitor app that doesn't try to make money ... there are plenty of such apps 21:22 Fixer tumeninodes: android is #1 in spying, like light years ahead 21:22 ShadowBot https://github.com/minetest/minetest/issues/1 -- GlowStone code by anonymousAwesome 21:23 tumeninodes Go Go Google 21:23 rubenwardy made the commit: https://github.com/minetest/contentdb/commit/4dc2424c3bc41c206861516131425b076fab30bf 21:24 IhrFussel BTW my issue was the little "cd;" command in Calinou's script...I want to compile 5.0.0-dev in "mt5" dir but that cd jumped back to home and compiled in "minetest" which already had 0.4.17-dev...so solved 21:24 tumeninodes I say build a separate native android version, put it up for $1 like everyone else and be done with it 21:24 rubenwardy deploying that will reduce the cost per page from 5MB to something hopefully 9x smaller 21:24 IhrFussel Is it live right now? 21:26 rubenwardy no, because it would make fixing this harder 21:28 IhrFussel Will fixing it require a newer Android build? I'd guess not cause it's a download 21:29 rubenwardy that commit can be done independently of Android versions 21:30 rubenwardy but the proper fix for this bug is likely to be needed in the app itself 21:37 IhrFussel Uhh 21:37 IhrFussel Where is the cdb menu in the PC version? 21:37 IhrFussel I don't see it 21:39 rubenwardy Content 21:39 rubenwardy sounds like you still have 0.4.x 21:39 rubenwardy if you have minetest installed globally, then make sure to build with RUN_IN_PLACE=1 21:39 rubenwardy cmake . -DRUN_IN_PLACE=1 21:42 IhrFussel The header said 5.0.0-dev but the main menu was still 0.4.X it seems ... I didn't install it I only run it by ./bin/minetest ... but I didn't specify that flag 21:43 IhrFussel So maybe it grabbed the wrong builtin? 21:53 rubenwardy if you have minetest installed globally, then make sure to build with RUN_IN_PLACE=1 21:54 rubenwardy the 5.0.0 engine is using 0.4.17 builtin 21:55 garywhite tumeninodes: Then it would only be fair that whoever already has Minetest in their Play Store account should get the 5.0 version for free if they were to do that 21:58 IhrFussel rubenwardy, it's fixed now thanks... when closing the game while you are in the cdb menu an error appears "2018-12-21 22:57:24: ERROR[Main]: Bitte wählen Sie einen Namen!" 21:58 IhrFussel "Please choose a name" 21:58 Krock old news 21:59 IhrFussel memory usage in the PC version is really only 50 or so MB in the cdb menu, so something with Android is faulty 22:20 Calinou IhrFussel: I should turn my script into a full-fledged thing people can download and run, this way it could be more robust :P 22:23 IhrFussel Calinou, your script works 100% flawless if the user didn't have MT installed globally before and wants the newest dev version... things like "don't go to home before cloning" and "specify run in place" are just more advanced things you can't take into account 22:27 tumeninodes garywhite: Only if they agree to pitch for the game j/k. Android is it's own beast, and best dealt with natively. And it is very hard to find anyone who is specifically skilled with android to do all which is required, without compensation. 22:28 tumeninodes Most who work with android on a reg basis either make $ from their apps or the ads which get space on their apps 22:29 tumeninodes android has become the myspace of OSs all glitter and poppy bubbles 23:15 Calinou IhrFussel: it could install something more integrated (with a shortcut icon, presence in the `PATH` and stuff like that) 23:15 Calinou that's all doable without requiring any more permissions than currently (they're only required to install dependencies currently) 23:39 IhrFussel I tested stu's newest apk with no cdb screenshots loaded...still drains memory and crashes 23:42 IhrFussel Hi paramat 23:54 paramat hi, can you open an issue for this? 23:56 paramat it'll be a blocker