Time Nick Message 00:07 hecks hi, what does the content_cao.cpp / GenericCAO::getPosition() / if(m_matrixnode) block do? 00:09 hecks I commented it out and it fixed #9139 , everything seems to work without it 00:09 ShadowBot https://github.com/minetest/minetest/issues/9139 -- Server receives an incorrect player position after attaching to an entity 00:16 hecks https://github.com/minetest/minetest/blob/c1e01bc638637efa788b5698238a465406bc3f5e/src/client/content_cao.cpp#L412 00:18 hecks hm okay, so the detach point is now wrong if the "vehicle" moves, but the yanker is definitely fixed 00:31 hecks ah, it's for moving the player with the parent, but it's also responsible for the bug somehow 01:49 appguru picked #9727 up again and implemented my APIs from the last post, will polish & push later 01:49 ShadowBot https://github.com/minetest/minetest/issues/9727 -- Implement generating & storing of inventory images by appgurueu 01:53 paramat skyliner_369 please stop posting off-topic in this channel. this channel is very strictly for MTEngine and MTGame development only, not modding, advice or questions. if in doubt ask in the other channels. sorry to be strict but, we have to be 10:22 ANAND Core-dev inputs required: https://github.com/minetest/minetest/pull/7924#discussion_r436252386 11:44 Krock ANAND: three lines could be added to make it backwards compatible 11:50 Krock other than and the possibly superfluous examples in the documentation, that the PR looks good. 11:53 Krock ANAND: https://github.com/minetest/minetest/pull/7924/files#r436096986 12:18 nerzhul then, i'm lost, the event published to MT is exactly the same in wayland & X11 BUT mt (ingame) doesn't properly recognize keys. when i add some debug code i see that keycode is pushed to isKeyDown std::list, but on the future get it's not in the list, very disapointing 12:29 Krock ANAND: I don't understand " and the example is required at all". Are you saying that the example is a requirement? 13:09 jas_ #7924 13:09 ShadowBot https://github.com/minetest/minetest/issues/7924 -- Allow binding dig, place actions to keys; remove LMB/RMB hardcoding by ClobberXD 13:09 jas_ thanks for the approval rubenwardy. 13:12 rubenwardy #10000 13:12 ShadowBot https://github.com/minetest/minetest/issues/10000 -- Recalc mesh normals for CAOs by dcbrwn 13:22 sfan5 now that's a good pr 13:34 ANAND :tada: 13:34 ANAND Krock: Sorry, I meant "not required at all" :) 13:36 Krock ah okay. In this case it might be better to shorten it to a brief list. Also to keep the Lua API file a bit smaller in size 13:37 ANAND Agreed 13:42 ANAND Shall I keep the joystick LMD/RMB unhardcoding (which I said I'll complete in 7924) for later? I haven't ever seen the joystick code, and it'll take me some time to get acquainted enough to work with it. 13:42 ANAND Ideally, I'd prefer someone who's already familiar with joystick handling to take it up if possible. 13:43 ANAND But if not, I can try. 13:44 Krock joystick would fit better into another PR 13:44 ANAND Ok, thanks 13:46 ANAND The original control_bits docs was cluttered, IMO, and hence I added line-breaks for each bit. Shall I revert those changes? 13:49 Krock up to you. fine either way 13:53 ANAND I'll leave it modified then 13:54 ANAND Krock: I've attended to both your requested changes (docs and rundata.punching) 13:54 ANAND both of* 13:58 ANAND Btw, did we come to a consensus regarding the breakage of `repeat_rightclick_interval` setting yet? 14:06 Krock meh, probably fine. 14:06 ANAND Ok, fine 14:08 Krock ANAND: I cannot eat apples 14:09 ANAND checking 14:11 Krock also reported here https://github.com/minetest/minetest/pull/7924#issuecomment-640058175 14:15 rubenwardy lool 14:16 rubenwardy my testing involved playing a game for a while, didn't test on_use 14:17 Krock rubenwardy: how much do you know about Android development? I was told that this could be updated using a merge request https://gitlab.com/fdroid/fdroiddata/-/blob/master/metadata/net.minetest.minetest.yml 14:18 rubenwardy I was a professional Android dev for 10 months :D 14:18 rubenwardy well, part time 14:19 rubenwardy that doesn't look like Android, looks like f-droid specific confis 14:20 Krock "yml" file. could be some sort of build configuration 14:21 Krock looks like it also needs a versionCode bump from 24 to 26 in the source code 14:22 rubenwardy we already done 14:22 Krock nice 14:22 rubenwardy it's done by the release script 14:24 rubenwardy Meeting today? 14:24 rubenwardy also, 5.3.0 freeze date decision 14:25 Krock yes 14:29 ANAND Wait, what? Talking about feature freeze already? 14:29 ANAND I thought the 5.3.0 release was months later :O 14:29 rubenwardy no, it was planned 1 or 2 months after 5.2.0 14:29 rubenwardy due to Android support 14:30 ANAND Ahh, ok 15:45 appguru Is there something like mod security for CSM? 15:52 rubenwardy yes, it;s also much more restructive 15:52 rubenwardy there's no whitelisted mods 15:53 rubenwardy it also ignores secure.enable_security 15:53 rubenwardy worth checking those things though 15:55 appguru So where can clientmods write? 15:55 rubenwardy no where 15:55 appguru I have implemented the storing thing for my inventory image generation PR using CSM, but I want to "secure" it somehow. 15:56 rubenwardy they have mod storage 15:56 rubenwardy but can't directly read or write 15:56 appguru :/ 15:56 appguru So I'll guess this will be a bit harder than expected 15:56 rubenwardy merging #9998 in 10 15:56 ShadowBot https://github.com/minetest/minetest/issues/9998 -- Add HTTP API to main menu by rubenwardy 15:56 rubenwardy sfan5, Krock: technically, we've never exceeded a merge time line 15:57 rubenwardy we never say what base "10" is in 15:57 rubenwardy merged an hour later? 10 is obviously base 60 15:57 appguru I think I'll just let 'em write to their mod folder 15:57 sfan5 rubenwardy: hmmm 15:57 rubenwardy CSMs don't have a mod folder at run time iirc, they're loaded into a virtual filesystem 15:57 rubenwardy iric 15:58 sfan5 correct 16:02 appguru IRIP 16:02 appguru Can I somehow determine the currently running CSM? 16:04 appguru The engine usually does some register magic for mods but this doesn't seem to work for CSM 17:00 Krock Today's meeting points: https://dev.minetest.net/Meetings <- nore, rubenwardy, sfan5, ( paramat ?) 17:00 rubenwardy trivial bug fix: 17:00 rubenwardy !title https://github.com/rubenwardy/minetest/commit/3c9d2cc8f608f500b11f65c5162381bf3ea48d4a 17:00 ShadowBot rubenwardy: Use correct extension for ContentDB thumbnails · rubenwardy/minetest@3c9d2cc · GitHub 17:01 Krock lgtm 17:01 rubenwardy merging in 10... 17:01 rubenwardy (that helper function isn't technically correct if the filename doesn't contain a dot, but it will always contain a dot) 17:01 Krock Meeting in 30 minutes, perhaps? 17:03 rubenwardy actually, merging a bit later 17:03 rubenwardy afk 17:03 rubenwardy 30 minutes is fine 17:04 rubenwardy I should probably refactor the CDB code in the main menu at some point, and separate the logic from the dialogs :? 17:07 Krock if it's generic code, yes. else I don't see much need for such efforts 17:11 sfan5 > Translations and minetest.conf.example need updating. Volunteers? 17:11 sfan5 what does this mean? 17:11 Krock well, so that translators get some time to translate the new strings until the next release happens 17:12 sfan5 oh that 17:12 Krock and minetest.conf.example can be done in the same step 17:12 sfan5 that's usually done when freeze starts 17:12 sfan5 and now that the wiki properly outlines the steps anyone could do that 17:12 Krock then it'll serve as a reminder 17:12 sfan5 (but I don't mind being the one to do it) 17:14 appguru I currently have a working CSM store_texture(texture, path) function 17:14 appguru Which I want to make "secure" - it shouldn't be able to write anywhere except for the CSM modpath 17:21 Krock void ScriptApiSecurity::initializeSecurityClient() 17:21 Krock s_security.cpp 17:21 Krock for path protection, use bool ScriptApiSecurity::checkPath in your functions 17:23 appguru Krock: this doesn't work for CSM 17:23 Krock then you need to implement it 17:24 appguru Yeah but I have little clue 17:24 sfan5 the fundamental issue is that CSM should have no direct disk access at all 17:25 appguru It is not direct 17:25 appguru It's just retrieving textures from the TextureSource & storing 'em 17:26 appguru But still, I'd like to keep those stored textures where they belong 17:26 Krock or get the hash of the media and copy the cached files over 17:30 Krock >> Meeting check-in goes here 17:31 Krock rubenwardy: you might want to push your CDB patch 17:32 rubenwardy pushing 17:32 rubenwardy done 17:33 Krock > Allow or deny UTF-8 characters in the source code 17:33 Krock This is about a potential change of the code stylig rules 17:33 Krock if characters should be allowed - which ones? 17:36 Krock > Do not use characters in C++ files from outside the ASCII character set. 17:36 Krock I'd suggest "Prefer ASCII characters as long it does not compromise readability.". 17:40 Krock objections? 17:43 appguru Imma reopen my revert PR then huh 17:44 appguru #9828 17:44 ShadowBot https://github.com/minetest/minetest/issues/9828 -- Revert "Replace non-ASCII characters in gameui debug display code" by appgurueu 17:44 appguru I'd say: 17:44 appguru - extended ASCII should always be allowed 17:45 sfan5 what exactly is "extended ascii"? 17:45 appguru 8-bit ASCII 17:45 appguru https://en.wikipedia.org/wiki/Extended_ASCII 17:45 appguru It contains "°" for example 17:46 sfan5 so ISO 8859-1? 17:47 appguru Yeah, aka Latin-1 17:47 appguru Although there appear to be small differences from what the wiki article says 17:49 sfan5 There are many extended ASCII encodings (more than 220 DOS and Windows codepages). EBCDIC ("the other" major 8-bit character code) likewise developed many extended variants (more than 186 EBCDIC codepages) over the decades. 17:49 sfan5 """small""" differences 17:50 sfan5 but since source files are not encoded in latin-1 I don't think that proposal makes much sense 17:52 sfan5 IMO it'd make more sense to allow all unicode (or the inverse: only ascii) 17:52 appguru I'd allow all Unicode TBH 17:52 appguru Except for some emoji poorly supported by fonts probably 17:52 sfan5 unicode use could be restricted to the BMP to exclude e.g. emoji but it's not like anyone would use those anyway 17:53 hecks Not every programming or terminal font has every character 17:53 hecks and when trying to display one, programs will typically pick a fallback character from a different font, potentially ruining monospace 17:53 appguru A "terminal font" should at the very least have full Latin-1 support. 17:54 hecks They probably do; the situation is more sketchy with character sets like shift-jis 17:54 appguru My terminal fonts usually even support color emoji, but that's on a different note. 17:56 hecks "all unicode" involves not just emoji, but also RTL marks and zalgo 17:57 hecks a better question is, what do you need unicode in code for? 17:58 hecks unicode also has the wonderful property of having several ways to represent one character 17:58 hecks which has security implications 17:59 appguru We probably don't need Unicode to be fair. 18:00 hecks ｙｏｕ ｐｒｏｂａｂｌｙ ｄｏｎ＇ｔ． 18:00 jonadab I could see the argument for needing Unicode support in player names. 18:00 hecks But this is about source code 18:00 jonadab Right. 18:00 appguru There was just a case where "°" was used for outputting degrees, and I think it should not have been changed to cryptic UTF-8 escape sequences (hence the PR) 18:01 appguru Unicode in player names? What a nightmare. 18:01 jonadab appguru: Yes. 18:01 hecks and in source, just about the only legit use for unicode is to write people's names in comments correctly 18:01 jonadab I'm not saying it should be implemented, only that I could see the argument someone might make for it. 18:02 Krock so.. no changes so far? 18:02 appguru Just merge my PR 18:02 jonadab hecks: In comments, you can go ahead and use Unicode, and if the user's text editors supports it, fine, if not, oh well. It's a _comment_. 18:02 jonadab The compiler doesn't have to care. 18:02 appguru Unicode could allow using some neat symbols for certain messages 18:03 jonadab Regarding the degree symbol, make it a #define DEGREE_SYMBOL or something. 18:03 hecks The issue is that allowing unicode in source, and not rejecting it by policy, allows for sneaky underhanded code that looks fine on most fonts... maybe, potentially. 18:03 jonadab To get the ugly escape sequence out of the part of the code people are looking at. 18:03 Krock sfan5: so you're opposed changing the code style rules? 18:03 appguru "⚠️ This would be a warning" 18:03 jonadab hecks: Ah. 18:04 appguru Latin-1 should be safe tho. 18:04 hecks for example, j ~= ϳ but your font might display them as the same thing 18:05 sfan5 Krock: not really but it's hard to define what exactly should be allowed 18:05 Krock sfan5: exactly my thoughts, hence "ASCII as long readability is good, else UTF8" 18:06 appguru By that criterion, my PR shoulda be merged :) 18:06 Krock it would not be limited, but discouraged 18:06 Krock zalgo text would be rejected ofc 18:08 Krock rubenwardy: ? 18:08 rubenwardy it should be very limited, and only used in string literals or comments 18:09 rubenwardy ° and … are two examples of unicode characters that are better raw 18:16 Krock "Avoid non-ASCII characters in source files. Other UTF-8 characters may (only) be used in string literals and comments where ASCII would worsen readability. 18:16 Krock " 18:16 Krock ^ how about this? 18:17 sfan5 sounds good 18:17 Krock any additions, rubenwardy ? 18:18 Krock by the way "Do not use or, use ||." is really weird. which C(++) compiler accepts "or"? 18:22 sfan5 I'd guess all of them 18:23 Krock til 18:24 Krock * Feature freeze date 18:24 appguru 2 more weeks please 18:26 Krock I think it would make sense to browse through the PRs and issues to place milestones 18:26 Krock and remove those that aren't necessary -> https://github.com/minetest/minetest/milestone/16 18:27 Krock The visuals and android fixes aside, there isn't too much that is necessary for the next release 18:28 Krock #9807 is probably the most nasty among the bugs to fix 18:28 ShadowBot https://github.com/minetest/minetest/issues/9807 -- Bone animations broke between client versions 18:29 sfan5 right so do we revert that commit? 18:29 Krock that is https://github.com/minetest/minetest/commit/e1fc72c6f 18:30 Krock question is whether the mods relied on the buggy behaviour, or whether it's now finally fixed properly 18:30 Krock s/ or / and / 18:31 sfan5 also regarding the milestone: the PRs I sorted into there are mostly ones that could easily go into 5.3 and I consider useful 18:31 Krock alright 18:33 Krock in the worst case we can revert the bone rotation commit, so we should be ready for a freeze in two or three weeks? 18:33 appguru you could get #9903 merged - a single extra pair of eye is all it takes, it's really smol 18:33 ShadowBot https://github.com/minetest/minetest/issues/9903 -- Properly implement exposing the zoom key by appgurueu 18:33 Krock rubenwardy: any preferred freeze date? early or late? 18:34 sfan5 I'd rather see a freeze in one or two weeks 18:34 rubenwardy yeah 18:35 sfan5 in fact it might be better to decide on a set date or it might get delayed again 18:36 Krock so, next week, but spend a while on bugfixes? 18:36 Krock basically as usualy 18:37 Krock noted 13 Jun for planned freeze start 18:37 appguru two weeks please 18:38 Krock is #9592 related to the high packet sending count, sfan5 ? 18:38 ShadowBot https://github.com/minetest/minetest/issues/9592 -- Sometimes certain packets don't go from server to client 18:38 sfan5 yes 18:39 Krock okay, so not important for 5.3.0 18:39 sfan5 Krock: start freeze in 1 week, then 2 weeks bugfixes as usual I'd expect 18:39 Krock bugfix time will vary as usual 18:41 Krock * Wireframe glitch (graphics related) 18:42 sfan5 was determined to be a driver bug 18:42 Krock #9072 AMD messed up? All we can do is wait for it to be fixed 18:42 ShadowBot https://github.com/minetest/minetest/issues/9072 -- Faces show as if in wireframe mode if anti-aliasing enabled 18:42 sfan5 and doesn't happen on default settings so no action IMO 18:43 Krock indeed, disable antialias seems to work 18:43 sfan5 more interesting would be #9008 but nobody knows anything about that 18:43 ShadowBot https://github.com/minetest/minetest/issues/9008 -- The skybox stopped working. I think an update on my computer may have messed something up. 18:44 Krock perhaps transparency issues? 18:44 Krock >mod containing a texture "hud_image.png" with a flat 90% transparency and 18:44 Krock the author might need to look into this and troubleshoot their own system 18:46 Krock and to come to an end, I might continue working on #9914 to split the Input and Output streams so it can be added somewhen 18:46 ShadowBot https://github.com/minetest/minetest/issues/9914 -- POC: Introduce SerializeStream helper class by SmallJoker 18:46 Krock end of meeting. thanks for participating. 19:03 sfan5 I guess the network logging could be improved regarding the desync thing 19:08 Krock would be great 19:08 hecks does anybody here know CAOs well? I'm trying to nail #9139 but the way CAOs handle position is a little confusing to me 19:08 ShadowBot https://github.com/minetest/minetest/issues/9139 -- Server receives an incorrect player position after attaching to an entity 19:08 Krock the server does not handle attachment positions well 19:09 hecks the problem here is that getPosition reports the wrong value for one frame after attachment, and it's the client's fault 19:09 hecks and it has something to do with that matrixnode, that's the code path that causes it 19:10 Krock have a look at Camera::update and the world offset (200m steps) 19:10 sfan5 random idea: call m_matrixnode->animate() (since that's one of the things that run before/during rendering) 19:10 hecks sfan5: at the top of getPosition? 19:10 hecks or, well, in that block 19:11 Krock hecks: could you please provide a minimal testing mod for this? 19:11 sfan5 why not after initializing the matrix node? 19:11 sfan5 but try both places just in case 19:14 hecks Krock: added a testing mod to the issue 19:15 Krock thanks 19:19 hecks sfan5: m_matrixnode does not have animate(), it's a dummy transform 19:20 sfan5 oh 19:20 sfan5 well then 19:31 sfan5 Krock: do you have any opinion or thoughts on player transfer distance being higher than object transfer distance? On servers I have observed that carts disappear when far enough even while a player is still sitting in them (they then appear to be flying) 19:32 Krock does not matter 19:32 Krock fixed in 5.2.0-dev or so 19:32 Krock well, they no longer teleport after detaching 19:33 Krock they look like they're flying but what do you want to do instead? 19:33 sfan5 why is object transfer distance smaller than player transfer distance? 19:34 sfan5 or are player object messages in fact sent to everyone in the server even if in 30000 nodes distance? 19:34 Krock players might want to find each other easier on servers 19:34 hecks player transfer is infinite on some servers, isn't it? 19:34 sfan5 infinite is the default yes 19:34 Krock player object messages are handled like other object messages 19:34 hecks maybe for mixed object/player hierarchies, use the maximum of the two distances? 19:35 Krock except that the add/remove from the scene packets have different ranges 19:35 sfan5 hecks: sounds like a workable idea 19:35 sfan5 so does that mean players stop updating when they leave the object send range? 19:36 sfan5 I think I should just read the code... 19:36 Krock sorry? players are removed from the scene if they're out of range 19:36 Krock it's like an entity is killed and removed 19:37 sfan5 then my question is how large is that range? because evidently entities "unload" (referring to client-side) before players do 19:37 Krock entities unload when the server tells the client to do that AFAIK 19:38 Krock the server is in full control of what's visible by players 19:38 Krock s/players/clients/ 19:38 Krock and for out of range objects, no position updates are sent (obviously= 19:38 Krock serverenvironment.cpp or so 19:39 Krock also clientiface.cpp for some of the object id caching 19:43 sfan5 https://github.com/minetest/minetest/blob/master/src/server.cpp#L1892 19:43 Krock FWIW, sending only nametags for further away players would also work well, but is a bit more complicated to implement 19:43 Krock m_env->getRemovedActiveObjects 19:44 Krock SendActiveObjectRemoveAdd only keeps track of the client-visible entities 19:45 Krock ah. clientiface only provides the variable (plus container) 19:45 sfan5 so with default settings you have the following values in that function: radius = 4*16, is_transfer_limited = false, player_transfer_dist = 0, player_radius = 0, my_radius >= 4 * 16 19:46 sfan5 https://github.com/minetest/minetest/blob/master/src/server/activeobjectmgr.cpp#L155 19:46 sfan5 and as seen here this translates to unlimited transfer distance for players 19:47 Krock are your questions a follow-up of an issue, or just casually asking? 19:48 sfan5 something I happened to notice, no existing issue 19:48 Krock ah okay. I did not miss anything in this case :) 19:48 sfan5 ...and since this code only checks the known_by set https://github.com/minetest/minetest/blob/master/src/server.cpp#L755 19:48 sfan5 ...this must mean even players 30000 nodes away will get any and all object messages 19:48 Krock and it's known as soon it's added in getAddedActiveObjectsAroundPos 19:48 sfan5 right? 19:49 Krock buffered_messages are all object update messages, and there they are filtered per-player 19:50 Krock if an object is out of range of a player, it's removed from m_known_objects and messages are no longer sent 19:50 hecks okay, it's not the camera offset 19:50 hecks the matrixnode itself is in the wrong place for a frame 19:51 Krock player 30000 nodes away will not get any and all object messages. 19:51 sfan5 okay that was worded badly 19:51 Krock hecks: does that mean the matrixnode calculation happens at the end of a step rather than the beginning? 19:51 sfan5 a player 30000 nodes away will not get any and all player object messages 19:52 hecks maybe, also the wrong matrixnode position is, unsurprisingly, 0,0,0 19:52 Krock I hope it's clear what I meant :3 19:53 sfan5 a I messed it up again 19:53 sfan5 "a player 30000 nodes away will get any and all player object messages" is the conclusion I get from the code 19:53 sfan5 (changed will not -> will) 19:54 sfan5 this is bad for performance but also makes sense, the player object is what makes the client display a nametag 19:55 hecks okay, so this 0,0,0 position is in fact the attachment position 19:56 hecks attachment at 0,1,0 produced 0,0.1,0 which makes sense in minetest's weird coords 19:58 hecks https://github.com/minetest/minetest/blob/master/src/client/content_cao.cpp#L1464 19:58 hecks this would be where it gets parented 20:15 hecks alright, got a possible fix 20:15 Krock will merge #9906 in 10 minutes 20:15 ShadowBot https://github.com/minetest/minetest/issues/9906 -- Restore visual_scale support for nodeboxes (and allfaces) by numberZero 20:16 hecks bleh, never mind 20:17 Krock for #10000: does it suffice to spawn an entity that uses this mesh? https://github.com/minetest/minetest/issues/9985#issuecomment-638375144 20:17 ShadowBot https://github.com/minetest/minetest/issues/10000 -- Recalc mesh normals for CAOs by dcbrwn 20:17 hecks possibly, but I'd use both .x and a .b3d for testing 20:18 Krock .x would be the uploaded file there and .b3d some mobs or player model 20:18 Krock though latter do not have normals backed in 20:18 hecks pretty much 20:19 hecks but since this is about normal generation, you'd have to use specifically a mesh known to not have them 20:19 Krock you'd have to test both to ensure nothing breks 20:20 hecks do you want an .x with normals? 20:20 Krock would be amazing if you could share such a model 20:21 PGimeno does anti-cheat roll back player rotation as well as position? 20:21 sfan5 there is an example model inthe issue 20:22 sfan5 PGimeno: it might reset it but not roll it back to previous values 20:22 PGimeno hmm ok, never mind then 20:23 Krock only positions are affected 20:25 PGimeno hecks: maybe this helps understanding the matrix node? https://github.com/minetest/minetest/pull/8019#issuecomment-460072630 20:25 Krock merging.... 20:35 hecks I have anticheat off for this 20:35 hecks this is purely a client problem 20:43 hecks okay so it's the parent's position that's wrong 20:43 hecks and only if the parent is fresh 20:46 hecks Reusing the same object doesn't seem to break it 21:18 hecks goodness, there's layers to this bug 21:19 hecks I managed to get it to update its position but now the camera offset is missing during that frame 21:41 hecks nah, the camera offset is in fact applied *twice* 22:05 sfan5 merging game#2691 in 10m 22:05 ShadowBot https://github.com/minetest/minetest_game/issues/2691 -- creative: Update for compatibility with minetest.creative_is_enabled by sfan5