Minetest logo

IRC log for #minetest-dev, 2020-06-05

| Channels | #minetest-dev index | Today | | Google Search | Plaintext

All times shown according to UTC.

Time Nick Message
00:16 kollaps[m] joined #minetest-dev
00:38 proller joined #minetest-dev
00:56 oil_boi joined #minetest-dev
01:08 rubenwardy merging #9900
01:08 ShadowBot https://github.com/minetest/minetest/issues/9900 -- Show Toast in UI thread and fix unpacking on Android 10 by ubulem
01:08 rubenwardy ...in 10
01:11 ANAND joined #minetest-dev
01:13 Miner_48er joined #minetest-dev
01:46 Foz joined #minetest-dev
02:49 gorbachev_pizza_ joined #minetest-dev
03:16 gorbachev_pizza_ joined #minetest-dev
03:55 rubenwardy Dependency resolution in ContentDB: https://rwdy.uk/bkqJy.png
03:56 rubenwardy The algorithm is super super painful
03:57 rubenwardy source: https://github.com/rubenwardy/minetest/tree/cdb_deps
03:58 rubenwardy essentially, when installing a mod an API request is made to ContentDB to find out the dependencies
03:58 rubenwardy for each hard dependency, it first looks in the game mods, the installed mods, and finally available packages
03:58 rubenwardy the available package chosen will have a preference on exact mod match and highest rating
03:58 rubenwardy then it recurses for each chosen mod
03:59 rubenwardy resulting in an API request for each chosen mod
03:59 rubenwardy in most cases, there's no more than a few requests
04:00 rubenwardy perhaps the client should download all the information from ContentDB to begin with, not sure how big the download would be
04:45 Warr1024 your best option in the long run would be to do one request for each depth layer.  Get the direct dependencies, and then do one request to fetch the data for all of them. Then get their dependencies and do one request for all of them.  You might be able to fit an array of package names in a URL query, or it may make more sense to do a POST...
05:16 rubenwardy The server side is quite fast, so round trip time is more important - meaning that batching like you suggest might be a good approach
05:33 ANAND joined #minetest-dev
05:45 calcul0n joined #minetest-dev
06:09 pmp-p joined #minetest-dev
06:14 benrob0329 joined #minetest-dev
06:45 NetherEran joined #minetest-dev
07:16 Icedream joined #minetest-dev
07:36 erlehmann joined #minetest-dev
08:00 ShadowNinja joined #minetest-dev
08:03 oil_boi joined #minetest-dev
08:31 Beton joined #minetest-dev
08:40 lisac joined #minetest-dev
10:30 proller joined #minetest-dev
10:31 Fixer joined #minetest-dev
10:33 Darcidride joined #minetest-dev
10:40 ANAND #7587 is a-waiting ;)
10:40 ShadowBot https://github.com/minetest/minetest/issues/7587 -- Implement on_wielditem_change callback registration by ClobberXD
11:55 NetherEran joined #minetest-dev
12:18 sfan5 merging game#2675 in 10 minutes
12:18 ShadowBot https://github.com/minetest/minetest_game/issues/2675 -- Implement searching for translated names in creative inventory by sfan5
12:28 appguru joined #minetest-dev
13:18 ANAND sfan5: Thanks for the timely advice regarding #7924. I've removed all changes to guiKeyChangeMenu.* and rebased.
13:18 ShadowBot https://github.com/minetest/minetest/issues/7924 -- [NO SQUASH] Allow binding dig, place actions to keys; remove LMB/RMB hardcoding by ClobberXD
13:18 ANAND There's a little more to do, and the PR will be ready for review. :)
13:19 ANAND Wow, that was quick :O
13:21 rubenwardy thanks for the rebase
13:27 rubenwardy ANAND:
13:27 rubenwardy /home/ruben/dev/minetest/src/script/lua_api/l_localplayer.cpp:234:15: error: ‘const struct PlayerControl’ has no member named ‘LMB’
13:32 Taoki joined #minetest-dev
13:47 ANAND On it
13:47 ANAND Already on it* :)
14:03 Warr1024 rubenwardy: actually, instead of batching, you could just make the server do the entire recursion based on a single package name in a single request.  You'll have to guard against pathological cases, I suppose, like circular deps or specially-constructed chains.  Some of the dependency recursion info could also be precalculated on the server-side
14:03 Warr1024 if you find it gets bogged down...
14:08 rubenwardy that would require the client sending their installed mods to the server
14:08 rubenwardy it could assume that they have no installed mods and games, and resolve based on that
14:08 rubenwardy then the client would skip and deps that are installed
14:09 sfan5 wouldn't that mean it has to resolve every possible alternative?
14:10 rubenwardy only the most likely one, I guess
14:10 rubenwardy there's another issue here - what if the candidate for a dependency doesn't support the base game, but an alternative does?
14:11 rubenwardy currently, it chooses candidates from ContentDB without knowing the candidate's dependencies
14:14 ANAND I notice a new file test_bm.txt at the root of my repo, containing <IP>|test_username
14:14 ANAND I assume this is created when running unittests
14:14 ANAND Why isn't this removed afterward? o_O
14:18 sfan5 someone forgot to write the code
14:19 ANAND PR incoming
14:19 ANAND 1 line change, but I'll test nonetheless
14:22 Warr1024 Oh, I see what you mean, ruben: you don't want the server to waste time searching for mods to resolve dependencies the client already has ... well, maybe it wouldn't be so bad to let the server just do the work and throw it out?  It could probably be done efficiently enough and shouldn't affect the results too badly.
14:23 Warr1024 I assume that all things being equal, it's probably better to have the server send more data in a response than have the client send more in a request, since (1) a lot of people have asymmetrical internet connections, and (2) the HTTP protocol is sort of designed to make larger responses less awkward tha larger requests anyway.
14:23 ANAND sfan5: #9994
14:23 ShadowBot https://github.com/minetest/minetest/issues/9994 -- [Trivial] TestBan: Cleanup properly after completing test by ClobberXD
14:23 ANAND Tested; works
14:24 rubenwardy also, .gitignore too
14:26 sfan5 ANAND: can I request a "// delete leftover files" comment to be added?
14:30 ANAND Ok, and ok
14:31 ANAND Shall I add a test_* glob instead of adding individual files to .gitignore?
14:32 sfan5 nah
14:32 ANAND Hmm, fine
14:33 ANAND The file has already been added to gitignore by Unarelith in #7902
14:33 ShadowBot https://github.com/minetest/minetest/issues/7902 -- Client-specific files moved to 'src/client' by Unarelith
14:34 ANAND That explains why it was already ignored when I was working on other stuff
14:35 ANAND Done
14:35 ANAND (sfan5 ^)
14:37 ANAND In the future, we could maybe override file read/write functions to work in a tests/ directory when running unittests
14:37 ANAND The directory will be created and destroyed by TestManager
14:40 rubenwardy #9995
14:40 ShadowBot https://github.com/minetest/minetest/issues/9995 -- ContentDB: Add Update All and download queuing by rubenwardy
14:40 ANAND As a side-benefit, there'll be only one line to ignore all of the files created by unittests (tbh, even this isn't needed as the files would be deleted, but I think it can remain, just in case)
14:40 ANAND rubenwardy: Awesome!
14:43 Darcidride joined #minetest-dev
14:59 rubenwardy actually, I suppose I could have it populate the cache Warr1024
14:59 rubenwardy if I make it so the response is a dictionary of package to dependencies, it could intelligently return packages that may need to be resolved
15:00 rubenwardy like, it gives you the dependencies for the package you request, but also the most likely candidate for each dep
15:00 rubenwardy I don't even need to do this no
15:00 rubenwardy *now
15:04 Warr1024 You could have the MT client download and cache the entire dependency map, but I'd have some concerns about scalability.  I've had some success encouraging people to release more new content for MT, and that may make your job harder over time...
15:05 rubenwardy that's not what I was suggesting
15:14 ANAND https://github.com/minetest/minetest/pull/7924#issuecomment-639555952
15:14 ANAND I'm so happy :')
15:15 ANAND Please try to review this PR as soon as possible - rebasing always has the danger of breaking this feature :)
15:36 rubenwardy #9997
15:36 ShadowBot https://github.com/minetest/minetest/issues/9997 -- Add dependency resolution to ContentDB by rubenwardy
15:56 ANAND Is it a good idea to automatically update minetest.conf.example as a post-merge step instead of updating just before a release?
15:58 sfan5 that'd cause lots of unnecessary changes in git
15:58 sfan5 ..but well I guess that's the case even now when most people also edit the .example config in their PRs
15:59 ANAND That's wasted work, and I didn't even know these were auto-updated until paramat told me.
15:59 ANAND Also, these files would be updated only if the merged PR modified settingtypes.txt
15:59 ANAND Should've mentioned that earlier
16:01 ANAND Could #7587 and #7924 be added to 5.3.0 milestone? I'm assuming the release would be around September(?) so we have time.
16:01 ShadowBot https://github.com/minetest/minetest/issues/7587 -- Implement on_wielditem_change callback registration by ClobberXD
16:01 ShadowBot https://github.com/minetest/minetest/issues/7924 -- [NO SQUASH] Allow binding dig, place actions to keys; remove LMB/RMB hardcoding by ClobberXD
16:02 ANAND (I've removed [no squash] from PR title :P)
16:02 ANAND I'd also love to see #9923 added to the milestone
16:02 ShadowBot https://github.com/minetest/minetest/issues/9923 -- Fix inventory swapping not calling all callbacks by appgurueu
16:03 ANAND It's an important bugfix that allows mods to get rid of hacky workarounds to prevent swapping
16:08 ANAND Once 7924 has been merged, I'll start working on #9781
16:08 ShadowBot https://github.com/minetest/minetest/issues/9781 -- Pick node by middle click
16:08 rubenwardy nice
16:11 ANAND sfan actually advised working on 9781 first, but I'm afraid that'll push 7924 back into the rebase soup.
16:12 ANAND Thought I'd clarify as to why I'm putting off 9781 for later
16:17 sfan5 this comment appears to be unaddressed > https://github.com/minetest/minetest/pull/7924#discussion_r409953227
16:24 ANAND I remember doing something about that. I'll check...
16:27 ANAND I did add a method, but I forgot to add a follow up comment to that discussion
16:29 ANAND sfan5: https://github.com/minetest/minetest/pull/7924/files#diff-b5b39c3a29359759ac6f0293dc37a453R149
16:34 ANAND Oh wait, I haven't even added a single line of documentation yet, lol
16:35 ANAND settingtypes.txt doesn't counts ;)
16:35 ANAND -s
16:36 ANAND Correct me if I'm wrong, but as far as I can tell, only the new `dig` and `place` player controls need to be documented.
16:42 sfan5 ANAND: but shouldn't that line be using the isKeyReleased method?
16:49 Lunatrius joined #minetest-dev
17:02 ANAND joined #minetest-dev
17:07 ANAND Added documentation: https://github.com/minetest/minetest/compare/5261352892c12eeb1c1ea7c38ca7ed72d3bd6d43..b59809bbb778a793fd2cec9fba598b0ebca69ee7
17:11 ANAND #7924 is ready for review :)
17:11 ShadowBot https://github.com/minetest/minetest/issues/7924 -- Allow binding dig, place actions to keys; remove LMB/RMB hardcoding by ClobberXD
17:26 Lone_Wolf joined #minetest-dev
17:36 lisac joined #minetest-dev
18:03 sfan5 merging #9994
18:03 ShadowBot https://github.com/minetest/minetest/issues/9994 -- [Trivial] TestBan: Cleanup properly after completing test by ClobberXD
18:05 fluxflux joined #minetest-dev
18:07 behalebabo_ joined #minetest-dev
18:11 NetherEran joined #minetest-dev
18:13 appguru joined #minetest-dev
18:23 nerzhul joined #minetest-dev
18:24 nerzhul hi
18:24 nerzhul sfan5, just to keep you informed about my devs around wayland, it seems... interesting
18:24 nerzhul wayland use EGL
18:24 sfan5 x11 could use that too, but it doesn't
18:24 nerzhul no
18:24 nerzhul it doesn't use it, it use framebuffer and map GL directly on it if i correctly understand
18:24 nerzhul (irrlicht)
18:25 nerzhul and here is the opengl version returned after initializing egl layer on wayland
18:25 nerzhul OpenGL ES 3.2 Mesa 20.0.7
18:25 nerzhul hello opengl es
18:25 nerzhul it seems as irrlicht 1.8 has no opengl es support it will be quite painful
18:25 sfan5 uh
18:26 sfan5 https://github.com/minetest/irrlicht/pull/6/files#diff-66f340be52015b573518cca73c1b26aaR315 just change this line?
18:26 nerzhul i will try to cheat saying i'm a opengl 3.0 to begin to see what happened :D
18:26 nerzhul which one ?
18:26 nerzhul you didn't linked me a link :D
18:26 nerzhul a line*
18:26 sfan5 CIrrDeviceWayland.cpp line 315
18:27 nerzhul oh yeah right :D
18:28 nerzhul because i tried to say hey i'm open gl 3.0 and irrlicht is not very happy :p
18:28 sfan5 you might have to call eglBindAPI( EGL_OPENGL_API ); too
18:30 nerzhul okay now irrlicht doesn't complain about open gl
18:31 nerzhul or not :D
18:31 nerzhul anyway i will try to dig move
18:31 nerzhul anyway i will try to dig more
18:31 nerzhul but the detection is obviously a fail if egl return a big string
18:34 nerzhul ah interesting, i have no more error since i set bindapi
18:35 nerzhul i just need to fix some interesting X11 bindings on the opengl layer it seems :D
18:35 nerzhul 2020-06-05 20:34:26: WARNING[Main]: Client: Could not find X11 Display in ExposedVideoData
18:35 nerzhul 2020-06-05 20:34:26: WARNING[Main]: Could not find x11 display for setting its icon.
18:35 nerzhul those are the remaining warnings,
18:36 nerzhul the new version proposed by glGetString is "4.6 (Compatibility Profile) Mesa 20.0.7
18:36 nerzhul "
18:36 nerzhul far more better
18:50 nerzhul arf we have X11 call inside mintest
18:50 nerzhul but now my surface seems to be used or something :D
19:31 Krock will merge #9721 in 10 minutes
19:31 ShadowBot https://github.com/minetest/minetest/issues/9721 -- Attempt to fix HUD scaling by MoNTE48
19:40 Krock mergign
19:41 rubenwardy and... my fan has died
19:41 rubenwardy guess that means no more compiling
19:41 Lone_Wolf Restart didn't fix? o.o
19:41 Krock that's why you should have an all-noctua-build
19:42 rubenwardy I "fixed" it the other day by cleaning everything, but that's not doing it now
19:53 sfan5 thoughts on https://github.com/minetest/minetest/pull/9983#issuecomment-637546888 ?
20:00 Krock Reasonable fix
20:01 Krock but I wonder whether this breaks nodes were you can barely jump onto
20:02 Krock IIRC the step up feature helped to jump onto higher walls
20:10 Lone_Wolf joined #minetest-dev
20:11 nerzhul joined #minetest-dev
20:11 nerzhul back, wow nice
20:11 nerzhul i have flickering MT rendering on wayland :D
20:20 nerzhul if you want to test, i updated the PR
20:42 nerzhul flickering fixes haha, i had bad old code
20:43 nerzhul now i need to handle keys events
20:43 nerzhul and we will need client side decorations ourselves, haha
20:47 sfan5 gnome does not support server side decorations so unless you pull in gtk/qt/whatever, user experience will always suck
20:51 rubenwardy as requested: #9998
20:51 ShadowBot https://github.com/minetest/minetest/issues/9998 -- Add HTTP API to main menu by rubenwardy
20:52 sfan5 uh oh wonder who will open 10000
20:58 Krock please make it an issue, showing a timeline of all issues so far
20:58 rubenwardy issue: too many issues
21:00 Krock the urge to shitpost is so huge
21:00 nerzhul it depends if i can get the xdg decorations or not, we will see
21:01 sfan5 yes, but not on gnome, it doesn't support that protocol
21:01 nerzhul yep
21:01 nerzhul anyway it's not so important we do a game not a desktop app
21:01 nerzhul it can be avoided :p
21:04 nerzhul you wrote wayland apps sfan5 ?
21:08 sfan5 no but I know of another project's troubles with gnome wayland
21:08 nerzhul haha
21:15 gorbachev_pizza_ joined #minetest-dev
21:18 behalebabo joined #minetest-dev
21:41 nerzhul okay i have key press events, it's the first step :D
21:41 nerzhul next step tomorrow
21:41 nerzhul and the keymap is fine on wayland <3
22:07 lisac joined #minetest-dev
22:38 kaeza joined #minetest-dev
22:42 appguru !tell sfan5 I have fixed the invswap fix Pr
22:42 ShadowBot appguru: O.K.
23:24 gorbachev_pizza_ joined #minetest-dev

| Channels | #minetest-dev index | Today | | Google Search | Plaintext