Minetest logo

IRC log for #minetest-dev, 2022-02-21

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

All times shown according to UTC.

Time Nick Message
00:09 Alias2 joined #minetest-dev
00:22 erlehmann celeron55 why would it be slow to have a rendered image? i mean, you only have to render it *once*. prerendering all nodes sounds dumb though.
00:23 erlehmann unless it is done in development.
00:23 erlehmann but i don't get why an inventory scene should have to render more than one frame
00:23 erlehmann it's not like there are any lighting changes for the nodes in inventory
00:26 erlehmann the most cursed thing i have seen in that respect was imgui, some ghetto UI library where the author was proud of rerendering everything in every frame instead of, uh, tracking dirty rectangles or so.
00:27 erlehmann regarding the inventory, so far i can place all of those nodes in the world and get barely any slowdown, but no idea really what it is. i'll try to bisect it i guess. with all the pain that entails.
00:38 erlehmann has sprinting been removed from minetest?
00:39 erlehmann btw who decided i can't go left or right while autoforwarding? it's really bad :(
00:40 rubenwardy erlehmann: you then need to keep the image in memory
00:40 rubenwardy GPUs work with meshes
00:40 rubenwardy To render an image, they need to render a mesh with 4 vertices with a texture mapped onto it
00:41 erlehmann so?
00:41 erlehmann i mean it's certainly not too wasteful for the texture size of one inventory page
00:42 rubenwardy Minetest originally rendered to texture and then to screen
00:43 rubenwardy RBA changed it to just render the nodes, as he claimed it was faster
00:43 rubenwardy In practice, 🤷
00:44 erlehmann can you point me to a commit or PR?
00:44 erlehmann because if that's the culprit, it would be ridiculous (i can see hundreds of nodes, and have 30fps, but an inventory page …)
00:50 erlehmann btw, the x2048/alpha_blend_indexed branch has not lead to much slowdown so far. 20 to 30 fps in the woods. computer gets really hot though, no idea why.
00:50 erlehmann oh and sprint works, my mistake.
00:50 erlehmann but left/right was really disabled :(
00:51 erlehmann (during autoforward)
00:51 erlehmann makes boats kinda bad
00:57 MTDiscord <Jonathon> https://github.com/minetest/minetest/issues/12048 already known
00:59 erlehmann oh, i thought it was *on purpose*
00:59 erlehmann sorry
00:59 erlehmann thx
01:59 v-rob joined #minetest-dev
03:27 v-rob joined #minetest-dev
05:00 MTDiscord joined #minetest-dev
06:55 v-rob joined #minetest-dev
07:12 calcul0n joined #minetest-dev
08:10 sfan5 <erlehmann> twrightsman[m] given the unwillingness of mt devs to even upstream patches to irrlicht [...]
08:10 sfan5 can you please stop pulling things out of your ass
08:12 sfan5 Irrlicht has been unmoving since 5 or more years, various users have tried to upstream fixes with mixed success
08:16 sfan5 besides: making the irrlicht devs aware of some of our fixes that are upstreamable is on my TODO list
10:44 appguru joined #minetest-dev
10:59 erlehmann sfan5 “unmoving” as in no new technical development? cutealien commits stuff from time to time, but releases seem super rare …
11:00 erlehmann sfan5 i'll retract my assertions and want to apologize
11:00 erlehmann sfan5 i had just looked at recent irrlicht history and that you had fixed some stuff that then never appeared in irrlicht proper
11:22 calcul0n joined #minetest-dev
11:23 proller joined #minetest-dev
11:31 appguru joined #minetest-dev
11:33 erlehmann twrightsman[m] regarding the upstream comment by sfan5, take note pls
11:37 appguru1 joined #minetest-dev
11:56 MTDiscord <Benrob0329> aren't we going to be absorbing Irrlicht at some point?
11:58 proller why not urho3d or something alive?
12:00 MTDiscord <Sublayer plank> urho3d looks to be way too much for minetest
12:03 rubenwardy Hugues Ross: it's mildly confusing that you use "give it a pass" to mean give it a pass / look over, when it can be interpretted as pass on it
12:13 HuguesRoss joined #minetest-dev
12:15 erlehmann Benrob0329 fixes to irrlicht can and should still be upstreamed, regardless of roadmap.
12:15 HuguesRoss joined #minetest-dev
12:52 olliy joined #minetest-dev
13:10 sfan5 well as soon as Irrlicht is absorbed many fixes stop being directly applicable
13:12 erlehmann well, if you give me a list of commits that are, i can offer to travel to the wasteland that is sourceforge and leave the package on the ground in front of the appropriate mudhole
13:13 erlehmann i mean existing ones, from the past. i wish i knew a better solution for the image loader fixes btw.
13:13 erlehmann (where you limited the dimension)
13:14 erlehmann btw, any idea why only the 0.4 gamma PNG texture seems so messed up?
13:14 erlehmann gimp also has gamma bugs (or so it seems), but there every gradient looks different
13:14 erlehmann with the lower values having darker gradients overall
13:15 erlehmann (which is consistent with not just doing gamma i guess)
13:32 appguru joined #minetest-dev
14:08 MTDiscord <luatic> erlehmann: can you upgrade libpng? I sometimes get PNG warnings which seem to stem from using an outdated libpng version
14:14 erlehmann luatic what kind of warnings are these?
14:16 erlehmann luatic if it is “iCCP: known incorrect sRGB profile” then the thing is that *newer* libpng is stricter about checking this
14:16 erlehmann and you need to fix the texture
14:16 erlehmann and by “newer” i mean, like, this warning goes years back
14:57 ronoaldo joined #minetest-dev
14:57 ronoaldo Hi! I have a crash at my server that is hard to debug... something happened with either the map/mod/modstorage after a recent update of my mods.
14:57 ronoaldo Server is crashing and producing a core dump file, which after a gdb bt shows this info: https://paste.debian.net/1231735/
14:59 ronoaldo Logs on verbose don't show exactly what is causing the crash, and don't show full warnings either. Logs on action shows these warnings: https://paste.debian.net/1231737/ (when I join, with the account ronoaldo,)
15:00 ronoaldo it is happening only when I'm at a specific position... if I teleport to spawn and join the server, no crashes. If any player goes to a specific position, after a few seconds it crashes (perhaps when loading a problematic block from database?)
15:00 ronoaldo any tips on debugging the issue to understand how to fix it would be apreciated.
15:01 ronoaldo Error happens on both 5.4.1 and 5.5.0
15:05 tekakutli joined #minetest-dev
15:06 proller joined #minetest-dev
15:23 Fixer joined #minetest-dev
15:55 v-rob joined #minetest-dev
16:10 v-rob joined #minetest-dev
16:36 sfan5 what's the abort message? neither the gdb backtrace nor the log files show it
16:37 erlehmann sometimes minetest just gives up though, so maybe there is none?
16:37 sfan5 it gets printed to stderr on crash so it might get lost if you don't have logging set up for that
16:38 ronoaldo sfan5: it looks like an issue with Draconis mod update from 1.1 to 1.2; reverting that one fixed the crash. I'm probably not having the stderr output logged then! It does not go to debug.txt right?
16:39 sfan5 correct
16:39 ronoaldo perfect, this is useful to know, going to setup for the server! thank you!
16:40 sfan5 if it's just for testing you can also start the server in a screen/tmux
16:42 ronoaldo good point!
17:11 v-rob joined #minetest-dev
18:07 tekakutli joined #minetest-dev
18:23 appguru joined #minetest-dev
18:38 MTDiscord <luatic> @Based and Tax Fraud-pilled draconis has a bug
18:39 rubenwardy ...wrong channel
18:47 MTDiscord <Jonathon> technically correct as it refers to above chat
18:47 MTDiscord <Jonathon> *correct channel
18:56 proller joined #minetest-dev
19:10 proller joined #minetest-dev
19:24 v-rob joined #minetest-dev
19:28 v-rob joined #minetest-dev
19:31 ROllerozxa joined #minetest-dev
20:16 v-rob joined #minetest-dev
20:44 olliy1or joined #minetest-dev
21:13 tekakutli joined #minetest-dev
21:31 v-rob joined #minetest-dev
22:13 proller joined #minetest-dev
22:28 MTDiscord <MisterE> I thought he was muted semipermanently
22:28 MTDiscord <MisterE> So wouldn't be able to respond
22:28 rubenwardy run-vcpkg is such a dumb plugin
22:29 rubenwardy relying on these plugins for CI is a lock-in
22:29 rubenwardy on GitLab, you can freely specify cache directories. So you'd make vcpkg a cache dir, and then call vcpkg directly
22:31 proller joined #minetest-dev
22:32 proller joined #minetest-dev
22:32 MTDiscord <MisterE> Hey, devs, is there any plans for a ver. 3 of dynamic media that allows you to send temporary media that does not fill up the cache because it can be forgotten? So you could do things like live television (showing the viewpoint of a server-controlled playerbot or smth)
22:32 MTDiscord <MisterE> That might be a bad ex
22:32 MTDiscord <MisterE> Anyhow media that you want to send often and then forget it.
22:33 sfan5 rubenwardy: you can do that with github actions (only builtin tools) too
22:33 MTDiscord <MisterE> Live livesteam music maybe
22:33 sfan5 that's also a bad example
22:33 rubenwardy does it support cache dirs?
22:34 sfan5 a list of dirs that you tell it to cache and it automatically restores?
22:34 rubenwardy there's actions/cache
22:35 sfan5 yes
22:35 sfan5 but I still don't know what "cache dirs" means to you
22:35 rubenwardy "This action allows caching dependencies and build outputs to improve workflow execution time."
22:36 rubenwardy exactly this
22:38 rubenwardy would you object to just using the old version of run-vcpkg that works
22:39 sfan5 the new one doesn't? but I don't mind either way
22:39 erlehmann MisterE i do not know how often you want to update your live television, but given that you can lag out players with hud updates, i think you should *probably* figure out if you really want that.
22:39 rubenwardy in order to use vcpkg triplets, the new version requires changing cmake and running vcpkg from cmake
22:40 MTDiscord <MisterE> every 1/4 sec would be nice
22:40 erlehmann MisterE if you doubt my assertion about hud updates, go on a mineclone2 0.71 server and stand in a fire until your connection is saturated.
22:41 sfan5 ?!?!?!?!?!?!
22:41 sfan5 who comes up with that
22:41 erlehmann the implementation is cursed in exactly the way live television would. we fixed it in mineclonia.
22:41 rubenwardy replying to me or MisterE/erleh
22:41 sfan5 you
22:41 rubenwardy erlehmann, MisterE:  #minetest please
22:41 erlehmann ok!
22:42 rubenwardy yeah it's odd. I'm not a huge fan of putting even more stuff into cmake
22:42 rubenwardy oh, there's an env variable that might be able to be used instead
22:43 rubenwardy still, it seems to be ignoring the path I give to the vcpkg.json file
22:45 v-rob joined #minetest-dev
22:46 MTDiscord <luatic> @MisterE, Modder of Sortz no, there is currently no way to make clients forget sent files during the session; ephemeral media is however forgotten after the session (and not cached on disk) at least.
22:55 MTDiscord <MisterE> Yes, it would be very useful to be able to forget a media that was no longer needed. This would provide a few advantages: in my music mod, there can be so many tracks that it is unlikely the same one will be used twice. It would be nice to be able to forget tracks after they are played so the cache isn't filled. Also, a blocker to having different textures of the same name being sent to different players (to make an object invisible to
22:55 MTDiscord some players and not to others ) is that once you send a texture, that namespace is used up. A workaround is possible, but with enough changes of the texture you will fill up the cache. Better to be able to send a new texture of the same name and then forget the old one
22:58 erlehmann i think you are doing something incredibly cursed and i am not sure if i would try the same approach
23:00 erlehmann misterE if i send different textures with the same name to different people i can give an object a different appearance, is that correct?
23:02 sfan5 that is correct
23:19 tekakutli joined #minetest-dev
23:22 erle joined #minetest-dev
23:29 MTDiscord <MisterE> I just said can have that many tracks, it doesn't have to
23:30 MTDiscord <MisterE> ok, here is a good example. Procedurally generated music
23:30 MTDiscord <MisterE> no its not, that would have to be a csm
23:40 SOMBRIO joined #minetest-dev
23:46 erlehmann mister E for music just a way to give an ogg stream would be enough
23:47 erlehmann can be generated on server
23:47 erlehmann put in streaming file
23:47 erlehmann done
23:49 MTDiscord <MisterE> True
23:55 SOMBRIO joined #minetest-dev

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