| Time |
Nick |
Message |
| 00:45 |
|
tekakutli joined #minetest-dev |
| 00:56 |
|
AliasAlreadyTake joined #minetest-dev |
| 02:58 |
|
v-rob joined #minetest-dev |
| 03:28 |
|
queria^clone joined #minetest-dev |
| 03:33 |
|
queria^clone joined #minetest-dev |
| 04:00 |
|
MTDiscord joined #minetest-dev |
| 06:49 |
|
tekakutli joined #minetest-dev |
| 08:05 |
|
hendursa1 joined #minetest-dev |
| 09:19 |
|
specing_ joined #minetest-dev |
| 09:34 |
|
olliy joined #minetest-dev |
| 10:57 |
|
calcul0n joined #minetest-dev |
| 11:49 |
|
calcul0n_ joined #minetest-dev |
| 12:30 |
|
Fleckenstein joined #minetest-dev |
| 12:32 |
Fleckenstein |
is there an issue for the misc problems with shaders disabled yet? like sprite animations displaying black and some animated 3d models displaying white? it's really an annoyance, not sure whether this is already being worked on or whether i should try to fix it myself |
| 12:32 |
Fleckenstein |
i believe its a problem with irrlicht, or am i mistaken? |
| 13:12 |
sfan5 |
those issues are news to me |
| 13:25 |
MTDiscord |
<Sublayer plank> um, wasn't it that bug erlehmann was talking about in regards to mineclone's enchanting table book model having a glitched texture in 5.5.0-dev? |
| 13:38 |
Fleckenstein |
most likely yes |
| 13:38 |
Fleckenstein |
its an older bug |
| 13:38 |
Fleckenstein |
i _somehow_ resolved it for chests and shulkerboxes by exporting the model in blender in a very specific way but it was quite a pain and I don't know how exactly i did it anymore |
| 13:39 |
Fleckenstein |
the thing really is, if shaders are enabled, everything is fine |
| 13:55 |
Fleckenstein |
so, am i right assuming that there has not been done any work yet? is there anything i should know trying to fix it? |
| 13:58 |
MTDiscord |
<luatic> WDYM by "glitched texture"? |
| 13:59 |
MTDiscord |
<luatic> Might be related to normals |
| 14:04 |
|
appguru joined #minetest-dev |
| 14:07 |
|
olliy joined #minetest-dev |
| 14:27 |
|
proller joined #minetest-dev |
| 14:30 |
|
tech_exorcist joined #minetest-dev |
| 15:08 |
MTDiscord |
<josiah_wi> The README does not say how to run the unit tests. I haven't run them before; should I know anything about this? |
| 15:32 |
Fleckenstein |
its simply just white or black |
| 15:32 |
Fleckenstein |
entirely |
| 15:52 |
MTDiscord |
<josiah_wi> I can't figure out how to run the tests. Please advise. |
| 15:54 |
MTDiscord |
<luatic> Which tests? Lua or C++ ones? |
| 15:54 |
MTDiscord |
<josiah_wi> C++ |
| 15:54 |
Fleckenstein |
isnt it --run-unittests? |
| 15:55 |
MTDiscord |
<josiah_wi> Ah, thank you. |
| 15:56 |
MTDiscord |
<josiah_wi> Well, this is quite something. |
| 15:56 |
MTDiscord |
<josiah_wi> I caused an overflow error in the pseudorandom number generator. |
| 15:57 |
MTDiscord |
<josiah_wi> Should I even open an issue for this... |
| 16:00 |
Fleckenstein |
oh no, in 5.5-dev the bug is even there with shaders enabled |
| 16:00 |
MTDiscord |
<josiah_wi> I'll bet I have ubsan still compiled in. I'll just disable that. |
| 16:01 |
MTDiscord |
<savilli> better to open a PR XD |
| 16:01 |
sfan5 |
so did it not happen with shaders disabled? |
| 16:02 |
sfan5 |
https://github.com/minetest/minetest/issues/11146 might be related btw |
| 16:02 |
sfan5 |
though I haven't seen anyone who is able to reproduce it |
| 16:02 |
MTDiscord |
<josiah_wi> savilli, whover wrote this pretty clearly accepted it would have UB. Fixing this mess is something nobody wants to deal with. |
| 16:02 |
|
Extex joined #minetest-dev |
| 16:03 |
|
hendursaga joined #minetest-dev |
| 16:03 |
MTDiscord |
<josiah_wi> I looked at the code and it's blaringly obvious. |
| 16:03 |
MTDiscord |
<savilli> or they just forgot to cast to unsigned |
| 16:03 |
MTDiscord |
<Sublayer plank> erlehmann seems to be able to reproduce it, he was talking about it as late as 4 days ago |
| 16:03 |
MTDiscord |
<Sublayer plank> https://irc.minetest.net/minetest-dev/2021-10-31#i_5892518 |
| 16:04 |
sfan5 |
keeping track of bugs that only get discussed but not properly filed is kinda hard... |
| 16:09 |
Fleckenstein |
in 5.5-dev, its white if shaders are off, black if shaders are on |
| 16:09 |
Fleckenstein |
kinda funny ngl |
| 16:10 |
Fleckenstein |
tried to bisect it but git bisect decided to SEGFAULT lmao |
| 16:13 |
MTDiscord |
<josiah_wi> How many of the Minetest CI builds run unit tests? |
| 16:14 |
MTDiscord |
<Sublayer plank> looks to be all linux compilers and the macos build |
| 16:15 |
MTDiscord |
<josiah_wi> Thank you. |
| 16:22 |
|
fluxionary joined #minetest-dev |
| 16:23 |
|
v-rob joined #minetest-dev |
| 16:24 |
MTDiscord |
<josiah_wi> "test them all" he said. Guess what, exhaustively testing every possible noise2d input is incredibly slow. |
| 16:25 |
MTDiscord |
<josiah_wi> I estimate around 4096 minutes to complete, give or take. |
| 16:25 |
MTDiscord |
<josiah_wi> Wait no. |
| 16:25 |
MTDiscord |
<josiah_wi> 4096 * 4096 minutes |
| 16:26 |
|
Fixer joined #minetest-dev |
| 16:26 |
MTDiscord |
<josiah_wi> I need answers, erlehmann. |
| 16:53 |
|
Extex joined #minetest-dev |
| 17:51 |
|
v-rob joined #minetest-dev |
| 18:03 |
|
longerstaff13 joined #minetest-dev |
| 18:32 |
|
v-rob joined #minetest-dev |
| 18:54 |
|
erlehmann joined #minetest-dev |
| 18:55 |
erlehmann |
i just read a bit of the logs. Fleckenstein is right, the bug is there in 5.5-dev with or without shaders. without shaders, the textures are white. with shaders, textures are black. |
| 18:56 |
erlehmann |
i did not have any success bisecting the issue. |
| 19:00 |
erlehmann |
Fleckenstein btw, it is a good idea to put all build scripts for (binary) artifacts in the repository. i have no idea, for example, how colors.json was created or how to recreate the blender exports for anything. |
| 19:01 |
erlehmann |
oh, wrong channnel sorry |
| 19:27 |
|
olliy joined #minetest-dev |
| 19:43 |
|
Extex joined #minetest-dev |
| 20:04 |
|
x2048 joined #minetest-dev |
| 20:08 |
x2048 |
I'm fixing rendering of shadows on entities, and I see this behavior that Minetest/irrlicht loads some models with insane normals, while other models are fine. Does anyone know the reason? |
| 20:09 |
sfan5 |
normals are only recalculated if they seem wrong |
| 20:09 |
sfan5 |
also make sure you have the latest irrlichtmt |
| 20:10 |
x2048 |
good catch, I'll try with the latest head, thanks sfan5 |
| 20:11 |
x2048 |
I've tried forcing the recalculation, but (a) for some reason MT asks to produce 'smooth' normals which makes little sense, and (b) it does not seem to change a lot. digging deeper... |
| 20:14 |
x2048 |
side notice, it looks like MeshCache in irrlichtmt does not work, at least in debug build. Had to skip through 25 loads of stairs.obj |
| 20:14 |
sfan5 |
that's intentional |
| 20:15 |
x2048 |
is that because of texturepacks? |
| 20:16 |
sfan5 |
it's because the code needs to modify the instance of the model it has or something |
| 20:22 |
|
v-rob joined #minetest-dev |
| 20:26 |
x2048 |
Maybe we could tweak it to use the cache and clone the mesh on get, would save parsers some work. |
| 20:26 |
x2048 |
Latest irrlicht, force mesh recalcualtion: https://user-images.githubusercontent.com/4933697/140414550-14e6fe37-6f4a-4b43-9070-e6215d9a645c.png |
| 20:26 |
x2048 |
sorry, s/mesh/normals/ |
| 20:27 |
sfan5 |
I did sort of but the only thing it did is increase ram usage #11652 (unclear if it's faster) |
| 20:27 |
ShadowBot |
https://github.com/minetest/minetest/issues/11652 -- Preload meshes at startup by sfan5 |
| 20:29 |
sfan5 |
blue are supposed to be normals? |
| 20:30 |
x2048 |
Yes, added this to show them: m_animated_meshnode->setDebugDataVisible(scene::EDS_NORMALS | scene::EDS_MESH_WIRE_OVERLAY); |
| 20:31 |
sfan5 |
looks like total nonsense |
| 20:31 |
sfan5 |
but how would it get those wrong |
| 20:33 |
x2048 |
One thing I found is this potential bug in exporter for blender: https://github.com/joric/io_scene_b3d/blob/master/export_b3d.py#L1136 |
| 20:33 |
x2048 |
Notice the order of x, y, z |
| 20:33 |
sfan5 |
if you force recalculation that shouldn't matter |
| 20:33 |
|
proller joined #minetest-dev |
| 20:36 |
x2048 |
I agree, will debug the mesh manipulator now to find out what's going on |
| 20:36 |
x2048 |
I'm also puzzled why player's model is perfectly fine |
| 21:18 |
|
specing_ joined #minetest-dev |
| 21:21 |
|
v-rob joined #minetest-dev |
| 21:51 |
erlehmann |
x2048 any luck bisecting it? |
| 21:55 |
sfan5 |
if I had to guess it happens even on the very first revisions that use IrrlichtMt |
| 21:55 |
erlehmann |
why though |
| 21:57 |
x2048 |
Kind of... models that are affected reuse vertices in the vertex buffer. Mesh manipulator will not calculate the normals for sharp edges correctly, because the vertex can only store one normal vector. |
| 22:07 |
x2048 |
They are also animated, and I'm looking for how this affects the normals. |
| 22:25 |
sfan5 |
huh |
| 22:26 |
sfan5 |
erlehmann: simply going from irrlicht 1.8 to their indev version broke a few things, wouldn't be surprised if this is one |
| 22:48 |
|
Taoki joined #minetest-dev |
| 23:05 |
|
v-rob joined #minetest-dev |
| 23:34 |
|
v-rob joined #minetest-dev |
| 23:43 |
|
YuGiOhJCJ joined #minetest-dev |
