Time Nick Message 00:20 repetitivestrain sfan5: hmm, but any idea why it scales so poorly? 08:10 sfan5 you mean why more threads cause so much more memory leaks? I don't know 08:12 repetitivestrain sfan5: yeah, pity 08:12 repetitivestrain Is it expected for the same block to be generated more than once? 08:12 repetitivestrain MapBlock, i mean 08:13 repetitivestrain as with multiple emerge threads I'm observing invocations of on_generated for mapblocks that have already been generated 08:21 sfan5 how are you determining that 08:24 repetitivestrain a flag assigned to every mapblock between minp and maxp is set on the first invocation of on_generated 08:25 repetitivestrain and if a subsequent call detects the flag, a warning is logged 08:25 repetitivestrain https://codeberg.org/mineclonia/mineclonia/src/commit/mineclonia_mapgen/mods/MAPGEN/mcl_levelgen/post_processing.lua#L327 08:25 repetitivestrain post_process_mapchunk is the on_generated callback 08:35 sfan5 can't reproduce here https://paste.debian.net/1377024/ 08:35 sfan5 using /deleteblocks causes this code to log, as expected 08:37 repetitivestrain there are three factors your code probably doesn't reproduce: this lua map generator is far slower than v7, i have six emerge threads enabled, and you have to fly about in an erratic pattern before it triggers 08:42 sfan5 https://paste.debian.net/1377026/ still no 08:44 repetitivestrain i pasted your code into my mapgen and after flying a hundred nodes from -20 to 20 for 30 seconds i received: 08:45 repetitivestrain https://paste.debian.net/1377027/ 08:45 repetitivestrain it's not special otherwise than in the sense of being a lua map generator 08:45 repetitivestrain i.e., singlenode 08:46 sfan5 sounds like its working then 08:46 sfan5 I am testing this in a standard mgv7 MTG world 08:47 repetitivestrain Singlenode is supposed to generate mapblocks more than once? 08:47 sfan5 no 08:47 sfan5 "it" = the test code 08:47 repetitivestrain Ah 08:47 sfan5 in any case if you can isolate the problem into reproduction steps that aren't "install mineclonia" please report a bug 08:48 repetitivestrain One moment 11:20 [MatrxMT] huh, how come we have two nether mods (+mods for those), but no aether? 11:21 [MatrxMT] okay, it just didn't make it to CDB 11:21 [MatrxMT] https://forum.luanti.org/viewtopic.php?t=22055 11:29 MinetestBot 02[git] 04grorp -> 03luanti-org/luanti: Minor lua_api.md improvements (#16169) 13986cd32 https://github.com/luanti-org/luanti/commit/986cd32f289b4d315dcb645180db8abfc0d2ae86 (152025-05-28T11:29:03Z) 11:29 MinetestBot 02[git] 04sfan5 -> 03luanti-org/luanti: Move one CI run to be on 64-bit ARM 13ae35f37 https://github.com/luanti-org/luanti/commit/ae35f37bc30ccddb0b9ae69341e5d7b4b876099a (152025-05-28T11:29:20Z) 11:31 MinetestBot 02[git] 04sfan5 -> 03luanti-org/luanti: Do not allow vector components to be nil 13a5263dc https://github.com/luanti-org/luanti/commit/a5263dc7ed4108b7ea57c08f82555bc7b2b4053c (152025-05-28T11:29:30Z) 11:31 MinetestBot 02[git] 04sfan5 -> 03luanti-org/luanti: Add unit tests for Lua vector reading 13ec16fb3 https://github.com/luanti-org/luanti/commit/ec16fb33d0fb224220d323120986711b99d2e19e (152025-05-28T11:29:30Z) 11:31 MinetestBot 02[git] 04sfan5 -> 03luanti-org/luanti: Reject NaN and Inf in check_v3d() too 136ca9d75 https://github.com/luanti-org/luanti/commit/6ca9d75f0bacfe71a232be4f590cc4f5d15ed068 (152025-05-28T11:29:30Z) 11:31 MinetestBot 02[git] 04sfan5 -> 03luanti-org/luanti: Change nil-component error to deprecation warning 135c25993 https://github.com/luanti-org/luanti/commit/5c2599315cb6295196a5dc8f96debc930f28b8ad (152025-05-28T11:29:30Z) 12:45 repetitivestrain sfan5: the culprit appears to be looking up node metadata in on_generated 12:47 repetitivestrain if a generated mapblock is somehow unloaded when metadata is requested, the same mapchunk is emerged again 12:51 repetitivestrain https://paste.centos.org/view/f94d863c 12:51 repetitivestrain here's a log generated by "minetestserver --info" 12:52 repetitivestrain where it can be observed that after the chunk at -52,8,-307 is generated, Map::getNodeMetadata prompts it to be generated again, and this second emerge triggers my assertions upon completion 13:48 sfan5 sounds like a bug 13:51 repetitivestrain https://paste.centos.org/view/e38732ae 13:51 repetitivestrain yeah and it appears much worse 13:52 repetitivestrain observe how the chunk between (48,-12,-312) and (52,-8,-308) is generated twice prompted solely by ordinary mapblock emerging 13:53 repetitivestrain by the same emerge thread too 14:10 repetitivestrain if it helps--this always follows a failure by the loop at the conclusion of ServerMap::finishBlockMake to retrieve a generated MapBlock 14:11 repetitivestrain e.g. if I apply this patch: 14:11 repetitivestrain https://paste.debian.net/1377072/ 14:11 repetitivestrain https://paste.debian.net/1377073/ 14:12 repetitivestrain then such messages always immediately precede the problematic redundant generations 14:13 repetitivestrain i suppose one or more mapblocks in the chunk is never generated with the result that the next attempt to load such an ungenerated mapblock winds up regenerating the entire mapchunk once more 14:19 repetitivestrain i can locate no code preventing mapblocks in generation from being unloaded 14:19 sfan5 I don't plan to investigate this without a reproducer 14:26 repetitivestrain sfan5: which i cannot provide short of installing mineclonia, but as i can reproduce the issue quite consistently, i can attest that this patch https://paste.debian.net/1377076/ eliminates it completely 14:26 repetitivestrain whether this change is TRT is another question altogether 14:26 repetitivestrain but i'm not too familiar with minetest/luanti's internals 14:50 repetitivestrain https://codeberg.org/halon/Minetest/commit/2dd6fab54ab06edef2663ad7e594e5287d3c0fb4 14:50 repetitivestrain i'm going to apply this locally for the present 16:31 [MatrxMT] gltf was worth it, in fact aloan says it was #1 :D https://forum.luanti.org/viewtopic.php?p=445001#p445001 16:31 ShadowBot https://github.com/luanti-org/luanti/issues/1 -- GlowStone code by anonymousAwesome 16:53 MinetestBot 02[git] 04ZenonSeth -> 03luanti-org/minetest_game: Workaround for sign regression with 5.12.0 13a6bf9dd https://github.com/luanti-org/minetest_game/commit/a6bf9dd526e51b25f1ca08d6d214c5e8442b503b (152025-05-28T16:53:44Z)