Time Nick Message 10:25 sfan5 merging #16524, #16522, #16513, #16512 in 15m 10:25 ShadowBot https://github.com/luanti-org/luanti/issues/16524 -- Mapgen: Correct border block criteria by SmallJoker 10:25 ShadowBot https://github.com/luanti-org/luanti/issues/16522 -- Update comments referring to obsolete TOCLIENT_INIT by fetsorn 10:25 ShadowBot https://github.com/luanti-org/luanti/issues/16513 -- Android: Update to SDK 35 by sfan5 10:25 ShadowBot https://github.com/luanti-org/luanti/issues/16512 -- Update builtin locale by Wuzzy2 11:17 [MatrxMT] ping sfan5 11:20 sfan5 right 14:20 sfan5 has anyone looked at these leaks? https://0x0.st/KBYz.txt 14:21 sfan5 I'm most concerned about the CAnimatedMeshSceneNode one 16:08 sfan5 also who is updating the credits? 17:27 Krock will push https://pastebin.com/raw/yWn7xhGK in 20 minutes. Context: https://irc.luanti.org/luanti/2025-09-28#i_6287417 17:29 Krock sfan5: I cannot open 0x0.st as of recently. PR_END_OF_FILE_ERROR. wget via GnuTLS also refuses to download it 17:30 sfan5 system up-to-date? 17:30 Krock yes. Debian 13 with only gimp pinned to version 2.x 17:31 MTDiscord I was confused about what you meant when you said Debian 13 was gimp pinned until I realized you were talking about the image editor. 17:32 MTDiscord ( ͡° ͜ʖ ͡°) 17:32 Krock Oh it's an actual English word. I did not even know. 17:32 MTDiscord Yes, with varying levels of kinkiness depending on how you use it. 17:33 Krock > * TLS connect error: error:0A0003E8:SSL routines::reason(1000) 17:33 rubenwardy in the UK, it's used as an insult for an idiot 17:33 sfan5 i was thinking it might have a very strict SSL config but according to https://www.ssllabs.com/ssltest/analyze.html?d=0x0.st&s=168.119.145.117&latest everything's fine 17:33 Krock (output according to curl -v) 17:48 MTDiscord I'm sorry, but I just couldn't get the idea out of my head. So I present to you Gimp Tux. 17:48 MTDiscord https://cdn.discordapp.com/attachments/747163566800633906/1422278891091005531/tux_hat_edit.webp?ex=68dc17f0&is=68dac670&hm=300fcd5b7dc6721c23a0f7dc87430e94eb5e2398c37fdbec61267ba6f3fc73c3& 17:52 Krock changed the default DNS and we're back to business 17:52 Krock nathan4220776: smooth criminal 17:52 Krock pushing the commit..... 17:54 Krock sfan5: I am pretty sure I never saw those leaks in valgrind. Proves that it's not that great. 17:55 sfan5 if your ISP's DNS blocks 0x0.st I'm sure that would interest the owner 17:55 Krock anyway. CAnimatedMeshSceneNode.cpp:557 is probably the cause for this leak. it does not delete the pointers. 17:55 luatic sfan5, Krock: try https://gist.github.com/appgurueu/37afa7b109ef5a719dd71713ac589fd8 17:55 Krock * :548 17:56 Krock luatic: please add a safety check to L548 (vector clear). It's quite risky to do that. 17:57 luatic ? 17:57 Krock oh wait. never mind. 17:57 luatic i looked into the memleak a moment ago, the diff is what should fix it 17:57 Krock It's important to drop them before clear, in any case. 17:58 luatic yes, which is what the diff does 17:58 Krock and CAnimatedMeshSceneNode::addJoints() is guaranteed to have either an empty vector, or one with dangling pointers to its disposal? 18:00 Krock ::removeChild does zero the pointer... which seems good 18:00 luatic it's probably best if i just employ irr_ptr 18:01 Krock yes, I believe that would be a more future-proof solution. 18:06 luatic updated https://gist.github.com/appgurueu/37afa7b109ef5a719dd71713ac589fd8 18:09 sfan5 let's put that in a PR rather than doing it ad-hoc 18:14 luatic #16530 18:14 ShadowBot https://github.com/luanti-org/luanti/issues/16530 -- Fix a memory leak: Use irr_ptr for bones by appgurueu 18:16 Krock looks good. Are the leaks gone? 20:01 luatic Krock: I think so. 20:01 luatic ./bin/luanti 2>| grep AnimatedMeshScene yields no results on a build with asan (which includes leaksan). 20:06 luatic ./bin/luanti 2>&1 | grep CAnimatedMeshSceneNode* 20:06 luatic it yields results without the commit 20:07 luatic (there are some more unrelated leaks)