Time Nick Message 12:37 definitelya https://github.com/NVIDIA/open-gpu-kernel-modules 12:37 definitelya Open-source Nvidia drivers go brrrr. 12:38 definitelya Idk if that's the right repo, I'm too excited ahah. 12:43 MTDiscord zamn 12:44 MTDiscord it was a hell of a thing to wake up to, news about the nvidia drivers going open source 12:45 definitelya It might have something to do with the recent massive hack that released most of the source code of their drivers. 13:01 Oblomov everybody is getting way too pumped about this 13:02 Oblomov with mostof the functionality moved to venor-locked, signed firmware and still propietary userspace tools it's more of a mockery than a really meaningful piece of news 13:06 definitelya Oblomov: One hack at a time, eheh. 13:24 MTDiscord definitelya: the lapsus hack? ehh I'm not confident that was exactly the case, nvidia has probably been planning this for a while 13:25 definitelya I see, yeah I don't know much about it. 14:13 MTDiscord Yeah they've been working on it for over a year 14:18 sfan5 Oblomov: the real news is that this will allow nouveau to get power management / reclocking and that they're working on allowing Mesa to use the nvidia kernel driver in the long term 14:18 Oblomov sfan5: OH MY GOD POWER MANAGEMENT IN NOUVEAU 14:18 Oblomov it's still BS 14:18 Oblomov honestly, don't get an nvidia card if you're on linux, period 14:18 erle is this a gift from LAP$U$? 14:19 Oblomov or use the propietary drivers 14:19 erle nvidia drivers have wasted so much time 14:19 Oblomov nouveau has been basically useless for three generations now 14:19 erle intel integrated graphics forever :3 14:19 MTDiscord AMD iGPUs :) 14:21 erle open source nvidia drivers, cats and dogs living together, what's next? a sauna in every NATO base? 14:21 definitelya Oblomov: Certainly 14:22 erle a guy once told me that the matrix movie is set in 1999 because that was the last good backup they could find hahaha 14:22 Oblomov erle: lol 14:22 sfan5 so is it a bad thing that nvidia will make nouveau work again or what is your point? 14:23 erle isn't it that nvidia has a lot of user space stuff still to open source 14:26 erle “the current codebase does not conform to the linux kernel design conventions and is not a candidate for linux upstream” 14:26 erle t. nvidia 14:28 erle ig if they upstreamed something AND released the userspace component of their driver solution, i'd be much more cheerful about it 14:30 erle but i bet if they did that some assclowns would immediately modify it to mine more buttcoins :( 14:30 erle i hate crypto bros 14:32 Oblomov sfan5: my point is «too little, too late» 14:43 definitelya Oblomov: Like with Minetest or any other FOSS project, you can contribute to it yourself, if you find its features lackluster. 14:44 MTDiscord definitelya: you literally can't 14:44 Oblomov definitelya: yeah because I'm going to waste my time t contribute to a shitty company's bottom line for free 14:44 definitelya Another way is to donate hardware to them I think, I'll do it when I switch PC. 14:44 MTDiscord this requires technical expertise only NVIDIA workers will have 14:44 definitelya luatic: I meant improving nouveau. 14:45 MTDiscord Still probably requires a lot of reverse engineering hardware if you don't work at NVIDIA 14:46 Oblomov definitelya: nouveau is crippled by nvidia requiring signed firmware as a first issue 14:46 Oblomov that simply cannot be solved without nvidia's support 14:47 Oblomov really, we are at a point in time where there is absolutely no need to support shitty companies that don't play nice 14:47 definitelya That's why sfan said, regarding mesa support, "in the long term". 14:47 Oblomov amd has solved all these issues a decade ago 14:48 Oblomov documenting their hardware, building their proprietary stack on top of the open source one and actually hiring developers to work on linux support 14:48 Oblomov intel the same 14:48 Oblomov tell me again why anyone should give a rat's ass about nvidia yet unless you're vendor-locked for cuda? 14:49 Oblomov (in which case the floss driver is meaningless AGAIN) 14:49 MTDiscord because people need way's to waste their money ;) 14:49 MTDiscord ways* 14:49 definitelya Not their fault, really. They all have a bottom line to reach. Regulations for this are always going to be falling behind, IMO. 14:49 definitelya That's the issue. 14:50 Oblomov regulation? 14:50 erle “not their fault, it is slightly less profitable to be much less of an assclown” 14:51 Oblomov well yeah when you've been ripping off your customers by making them pay 3x extra for stuff that is just a bitflip in your proprietary driver 14:51 definitelya Oblomov: Yeah, security by obscurity is, for example, harmful to your and my freedom. 14:51 definitelya But this is getting offtopic lol. 14:51 Oblomov the point stands: nvidia's “open sourcing” is more of an “open washing” 14:52 Oblomov «hey look we open source stuff too» 14:52 erle which regulations exactly are saying you need to create shit drivers 14:52 Oblomov linus-nvidia-fuck-you.gif 14:52 definitelya It's still a win in my book. 14:52 definitelya xD 14:53 Oblomov “hey they stopped beating and raping me every day, now they only beat me every other day” is still a win too 14:53 definitelya That was unnecessary but ok. 14:53 Oblomov everybody is making a huge fuss for something that literally gives us nothing concrete now and should have been done a decade ago 14:55 Oblomov definitelya: people are getting head of heels because nvidia is marginally less shitty, how is my metaphor inappropriate 14:55 Oblomov when nvidia start playing nice with the rest of the floss ecosystem my attitude towards them will change 14:55 definitelya No it's appropriate, it just came out a bit strong to me ahah. 14:56 Oblomov it was intentionally strong, although I can see how it would be offensive for some 14:56 Oblomov should have put a content warning in front 14:56 definitelya ahah it's aight. 15:00 ROllerozxa o.o 15:01 definitelya s/came out/came off 15:30 * jonadab still misses Matrox graphics cards. 15:35 erle LIBGL_ALWAYS_SOFTWARE=1 forever 15:36 erle Oblomov it was inappropriate because the bodily and emotional harm from bad graphics drivers is *probably* about zero 15:39 Oblomov true 16:03 jonadab You say that, but if a bad graphics driver crashes your X server, that takes all your GUI applications with it, which can result in dataloss if you weren't smart enough to do everything in terminal multiplexer sessions. 16:33 Oblomov jonadab: don't. buy. nvidia. 16:34 jonadab I generally just use onboard video. 16:34 jonadab Minetest is pretty much the most graphics-intensive game I play. 16:34 jonadab *Most* of my gaming time, goes to games that run in a terminal. 16:37 ROllerozxa I honestly wouldn't be surprised if dedicated GPUs actually run Minetest worse than integrated xD 16:56 erle rubenwardy wsor rejected my package “crashtest” based on your comments about crashing not being useful, but it now contains a crash mitigation, i.e. if you install it your server does not crash from a thing that otherwise does crash it (invalid emerge). i plan to add any crashes with a suitable fix. is that allowed on cdb? 16:57 erle rubenwardy basically, instead of “this is a package full of crashes”, i want to pivot to “this is a package full of crashfixes and their test cases” 18:09 BuckarooBanzai sfan5: thank you for the latest prometheus PR, the new metrics are really useful now: https://monitoring.minetest.land/d/BRwIg8xMz/engine-metrics?orgId=1&var-instance=test.pandorabox.io:443 ;) 19:47 sfan5 BuckarooBanzai: nice 19:48 sfan5 did you figure out how to get meaningful numbers from the env step time? 19:48 sfan5 I briefly tested rate(minetest_whatever_the_name_was)[1m]/1000 but the numbers didn't match what the step time should be 19:52 sfan5 https://grafana.com/blog/2020/09/28/new-in-grafana-7.2-__rate_interval-for-prometheus-rate-queries-that-just-work/ btw 19:52 sfan5 you don't seem to be using that 22:11 MTDiscord erle, continue the package approval threads. Also, @wsor 22:20 MTDiscord What? 22:27 MTDiscord This (irc, this is a reply to erle's post above 22:40 MTDiscord Yeah, I saw 22:41 MTDiscord The core con ept of the mod doesn't fit concentdb. It's better off as a test mod for a issue 22:41 MTDiscord *concept 22:41 MTDiscord *contentdb 22:41 MTDiscord I hate mobile 22:46 MTDiscord conceptdb sounds fine too 22:47 erle Jonathon the core concept is “prevent crashes” 22:47 erle how is this a problem 22:47 erle this is not a test mod for an issue 22:47 erle in fact, installing this modpack is currently the ONLY way to prevent a server crash if a client triggers a malicious emerge 22:48 erle am i punished for doing TDD here? 22:49 rubenwardy TDD doesn't tend to include broken code in production 22:49 erle but i changed the mod after you complained 22:49 rubenwardy there's no dependencies between the two mods, running the crash in production is also weird 22:50 erle there are no dependencies because otherwise you can not separately enable or disable them 22:51 erle rubenwardy should the crash fix mod depend on the crash mod in your opinion? 22:51 erle look, originally i wanted to make two modpacks. one that contains all the crashes, one that contains all the fixes. 22:52 erle since i was told crashes are not acceptable as the sole content on cdb, i made one that contains the crash as a test case and contains the fix. i have a bunch of other fixes like that. 22:53 erle as far as i see, this is the only easy way to deliver fixes for these crashes to ordinery users running minetest versions in which this stuff is not fixed (for this emerge hang, this is all released versions). 22:54 erle rubenwardy should i add some minor playable content or a command so that cdb lets me distribute security patches for versions of minetest that are no longer supported? 22:55 MTDiscord contentdb is not a bug reporting place 22:55 erle oh but this is not for bug reports 22:55 erle you misunderstand the purpose if you think that 22:56 erle consider the following, you run minetest 5.5, some griefer crashes your server using a thing, you install this mod, it is safe. 22:56 erle there are even more crashes in older versions that can be protected against using lua. 22:57 rubenwardy There should be an otpional dependency between hang_emerge and hang_emerge_fix. However, hang_emerge isn't something that should be on CDB 22:57 rubenwardy what about making one mod with a setting to change between crash and fix, default to fix? 22:57 erle rubenwardy if hang_emerge is not on cdb, then how would a user of, say minetest 5.4.1 in debian figure out if debian has fixed it or not? 22:59 erle “There should be an otpional dependency between hang_emerge and hang_emerge_fix” … okay, so you mean a mod should depend on its test cases. i can do that, so that hang_emerge_fix depends on hang_emerge. 23:01 rubenwardy it would be the other way around, as you don't want Minetest to crash if both are loaded 23:01 rubenwardy you want the fix applied first 23:01 erle it can not crash if both are loaded, because emerging is not possible during mod loading stage and i use minetest.after(0, […]) here 23:01 rubenwardy ah 23:01 erle also the problem is, if hang_emerge depends on hang_emerge_fix then you CAN NOT test if your version of minetest is vulnerable 23:02 erle my goal is to a) make it possible to install this thing and not worry b) toggle single mods to figure out if it is vulnerable 23:03 erle and i don't get why i should add some hard-to-find setting if you can just toggle mods in modpacks. or add entirely superfluous dependencies that are in no way technically necessary for this to work. 23:03 erle by default, all mods in a modpack are enabled, am i wrong? 23:03 rubenwardy only if you enable the modpack 23:03 erle technically correct, the best kind of correct ^^ 23:04 erle but yeah, given the fact that i want to collect *all* the crash fixes (e.g. stuff from mineclonia) so that the wider minetest community can benefit, what exactly needs to change now? 23:04 erle maybe the description should emphasize that it fixes crashes 23:04 erle instead of that it can cause crashes haha 23:06 erle the thing is, because minetest coredevs are never ever going to make re-releases of old versions of minetest i see no other option to deal with stuff, given that e.g. linux distributions continue to ship old minetest for months or years.