[2026-03-12T05:12:55Z] honestly i'm at a point where i'm convinced the KCMA-D8 i got is fried [2026-03-12T05:13:02Z] it doesn't even post [2026-03-12T06:23:03Z] morning [2026-03-12T06:23:35Z] I got a question for you guys who use DEs/WMs where applications are run by launchers [2026-03-12T06:24:13Z] the question is this: is the launcher spawned each time you need it anew or is it a static process which then becomes the PPID of each process it starts? [2026-03-12T06:35:04Z] spawned anew each time. I use dmenu, so it cant really be something persistent [2026-03-12T06:35:31Z] dmenu then exec's whatever you just chose iirc [2026-03-12T06:54:56Z] sad_plan: that's good. because that not being the case is somewhat of a problem when trying to make the DE more resilient against running out of memory or process priorities [2026-03-12T06:56:32Z] for exmaple if I set a higher priority for plasmashell that gets inherited to child processes. [2026-03-12T06:56:55Z] which is every process started via krunner or the start menu [2026-03-12T06:59:18Z] the reason why that is an issue for me is because for various reasons I sometimes run into heavy swapping which basically slows my DE to a crawl which then makes it so that I can't even kill the offending process. [2026-03-12T06:59:54Z] using MaxSwapBytes=0 in the .service unit would disallow swap for the DE components, effectively stopping that from happening [2026-03-12T07:04:12Z] Ozymandias42: you should use earlyoom [2026-03-12T07:04:33Z] the kernel's oom killer is famously dogshit [2026-03-12T07:04:53Z] this would help but this feels as if it doesn't solve the issue at the source. it just pre-empts not well working actual oom [2026-03-12T07:05:07Z] you can write a patch for linus if you'd like :P [2026-03-12T07:06:29Z] I mean I don't think the kernel's oom is necessarily at fault. the way it works it needs proper oom-scores and info about whether a process may use swap at all or not. [2026-03-12T07:06:57Z] I don't think most linux distros use even half the available config tweaks available to actually optimize resource allocation [2026-03-12T07:07:20Z] I will install earlyoom though [2026-03-12T07:07:22Z] good tip [2026-03-12T08:46:36Z] Ozymandias42: do you have low amount of ram? you could look into zram or zswap in this case [2026-03-12T08:46:57Z] I use zram atm, but I might aswell just use zswap instead [2026-03-12T08:48:13Z] my ratio is a little ridiculous [2026-03-12T08:48:28Z] i have 48gb of memory in my desktop and i have 24gb of zram because why not [2026-03-12T08:50:21Z] nice kris_. I cant mount all of my zram stuff, because of sbase. so in my case, Im kinda leaning torwards going back to zswap, but configure it to use 50%. I instead have a 10G /tmp using zram. [2026-03-12T08:50:37Z] not sure this is very optimal on my end though. but eh.. what the hell. it does atleast work [2026-03-12T08:57:10Z] is that even compressed [2026-03-12T08:57:13Z] the point of zram is compression [2026-03-12T08:57:47Z] on that note, apparently lz4 is about to get a ~50% performance increase [2026-03-12T08:58:07Z] https://www.phoronix.com/news/Linux-ZRAM-50p-Compress-Boost [2026-03-12T09:00:59Z] I do have enough memory. 16gigs. I also do use zram. the issue is that some services I use at work are garbage memory wise and sometimes lead to the issues described [2026-03-12T09:03:41Z] kris_: both zram and zswap is compressed. and my /tmp is comrpessed. although I havent verified it though [2026-03-12T09:03:49Z] damn, thats a lot kris_ [2026-03-12T09:04:03Z] Ozymandias42: ah, I see [2026-03-12T09:05:07Z] zram and zswap can be combined which is kinda cool [2026-03-12T09:05:39Z] sad_plan: yeah the question i had was more about whether or not your /tmp had compression [2026-03-12T09:05:39Z] or zram and normal swap because apparently there's swap priority for various swap devices [2026-03-12T09:05:58Z] your /tmp is compressed? how? [2026-03-12T09:08:00Z] kris_: it should be, seeing as I mount my /dev/zram0 to /tmp [2026-03-12T09:08:06Z] Ozymandias42: by mounting your zram as /tmp [2026-03-12T09:08:27Z] i can't tell if that's the hackiest thing ive ever heard or the most genius thing i've ever heard [2026-03-12T09:08:32Z] both, probably [2026-03-12T09:08:43Z] also, you *caan* use zram, with zswap, but it doesnt work like you expect. gentoo has some info about it [2026-03-12T09:08:53Z] kris_: its actually listed in the kiss wiki :p [2026-03-12T09:09:12Z] how do you mount zram as /tmp first time I'm hearing about this. I only know about using a tmpfs as /tmp [2026-03-12T09:10:13Z] Ozymandias42: just echo $size to /sys/block/zram0/disksize, which sets the size of the partition. mk2fs -q /dev/zram0. mount /dev/zram0 /tmp [2026-03-12T09:10:15Z] tada [2026-03-12T09:10:36Z] now /dev/zram0, which is compressed zram, is mounted as /tmp [2026-03-12T09:10:44Z] ah so you need two zram devices. one for use as compressed ramdisk storage and one for use as swap? [2026-03-12T09:11:19Z] i think he's saying he puts his swap on /tmp since he can't use zram normally due to sbase [2026-03-12T09:11:31Z] in all fairness the filesystem overhead there should be extremely low [2026-03-12T09:11:40Z] like thats not too bad its just unhinged [2026-03-12T09:11:45Z] no, zram and zswap isnt compatible with eachother. iirc, zswap will take precedent here, and make zram useless. but it might depend on the usecase or configuration. but the documentation says so atleast [2026-03-12T09:11:46Z] sbase? [2026-03-12T09:11:55Z] suckless' coreutils [2026-03-12T09:11:58Z] ah [2026-03-12T09:12:30Z] I actually wonder if zram is all that smart to use at all on modern devices with nvmes. [2026-03-12T09:12:43Z] it's muuuuuch faster [2026-03-12T09:12:49Z] even with nvmes [2026-03-12T09:12:54Z] zram certainly made sense on slower flash or mechanical hard drives but with how fast flash is today zswap should be totally sufficient [2026-03-12T09:13:01Z] kris_: I didnt touch my swap. I hae a 8G swap partition. I wanted to use zram for several devices, because the documentation says using more than one huge one is better. but I cant mount the rest for some reason. I can only mount one as /tmp [2026-03-12T09:13:31Z] Ozymandias42: ram is immensly faster than even a fast nvme disk, so noo. ram is still favorable [2026-03-12T09:13:40Z] of course it still is orders of magnitude faster but keep in mind the kind of data that actually gets put into swap [2026-03-12T09:13:54Z] all kinds of data gets put into zram for me even when theres no real memory pressure [2026-03-12T09:14:07Z] swapon --discard --priority 100 /dev/zram0 [2026-03-12T09:14:09Z] i use prio 100 [2026-03-12T09:14:15Z] that's due to your sawappiness value [2026-03-12T09:14:17Z] kris_: also, regarding swap again, I cant use zrap as addition swap, because sbase's mkswap wont let me create, or mount the swap partition because it errors on size being too small. [2026-03-12T09:14:32Z] ive tried to fix it and just make it not error on size, but it didnt matter :p [2026-03-12T09:14:35Z] given i have 20 cores i like the idea of just having it compress stuff [2026-03-12T09:15:27Z] lz4 compression is also really fast, so I think the supposed performance hit on compressing/decompressing the zram stuff is negligable [2026-03-12T09:16:46Z] i cant remember the last time my memory usage went over 12gb honestly [2026-03-12T09:16:59Z] so it's pointless for me but it's also transparent because there's never a time where all 20 of my cores are pegged [2026-03-12T09:17:09Z] i compile with j18 usually so my machine remains usable while im compiling [2026-03-12T09:17:57Z] I only ever use up most of of my ram if I rebuild my system, but thats because kiss doesnt delete the sources in the mean time. and if I use conty at the same time, I can run out of memory. so in those cases, I have to manually delete stuff in /tmp to avoid running out :p [2026-03-12T09:18:20Z] kris_: yeah not using all of your cores does let you do that. I comopile with -j9, and I have a 4c8t cpu [2026-03-12T09:18:50Z] my system is still somewhat usable, but I cant view videos in fullscreen without it missing LOTS of frames lol [2026-03-12T09:18:50Z] i bought so much memory because at the time i was regularly compiling android [2026-03-12T09:19:01Z] where 32gb is a minimum [2026-03-12T09:19:22Z] i'm not doing that anymore though because the lineageos people came to their senses and allowed microg signature spoofing in mainline [2026-03-12T09:19:25Z] so i dont have to maintain a fork anymore [2026-03-12T09:19:37Z] ah. ive never actually compiled anything android related. I was thinking about setting up an env for this, but I kinda never did, as I dont really need to [2026-03-12T09:19:58Z] did they now? I remember having to patch LOS for this some years ago :p [2026-03-12T09:20:08Z] yeah you can just install it from fdroid and go about your day now [2026-03-12T09:20:21Z] ah, nice [2026-03-12T09:20:23Z] because a lot of us got very loud and very pissed off [2026-03-12T09:20:42Z] ah, yes, someone just got tired of all the yelling about it then :p [2026-03-12T09:20:44Z] i don't like compiling android on a regular basis and i also didnt want to use lineageos4microg [2026-03-12T09:21:21Z] LOS4microg would be redundant now with LOS incliding this upstream. is that project even still active? [2026-03-12T09:21:42Z] for some reason yeah [2026-03-12T09:21:51Z] a little bit ago, AS FAR AS I KNOW, they leaked their signing keys on accident and still havent changed them [2026-03-12T09:21:58Z] which just gives me more ammo for the "i dont want to use some random fork" situation [2026-03-12T09:22:07Z] which is why i was maintaining my own [2026-03-12T09:44:54Z] sad_plan: just do `nice -n+19 make -j9` or for good measure `ionice -c idle nice -n+19 make -j9` [2026-03-12T09:45:44Z] doing that makes it so all other tasks have priority in terms of process scheduling and IO. so watching videos will stay fluid without framedrops [2026-03-12T09:47:49Z] kris_: well shit. thats terrible. reminds me of termux's situation. they just use a debug key, and at that point, using fdroid is actually a better solution from a security standpoint. which is kinda silly, when fdroid isnt very optimal from the security standpoint.. [2026-03-12T09:48:50Z] Ozymandias42: yeah, ive messed with the niceness at times, but mpv still cant play the video properly if I run with -j9 :p I dunno why that is. so mostly I just view the video in a smaller window size instead :p [2026-03-12T09:49:11Z] why use -j9 at all on a 8thread cpu then? [2026-03-12T09:49:14Z] just do -j7 [2026-03-12T09:51:48Z] its recommended to use +1 on the number of threads :p [2026-03-12T09:52:12Z] that is if you want it to run as fast as possible [2026-03-12T09:53:17Z] which I guess you don't want if you want to be able to do other stuff while it's compiling [2026-03-12T09:57:45Z] well, sure. sometimes I just run -j4 instead to not be bothered by it. [2026-03-12T09:58:04Z] but if im rebuilding everything under kiss, that would take forever. so I rather run with -j9 instead. [2026-03-12T09:58:47Z] however, I often force remove stuff like webkitand other packages taking forever to build, so that its more controlled in case things fail on the way [2026-03-12T09:59:21Z] which kinda alludes to the reason for me using oasis. its much more stable, as all builds are reproducable [2026-03-12T09:59:44Z] ive ported several things to oasis, and ill probably keep porting things to oasis as time goes by [2026-03-12T10:08:29Z] also, regarding number of cores - this is interesting https://blogs.gentoo.org/ago/2013/01/14/makeopts-jcore-1-is-not-the-best-optimization/ [2026-03-12T10:08:40Z] now I need to test this myself lol [2026-03-12T13:33:35Z] https://jf.si/posts/2024-07-14-openbsd-7-5-router-with-vlans/ [2026-03-12T13:33:39Z] guess ive found my spring break plans [2026-03-12T13:33:48Z] manual vlans with openbsd isn't even remotely as annoying as i thought it was [2026-03-12T13:34:04Z] going to be so satisfying not having to deal with openwrt on typical x86 hardware anymore [2026-03-12T13:34:11Z] (i love openwrt but good god) [2026-03-12T17:57:54Z] part of my spring break plans anyway i wanna gamering and get a bit into C++ [2026-03-12T17:58:13Z] ive been trying to muster the courage and also have the free time to get into C++ for a while but that language looks like a genuine dumpster [2026-03-12T19:48:15Z] Yeah, C++ is awful. The only language that's worse is Rust (even more complex, even less bootstrappable).