10[00:20:57] <jhutchins> I am so glad that no-one is getting me up at 02:00 to patch a critical vulnerability on 750 servers and fix all the ones that didn't take it well.
88[01:19:45] <alex11> how can i fix links i click trying to open in thunderbird instead of firefox-esr? my web browser is set to firefox-esr in xfce's preferred applications, so that's not it
89[01:20:14] <oxek> alex11: links you click in which application?
121[01:37:51] *** Quits: JordiGH (jordi@replaced-ip) (Remote host closed the connection)
122[01:38:34] <mangix> cws: Debian 10 Buster has a bug with libstdc++ 8.3 that was fixed in 8.4. std::filesystem does not support files above 2GB in size
554[10:49:49] <jmd> How do I change my keyboard mapping?
555[10:50:06] <jelly> !keymap
556[10:50:06] <dpkg> Run "dpkg-reconfigure keyboard-configuration" to change both your default console and X keymap; the setxkbmap utility can be used to adjust keymap settings during X operation. For setting up X keymaps, see xkeycaps, xev and ask me about <multimedia keys>.
560[10:51:17] <jelly> jmd, a desktop enviroment like Gnome, KDE, Xfce might have its own settings stored on per-user basis somewhere else
561[10:51:42] <tete_> hi, i use debian 11 with xfce. not sure if this is the correct channel or if i should go to #xfce. at the moment i have a panel which has buttons for each open application, but they are so small and stacked (always 2 in a cell). if possible i would have all side by side
587[10:55:26] *** tnewman9 is now known as tnewman
588[10:55:26] *** zamnedix_ is now known as zamnedix
589[10:56:33] <jelly> ZenWalker, possibly your provider to blame. My outgoing IP address is in Germany fwiw.
590[10:56:35] <ZenWalker> jelly: here tor is using usa server
591[10:56:35] <tete_> jelly, thanks i already checked the configuration for that but i was unable to find anything that relates to the size of the buttons
592[10:56:49] <ZenWalker> line17: which is your country?
633[11:12:13] <Ede|Popede> tete_: should get smaller with the size setting in the same dialog. 24px in 2 rows is neat on my 21“ 1080p display
634[11:13:12] <tete_> ok that was it... i set the size too high, wow thats wierd. i had it at 54 then they were side by side, now with 43 its as i expected
635[11:13:27] <tete_> thanks!
636[11:13:49] <tete_> i have 27" wqhd
637[11:14:13] <jelly> I usually had a 128px panel on the right edge, and then tasks would get oriented either vertically or across the panel, which was nice
638[11:14:41] <Ede|Popede> nice size. and better for 2-sided PDF i think, even 1080 isn't optimal if borders can't properly be removed.
639[11:14:46] <jmd> Does anyone know how Qt works on debian? I have installed qt5-devtools. It seems to work, but I have to give an explicit path to start anything.
640[11:14:59] <jelly> wasting vertical space on panels on 16:9 monitors is just nonsense
641[11:15:52] <jmd> For example I can start linguist with /usr/lib/qt5/bin/linguist - that works fine. But if I just type "linguist" I get an error message "could not find a Qt installation of ''"
642[11:16:44] <Ede|Popede> jmd, you type this in the shell or in some qt dev thing?
648[11:17:45] <cyberjunkie> Just curious, I installed Debian in a minimal way, apt doesn't download/install the dependencies automatically
649[11:17:51] <Ede|Popede> ah, linguist is a link to qtchooser. didn't even know i have it installed.
650[11:17:59] <cyberjunkie> Is this normal, or am i doing something wrong?
651[11:18:18] <TheBigK02> can u share ur console output ?
652[11:18:39] <cyberjunkie> Me?
653[11:18:45] <Ede|Popede> jmd, did you play with its options?
654[11:18:46] <TheBigK02> cyberjunkie: yes
655[11:19:23] <cyberjunkie> TheBigK02: It's a different system, and it's meant to be a minimal headless server. I'll type out a gist of it. This hapens when install openssh-client
657[11:19:54] <cyberjunkie> Some apckages could not be installed, this may mean you have requested an impossible situation, or if you are using the unstable distribution that requires some packages have not been creatd or moved out of incoming
658[11:19:57] <cyberjunkie> is it my repo list?
659[11:20:04] <Ede|Popede> jmd, what's your QTTOOLDIR ?
660[11:20:04] <jmd> Ahh I have to pass -qt=5
661[11:20:10] <TheBigK02> cyberjunkie: you cant ssh to it?
662[11:20:16] <jmd> Is there a way to make it default to that?
663[11:20:24] <cyberjunkie> i have to install open ssh server which i can't :P
664[11:20:39] <jmd> Ede|Popede: Unset
665[11:20:50] <ratrace> cyberjunkie: which debian did you install, what ISO filename did you use?
672[11:21:48] <cyberjunkie> but i used the non-graphical install
673[11:21:56] <cyberjunkie> TheBigK02: Ok
674[11:22:06] <ratrace> cyberjunkie: and what options in the Software section did you select? If you didn't select openssh server, you won't have it, no
675[11:22:41] <ratrace> also, like TheBigK02 suggested, check the sources.list, make sure they're all set:
676[11:22:43] <cyberjunkie> but would it stop apt from installing dependencies for any package
677[11:22:44] <ratrace> !buster sources.list
678[11:22:45] <dpkg> A suitable /etc/apt/sources.list for "Buster" has the lines: "deb replaced-url
679[11:22:56] <cyberjunkie> ok
680[11:23:10] <ratrace> cyberjunkie: missing sources.list entries would, yes. make sure you have these, run `apt update` and try installing what you need
681[11:23:22] <cyberjunkie> I deleted it, doing theupdate now
682[11:23:33] <cyberjunkie> Worst case, I can reinstall it and let openssh install
683[11:23:51] <ratrace> if you have console access, probably no need to reinstall
684[11:23:55] <cyberjunkie> But it's not the standard is whatI wanted to know
685[11:23:58] <TheBigK02> cyberjunkie: i dont think thats necessary when the installer ran through properly
686[11:24:15] <ratrace> cyberjunkie: "standard" is what you choose on the software selection screen
687[11:24:35] <ratrace> I don't recall which of those is selected automatically, I don't think openssh-server is ,but a desktop would be. could be wrong.
688[11:24:45] <line17> zenwalker: good
689[11:25:22] <cyberjunkie> Ok, I remmeber unselecting the X/XFCE/desktop env, and i can't remember if i unchecked something else
690[11:26:24] <cyberjunkie> I think it asked me something about the repos during the install, but I assumed it wanted to download a bunch of things, so I may have skipped it
691[11:26:32] <cyberjunkie> Like, where it offers me a list of nearby repos
692[11:27:07] <TheBigK02> that doesnt install anything. that just includes a proper sources.list if i remember correctly. then sources.list would be a good starting point and see if the update works fine
693[11:27:31] <cyberjunkie> apt-get update leaves a blank sources.list now
694[11:27:37] <TheBigK02> that just does a proper apt-cache ... nothing that bloats ur debian out
695[11:27:57] <TheBigK02> u have no mirrors in /etc/apt/sources.list ?
696[11:28:29] <cyberjunkie> yep, manually entered a sources.list and it's installing fine :P
700[11:28:48] <TheBigK02> cyberjunkie: that was easy. good to hear u got it to work
701[11:29:01] <TheBigK02> deb-src not needed unless u want to recompile packages
702[11:29:13] <cyberjunkie> ok, and there's something about security.debian.org
703[11:29:17] <cyberjunkie> Is that needed
704[11:29:29] <cyberjunkie> first things first, openssh-server
705[11:29:31] <TheBigK02> oh yes, thats probably the most important one :)
706[11:31:28] <cyberjunkie> Ah good, it's been 2 weeks since I've been messing with Proxmos (ubuntu server, freebsd, nixos) and finally, i've made up my mind, headless debian for this used thinkcentre nano that was at some corporate setup
707[11:32:22] <TheBigK02> debian is like a swiss army knife. u can do a lot of stuff with it :)
708[11:32:36] <cyberjunkie> Nice, am an arch user on the laptop, fedora on the desktop
709[11:33:28] <TheBigK02> i never gave arch a go... but it reminds me a lot of gentoo in a way... and it was annoying to maintain in the long run
710[11:33:58] <cyberjunkie> lmao
711[11:34:19] <cyberjunkie> Ah. My first distro was Slackware in 96. Ended up losing my Windows partition.
712[11:34:40] <cyberjunkie> for some weird reason, I thought Debian still demanded you install every dependency manually
714[11:36:03] *** Quits: macrobat (~beep@replaced-ip) (Read error: No route to host)
715[11:36:29] <TheBigK02> i started with suse 7.2 if i remember correctly. never got internet to work and then ditched it... and came back to linux with debian 3.1... since then it was my server distribution of choice... on desktop i switched to ubuntu... and i feel a little ashamed for it... :p
747[12:08:58] <queip> there was a recent breakthrough in breaking RSA
748[12:09:41] <queip> could debian move to EC signatures, and/or lamport (which is pretty much guaranteed unbreakable and btw QC resistant) in signing packages, releases, install images?
749[12:10:01] *** Quits: dselect (~dselect@replaced-ip) (Quit: ouch... that hurt)
750[12:10:20] <ratrace> queip: has that paper been published and peer reviewed?
759[12:11:44] <ratrace> publish, peer review, and lesse what Schneier has to say.
760[12:11:46] <queip> ratrace: source to it is in the article
761[12:12:22] <ratrace> yes, "work in progress". so not final publication, no peer review.
762[12:12:52] <queip> ratrace: it takes a while to peer review. author is Schnorr, as in "schnorr signature" so he probably knows what he talks about
763[12:13:08] <ratrace> I'm not saying the paper is not true. I'm just saying I'm delaying hitting the panic button until those conditions have been satisfied and/or Schneier lit the Beacons of Gondor.
764[12:13:13] <mrjpaxton[m]> I can't deny that ECDSA is excellent, too. But RSA has been foolproof for a long time.
765[12:13:21] <queip> debian is not exactly known for rolling up massive updates too fast so better start thinking about it already
766[12:13:28] <queip> not when we confirm that rsa is borken
767[12:13:55] <queip> mrjpaxton[m]: yeah that paper desribes totally new mathematical weakness
768[12:13:59] <ratrace> worst case, we manually bump all RSA to EC where applicable, defaults are anyway just defaults
769[12:14:11] <ratrace> that said.... which EC :) that's _another_ field of contention
770[12:14:27] <queip> for example, Lamport is very easy to implement, and also is basically guaranteed to not have unknown mathematical problems - and as bonus is QC resistant
771[12:14:45] <queip> we might add additional lamport signs right now (while leaving GPG/RSA as is)
778[12:17:05] <ratrace> not all ECs are creataed equally.
779[12:17:14] <queip> yeah I know. secp256k1
780[12:17:23] <ratrace> NIST curves have traditionally been something to eye suspiciously.
781[12:17:38] <queip> this one literally secures billions USD that is one break away from getting it, so I would trust that on
782[12:17:44] <mrjpaxton[m]> Also, if you're asking about quantum computing resistant stuff, then we would need to hire a post-Doctorate CS professor AND a post-Doctorate cryptanalyst working for the government, who actually have access to a QCs in order to really know if the encryption "really is resistant". And even then, who really knows?
783[12:17:52] <queip> ceertainly do not use secp256r1 which is probably backdoored
784[12:18:01] *** debhelper sets mode: +l 1002
785[12:18:28] <mrjpaxton[m]> I don't wanna hear them conspiracy theories tho.
786[12:18:35] <ratrace> mrjpaxton[m]: they aren't
787[12:18:46] <mrjpaxton[m]> But some of us will believe anything. :p
788[12:19:37] <ratrace> just like "underhanded C" competition, complex EC math, especially when there's "magic numbers involved", could hide backdoors. nothing "conspiracy theoretical" about it, it's mathematical facts.
792[12:20:33] <TomyWork> ratrace, any conspiracy theory is possible. you outlined how it's possible, not how it's plausible
793[12:20:41] <queip> mrjpaxton[m]: secp256r1 and that double RGB something are confirmed backdoored, it's well known since long time
794[12:20:45] <ratrace> it's not in vain that everyone who understands cryptography will warn that one should never, ever, invent their own crypto. the reason is very simple. a LOT of possibilities to create weak crypto unintentionally.
795[12:20:58] <TomyWork> ok i take that back, some conspiracy theories aren't even possible
798[12:22:04] <mrjpaxton[m]> Oh yeah, I always trust in math, just for the record. At least the math I can already understand.
799[12:22:23] <queip> mrjpaxton[m]: one can assess algos to be vulnerable to QC, or likely resistant, without need to access real QC - it's well known what algorithms QC can process. for long time, the question is only do we/gov yet have computers that can do that theoretical operations
800[12:22:24] <mrjpaxton[m]> I wish I finished Calc 3 and Linear Algebra, at least.
801[12:22:29] <ratrace> "Once you eliminate the impossible, whatever remains, no matter how improbable, must be the truth." -- Sherlock Holmes or A. C. Doyle...
803[12:23:10] <queip> ratrace: secp256k1, and ed25519, are nothing-up-my-sleeve
804[12:23:36] <ratrace> TomyWork: technically, _any_ theory is possible. sufficiently advanced technology is indistuingishable from magick. A. C. Clarke
805[12:23:41] <queip> and lamport has no constants in it at all. (underlying sha/keccak will have but not in such way, just like S-boxes constants)
839[12:37:22] <ratrace> regardless of the RSA being broken, openssh have been warning about not using default RSA (SHA1) keys for some time now, for example
840[12:37:56] <mrjpaxton[m]> SHA1 is really only good for checksums now, if even that.
848[12:42:24] <TomyWork> i know that's the allegiation
849[12:43:27] <queip> this particular RSA break 2021/Shnorr, might turn out to not be practical. Still I would propose to add secondary signatures from a "distant" crypto system like ec, and also for one that is extremely different and btw is QC resistant
850[12:43:38] <queip> if it turns out we have still much more time for it, then the better
851[12:44:06] <TomyWork> and i guess based on the lack of transparency, it is sensible to reject that cipher, but you used the word "confirmed" earlier, which implies something stronger
852[12:44:38] <queip> TomyWork: ok better call it "is widely believed to be backdoored"
853[12:44:50] <TomyWork> :thumbsup:
854[12:47:27] <TomyWork> how does that actually work, btw? they generate a shitload of ciphers randomly, subject them to (supposedly automatic) tests, then subject the best candidates to more tests, until they're confident they have a good cipher?
855[12:48:00] <ratrace> by "ciphers" you mean curve parameters?
856[12:48:01] *** debhelper sets mode: +l 1008
857[12:48:10] <TomyWork> i suppose
858[12:48:52] <ratrace> well there's a range of values which makes sense due to P!=NP, how exactly they test them... not sure
908[13:44:50] <zykotick9> regarding GuiltySpark343's question, I've used the make-jpkg to install oracles's JRE in the past - does it also work for JDK like the OP wanted?
943[14:12:50] <ratrace> Lope: it's actually preferred, if you ask me. Unlike with individual files that get corrupt, you can't just revert from backup individual rows that correspond to corrupt places in a table file
944[14:13:16] *** Quits: szorfein (~daggoth@replaced-ip) (Remote host closed the connection)
945[14:14:04] <ratrace> then again, depends if your use case is OLTP-centric or OLAP-centric, and if the latter tolerates whole-db reverts in case of trouble
983[14:46:13] <ratrace> and while griping... the new UI elements of the current firefox-esr are atrocious. That awesomebar is popping out and makes the cues for active tab almost invisible.
984[14:47:45] *** Quits: Nex8192 (~Nex8192@replaced-ip) (Remote host closed the connection)
985[14:48:01] *** Quits: Filohuhum (~filohuhum@replaced-ip) (Remote host closed the connection)
1003[14:53:33] <jmonrods> Ok it suddenly changed, now -1430161280
1004[14:53:43] <digdilem> Sounds like a bug
1005[14:53:43] *** Quits: Filohuhum (~filohuhum@replaced-ip) (Remote host closed the connection)
1006[14:53:48] <jelly> jmonrods, your version of smartctl might be misinterpreting the attribute, or the firmware might be using a known attribute for unknown purposes
1007[14:54:40] <ratrace> jmonrods: surprisingly a lot of people use Edge, me included in the windows-gaming-vm-i-dont-really-care-about-anything-other-than-steam-but-i-need-a-browser-to-look-up-stuff-when-i'm-stuck.
1008[14:54:43] <jelly> jmonrods, which debian release is this and does smartctl -a /dev/whateverdisk show the same value?
1009[14:54:43] <jmonrods> thanks! it's a really old PC so probably the SSD is not supported by the firmware
1010[14:55:04] <ratrace> oops, taht was for digdilem :)
1011[14:55:05] <jelly> I meant the ssd firmware on the ssd.
1012[14:55:20] <jelly> it. self.
1013[14:55:24] <jmonrods> it says smartctl is not installed
1014[14:55:28] <jelly> install it
1015[14:55:33] <ratrace> with --no-install-recommends
1016[14:55:35] <jmonrods> ok ok will see
1017[14:55:36] <jelly> package smartmontools
1018[14:55:38] <ratrace> or else you pull in halfa mailing suite
1019[14:55:51] <jelly> oh dear. Yeah, what he said
1020[14:55:58] <ratrace> of Exim persuasion. The ugliest of them all.
1021[14:56:02] <jmonrods> hahaha ok
1022[14:56:11] <jelly> or install dma first, then install smartmontools
1023[14:56:21] <jaggz> ratrace, the ubuntu version, thankfully, installs without error (and tensorflow-gpu loads it) ..
1027[14:57:52] <jmonrods> done, thanks for the help! now it reads a 234 Unknown_Attribute of 32969851840
1028[14:58:15] <ratrace> jaggz: also in the post container world, companies might decide not to bother and simply defer to a snap or flatpak
1029[14:58:28] <jmonrods> but I wouldn't worry too much, data is backed up anyways
1030[14:58:35] <ratrace> which are thankfully installable on debian.
1031[14:58:47] <jelly> that's completely different :-D
1032[14:58:58] <jelly> jmonrods, you can also try this weird thing
1033[14:59:01] <jelly> !hdsentinel
1034[14:59:02] <dpkg> A gratis but closed source <SMART> client with slightly more user-friendly output than <smartctl> and ability to read more USB devices than smartctl. Binaries for x86 and ARM at replaced-url
1035[14:59:20] <jmonrods> oh nice! I'll look into it!
1036[14:59:23] <ratrace> ECC uncorrectible is never alone, usually correlates with reallocations or pending sectors
1037[14:59:35] <[O-O]> !start a race war
1038[14:59:36] *** spal_ is now known as spal
1039[14:59:41] <[O-O]> !start a de war
1040[14:59:42] <dpkg> KDE makes me vomit!
1041[14:59:56] <jmonrods> also it's not an old disk, it's been powered for only 3 months
1042[14:59:58] * jelly eyes [O-O]
1043[15:00:14] <ratrace> jmonrods: statistically, they have increased probability of dying within first 6 months
1044[15:00:30] <jmonrods> ah that's true it may fail at the beginning or at the end
1045[15:00:31] <ratrace> then that probability tapers off and starts increasing again with age
1046[15:00:44] <jelly> ratrace, that works for ssd too?
1047[15:00:59] <ratrace> I don't have sufficient data points for that
1048[15:01:00] <jelly> HDD have that bathtub curve
1049[15:01:11] <ratrace> oh right, this is SSD ....
1050[15:01:25] <jmonrods> I think it works that way for any product
1051[15:01:43] <ratrace> but look, why not. production imperfections, cold solders, stuff should show up in first few months of operation or stfu until the age gets to bite it
1054[15:02:48] <ratrace> sometimes the produc is so cheap and of low quality that the age factor overlaps with those "first few months" and it just dies :)
1055[15:03:34] <jmonrods> lol don't scare me, it's a chinese ssd so that's why I panicked with the negative number
1056[15:03:55] <jaggz> i wonder how it'll all interoperate between downloaded code -- in this case python, a python env with its dependencies, and their system-dependencies/libs
1057[15:04:10] <ratrace> the raw SMART values are rarely meaningful. in many attributes thoes values are bitmasks or other, nonlinear data
1060[15:04:38] <ratrace> what makes most sense is looking at the normalized values, that's what the firmware "thinks" is the relative state of the value in terms of "good" or "bad".
1065[15:05:54] <ratrace> smart data is mostly false negatives. if it shows issues, then you most likely have issues. if it doesn't, ... it doesn't mean you don't have issues.
1066[15:05:58] <jmonrods> I see. Yes they probably have built their own controller into their chips
1067[15:06:13] <jelly> hdsentinel provides interpretations for some of the custom attributes
1068[15:06:16] <jaggz> (because diagnostics on "drive healthy?" is such secret stuff .. don't want other manufacturers to know how you communicate that)
1069[15:06:29] <jelly> smartctl may have caought up with it or not
1071[15:06:49] <ratrace> personally I woulnd't bother. I'd stuff enough redundancy under ZFS or BTRFS and let them eat cake. oh the disk died.... boo. *replace*.
1072[15:07:03] <jelly> For SSD, I guess "TB written" is the most common indicator of age
1074[15:07:11] <jmonrods> Yeah! they are cheap enough. I'll back up every 2 weeks and let it be
1075[15:07:18] <jaggz> I do understand you might want to make advances to things, and not be limited to whatever's offered .. but keeping it proprietary is not necessary (I think)
1076[15:07:41] <jaggz> my issue is my seagates are slow as hell.. just tried them after a LONG time of using wd's mostly
1077[15:07:45] <ratrace> jmonrods: I back up enterprise grade EVO pros daily :) if I had cheap chiense things, I'd back them up.... minutely :)
1078[15:08:29] <ratrace> jaggz: thing with seagates is that unless you get the enterprise drives, they're probably SMRs
1079[15:08:39] <ratrace> and SMR in "normal" workloads is SLOOOOOOOW
1080[15:09:28] <jaggz> SMH
1081[15:09:42] <jaggz> and on top of it all, I think my 12gb x 2 GPU is spying on me
1085[15:10:47] <ratrace> jelly: yes, and some even come with a "health" indicator as SMART attrib, that creeps toward 0 (normalized) as time passes
1086[15:11:23] <jaggz> monitoring memory to determine the types of http(s) traffic I do, sites I use, etc. then the desired data they want -- they use an encoder to reduce the data down to tiny vectors made to follow normal patterns of behavior for me, with the variations needed to encode the data
1087[15:11:25] <jaggz> and then send it out
1088[15:11:33] <ratrace> oh, another nonobvious thing with SSDs. if they're unpowered, they start losing data. in weeks even. it's not medium degradation, powered back on it'll work normally, but data might be lost.
1089[15:12:21] <jaggz> or they change the system behavior so its EM emissions have a nice clear carrier signal they pick up with their satellites
1090[15:12:43] <imMute> ratrace: weeks? I'm gonna need a citation on that...
1092[15:13:05] <jmonrods> is it that severe? I know it's charge at the floating gates, but flash memory is the same
1093[15:13:16] <jmonrods> wow 7 days
1094[15:13:33] <jaggz> u.s. govt network security stuff points out that systems in supposed-to-be-isolated areas need to be careful of things like, say, metal pipes that can pick up the EM signals and carry those through to unshielded areas
1095[15:14:29] <jelly> ratrace, hdsentinel got a synthetic health % result as well! It's bogus but better than having to think about smart on your own
1096[15:14:30] <jaggz> didn't know that about ssd's.. wow..
1097[15:15:01] <jmonrods> best back up will be always tape
1104[15:17:52] <jaggz> (I think they're using the highly-sensitive nature, of these mass-produced (commodity I think they call it) magnetic sensors, for reading B fields of the body now.. very cool)
1105[15:18:09] <jaggz> in the past one typically needed a SQUID device for that
1179[16:27:58] <TomyWork> zykotick9, jelly why would you use amazon if there's adopt, right there. I use it on my ubuntu desktop, I use it with our software and a ton of 3rd party software. we ship it to our customers, they use it with our software and some 3rd-party software that we also repackage and ship, we had literally zero issues with it
1188[16:33:17] <Vanfanel> Hi! Is there a way to check for a lib but avoiding the messages printed by AC_CHECK_LIB? I need to use my own AC_MSG_CHECKING and AC_MSG_RESULT messages
1190[16:34:59] <jelly> TomyWork, long term worries. a) who maintains adoptjdk b) are they more likely or less likely to keep maintaining this stuff than amazon
1246[17:39:11] <line17> i recovered my files with testdisk but all files have lock icon and cannot open them. i tried "chown jack FOLDER" but not worked. also did chmod 777 but same not worked
1247[17:40:42] <flrnd> if you do a "file <something.ext>" what output do you get?
1336[18:48:33] <cws> Yeah, tons of buffers/caches.
1337[18:48:45] <cws> You've actually got some 5.odd gigs of RAM free.
1338[18:48:51] <cws> And barely any swap usage. All seems normal.
1339[18:49:07] <cws> Swap usage isn't a bad thing. That's the system doing its job.
1340[18:49:33] <NetTerminalGene> but it fills whenever i watch big sized video
1341[18:49:41] <greycat> that's normal
1342[18:49:44] <greycat> !free ram
1343[18:49:44] <dpkg> Unlike information, your computer's memory does *not* want to be free. Free RAM is wasted RAM! Linux tries to use free physical memory for caching files from disk which speeds up disk access considerably. Linux releases RAM from these caches if programs need it. If you want to know how much physical memory the free(1) tool says you have left for program use, it's 'free' + 'buffers' + 'cache'. Also ask me about <swapwake>.
1344[18:49:53] <cws> It doesn't "fill." Data gets cached.
1345[18:50:09] <greycat> !linuxatemyram
1346[18:50:11] <cws> If the system needs to allocate more memory for ACTUAL usage, caches get dumped.
1347[18:50:20] <NetTerminalGene> cws: after that mpv crashes, but vlc doesn't
1348[18:50:31] <cws> The only number that actually matters is the one under "available" in the "free" output.
1377[19:10:43] <Kobaz> greycat: not entirely accurate. a memory leak is when a process allocates memory that it does not then later free, without having to terminate the program
1379[19:11:20] <greycat> Two ways of saying the same thing.
1380[19:11:20] <ratrace> ... and doesn't reuse it.
1381[19:11:27] <Kobaz> and it doesn't have to be personal use either, it could be shared memory
1382[19:11:49] <Kobaz> allocates and then doesn't free is not 'unbounded'
1383[19:11:49] <greycat> The point is, if LINUX caches a file, that is not a memory leak. Not even remotely close.
1384[19:11:59] <Kobaz> well yeah, that's entirely different
1385[19:13:20] <Kobaz> allocating an unbounded amount of memory is a type of leak... maybe? depends how far you stretch the definition
1386[19:14:28] <greycat> If there's a bound, it means there isn't a leak. Allocating 200 GB to store a user's password is technically not a memory leak, since the amount is finite.
1387[19:14:33] <Kobaz> unbounded memory more falls under the definition of 'space leak'
1393[19:15:49] <Kobaz> So you're saying any time someone does malloc(100), it's never going to a leak, that's just patently false
1394[19:15:56] <Kobaz> That's bounded, so... it can't be a leak?
1395[19:15:57] <greycat> ...
1396[19:16:06] <greycat> IF YOU DO IT ONE TIME, IT IS NOT A LEAK
1397[19:16:10] <Kobaz> Sure it is
1398[19:16:15] <greycat> *plonk*
1399[19:16:37] <ratrace> leak is increasingly reserving and not reusing memory. so you hit memory limits, the kernel return ENOMEM but the program isn't reusign what's already reserved.
1400[19:17:18] <ratrace> reserving and not freeing for the duration, but reusing, isn't a leak. the key here is _not_ (re)using reserved ram and constantly allocating more
1401[19:17:56] <Kobaz> memory leaks are pretty simply and specifically defined as, unreachable memory or unfreed memmory once allocated
1406[19:18:48] <ratrace> unfreed? no. python is never returning memory to teh OS, onc allocated. that's not a leak, that's design. python does internal memory management and reusing already reserved blocks.
1407[19:19:22] <ratrace> so the key is: the program isn't using what's reserved and keeps reserving more. you both are correct, you just need to merge your truths into a single one :)
1408[19:19:26] <Kobaz> okay well, 'released' in whatever method your particular running environment needs for memory management
1409[19:20:23] <NetTerminalGene> man, caching is leak or not?
1410[19:20:29] <NetTerminalGene> just tell me this
1411[19:20:34] <Kobaz> haha, no it's not
1412[19:20:40] <Kobaz> we got into a little bit of a side thing here
1420[19:22:24] <NetTerminalGene> Kobaz: but is watching video cause that?
1421[19:22:25] <Kobaz> if you suddenly need more memory, and you don't have enough in the 'free' pool. it will page out the cache or dump it entirely and give you free memory
1422[19:22:28] <flrnd> I wonder why every example of memory leak is either C or C++ :?
1424[19:22:59] <ratrace> Kobaz: I'd add here: in general.
1425[19:23:15] <flrnd> NetTerminalGene: what happened exactly? I'm late and curious
1426[19:23:16] <Kobaz> ... in general
1427[19:24:26] <NetTerminalGene> flrnd: memory fills if i watch something like 5-6 GB video with mpv or vlc
1428[19:24:37] <Kobaz> NetTerminalGene: if watching a video makes your memory go up and your video is studdering or having other issues... possible problems: 1) your computer is too slow, 2) your video card is too slow, 3) you don't have enough memory... 4) video is corrupted?
1429[19:24:40] <greycat> NetTerminalGene: so what?
1430[19:25:10] <NetTerminalGene> Kobaz: neither.
1431[19:25:18] <Kobaz> 5-6GB video is a fairly large video... large videos typically need more ram to watch then say... a small video
1432[19:25:21] <NetTerminalGene> i have 8GB of ram. i think it is anough
1436[19:25:46] *** Quits: beelzebuzz (~rasputin@replaced-ip) (Remote host closed the connection)
1437[19:25:55] <flrnd> 8GB is bare minimum for today's standards
1438[19:26:00] <Kobaz> 8gb enough? depends what else the OS/APPS are using, and what type of video technology you're using, and how much ram it needs to operate
1439[19:26:01] <NetTerminalGene> do TVs have more memory?
1440[19:26:23] <NetTerminalGene> i don't think so
1441[19:26:28] <greycat> 14:20 NetTerminalGene> man, caching is leak or not?
1442[19:26:29] <flrnd> er
1443[19:26:31] <Kobaz> Stop 'thinking'
1444[19:26:35] <flrnd> you are comparing pears with apples
1445[19:26:37] <Kobaz> Read the data sheet and find out
1446[19:26:40] <greycat> You *WERE* told this already. Repeatedly
1447[19:27:02] <Kobaz> Also, TV's do a lot of decoding in specialized hardware
1448[19:27:31] <apaz_> What TVs do is not applicable here
1449[19:27:32] <flrnd> anyway, you asnwered yourlsef. You have the exact same problem with mpv AND with VLC
1450[19:27:42] <flrnd> yourself*
1451[19:27:58] <greycat> Linux uses idle memory to cache files. If the memory is needed, Linux drops the cached file and uses the memory for something else. You were given MULTIPLE explanations of this. You were given a web page.
1452[19:28:20] <Kobaz> NetTerminalGene: buy more ram?
1453[19:28:35] <flrnd> Kobaz: no becoz tvs have less memory :P
1454[19:28:39] <Kobaz> haha
1455[19:28:51] *** Quits: shtrb (~shtrb@replaced-ip) (Remote host closed the connection)
1462[19:31:57] <Kobaz> NetTerminalGene: Smart TVs tend to have a minimum of 2GB of ram... but.. as mentioned. TVs are NOT PCS. They are highly specialized pieces of hardware that are dedicated to processing video and have highly optimized methods to do so in as little footprint as possible. You cannot compare a TV to a PC in terms of memory usage.
1463[19:32:08] <NetTerminalGene> even my ancient phone with 1GB of ram can play 1080p video
1464[19:32:09] <Kobaz> If you want to watch video using no ram on your PC, get a TV
1465[19:33:11] <Kobaz> You can ask all the questions you like. Again... an 'ancient phone' is not a PC
1466[19:33:24] <ratrace> chances are they're linux based ARM thingies with custom userland
1467[19:33:33] <ratrace> android even
1468[19:33:43] <Kobaz> They also have specialized hardware for handling video. And that 1080p is downsampled I'm sure
1503[19:38:18] <apaz_> You allocate as many framebuffers as you're going to need, then you make your own pool of them in a circular buffer and re-use them. You only have to allocate a relatively small amount of memory at the beginning, then you free it when the application is closed or you exit the video.
1504[19:38:57] <Kobaz> NetTerminalGene: and your buffers go up?
1505[19:39:09] <Kobaz> buffers/cache ?
1506[19:39:25] <apaz_> Then when you're reading the file, you decode and shove it into the framebuffers. You don't have to hold on to encoded video.
1507[19:39:29] <flrnd> NetTerminalGene: but I don't see the problem here. does mpv or vlc crashes mid play or what?
1508[19:39:44] <Kobaz> buffers/cache going up is again... NORMAL
1509[19:39:54] <Kobaz> generally you will get that memory back if your system needs it
1510[19:40:38] <Kobaz> when you zoom back 30 frames, it still has a good chunk of that in memory (depending on the implementation), so seeking back will be really fast
1511[19:40:57] <Kobaz> And yes, preload, if you zip forward 30 frames, it probably already has that in memory/cache
1512[19:41:17] <apaz_> If you were to load the whole thing and hold it in memory, that would be bad.
1513[19:41:42] <apaz_> You could stash on disk, but like... meh.
1514[19:41:45] <Ede|Popede> apaz_: it does not. try to feed a .ts file (or a stream of segments) to stdin, you won't be able to rewind too much.
1515[19:42:02] <greycat> To clarify: if an *application* holds a full video in memory, that's bad. If LINUX does it, as part of a general file caching strategy, that's not.
1516[19:42:07] <mentor> transport streams aren't indexed
1528[19:44:03] <Ede|Popede> apaz_: "I would be surprised if"
1529[19:44:42] <Ede|Popede> mentor: `mpv $file` vs `mpv <$file` (if that's the correct redirection xD)
1530[19:44:54] <apaz_> Ah, gotcha
1531[19:45:31] <apaz_> Technically mpv could just read the stream, shove it into a temp file, then do whatever it wanted with that temp file.
1532[19:45:47] <Ede|Popede> or it may be related to ffmpeg, one of the things they did afaik is to remove code from the player which doesn't belong into it
1609[20:24:38] <Kobaz> if the pcap goes above 100MB, wireshark stops playing individual strams... if you filter out the stream, write it to a new pcap, it works fine
1686[21:27:40] *** Quits: Nitori_ (~kappa@replaced-ip) (Quit: "I hate eggs with no yolk. My friend Herensuge was beaten to death by one.")
1687[21:27:56] <flrnd> I wonder why he wants to let only the bios handle the thermal. AFAIK one of the reasons that thermald exist is to prevent the bios to throttle as much as possible the cpu, or am I missing something?
1784[22:26:26] <dpkg> pm is, like, private message (something you don't do without asking for permission first, see <private>), or post meridiem, or prime minister, or perl mongers, or pathetic moron, or provocative/preventative/percussive maintenance, or power management (see <pm debug>)
1785[22:27:52] <karlpinc> !firmware images
1786[22:27:53] <dpkg> There are <live> system and <netinst> and DVD images containing non-free Debian <firmware> packages available from replaced-url
1793[22:34:15] <flrnd> tosted: what you are trying to do is a bit more complex than simply change write permission to the ISO, I'm tempted to let you continue with that path, but *good person* me is just crying
1794[22:36:03] <flrnd> first off all, use the non-free iso image. If that image doesn't have the drivers you need, replaced-url
1795[22:36:50] <flrnd> or simply copy the driver to a second USB and load it (you have an option during the install to load additional drivers)