freenode/#devuan/ Friday, 2020-02-14

tuxd3vhello, is there any possibility that gnu compiler 8.3 comes to ascii i386 os?02:04
tuxd3vI need to install Cuda toolkit02:05
tuxd3vand it complais about 6.3..02:05
gnarfacetuxd3v: 0 possibility02:17
gnarfaceas a general rule, Debian doesn't do major version updates to stable/oldstable02:19
gnarfacei see 8.3 is in beowulf though02:19
tuxd3vgnarface, yeah, but it would be nice to have them in ascii-backports, don't you think? :)02:36
tuxd3vits a mission to install cuda toolkit, in a machine with 2 diferent vendors graphics, in which both need libGL.so..02:37
tuxd3vthe gcc problem was surpassed easilly, now I need to do some black magic to have both, mesa-glx, and at same time nvidia-glx..02:38
gnarfacei don't think it's ever coming to backports, either.  it probably needs a newer glibc or something02:41
gnarfacei think they only added it to beowulf recently02:42
gnarfaceyou can try backporting it yourself, but as i've said before - by the time you get down to backporting all the dependencies, and those include core system components like glibc, usually you might as well have just upgraded to the next release by then02:43
tuxd3va bin-utils it need for sure, because of ls, ar, ar, and yeah a lot of other things are related..02:44
tuxd3vbut those packages could be in ascii-backports, it would be very nice to have them in i38602:45
tuxd3vMy opinion is that a i686 makes more sense than  i386, but I don't know how i386 was created.. it can have features from i686, i don't know..02:46
tuxd3vI am saying this because i feel that i386 is a lot less memory contrained than amd6402:47
tuxd3vI for instance are comming from amd64 to i386, due to 4GB memory contraints..02:47
tuxd3vand when I installed nvidia drivers after reboot, my machine consumes already around 450MB02:48
tuxd3vyeah, having dual hraphics make me suffer a bit, maybe the motif for migration, doesn't exist after all, since I get similar cold boot Ram usage..02:49
tuxd3v motif -> reason02:50
onefangIn general running 32 bit software uses less memory than 64 bit software, coz the memory pointers are half the size, and there might be a lot of memory pointers, so it adds up.02:51
onefangAnd then moving smaller amounts of memory to and from the CPU means tings might run a little faster as well.02:52
tuxd3vyeah, and I notice that, but after install nvidia drivers..I see that ~200MB are consumed onde by it.. so long story short 'maybe the migration to beowulf is not so hungry in 386..'02:53
GyrosGeierno02:53
GyrosGeierx86 32-bit does not have PC-relative addressing modes02:53
tuxd3v onde by -> only by02:53
GyrosGeieremulating these eats up any performance gains you get from the smaller pointers02:54
GyrosGeierthe only place where you don't need PC-relative addressing is statically linked programs02:54
tuxd3vGyrosGeier, it could be true at least at some extent, but i386 code density relies less in simd intructions than amd6402:54
tuxd3vand simd instructions duplicates the amout of registers at each iteration02:55
GyrosGeiersimd instructions are unaffected by pointer size02:56
tuxd3vi386 has less simd intructions, ofcourse the ones it has are not so powerfull, but hey.. they also consume less memory than the amd64 counterparts..02:56
GyrosGeieryou have wider SIMD instructions that have longer opcodes, but do more work at once02:56
tuxd3vGyrosGeier, ofcourse they are, but the point is not about pointers here, but about the amount of registers you will be using when simd is in place02:57
tuxd3vremember that each process will the schedules again02:57
GyrosGeierthat is not a relevant performance impact02:57
tuxd3vpreemptive or not, it will have to save its context, each time it goes of cpu02:57
tuxd3vand they ofcourse are saved to ram memory02:57
tuxd3vthat is not performance, that I am talking about02:58
tuxd3vI am talking about code density/Ram usage02:58
GyrosGeierand even then, SIMD registers are saved lazily02:58
tuxd3vhumm02:58
GyrosGeieronly when actually switching to a different task and if they were actually used02:58
tuxd3vyes, but the thing is..02:59
GyrosGeierthe kernel doesn't use MMX/SSE/AVX registers02:59
GyrosGeierso taking the timer interrupt is cheap02:59
tuxd3vthe userpace uses02:59
tuxd3vand you now, when simd comes in place at each iteration it duplicates the amount of registers..03:00
tuxd3vofcourse, its faster!03:00
tuxd3vbut ofcourse it uses more memory too03:00
GyrosGeierand tasks initially start out with SIMD support disabled, so the first SIMD instruction traps, then the kernel sets a flag to actually save and restore the registers03:00
tuxd3vso code density in i386 is better than amd6403:00
GyrosGeierno03:00
tuxd3vsimd starts disabled because simd is like a plague03:01
tuxd3vit consumes tons of power03:01
GyrosGeierno, power management is automatic03:01
GyrosGeierthe CPU would use the same if it wasn't disabled03:01
GyrosGeierthat is just to get the lazy switching03:02
onefangI said "in general", and should have added "if in doubt, benchmark".  No need to get into nitty gritty details of a specific CPU architecture here, there are better places for that.03:02
tuxd3vits exactly because of that they are disabled, and when a program requires simd a exception is raized and the kernel activates them03:02
GyrosGeierthe memory to store the registers is also always reserved03:02
GyrosGeiernot using them doesn't save anything03:03
tuxd3vyes, but i386 usesa lot less registers :)03:03
GyrosGeieryes03:03
tuxd3vso a lot less memory is reserved..03:03
GyrosGeierwhich degrades code density03:03
GyrosGeieri386 always spills registers to the stack03:03
tuxd3vwith simd at some extent it could be03:03
DarwinElfif a package I'm rebuilding needs debhelper 12, what can I do?  Debian 2 has debhelper 10...03:03
GyrosGeierDarwinElf, there should be a backport of debhelper03:04
GyrosGeieralso, that shouldn't have too many dependencies03:04
onefangDebian 2?03:04
DarwinElfoops, Devuan 203:04
DarwinElfi haven't used backports yet...03:04
GyrosGeierDarwinElf, also, I recommend pbuilder for compiling packages03:05
DarwinElfi'm not compiling more than one, so at this point I just need to learn how to install a backport on Devuan...03:06
GyrosGeierstill a good idea to use pbuilder -- that way, you don't need to install anything on your real machine03:07
DarwinElfwell, I will need to install it03:07
GyrosGeiertrue03:07
GyrosGeierbut you don't need any build dependencies03:07
DarwinElfactually I did03:08
GyrosGeierand you can often do your own backports03:08
GyrosGeierpbuilder is like "here's a source package, please set up a chroot, install all the builddeps inside, compile the package and give me the result"03:08
GyrosGeierall in one command03:08
DarwinElfi don't really want to do that until I've done it myself without any tools03:09
GyrosGeierdebhelper 12.1.1 is in beowulf03:10
* GyrosGeier sets up ascii environment03:12
GyrosGeierhmmk03:15
GyrosGeierdebhelper backport is complicated03:15
GyrosGeieryou can install dh-autoreconf and debhelper from Debian buster03:17
GyrosGeierthat works03:17
GyrosGeierthose are the same packages as in beowulf03:17
GyrosGeiersanest way is to add beowulf sources, and set up an apt_preferences file to set the default release to ascii03:18
GyrosGeierthen you can use "apt install dh-autoreconf/beowulf debhelper/beowulf"03:19
gnarfaceyes, that might work though it's important to note that mixing distros and versions is explicitly advised against by Debian03:21
GyrosGeieryes03:21
gnarface(and all Debian derivatives inherit the risks they are warning of, despite that Ubuntu leadership lies and sweeps them under the rug)03:21
GyrosGeierfor those two packages I'd say it's mostly harmless03:22
DarwinElfyou mean Ubuntu Windows?03:22
GyrosGeierbut yes, pbuilder would be preferable :)03:22
DarwinElfwell, I already have it from Devuan and am using it now03:22
gnarfacenever even touched it.  i'm just talking about Ubuntu PPA's... they're very risky03:22
GyrosGeierPPAs avoid mixing03:23
GyrosGeierpeople upload source packages, and the PPA compiles once per release03:23
gnarfacethey do not, that's the whole point of PPA's; they're community-contributed and unregulated03:23
gnarfaceyou have no idea what they're doing behind the scenes there03:23
GyrosGeierthey run pbuilder :)03:23
gnarfaceUbuntu has been very dishonest and misleading about that03:23
chillfanhttps://www.fsf.org/blogs/rms/ubuntu-spyware-what-to-do03:25
chillfanall you need to know about ubuntu ^^03:26
* tuxd3v Zas... mission acomplished :)03:26
* tuxd3v cuda toolkit installed in a unsuported os, unsuported compiler, unsuported glx infraestructure :)03:26
onefangCongrats, now don't come here asking for support for that lot.  B-)03:27
onefangTongue firmly in cheek, for those that didn't spot that.  lol03:27
GyrosGeierthe ubuntu setup for PPAs is rather easy to replicate03:28
GyrosGeier# pbuilder create --basetgz /tmp/xenial.tgz --distribution xenial --mirror http://archive.ubuntu.com/ubuntu03:28
GyrosGeiernow all you need is to replace "xenial" by a shell variable and make a for loop around that03:28
GyrosGeierand also run a for loop for building packages03:29
tuxd3vonefang, I followed my bisect intructions.. the hold perl install utility saves a man sometimes lol.. I extracted the package and installed some things by hand :), after that easy peasy03:30
tuxd3v hold  -> old03:31
DarwinElfturns out I won't need this package because a good enough one is in backports, and some of the results were confusing (like built ones for a newer and older version)... but at least I learned how to do it... would rather have learned it without debuild (or even equivs, but didn't use it) but maybe that's not necessary03:56
DarwinElfthanks for all the replies when I was asking questions about that03:56
Digitwhat packages do i install to get the go compiler stuff?  (needed for pixterm https://github.com/eliukblau/pixterm (like termpix)~(neither in repo)).   or is it not packaged either?  should just get it manually?  or are there other pixterm/termpix alternatives in the repos already?07:41
tarzeauDigit: https://go-team.pages.debian.net/08:17
psarriahi, are there any plans or progress for the next "Debian 10" based Devuan ?13:21
onefangIt's almost ready.13:23
psarriagreat news then13:24
hemimaniacoh good, I can't wait, I broke my devuan real bad16:38
hemimaniacbut it wasn't devuan's fault16:39
James1138How did you break Devuan?!  I do not want to make same mistake please.16:59
hemimaniactried upgrading/updating it without following the posts in the forums. tried to do it in 2 steps instead of 5, got b0rked17:00
James1138Ahhh17:01
hemimaniacoh and tried to force in a couple PPA's (which they explicitly tell you not to do) and ya, broken17:01
debuankernelWhat is advised how to compile PM_TRACE in to the kernel on devuan, and for acpi resume debugging?19:28
debuankernelMay be a pointer?19:28
debuankernelThanks in advance19:28
debuankernelOr: Which source do I take?19:29
fsmithreddebuankernel, I think this still works: http://forums.debian.net/viewtopic.php?f=16&t=3652519:37
debuankernelthanks19:45
systemdletedevuan is now #16 on distrowatch for non-systemd distros?   Is this because of a possible re-merge with Debian?21:31
systemdlete(am I dreaming?)21:31
djphdevuan merged back in with Debian!?21:34
djphABANDON SHIP21:35
golinuxNo it's because we kick ass21:35
djphgolinux: oh, sweet.  continue as normal then.21:35
systemdleteIf we "kick ass" why only 18th in the rankings?21:35
systemdleteIIRC, devuan was much higher way back21:35
djphbeowulf is nice; moved up to that on my apache / postfix box.21:35
systemdletehas beowulf been released?21:36
golinuxI don't follow numbers.  If you want that number to go higher, contribute or be quiet.  :D21:36
golinuxSoon.  Putting on the final touches21:37
systemdleteSome people here talk about beowulf as if it had been released already.  OK, thanks.   I am looking forward to Beowulf.21:37
systemdlete(I know how to "make" a beowulf and have done it a few times already)21:38
MinceRthose rankings reflect interest on the distrowatch site, nothing more21:38
MinceRyou can expect more page views when there's more hype21:38
systemdleteIt's a heuristic for me.  Just makes me wonder.  I know it is not scientific.21:39
djphI had to update to testing -- needed PHP 7.3 to get off Owncloud21:39
djphlessee -- MX, antix, gentoo, slackware ... all long-standing releases (IIRC)21:39
systemdleteSo you think it is due to devuan being relatively new?21:40
systemdletethat would make sense21:40
djphnot "new" so much as going up against some considerably older names21:41
systemdlete(I DID say "relatively" djph)21:41
djphwhat I mean is it's not that devuan itself is relatively new (in general) -- but that the comparisons have been around for _AGES_ -- and the people using them aren't likely to leave without their chosen project making some big missteps21:42
systemdleteI thought I read a link from here a month or two back saying that the debian project was looking at not being so systemd-centric.  Someone here posted it.21:42
djphyeah, their last GR / vote or something like that21:42
djphI dunno what actually came of it21:42
systemdlete(djph, I think we are saying the same thing, with different words)21:42
systemdletethere were umpteen proposals for the exact wording, yes.21:43
djphsystemdlete: the other ones "above" dev1 all appear to be rather specialized (and again, popular in their nicehs)21:43
djph*niches21:43
systemdleteanti-x and MX linux are not "specialized at all really.  I am using both of them.  For me "specialized" means, like, ipfire or openmediavault21:44
systemdletedjph:  "relatively" means precisely what you are saying.  "in comparison to..."21:45
golinuxMaybe this should go to #debianfork. It is hardly a support discussion.21:45
systemdleteit has degraded into a social discussion, yes.21:46
systemdletelet's take it there21:46
golinuxThanks21:47
Akulihow do i search for non-systemd distros on distrowatch?22:16
Akuli(hmm, this isn't a support question either)22:17
Akuli(now it's a crosspost on #debianfork too and everyone are even more annoyed because i cross-posted)22:17

Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!