#archlinux-ports | Logs for 2018-01-16

[00:07:32] -!- p71 has quit [Read error: Connection reset by peer]
[00:26:02] -!- jp has quit [Read error: Connection reset by peer]
[00:31:43] -!- jp has joined #archlinux-ports
[00:57:15] -!- p71 has joined #archlinux-ports
[01:03:03] -!- City-busz has quit [Read error: Connection reset by peer]
[01:03:13] -!- eschwartz has quit [Read error: Connection reset by peer]
[01:03:25] -!- City-busz has joined #archlinux-ports
[01:03:40] -!- anyone has joined #archlinux-ports
[01:12:47] -!- guys has quit [Ping timeout: 248 seconds]
[01:32:11] -!- guys has joined #archlinux-ports
[05:21:31] -!- hringriin_ has joined #archlinux-ports
[05:24:57] -!- hringriin has quit [Ping timeout: 264 seconds]
[05:24:57] hringriin_ is now known as hringriin
[05:36:46] -!- MrBIOS has joined #archlinux-ports
[05:38:05] <MrBIOS> hi folks, anybody around?
[05:38:18] <MrBIOS> Is there a mailing list for archlinux32?
[05:38:29] <guys> Yes
[05:38:48] <guys> https://lists.archlinux.org
[05:38:49] <phrik> Title: arch-ports Info Page (at lists.archlinux.org)
[05:39:15] <guys> Also see the "Mailing List" link on the homepage: https://archlinux32.org
[05:39:16] <phrik> Title: Arch Linux 32 (at archlinux32.org)
[05:44:35] <MrBIOS> Thanks. is anyone here involved with it, out of curiosity?
[05:49:18] -!- deep42thought has joined #archlinux-ports
[05:50:10] <deep42thought> MrBIOS: Involved with the mailing list?
[05:50:28] <MrBIOS> with the 32-bit port, generally
[05:51:40] <deep42thought> here I am :-)
[05:52:11] <deep42thought> but others idling around are also involved ;-)
[05:54:12] <guys> MrBIOS: Well, deep42thought is the main person in charge, basically. ;)
[05:54:40] <guys> Also tyzoid and abaumann
[05:55:12] <guys> I mostly lurk to answer questions and offer tips, most of my time is busy with the x86_64 mainline
[05:55:12] <deep42thought> infra is also from rewbycraft and Polichronucci
[05:55:35] <deep42thought> and the whole concept is from City-busz
[05:59:19] <MrBIOS> got it. A couple of years ago, I purchased about 180 OLPC XO-1's, which are 32-bit, single-core 433MHz AMD Geode based machines.
[05:59:57] <MrBIOS> I've owned a couple of OLPC XO-1's since the Give One, Get One program was launched by the OLPC foundation ten years ago.
[06:01:10] <MrBIOS> I mostly purchased these to keep them out of a landfill, and have refurbished and sold some to some non-profits who manage existing deployments of XO-1 OLPC laptops, but I still have a ton of them.
[06:01:55] <MrBIOS> anyways, the userland has been stalled at Fedora 18 for many years.
[06:03:21] <MrBIOS> deep42thought, out of curiosity, are you familiar with this device, even superficially?
[06:03:26] <MrBIOS> specs at http://wiki.laptop.org
[06:03:27] <phrik> Title: Hardware specification - OLPC (at wiki.laptop.org)
[06:03:46] <deep42thought> MrBIOS: I'm just reading into it
[06:06:12] <deep42thought> ah, it's i586
[06:07:14] <MrBIOS> yes, ihs
[06:07:15] <MrBIOS> ish
[06:08:48] <deep42thought> so you'll be interested in abaumann's project: the porting to i486 :-)
[06:09:14] <MrBIOS> well, I would be willing to bet that most packages would run unmodified
[06:09:47] <MrBIOS> worth a try anyways. I'd be using a custom-built kernel
[06:10:01] <MrBIOS> abaumann_ ping :)
[06:10:16] <deep42thought> he's probably still asleep
[06:10:23] <deep42thought> (I'm up a little early)
[06:10:40] <MrBIOS> you're in Germany, how about him?
[06:10:46] <deep42thought> Swiss
[06:10:49] <guys> I'm up a little late :)
[06:10:53] <deep42thought> :-D
[06:11:00] <deep42thought> at least we have some overlap ;-)
[06:11:03] <MrBIOS> I'm on the west coast of the US, so it's still evening here
[06:11:48] <guys> MrBIOS: technically it is evening here too, it's only midnight after all...
[06:12:09] <deep42thought> I find it mind numbing, that this should be all the same sun - here it's about to rise, on your place it just set ...
[06:12:29] <guys> "angles"
[06:12:30] <deep42thought> I should convert to flat earth creationism ;-)
[06:12:39] <guys> also possibly quantum
[06:12:49] <deep42thought> relativity makes that, I guess
[06:13:43] <MrBIOS> anyways, as an aside, I have a strong interest in getting functional OLPC XO-1's into the hands of folks such as yourselves, so if anyone wants one, let me know.
[06:14:07] <deep42thought> thanks
[06:14:13] <MrBIOS> They can boot from the embedded 1GB NAND, as well as the full size SD card, or from USB
[06:14:33] <deep42thought> I think abaumann could use one (he struggled to get real i486 hardware)
[06:15:03] <deep42thought> my wife specifically forbid me to set up old hardware for archlinux32 :-(
[06:15:33] <MrBIOS> well, the OLPC is pretty small :)
[06:15:45] <deep42thought> it's not the size, but the time, I think :-D
[06:16:24] <MrBIOS> understood ;)
[06:23:30] <guys> It is the kind of thing where I'd like to say I want one, except I'm not sure what I'd do with it...
[06:24:03] <deep42thought> !grab guys
[06:24:04] <phrik> deep42thought: Tada!
[06:26:42] <TomTheDragon> deep42thought: you don't have any HW to test?
[06:27:11] <deep42thought> I have a i686 box and a not-yet-archlinux, almost-i686 box
[06:27:21] <deep42thought> "in production"
[06:27:27] <TomTheDragon> almost-i686?
[06:27:43] <deep42thought> i586, but has all but one commands from i686
[06:27:51] <TomTheDragon> ah, that
[06:28:51] <deep42thought> e.g. the minimum, debian runs on
[06:30:02] <TomTheDragon> yeah. I'm definitely interested in real i586 port... wouldn't mind finding a way to port debian since that's what I already use, but it looks like their build system is even more arcane.
[06:30:14] <MrBIOS> https://support.amd.com
[06:30:24] <MrBIOS> I'm pretty sure Geode LX is much closer to i686 than i486 :)
[06:30:56] <deep42thought> but it's closer to i586 than to i686
[06:31:11] <deep42thought> at least closer, than what I got
[06:31:20] <MrBIOS> is it? I'm not sure which instructions it'd be missing
[06:31:22] <TomTheDragon> Yeah... definitely. But one missing instruction means it's just as incompatible as a few tens to hundreds.
[06:32:14] <TomTheDragon> Barring opcode emulation, which is always an option for a few missing opcodes.
[06:34:20] <MrBIOS> http://wiki.laptop.org says it's i686
[06:34:21] <phrik> Title: Geode instruction set - OLPC (at wiki.laptop.org)
[07:03:15] -!- deep42thought has quit [Quit: Leaving.]
[07:32:54] -!- oaken-source has joined #archlinux-ports
[07:34:31] -!- MrBIOS has quit [Quit: Going offline, see ya! (www.adiirc.com)]
[08:23:25] -!- deep42thought has joined #archlinux-ports
[08:52:53] -!- abaumann has joined #archlinux-ports
[08:53:18] <deep42thought> Hi abaumann
[08:53:21] <abaumann> Hi there. Just got to work.
[08:53:37] <deep42thought> It looks like we sneaked sse2 instructions into our packages
[08:53:42] <abaumann> MrBIOS: those machines have a AMD Geode LX, so it's the same CPU as in the Alix board.
[08:54:03] <abaumann> So I would try the standard Archlinux32 ISO.
[08:54:37] <abaumann> Yeah. SSE2 is nowadays generated automatically. And I think only last generation i686 have SSE2 at all.
[08:55:01] <abaumann> So we are back to the opcode sniffer :-)
[08:55:26] <deep42thought> is it possible to pretend not to have sse2 in the build environments?
[08:55:38] <deep42thought> e.g. fake /proc/cpuinfo
[08:56:10] <abaumann> aeh. If you have qemu yes.
[08:56:20] <abaumann> Then you can pretend to have a 486 with SSE2, if you like :-)
[08:56:39] <abaumann> hence I don't like the chrooting method for builds.
[08:56:59] <abaumann> Something like the uname hack of LFS could be done, faking /proc/cpuinfo.
[08:57:11] <abaumann> This would mean to clone the procfs code.. :->
[08:57:39] <abaumann> MrBios: make sure you set Architecture=i686 explicitely in /etc/pacman.conf, uname -a recognizes the AMD Geode LX as i586, which is not entirely correct.
[08:58:10] <abaumann> And speaking of a true i486: https://bbs.archlinux32.org
[08:58:11] <phrik> Title: Unisys CWD-4002 / Artwork, Screenshots & Setups / Arch Linux 32 Forums (at bbs.archlinux32.org)
[08:58:20] <abaumann> ISA bus, NE2000 card, 32 MB RAM.
[08:58:53] <abaumann> But I'm not oposed to get another old laptop, tried myself to get one of those "laptop for children" myself..
[09:04:16] <deep42thought> we could also bind-mount a different /proc/cpuinfo
[09:04:35] <deep42thought> if that's the only place where the compiler can get the information if the cpu knows sse2
[09:05:08] <abaumann> it's not the compiler, I think?
[09:05:13] <abaumann> more some macros in autoconf or cmake?
[09:05:35] <abaumann> Well written software has a --without-ssl2 or so. :->
[09:06:12] <deep42thought> ah, I thought if you pass "-O1000", the compiler tries to squeeze out the last femtosecond by using all available instructions
[09:08:16] <abaumann> not if you pass -march
[09:08:40] <deep42thought> ok, so -march=i686 should avoid sse2
[09:08:43] <deep42thought> ?
[09:09:14] <abaumann> Yes, I think so, because SSS2 was standard for early 64-bit CPUs when it got a requirement for being a 64-bit Intel CPU.
[09:09:44] <abaumann> But some packages have a hand-crafted sniffer and some GAS/NASM code, for instance OpenSSL
[09:09:55] <abaumann> so you have to know how to turn of those things.
[09:10:13] <abaumann> Which packages have SSE2 reports now?
[09:10:39] <deep42thought> https://bbs.archlinux32.org
[09:10:40] <phrik> Title: Too many coredumps after update. / Pacman / Pacman Upgrades / Arch Linux 32 Forums (at bbs.archlinux32.org)
[09:10:40] <abaumann> I myself know about Vivaldi (which is a closed-source SSE2 "infected" binary).
[09:11:06] <deep42thought> qt5-base seems to be the biggest issue :-/
[09:11:39] <abaumann> and librsvg is gnome?
[09:11:53] <deep42thought> looks like it is
[09:12:00] <abaumann> yeah. they most likely do some trickery to enable whatever is available on the machine.
[09:12:10] <abaumann> a normal configure/make/make install package usually works just fine.
[09:12:13] <deep42thought> we broke the main two desktops!
[09:12:23] <abaumann> uh! this is cool. :-)
[09:12:44] <abaumann> and there is a x265 problem too around that corner.
[09:12:45] * deep42thought congratulates everyone with a minimalistic desktop - like he himself uses
[09:12:55] <deep42thought> just solved that
[09:12:55] <abaumann> I have X and notion. :->
[09:13:06] <deep42thought> fluxbox
[09:13:15] <abaumann> very minimalistic, indeed :-)
[09:13:34] <abaumann> well.. we se the positive side effects..
[09:13:37] <abaumann> see
[09:15:04] <abaumann> Sadly I cannot do much till this evening.. work.. moving around data..
[09:15:32] <deep42thought> yeah, np
[09:15:33] -!- yar has quit [Excess Flood]
[09:18:50] -!- yar has joined #archlinux-ports
[09:45:18] -!- abaumann has quit [Quit: Page closed]
[10:27:08] <buildmaster> haskell-hedgehog is broken (says buildknecht3).
[10:56:07] -!- deep42thought has quit [Quit: Leaving.]
[11:38:45] <buildmaster> pandoc is broken (says buildknecht3).
[13:08:16] -!- deep42thought has joined #archlinux-ports
[13:26:11] * buildmaster goes insane.
[13:28:41] <guys> omg
[13:29:54] <deep42thought> well, consul decreased its pkgrel from 2 to 1
[13:29:59] <deep42thought> in commit c8c56f2a0a8a
[13:30:58] * buildmaster resumes sanity.
[13:31:19] <deep42thought> buildmaster: good boy!
[13:42:11] <guys> You caught us! 17 minutes of shame!
[13:42:31] <deep42thought> it updates every 5 minutes - you need to be faster :-)
[13:42:56] <deep42thought> no, every 10 minutes - still too fast
[13:43:51] <deep42thought> the sub-pkgrel system also pays off: we now have consul-1.0.2-1.1 in community-testing and consul-1.0.2-1.0 in community :-D
[13:49:45] -!- guys has quit [Ping timeout: 276 seconds]
[13:56:23] -!- alyptik has quit [Ping timeout: 256 seconds]
[13:57:24] -!- alyptik has joined #archlinux-ports
[14:36:31] -!- guys has joined #archlinux-ports
[15:00:04] anyone is now known as eschwartz
[15:59:27] -!- deep42thought has quit [Quit: Leaving.]
[16:18:30] -!- insecta has quit [Quit: WeeChat 2.0.1]
[16:21:38] -!- insecta has joined #archlinux-ports
[17:40:55] -!- deep42thought has joined #archlinux-ports
[17:41:34] -!- abaumann has joined #archlinux-ports
[17:42:56] <deep42thought> abaumann: the ospray error looks like it might be some toolchain issue, but this is more your expertise
[17:43:14] <abaumann> ui: ok.
[17:43:28] <abaumann> for disabling SSE2 in qt5
[17:44:46] <deep42thought> unfortunately, I have to leave in a few minutes
[17:45:29] <abaumann> I'll have a look what I can do.
[17:45:39] <abaumann> priority has the SSE2-problem, I think.
[17:45:50] <deep42thought> your decision
[17:45:53] <abaumann> ok.
[17:47:30] <abaumann> /usr/lib/gcc/i686-pc-linux-gnu/7.2.1/include/xmmintrin.h:853:1: error: inlining failed in call to always_inline ‘void _mm_setcsr(unsigned int)’: target specific option mismatch
[17:47:36] <abaumann> mmh. sweet. :-)
[17:47:48] <abaumann> but currently it affects only the compilation of a package in community.
[17:47:57] <abaumann> having Qt5 back is more important.
[17:48:03] <deep42thought> yeah
[17:48:16] <deep42thought> I just wanted to mention - in case the errors are connected
[17:48:28] <abaumann> I don't think so.
[17:48:30] <deep42thought> ok
[17:48:47] <abaumann> gcc 7.2.1 start to break some stuff, like inline tricks and assembly hints.
[17:49:03] <abaumann> has nothing to do with SSE2.
[17:49:12] <abaumann> .. I hope. :-)
[17:53:48] <abaumann> # FS#38796:[[ "${CARCH}" = "i686" ]] && SSE2="-no-sse2"
[17:53:59] <deep42thought> trivial
[17:54:04] <abaumann> yep.
[18:01:30] -!- deep42thought has quit [Quit: Leaving.]
[18:25:00] <brtln> told you to grep for things like these before deprecation :P
[18:41:09] <abaumann> true :-)
[18:41:11] -!- isacdaavid has joined #archlinux-ports
[18:42:46] <buildmaster> qt5-base is broken (says buildknecht).
[18:42:59] <abaumann> on the other hand, such problems will pop up also in the future, so having an opcode sniffer for SSE2 is a good thing anyway.
[18:43:03] <abaumann> *darn* (TM)
[18:43:40] <abaumann> qt5-base broken, there is also the possibility that SSE2 is now required (I have seen bug reports pointing into that direction).
[18:44:58] <abaumann> ERROR: Unknown command line option '-no-ssl2'.
[18:45:01] <abaumann> very funny. :-)
[18:45:04] -!- oaken-source has quit [Ping timeout: 256 seconds]
[18:45:06] <abaumann> it's really late..
[19:34:24] <buildmaster> ovmf is broken (says buildknecht2).
[19:36:24] <buildmaster> pandoc is broken (says buildknecht).
[19:50:17] <abaumann> qt5-base-5.10.0-3.1 seems ok again. KDE starts.
[19:51:27] <abaumann> at least the splash screen is.. :-)
[19:51:38] <abaumann> now I have a really black background..
[19:53:29] <abaumann> "All shell packages missing. this is an installation issue, please contact your distribution". :-)
[19:55:21] <abaumann> aha: pacman -S plasma-desktop
[19:56:21] <abaumann> courtesy to: https://bugs.debian.org
[19:56:25] <phrik> Title: #851175 - plasma-workspace - All shell packages missing. This is an installation issue, please contact your distribution - missing dependency on plasma-desktop-data - Debian Bug report logs (at bugs.debian.org)
[20:07:47] -!- abaumann has quit [Quit: leaving]
[20:11:02] <buildmaster> virtualbox is broken (says buildknecht2).
[20:12:32] -!- abaumann has joined #archlinux-ports
[20:13:09] <abaumann> ospray: OSPRay SSE4 ISA target enabled OSPRay AVX ISA target enabled OSPRay AVX2 ISA target enabled Found Intel SPMD Compiler (ISPC): /usr/bin/ispc, aeh, this sound all quite unhealthy for a i686 :-)
[20:13:14] <abaumann> the toolchain is fine.
[20:19:32] <abaumann> ospray: Your Embree build needs to support at least one ISA >= SSE4.1! in cmakelist.txt
[20:19:37] <abaumann> so, yes. blacklisted.
[20:19:54] <abaumann> don't expect any Intel library to work on old hardware..
[20:20:07] <TomTheDragon> ah... packages which are depending on a prebuilt lib?
[20:20:09] <abaumann> new Intel library (that is)
[20:20:25] <abaumann> no. embree has a ISA switch and is built for a minimal ISA
[20:20:36] <abaumann> the lowest one is SSE2 anyway (I think).
[20:20:46] <abaumann> now ospray needs at least SSE4.1 from Embree.
[20:20:56] <abaumann> TomTheDragon: hi, BTW. :-)
[20:22:49] <abaumann> virtualbox: cc1plus: out of memory allocating 2524 bytes after a total of 8503296 bytes
[20:23:10] <abaumann> deep43thought: I kicked virtualbox onto my slave..
[20:24:06] <abaumann> s/deep43thought/deep42thought/g :-)
[20:29:03] <TomTheDragon> hi abaumann
[20:29:17] <abaumann> hi.
[20:29:42] <abaumann> deep42thought: I blacklisted ovmf (EFI for virtual machines). Doesn't build for 32-bit.
[20:31:41] <TomTheDragon> yay, OOM while compiling. that's always fun.
[20:32:44] <abaumann> virtualbox has issues when glueing together tons of SOAP object files.. funny thing: it works on my slave, but not on the others.
[20:38:47] <abaumann> virtual memory exhausted: Operation not permitted
[20:38:55] <abaumann> on virtualbox: jippie.
[20:39:11] <abaumann> we will see the time when an 32-bit address space is not enough to start a g++ compiler ;-)
[20:46:32] <abaumann> aha: my slave is now also set up properly (with a build user) and is no longer running as root.
[20:46:39] <abaumann> that's the only difference I can find..
[20:47:18] <TomTheDragon> is the compiler just getting more and more bloated, or something? I've noticed compiles seem to take more RAM than they used to, by far.
[20:47:25] <TomTheDragon> or is it usually the application software that's the limit?
[20:47:52] <abaumann> In C++ you do a lot more templating and stuff. this eats memory.
[20:48:12] <abaumann> the C part of gcc is still very small and can compile on as little as 16 MB RAM.
[20:48:32] <abaumann> C++ needs 64 MB RAM to run confortably.
[20:48:34] <abaumann> gcc that is.
[20:49:14] <abaumann> another thing which was trendy is to to amalgamation of source code before LTO was invented.
[20:49:36] <abaumann> optimization across multiple artifacts is nice, but again, needs more memory.
[20:51:14] <TomTheDragon> I've seen way more than that much RAM used
[20:54:37] <TomTheDragon> Especially with C++
[20:55:00] <abaumann> well, it depends on the complexity of the software too..
[20:56:08] <abaumann> weird: my systemd-nspawn runs on a 64-bit machine with 8 GB of RAM. Ok, I'll add some big swapfile to it..
[21:13:09] -!- abaumann has quit [Quit: leaving]
[21:16:03] -!- abaumann has joined #archlinux-ports
[21:16:11] <abaumann> tmpfs 4052440 24080 4028360 1% /tmp
[21:16:22] <abaumann> ok. Let's mount /tmp on some big platter.
[21:26:04] <abaumann> 16K /tmp
[21:26:09] <abaumann> weird. really.
[21:26:11] -!- abaumann has quit [Quit: leaving]
[22:09:57] -!- insecta has quit [Ping timeout: 240 seconds]
[22:12:09] -!- insecta has joined #archlinux-ports
[22:23:46] -!- isacdaavid has quit [Quit: Leaving.]
[22:30:36] -!- Polichronucci has quit [Ping timeout: 256 seconds]
[22:35:30] -!- Polichronucci has joined #archlinux-ports
[23:22:25] -!- insecta has quit [Ping timeout: 248 seconds]
[23:24:30] -!- insecta has joined #archlinux-ports
[23:47:38] -!- insecta has quit [Ping timeout: 265 seconds]
[23:49:51] -!- insecta has joined #archlinux-ports