Friday, 2023-08-18

*** gnucode <gnucode!~gnucode@user/jab> has quit IRC02:54
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC05:03
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc05:03
*** yambo <yambo!> has quit IRC06:56
*** ghostmansd <ghostmansd!> has quit IRC07:13
*** yambo <yambo!> has joined #libre-soc07:15
*** yambo <yambo!> has quit IRC07:20
*** yambo <yambo!> has joined #libre-soc07:21
lkclsadoon: added08:12
lkclfor the make.conf you can just commit it to the wiki then use "[[!inline somethingsomethingfilename raw=yes]]" there are a huge number of examples08:15
lkclif it's a *small* file do it inline08:15
lkclsadoon, bear in mind that the whole purpose is to create a "reproducible build from source".  any "convenience and shortcuts by giving out binaries" is (a) a massive legal liability that is completely unacceptable and (b) means that if there is a problem and you are not around then information is lost and we waste NLnet money trying to reproduce and re-discover what you did08:22
lkclalso could someone please cut/paste the conversation with michiel into the nlnet-ongoing EUR 100,000 bugtracker task08:23
openpowerbot[irc] <lkcl> cesar[m]1, programmerjake toshywoshy ^08:23
lkclmarkos_, ghostmansd[m] ^08:24
lkcli cc'd you all08:24
lkclsadoon, who wrote the wording that you used surrounding the devscripts?08:29
lkcl"they may look confusing" i have removed08:29
lkcland i have added a hard warning "it is your legal responsibility"08:29
lkclif we have anything that stupid on the wiki - failing to warn people of the consequences - the entire project and all members are exposed to massive legal liability08:30
lkcl"i ran the scripts they deleted my system i'm suing you for damages because you failed to warn me"08:30
markos_lkcl, that does not stand actually08:31
markos_people are expected to check what they run on their systems, with or without warnings08:33
markos_that especially holds for things like scripts08:34
markos_in the unlikely scenario that someone would like to hold libresoc accountable for having their files deleted for following the scripts08:34
markos_in any case, that's text for a distribution license, ie BSD, MIT, GPL, etc, not a simple warning, as long as there is a generic license text for the whole repository, we're good, and I believe there is one08:35
lkclmarkos_, i am not taking the risk08:41
markos_also, liability damages are possible only if there is an actual contract between the parties, not for stuff that is available on the internet freely and the user *chooses* to run the software by their choice08:41
lkclthere have actually been people who have destroyed their host system by failing to follow the instructions08:41
lkclthere were only 7 commands that they needed to run and they failed to run step (2) (creating and entering the chroot)08:42
*** ghostmansd <ghostmansd!~ghostmans@> has joined #libre-soc08:48
lkclnow, that said - i am interested to hear how other projects handle this - if there is a "standard way"08:49
lkcl(that sets legal precedent).08:49
lkclmarkos_, in different jurisdictions unfortunately that's not 100% true. the U.S. has many legal precedents surrounding the right of e.g. burglars to sue for damages after injuring themselves during a theft08:51
*** ghostmansd <ghostmansd!~ghostmans@> has quit IRC08:53
markos_this is a different use case :)08:53
markos_I'm not saying to not put a wording, I'm just saying that the lack of a warning does not constitute a risk of massive legal liabilities against the members of the project :)09:11
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC09:30
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc09:30
*** octavius <octavius!> has joined #libre-soc10:07
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC10:30
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc10:49
openpowerbot[irc] <sadoon[m]1> <openpowerbot> "[irc] <lkcl> sadoon: added" <- Thanks!11:17
openpowerbot[irc] <sadoon[m]1> <openpowerbot> "[irc] <lkcl> for the make.conf..." <- Oh that's cool will edit it to use that11:17
openpowerbot[irc] <sadoon[m]1> <openpowerbot> "[irc] <lkcl> sadoon, bear in..." <- Unfortunately with Gentoo it's inescapable because of how Gentoo works11:18
openpowerbot[irc] <sadoon[m]1> <openpowerbot> "[irc] <lkcl> sadoon, who wrote..." <- It's copy-pasted from the microwatt guide but I might remove it anyway or rewrite it11:19
openpowerbot[irc] <sadoon[m]1> <sadoon[m]1> "Unfortunately with Gentoo it's..." <- Actually completely disregard what I said, I might be able to fix that issue. Give me some time11:29
openpowerbot[irc] <sadoon[m]1> Ok here goes: to make it completely reproducible, we can use a Gentoo stage3 tarball from a certain date and never change it, rebuild everything in it using our make.conf, and voila11:33
openpowerbot[irc] <sadoon[m]1> A stage3 is still binary, but not provided by us, just like a debian iso for example11:33
openpowerbot[irc] <sadoon[m]1> But that's Gentoo's official stage311:33
openpowerbot[irc] <sadoon[m]1> If you use the same stage3 with the same steps you are guaranteed to have the same code11:34
openpowerbot[irc] <sadoon[m]1> This simplifies the process while also guaranteeing reproducibility and meaning that we don't have to publish any binaries11:35
openpowerbot[irc] <sadoon[m]1> 1- download the stage3 2- add our make.conf 3- rebuild with a single line command11:36
openpowerbot[irc] <sadoon[m]1> If at any point in time we need software not available in the stage3, we pin a specific version of it11:38
*** ghostmansd <ghostmansd!> has joined #libre-soc12:06
*** ghostmansd <ghostmansd!> has quit IRC12:10
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC12:11
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc12:42
markos_sadoon, if using a gentoo stage3 tarball, might be good to mirror it as well, along with the url and checksums, just in case...12:45
openpowerbot[irc] <sadoon[m]1> Yeah12:45
openpowerbot[irc] <sadoon[m]1> They will take it off their mirrors at some point so good idea12:46
markos_with very old releases, debian puts them in the archive, and urls change12:46
markos_you can still find images for eg. potato but they're not in the original urls12:46
openpowerbot[irc] <sadoon[m]1> For debian that would be massive, or we could just mirror the chroot-necessary part of the repo only12:46
markos_no point in doing that for debian, but at some point we will need a package repo anyway12:47
markos_until it gets merged properly in debian at least12:47
*** ghostmansd[pc] <ghostmansd[pc]!> has joined #libre-soc12:54
*** ghostmansd[pc] <ghostmansd[pc]!> has quit IRC13:00
openpowerbot[irc] <sadoon[m]1> That's a totally different thing, sffs repo and eventually svp64 is needed13:03
*** ghostmansd[pc] <ghostmansd[pc]!> has joined #libre-soc13:03
openpowerbot[irc] <sadoon[m]1> Alright almost done with debian scripts13:04
openpowerbot[irc] <sadoon[m]1> They'll be in a separate branch on dev-env-scripts and I put a warning that they are WIP13:04
*** ghostmansd[pc] <ghostmansd[pc]!> has quit IRC13:13
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC13:13
*** ghostmansd[pc] <ghostmansd[pc]!> has joined #libre-soc13:13
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc13:14
*** ghostmansd[pc] <ghostmansd[pc]!> has quit IRC13:17
markos_btw, octavius toshywoshy, starting linux in verilator doesn't use more than 1GB of ram (actually much less), but I see about 40GB cached13:17
*** ghostmansd[pc] <ghostmansd[pc]!> has joined #libre-soc13:17
markos_btw didn't get to the prompt13:17
markos_reached here:13:18
markos_[    2.545271] printk: console [ttyS0] disabled13:18
markos_START_BIT error 434 30613:18
markos_                       [    2.570906] serial8250.0: ttyS0 at MMIO 0xc0002000 (irq = 16, base_baud = 3125000) is a 16550A13:18
markos_[    0.000000] printk: bootconsole [udbg0] enabled13:18
markos_and then it seemed to restart13:18
markos_after ~1h13:18
markos_is this expected/normal?13:19
openpowerbot[irc] <sadoon[m]1> You're getting close13:20
openpowerbot[irc] <sadoon[m]1> Restart you mean like the whole boot process?13:20
openpowerbot[irc] <sadoon[m]1> No that's not normal :(13:23
openpowerbot[irc] <sadoon[m]1> I pushed the debian scripts to new branch bug999, please guys don't run any of it just yet13:24
openpowerbot[irc] <sadoon[m]1> It's almost complete but I need to be back home to add the list of packages to build13:25
markos_ 41.71%  microwatt-verilator  [.] VlMTaskVertex::waitUntilUpstreamDone13:25
markos_  32.97%  microwatt-verilator  [.] VlWorkerThread::dequeWork13:25
markos_   8.72%  microwatt-verilator  [.] VlMTaskVertex::signalUpstreamDone13:25
markos_that's why it takes so long13:25
octaviusNot sure why it restarted, worked for sadoon as well13:33
octaviusAlso, I tried with Libre-SOC core, but interestingly enough it seems to reset after the message "Booting from BRAM at 0x600000", I probably didn't change an address or something like that13:34
markos_the above was profiling microwatt-verilator with perf top -p, this basically means that basically 80% of the time of execution is spent on spinlocks13:42
markos_even if running on single threaded, this is ridiculous, that's why it's so slow13:42
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC13:44
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc13:45
markos_so I propose 2 things: 1. verilator 5 is out for a while, maybe it's fixed there, 2. if not, then I suggest we fix this, for a few years I did exactly that, I optimized C software on NUMA systems with hundreds of cores and TB of RAM to scale well, we had our own spinlocks/mutex library and we optimized the hell out of everything, this is exactly such a usecase13:45
markos_doesn't even require assembly, just a better use of atomics13:47
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC13:58
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc13:59
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC14:34
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc14:35
markos_wow, just had a look and realized that for some packages we are using ancient versions, we need to move from buster to bookworm14:50
openpowerbot[irc] <sadoon[m]1> lkcl: so remember yesterday when I said the ULX hasn't shipped yet?14:51
openpowerbot[irc] <sadoon[m]1> It just did :D14:51
openpowerbot[irc] <sadoon[m]1> Looks like they finally got some back in stock14:51
openpowerbot[irc] <sadoon[m]1> As if they knew that I really need it now14:52
openpowerbot[irc] <sadoon[m]1> This is both the 12K and 85K models in one shipment, awesome!!14:53
markos_sadoon, are these FPGA boards?15:00
openpowerbot[irc] <sadoon[m]1> YES :D15:02
openpowerbot[irc] <sadoon[m]1> Even the board itself is libre15:02
openpowerbot[irc] <sadoon[m]1>
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC15:22
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc15:22
lkcl> This simplifies the process while also guaranteeing reproducibility15:37
lkclperfect, sadoon15:37
openpowerbot[irc] <sadoon[m]1> :)15:39
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC16:12
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc16:14
openpowerbot[irc] <cesar[m]1> The ECPIX-5 also seems to be back in stock:
openpowerbot[irc] <sadoon[m]1> Seems like the production bottleneck was with the FPGA itself16:37
openpowerbot[irc] <sadoon[m]1> Both use the same chip16:37
openpowerbot[irc] <sadoon[m]1> Same LUTs also16:37
openpowerbot[irc] <programmerjake> I found a patch that fixes float128 in glibc for microwatt:
openpowerbot[irc] <programmerjake> if you use that in debian/gentoo it will likely work better than trying to set long double to 64-bit16:39
openpowerbot[irc] <sadoon[m]1> Awesome!!16:40
openpowerbot[irc] <sadoon[m]1> Thanks16:41
openpowerbot[irc] <programmerjake> full commit
openpowerbot[irc] <programmerjake> all the commits for supporting microwatt
openpowerbot[irc] <sadoon[m]1> Perfect, these should be compatible with pretty much any version we need16:54
openpowerbot[irc] <sadoon[m]1> (just from a quick peek)16:54
openpowerbot[irc] <programmerjake> yay!16:54
openpowerbot[irc] <sadoon[m]1> I'm on a train but luckily my laptop already has the full source of glibc debian version, I'll make it a git repo and push it to libre-soc as well with the patches16:55
openpowerbot[irc] <sadoon[m]1> So I might lose internet but I'm working on it, thanks for sending this again!16:56
openpowerbot[irc] <programmerjake> k, have a safe trip!16:56
openpowerbot[irc] <sadoon[m]1> thanks16:57
openpowerbot[irc] <sadoon[m]1> :)16:57
openpowerbot[irc] <programmerjake> found the patches by reading microwatt's readme where it says they have a patched buildroot for microwatt...seems an obvious place to look in retrospecr16:57
openpowerbot[irc] <programmerjake> retrospect*16:57
openpowerbot[irc] <sadoon[m]1> With this, octavius recent work on simulation, and my FPGA shipping just in time, we might be ready to run Linux very soon I hope16:58
openpowerbot[irc] <programmerjake> yay!16:59
openpowerbot[irc] <sadoon[m]1> I did notice microwatt is missing some stuff in the cflags that we did add yet their system works just fine17:28
openpowerbot[irc] <sadoon[m]1> They don't have -mno-htm -mno-crypto in glibc for example17:29
openpowerbot[irc] <programmerjake> well, iirc crypto requires vmx/ -mno-altivec should disable it, we're passing the flag to be safe i guess17:30
openpowerbot[irc] <sadoon[m]1> ah17:30
openpowerbot[irc] <sadoon[m]1> and htm?17:30
openpowerbot[irc] <programmerjake> htm is deprecated, so glibc probably doesn't use it17:30
openpowerbot[irc] <sadoon[m]1> Interesting17:31
openpowerbot[irc] <sadoon[m]1> I don't have permission to create a new repo, can someone create a glibc repo so I can push it? I applied the patches but I can't test them just yet so keep that in mind17:37
openpowerbot[irc] <sadoon[m]1> Hmm actually I could test this stuff on my POWER8 server even so hehehe17:38
openpowerbot[irc] <programmerjake> @lkcl17:38
openpowerbot[irc] <sadoon[m]1> Time to put my money where my mouth is and test my scripts on my own server hheheheeh17:43
octaviussadoon, making some pretty swift progress, good work!17:54
openpowerbot[irc] <sadoon[m]1> Thanks :D17:54
openpowerbot[irc] <sadoon[m]1> I'm logged into trusty old Tyan to rebuild glibc and see how things go, hope I don't lose internet but I'm using tmux so it doesn't completely fail if I disconnect17:56
openpowerbot[irc] <programmerjake> if you use mosh it starts working again whenever you reconnect18:05
openpowerbot[irc] <programmerjake> though you do lose your scrollback buffer18:05
openpowerbot[irc] <sadoon[m]1> tmux also works well that way18:07
markos_tmux has the benefit of splitting the terminal in multiple windows18:08
markos_so if you have elaborate sessions with monitoring, compilation, testing, etc, tmux is indispensable18:09
octaviusFor some reason having bash in chroot (in tmux) made my arrow keys stop working (they became control chars instead)19:10
markos_wrong terminal not passed correctly from schroot to tmux, just upgrade tmux19:17
markos_I had that exactly, was fixed with an upgrade19:17
octaviususing tmux in a debian 10 chroot19:52
octaviusI guess I'd need to manually upgrade tmux19:53
openpowerbot[irc] <sadoon[m]1> octavius I don't think you need to upgrade19:53
openpowerbot[irc] <sadoon[m]1> Just need a good tmux.conf19:53
openpowerbot[irc] <sadoon[m]1> I'll send you mine if you want19:53
octaviusSure, that'd be nice!20:00
octaviusMaking configs is an art in itself20:00
openpowerbot[irc] <sadoon[m]1> Alright :)20:15
openpowerbot[irc] <sadoon[m]1> Please read it thoroughly before applying though20:15
openpowerbot[irc] <sadoon[m]1>
openpowerbot[irc] <sadoon[m]1> It's based on someone else's but I changed too many things, can't remember who exactly20:17
openpowerbot[irc] <sadoon[m]1> The specific line is the xterm-256colors I think that might fix your problem20:18
lkclULX3S on the way? hooray! and the reason i say "hooray" is because it has SDRAM. no bullshit DDR3 training20:35
lkcli use minicom20:35
lkcland the only thing you need to make sure of is *not* to use rts or cts. which is the default.20:36
lkclout-of-the-box minicom "just works"20:36
lkclsadoon i'll make a git repo "libc6"... done20:38
lkclit's empty20:38
lkclprogrammerjake: doh, well spotted20:40
openpowerbot[irc] <sadoon[m]1> Debian uses glibc so I'd prefer that but either works20:40
openpowerbot[irc] <sadoon[m]1> As a name20:40
openpowerbot[irc] <sadoon[m]1> Yes the RAM is a big bonus but it's only 32MB, you think it'll be enough?20:40
openpowerbot[irc] <sadoon[m]1> (Desolder and solder on 128MB for fun and profit)20:42
openpowerbot[irc] <lkcl> sadoon[m]1, programmerjake,;a=summary20:44
openpowerbot[irc] <lkcl> ok 1 sec...20:44
openpowerbot[irc] <lkcl> done20:45
openpowerbot[irc] <lkcl>;a=summary20:45
openpowerbot[irc] <sadoon[m]1> Awesome, might be a little busy tonight but I'll upload as soon as I can20:46
openpowerbot[irc] <lkcl> no, nobody is authorised to create arbitrary untrusted repositories, i make sure to explain to people that repositories are never to contain binaries or other auto-generatable output... and *then* create the repository for them20:47
openpowerbot[irc] <lkcl> plus there's only a 20GB allocation of the sponsored hosting from mythic-beasts, and 12GB of it is already used20:48
openpowerbot[irc] <sadoon[m]1> I think glibc is about 20MB or something compressed which is actually very shocking21:29
openpowerbot[irc] <sadoon[m]1> (Also I think I'm missing context on the untrusted repos thing, is the bridge acting up again?)21:30
*** octavius <octavius!> has quit IRC21:54
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC22:17
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc22:17
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC23:19
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc23:19

Generated by 2.17.1 by Marius Gedminas - find it at!