IRC channel logs

2018-10-29.log

back to list of logs

<joshuaBPMan>Hello, I'm playing around with guile-charting. To get it to work do I need to install guile-charting and it's dependancies: cairo guile-cairo? Or is guix smart enough to pull in the dependancies?
<atw>joshuaBPMan: it should be sufficient to install the package guile-charting
<joshuaBPMan>hmmm. it doesn't seem to be working
<joshuaBPMan>but that's what I thought
<atw>can you elaborate on how it's not working?
<joshuaBPMan>sure. Sorry. https://wingolog.org/projects/guile-charting/doc/charting/
<joshuaBPMan>There are some examples on that webpage
<joshuaBPMan>the scatter plot fails with this error:
<joshuaBPMan>ERROR: In procedure cairo-image-surface-create: Wrong type (expecting exact integer): 531.0
<joshuaBPMan>
<joshuaBPMan>I'm not sure why.
<atw>Could you share your invocation of make-scatter-plot?
<joshuaBPMan>It's the last example on this page: https://wingolog.org/projects/guile-charting/doc/charting/
<atw>joshuaBPMan: ah, sorry, I misunderstood. I can reproduce the error exactly, seems like a bug. I can't find a bug tracker for guile-charting, so I guess you should email Andy Wingo
<joshuaBPMan>atw: thanks! I'll do that!
<joshuaBPMan>he's said he may not be the best maintainer, so we'll see.
***janneke_ is now known as janneke
<g_bor>hello guix!
<tune>cannot build derivation `/gnu/store/gxzr0fm15n283fr481356qcrrds762ny-vinagre-3.22.0.drv': 1 dependencies couldn't be built
<tune>guix package: error: build failed: build of `/gnu/store/gxzr0fm15n283fr481356qcrrds762ny-vinagre-3.22.0.drv' failed
<g_bor>Hello tune!
<g_bor>The output you pasted shows that a dependency could not be build.
<g_bor>Which dependency was that?
<g_bor>You should see it earlier in the output.
<g_bor>That might alos give some clues, on how to resolve the issue.
<g_bor>s/alos/also
<roptat>hi guix!
<g_bor>hello roptat!
<g_bor>Any news on hydra?
<g_bor>From here, it still looks down...
<roptat>I was able to compile an empty class with my scala compiler :D
<roptat>I can't compile anything more complicated than that yet ^^
<g_bor>Wow, that's amazing :)
<g_bor>Can we somehow get involved?
<g_bor>What is missing?
<roptat>well everything else :p
<roptat>my parser is mostly done, except I can't parse every expression and I miss a few things in class declarations
<roptat>I have a small bytecode generator, but I miss every phase in between
<roptat>now I'm working on resolving imports
<g_bor>can we have a look at the work?
<roptat>not right now, sorry
<roptat>I'll give you a link this evening
<g_bor>roptat: after finding a stupid mistake of mine, icedtea reproducibility is getting into good shape.
<roptat>great!
<ng0>i need a short reminder. unless otherwise defined, mcron2 jobs runs as root, right?
<roptat>that reminds me I should push my openjdk packages
<g_bor>I hope to get the diffoscope output down to the size where it does not eat up all my memory :)
<roptat>wow, is it that bad?
<g_bor> https://www.gnu.org/software/mcron/manual/html_node/Guile-Syntax.html#Guile-Syntax
<g_bor>the top paragraph says, that all jobs not having user are run as the configuration_user global variable, which should be root as default on GuixSD I guess.
<ng0>except that this does not answer the question nor does our documentation. the "should" part of your reply is also what I assume, but having it written somewhere would be better
<ng0>"... defaults to running as root" or something
<ng0>since bitlbee is not invoked with a user and it runs as root, I think (don't want to read the code right now) that root is our default user (for mcron)
<roptat>where is mcron-job defined?
<ng0>wrong service. bitlbee has a user.
<ng0>okay, I think I can remind myself by just starting another server
<roptat>mcron documentation: “Mcron should be run by the user who wants to schedule his jobs”
<roptat>and in guix, the mcron service is run as root
<roptat>so I'd say by default the user is root
<ng0>yes. I asked because it is not obvious if you don't work with some of the services all the time.
<ng0>thanks
<roptat>does that mean as a user you can make mcron execute jobs as root?
<snape>is there a way to see https://bayfront.guixsd.org/.well-known/logs/ in the browser instead of having to download it?
<sneek>snape, you have 1 message.
<sneek>snape, rekado says: I’m still traveling and can only take a closer look at Berlin on Sunday.
<roptat>snape: I think it's server-side
<roptat>the mime-type it returns is "application/octet-stream", so the browser will only download that
<roptat>if it returned "text/plain" the browser would have opened it directly I think
<g_bor>I don't think you can run jobs as root.
<roptat>g_bor: ok, I guess you run with the user priviledges so it just fails
<g_bor>I also don't see configuration-user variable exposed, if I get it right that provides a default user.
<g_bor>roptat: most probably yes.
<snape>roptat: yes, too bad, thanks
<snape>rekado: Berlin returns 502
***specing_ is now known as specing
<kmicu>Hi folks, did you try to setup GuixSD on a “legacy BIOS” with a seperate GPT boot parition? I assume my BIOS doesn’t support GPT cuz it gives me ‘GRUB _’ prompt and nothing else but that could be also a misconfiguration. 🤷
<roptat>kmicu: you mean with two hdd?
<roptat>or one hdd with GPT partitions, one of which is your boot partition?
<kmicu>Hi roptat, one disk. Two partitions /dev/sda1 with boot flag and /dev/sda2 for root. But /dev/sda1 is GPT not MBR.
<roptat>you mean the disk, right?
<roptat>(GPT and MBR are partitionning schemes)
<roptat>I think you can't use GPT with a legacy bios
<kmicu>Usually you use BIOS/MBR, but I wanted to try GRUB plus a seperate BIOS boot partition.
<roptat>but we have support for EFI
<kmicu>You can, this is why I’m asking. But some older BIOSes cannot.
<nckx>kmicu: What you're trying to do is pretty standard & should work. I did it for years & 've never heard of it not working, but of course the "B" in BIOS stands for buggy...
<nckx>So yeah, no way we can know if your BIOS is substandard or not ☹
<kmicu>I already have EFI based GuixSD, now I want to revive my old laptop.
<kmicu>nckx: this is exactly why I’m asking. So you had a legay BIOS with a seperate GPT boot partition?
<nckx>kmicu: Sure. On 4 (different) machines.
<roptat>ah, so I guess I'm wrong then :)
<kmicu>nckx: thank you very much. That will greately simply troubleshooting.
<nckx>roptat: I'm sure there are buggy BIOSes that can't. I've never encountered one, though. I'd not give up hope on the fixable software side yet, kmicu.
<nckx>What does GRUB say when installing?
<kmicu>nckx: there was no examples for a GRUB, BIOS/GPT partitioning scheme in the docs. So it was better to ask.
<kmicu>Currently it says ‘GRUB _’ :)
<nckx>Note that I might have installed GRUB (to the BIOS boot partition) manually back then, my memory is legendarily bad. So try that too. You should only need to do that once.
<nckx>Hehe.
<nckx>During installation, I mean.
<kmicu>Installation like 'guix system init' step before a reboot?
<kmicu>(There was a message at the end: everything is ok, no errors. So I did not read the log.)
<nckx>Can't fault you for that. But maybe Guix is invoking GRUB slightly wrong for your use case.
<kmicu>Time to investigate. Though after happily using EFI/GuixSD/NixOS I totally forgot how to debug GRUB :P
<kmicu>How do you define <file-system> objects for ‘dependencies’ entry? Can I do it in place?
<tune>04:56:24 < g_bor> Which dependency was that?
<tune>it might be spice
<tune> https://my.mixtape.moe/ysubes.png here's a screenshot of a bit more output
<roptat>tune: ideally you would use a paste service (paste.debian.net for instance) to copy text: it's lighter and much easier to read for screen readers ;)
<roptat>(well maybe a monochrome png image is not that big ^^)
<roptat>tune: under the red line, there is a line telling you that the log is somewhere
<roptat>can you run "bzless <that file>" then search for "error"?
<roptat>tune: I can't reproduce your issue with a recent guix, what version are you using?
<tune>ah sorry. one moment
<ng0>whoever introduced changes to the UI should run more tests with it. I don't mean it in a negative way, but some build (no idea which one, I'm in a long series of builds) just broke the output in a way that I'm only getting gibberish for the renmaining tail of buildlogs. iirc vim builds (specifically: the testsuite) would do something similar to your open terminal.
<ng0>"but what do I mean by gibberish?"
<ng0>well:
<ng0>⎻▒▒⎽e ◆b┤☃┌d' ⎽┤cceeded ▒°├e⎼ 1↓3 ⎽ec⎺┼d⎽
<ng0>├e⎽├/e┼±☃┼e/├e⎽├_⎻⎺⎺┌↓⎻≤::P⎺⎺┌E┴e┼├⎽Te⎽├::├e⎽├_c⎺┼┼ec├_⎺┼_⎽┤b⎽e─┤e┼├┌≤_⎼ec⎼e▒├ed PASSED [ 2▮%]
<ng0>etc
<roptat>oh yes, vim tests :/
<roptat>is that with guix build or guix package?
<ng0>build
<roptat>I haven't seen anything like that before and I'm not familia with UI code
<ng0>wiht building vim you'd get no problems a while back. it would just continue building and only when builds finish make your terminal act funny
<roptat>can you report it?
<ng0>no time to do so.
<ng0>maybe in a couple of days
<snape>it's also with guix package
<snape>ng0: one bug report email is quicker to do than the whole discussion you just had :)
<ng0>I'm tired and busy elsewhere
<ng0>what I just pasted in was a sideeffect of 'busy elsewhere'
<snape>alright, I did the report
***sturm is now known as Guest58476
<tune>roptat: sorry for the delay, I was away from my computer and it was annoying to do some of it over ssh without a mouse. here's my guix version and some hopefully helpful info from that log: https://paste.debian.net/1049582/
<tune> https://paste.debian.net/1049583/ oh and this was the bottom part of that log, it contains 'error' but not sure if it's helpful or not
<roptat>tune: oh, that's the latest revision... I'll guix pull to that and try again, but it might be only a transient error (it happens with tests). Maybe it'll just work if you try again?
<roptat>the last paste is not useful, it just means that the command that was run didn't succeed (returned != 0)
<kmicu>Installing GuixSD when substitutes are not on build farms was a mistake ;P
<kmicu>(Or I’ve mistyped berlin pubkey.)
<snape>kmicu: no, they aren't available indeed
<kmicu>snape: should I execute 'guix pull' before 'guix system init' to get substitutes? What do you think.
<snape>kmicu: I don't know if there is a build farm that has built the latest commits yet
<kmicu>I only want to buil bare-bones.scm.
<snape>so either you have a powerful machine that can build them, or you can clone the Guix git repository, and checkout an older commit
<kmicu>Aka I don’t want to buil gcc 5.5.0 after each mistake during guix system init.
<snape>without 'guix pull', your Guix will be too old, and with 'guix pull' it will be too recent
<snape>but I'd 'guix pull', still
<kmicu>To clone I would need to install git, to instlal git I’d need to execute guix command which doesn’t have substitutes :P 🤷
<snape>right ;-)
<snape>so... guix pull, choose something small (bare-bones.scm, yes) and be patient :p
<kmicu>Do FSF admins bring Hydra with old cache? Maybe I should just wait.
<snape>there is a mirror with cache I believe (mirror.hydra.gnu.org), but I'm not sure how well it works
<snape>I don't think it works well without the real one
<kmicu>There is no harm to check with a dry-run (after I finish compiling gcc 5.5.0 for nth time ;P).
<kmicu>Fun fact: lightweight-desktop.scm example uses Qt and it takes over 12h to compile it on a legacy hardware. Don’t ask me how I know that 😺
<roptat>kmicu: try with "guix pull --commit=21c51ebd66faa0b47f0be223d97c3a371346f54b" it's from last friday and should have more substitute available than today's guix
<kmicu>Thank you roptat. That’s very helpful.
***jonjitsu[m]1 is now known as jonjitsu
<snape>indeed, I had forgotten about this
<kmicu>Heh, thank you folks. I’ve made a typo in berlin’s pubkey 😹
<g_bor>hello guix!
<g_bor>I need some assistance with icedtea reproducibility.
<g_bor>One of the differing files exhausts my memory when trying to compare using diffoscope.
<g_bor>It is ...icedtea-1.13.13-jdk/jre/lib/amd64/server/libjvm.so
<g_bor>Could someone have a look, and possibly send me back the diffoscope output?
<janneke>g_bor: have you tried finding the link command: *.o -> libjvm.so and run cmp/diff on those?
<g_bor>janneke: not yet. Thanks for the idea. Will have a look.
<g_bor>Now I have only 4 files differing, and two of them are timestamp differences only, one has a timestamp difference, and something else (maybe fixing the timestamp is enough, if it is a checksum) and this libjvm.so :)
<g_bor>all in all it is starting to look good :)
<bzp>hi guys
<bzp>I'm trying to install the base system, how do I solve this problem?
<bzp>or what is due?
<bzp> http://pasteall.org/pic/show.php?id=0b8da017283d02cdd9e70a05e9858d4a
***daviid is now known as Guest29881
***Guest29881 is now known as daviid
<roptat>bzp: hydra (our substitute server) is offline for maintenance, while berlin (our second substitute server) doesn't have substitutes for guix 0.15, nor for the last revision
<roptat>bzp: that means you'll need either to run with "--fallback" (building everythnig locally) or use "guix pull" firt to use a more recent revision that berlin has substitutes for
<roptat>but hm... I'm also getting 504 and using --fallback only leads guix to try and download things in a loop...
<kmicu>Before restarting I’ve checked grub.cfg and there there is ‘set root to the paritition that contains /gnu/store’. And my /gnu/store is not on the same subvolume as root fs. Let’s hope for the best and restart anyway.
<kmicu>bzp: when you set berlin.guixsd.org during 'guix system init --substitute-urls="https://berlin.guixsd.org" /mnt/etc/system.scm /mnt’ then you can get a bare-bones system with only a little compiling (e.g. texlive tiny or FontForge).
<kmicu>(You need to authorize berlin via guix authorize --archive < /path/to/berlin.pubkey before trying that though. Generally I think it’s better to postpone any installations.)
<bzp>when is the 'hydra' server available?
<kmicu>(FWIW https://raw.githubusercontent.com/guix-mirror/guix/master/berlin.guixsd.org.pub)
<kmicu>bzp: according to https://lists.gnu.org/archive/html/guix-devel/2018-10/msg00434.html should be ready soon.
<g_bor>bzp: it down due to maintenance, we thought it will be back at some time in the weekend, but apperantly it is still not in good shape
<kmicu>Don’t see any new updates at https://quitter.im/fsfstatus
<bzp>What steps should I take to be able to install guixsd?
<bzp>There are no mirrors, like other distributions?
<roptat>well, we have mirror.hydra.gnu.org, but I think it's down too...
<Laalf>bzp: there are other substitute servers. berlin.guixsd.org for example
<bzp>How do I add mirrors to continue the installation of guix?
<tune>roptat: my updates went through this time, so just trying it again worked
<g_bor>bzp: my last info is that berlin.guixsd.org works.
<roptat>bzp: we don't have other mirrors (hydra, mirror.hydra and berlin), but you should be able to finish the installation by building a few packages locally
<roptat>it's transparent
<kmicu>Laalf: berlin doesn’t have many substitutes for 0.15 ISO.
<roptat>bzp: you only have to pass --fallback as it's suggesting you to :)
<kmicu>(Even compiling basic ssh system w/o Xorg took some hours.)
<roptat>yeah, but at least you get the same bit-for-bit identical system as if you had substitutes :p
<g_bor>anyone knows what happened to berlin earlier?
<roptat>it seems it doesn't build substitutes anymore
<g_bor>at 18:45 snape wrote to the ml that berlin now works.
<kmicu>Iirc from the mailing list there were evaluations errors (civodul requested to use linters to prevent that) and now berlin’s operation is resumed.
<roptat>g_bor: oh cool :)
<kmicu>Could you help me with this riddle: you are after guix system init, but with broken grub. How to reconfigure the system from livecd?
<kmicu>init takes /mnt path. But I don’t see anything like that for reconfigure.
<roptat>kmicu: you should be able to use init again (it will replace your first generation instead of adding a new one)
<g_bor>roptat: I have a problematic file in icedtea-6, it is icedtea-1.13.13-jdk/jre/lib/amd64/server/libjvm.so
<roptat>g_bor: can you send me your patches? I'll try build it and run diffoscope
<g_bor>It is not reproducible, but diffoscope ate all my ram trying to find the difference.
<g_bor>I will send them, but I will rebase first to current staging. Is it ok to push them to a gitlab repo?
<roptat>sure
<kmicu>roptat: for me each time init does not reuse deriviations from /mnt/gnu/store.
<g_bor>roptat: https://gitlab.com/g_bor/guix/tree/fix-icedtea6-javadoc
<g_bor>there are still other differing files, I guess all of them have only timestamp issues.
<g_bor>I can run diffoscope on them just fine, but this libjvm.so is just too much :)
<kmicu>Could you paste your generated grub.cfg somewhere please? Especially if you are a legacy BIOS user.
<roptat>g_bor: that's on staging, right?
<roptat>nvm, I'll just clone your repo :)
<g_bor>roptat: ok, thanks
<roptat>g_bor: https://framagit.org/tyreunom/scabo
<rekado>ng0: I’d prefer if you didn’t blame an error on the work of other people prematurely.
<rekado>it’s more helpful to report a bug instead of assigning blame on IRC.
<kmicu>Any folks with grub.cfg for legacy BIOS around? ^^
<kmicu>(I don’t find any pointers to the boot partition in mine so I wanted to compare them.)
<roptat>g_bor: first build down, now checking
<roptat>kmicu: I think we recommend using the same partition for / and /boot
<roptat>kmicu: but I don't think it needs any pointer to your /boot
<kmicu>Yeah, in case of MBR the same parition is ok, but GPT needs a seperate /boot.
<nckx>kmicu: Do you mean /boot/efi?
<nckx>Oh no this is still BIOS.
<nckx>Then roptat is correct (well, it's more than 'recommend': separate /boot is not supported.)
<kmicu>There is that ‘set root to the paritition that contains /gnu/store’ comment in grub.cfg but there is nothing related to it.
<kmicu>nckx: oh, so you use legacyBIOS+GPT without a seperate boot parition?
<nckx>kmicu: *used
<nckx>Yes.
<kmicu>Thank you. That couled be my mistake from the beginning.
<nckx>Separate /boot is unnecessary.
<nckx>That was already true of most of my non-GuixSD systems too.
<kmicu>Unnecessary < not possible. Arch wiki has ‘On a BIOS/GPT configuration, a BIOS boot partition is required.’ so dunno.
<nckx>kmicu: Read that again ;)
<nckx> /boot != boot.
<kmicu>I’m sorry but I don’s understand the difference.
<nckx>The 'BIOS boot partition', confusingly, has nothing to do with /boot. It doesn't even contain a file system. It's a small (~1MiB) slab of disk for GRUB to store some raw data.
<nckx>That data used to fit in some unused space on MBR drives, but that same space was chosen to store the GPT partition table.
<kmicu>But later wiki explains to create ~mebibyte parition with BIOS boot type.
<nckx>Yep. That's the one.
<nckx>That's all you need to do. Create it. GRUB will use it.
<kmicu>So the point is that a separate partition is requries e.g. /dev/sda1 but I should not point to it in (bootloader …) config?
<nckx>*use it at installation time. There's nothing to configure.
<nckx>Yep!
<nckx>I usually used /dev/sda9 for it to keep it away from the 'real' partitions, since it's useless, but that's a matter of taste.
<nckx>(/dev/sda1 is always / in my world.)
*nckx must go. People need me.
<kmicu>Thank you nckx. ʕノ•ᴥ•ʔノ ︵ λ𝛌𝚲𝝀
<roptat>what's that? ^^
<kmicu>A bear throwing lambdas. And we all know that bears are better than BattleStar Galactica.
<pkill9>False. Bears eat beets.
<g_bor>roptat: thanks for checking, and also thanks for sharing scabo.
<roptat>I lack references to something I think... nevermind :)
<roptat>g_bor: there one empty class in the scala source, that's the only thing it builds right now ^^
<roptat>ok, icedtea built a second time now
<roptat>running diffoscope
<kmicu>roptat: it’s the reference to The Office episode https://youtu.be/WaaANll8h18?t=30 and I assume pkill9 is familiar with the topic 😺
<roptat>g_bor: OSError: [Errno 12] Cannot allocate memory
<roptat>:/ I have 16 GB of RAM though...
<roptat>what's going on?
<g_bor>roptat: Just the same here, but I only had 8G, so I really don't know any more :)
<g_bor>I will return to that tomorrow.
<g_bor>Good night!
<roptat>thanks, good night :)
<roptat>g_bor: I found the issue!
<roptat>I ran hexdump on both versions then diff'ed that
<roptat> https://paste.debian.net/1049625/
<roptat>g_bor: that's a timestamp :D
***ali is now known as Guest63633
<nckx>kmicu: Was that... rage?