IRC channel logs

2018-04-21.log

back to list of logs

<pkill9>is it possible to tell the 'wrap-program' function to not resolve symlinks?
<pkill9>specifically, i want the wrapper to execute /gnu/store/<package>/bin/.program-real
<pkill9>and .program-real is a symlink to /gnu/store/<package>/opt/bundle/binary
<pkill9>actually nvm
<pkill9>ignore my bikeshedding. ok a different question, how do i fix the error of the wrap-program function generating a script which a shebang of '#!#f' when using the trivial-build-system? I assume the #f is the guile 'false' value, i assume because it's not finding bash as an input, but it still creates that shebang when I provide bash as an input.
<pkill9>well, provide it in the 'inputs'
<pkill9>do I need to specify it as an input somewhere in the trivial-build script?
<buenouanq>interesting
<buenouanq>that would mean that the shell path value isn't set or set to #f
<pkill9>buenouanq: how do i set the shell path value?
<pkill9>ah found out how
<pkill9>it didn't work, I used (setenv "SHELL" "/bin/sh")
<buenouanq>I have no idea.
<buenouanq>sorry
<pkill9>i got it working =]
<pkill9>I set PATH to <bash package>/bin with setenv and it added it to the shebang
<siraben>I just realized using the substitute server https://berlin.guixsd.org made download speeds so much faster.
<siraben>From 70 KiB/s to > 1.0 MiB/s
<siraben>This is from South East Asia
***lo_mlatu_ is now known as lo_mlatu
<rekado>FYI: I have a bunch of R related upgrades that I’ll push tonight.
<pkill9>my ethernet connection seems to be unable to get a connection after waking up the laptop from suspend
***dddddd_ is now known as dddddd
<zybell_>pkill9:dhcp?
<pkill9>i tried running `sudo dhclient enp6s0 -v` and it just sends that packet to 255.255.255.255 and doesn't recieve anything
<zybell_>1packet? should be 2
<pkill9>i'll test it later
<pkill9>and upload the output to a pastebin
<zybell_>tcpdump
<apteryx_>Hello Guix!
<apteryx_>'guix package -u .' always creates a new generatinon, even when run consecutively. Is this a bug or expected?
<Sleep_Walker>apteryx_: is that happening for you after `guix pull`?
<Sleep_Walker>(IIRC there was bug which should be fixed, let me find it)
<Sleep_Walker>hm, I can't find it
<Sleep_Walker>but either you or someone here on IRC reported something similar
<apteryx_>Sleep_Walker: yes I read some mail once about someone having that problem. I don't remember the answer on the thread and can't seem to find it anymore.
<apteryx_>Sleep_Walker: yes I'm on a freshly pulled guix from yesterday.
<apteryx_>Sleep_Walker: maybe bug 29244 (guix update -u always wants to update: python-wrapper & python-ipython)
<apteryx_>looks like when one of your packages have propagated-inputs, this package always get updated. This is a simplification of the current update mechanism that could be fixed.
<Sleep_Walker>aha
<Sleep_Walker>makes sense
<janneke>it seems that gitlab's `.../repository/archive.tar.gz?ref=' download feature does not work anymore
<janneke>all i get is master, can someone confirm or share more information?
<apteryx_>Works here: https://gitlab.com/antora/antora/-/archive/master/antora-master.tar.gz
<apteryx_>but maybe they've added the branch name in the URL?
<janneke>apteryx_: yes, this new branch scheme download works
<janneke>we don't use this '/-/' anywhere, we use the (old?) ?ref= scheme
<apteryx_>the old scheme seems to work here: https://gitlab.com/daym/pyodbc-c/repository/archive.tar.gz?ref=v3.1.4
<janneke>apteryx_: in PKG-INFO it says: Version: 3.0.7
<janneke>how can I I verify that I'm getting 3.1.4?
<janneke>did you verify?
<zybell_>you can verify that/if you dont. dl with different refs should give different md5
<zybell_>(no need for sha here)
<janneke>zybell_: for my projects, i'm getting master; ?ref= is being ignored
<janneke>zybell_: tnx... apteryx_: try https://gitlab.com/daym/pyodbc-c/repository/archive.tar.gz?ref=v3.1.4 and https://gitlab.com/daym/pyodbc-c/repository/archive.tar.gz?ref=v3.1.3 https://gitlab.com/daym/pyodbc-c/repository/archive.tar.gz?ref=master; they're identical
<Sleep_Walker>mbakke: I'm a bit confused, the problem I had was missing PYTHONPATH in env - python-six worked as input correctly, it wasn't needed as propaated-input
<Sleep_Walker>why did you thought it is required?
<Sleep_Walker>(python-onetimepass package)
<mbakke>Sleep_Walker: oh okay, I guess onetimepass is an "end user" package, and not a library?
<Sleep_Walker>mbakke: it is library
<mbakke>Then six needs to be propagated, so that it ends up in profiles along with onetimepass. Otherwise it won't be found.
<mbakke>I guesd the reason it works now is because some other dependency already propagates six?
<mbakke>Try e.g. `guix environment -C --ad-hoc python-onetimepass python coreutils` and then `echo $PYTHONPATH`. If six is not propagated it won't be present and thus not found at runtime.
<mbakke>Unfortunately we don't have a better way to record Python dependencies than propagation and PYTHONPATH atm.
<Sleep_Walker>mbakke: thanks for explanation
<Sleep_Walker>mbakke: it shouldn't be hard to generate such dependencies programatically
<mbakke>Sleep_Walker: there's a lengthy discussion about it on guix-devel recently if you'd like to get your hands dirty.
<mbakke>The gist of it is that yes, you can generate .pth files with paths to all inputs, but if libary A and library B depends on different versions of library C weird things will happen. IIRC.
<Sleep_Walker>that sounds like another package relationship
<Sleep_Walker>anyway, thanks
<Sleep_Walker>I'll push the fix and read the thread
<OriansJ>Just a stupid idea but what if we forbid all sources that are not in version control. Thus when updating we need only pull the diffs and checkout the desired build source in a deterministic fashion
<efraim>load /gnu/store/..foo-1.0-checkout into /tmp and 'git checkout 2.0' and re-store in the store?
<OriansJ>efraim: perhaps we should have a standard like /gnu/source/foo be the default source repo for foo and have updates simply pull from version control into the folder and checkout the corresponding code
<soundtoxin> https://hastebin.com/iveqikiseg.scala anyone else getting a build failure like this? also, not sure if related, but I can't run urxvt since last reboot due to some libperl error. I went to run updates in case that fixed it, but then the updates couldn't finish.
<efraim>i'm not seeing anything in the paste
<soundtoxin>as in nothing of value or it's blank?
<efraim>Blank
<soundtoxin>if you've got noscript or something, that could be why. where should I paste instead?
<efraim>I see text in it now
<soundtoxin>ah okay
<efraim>It showed up without any pasted text before in icecat and links
<efraim>This is after the 'pretend Perl 5.26.2 is really 5.26.1' patch?
<soundtoxin>I don't know. Can I check somehow? I sorta just run updates every now and then.
<soundtoxin>I don't actually use perl, but doing 'perl --version' says perl: command not found
<pkill9>does 'https://mirror.hydra.gnu.org/guix/nar/gzip/65ph2ikqkjsw3qjrxqbk5dsjfi2pvpqc-docbook-xml-4.3...' fail to downlaod for anyone else?
<pkill9>it's just waiting on there
<pkill9>oh nevermind it worked now
<soundtoxin>I think that one was just giving me trouble
<pkill9>now udisks is giving me trouble
<pkill9>i think the server's having some troubles
<pkill9>oh it downloaded
<pkill9>not sure if the scrollbar has some troubles either
<pkill9>oh well
<matijja>hello everyone
<matijja>when i try to reconfigure system with "guix system reconfigure -- /etc/config.scm" returns me error: guix system: error: stat: No such file or directory: "system"
<matijja>what am i doing wrong?
<pkill9>does it work if you remove the '--'?
<matijja>same error
<thomassgn>Anyone of you have freefall fail to build? I keeo getting 'make: cc: Command not found'...
<thomassgn>when it starts phase build. Makes sense, but I can't see why... cc should be part of gnu-build-system and whotnot shouldn't it?
<thomassgn>This is with the 4.16 kernel that was recently introduced. I mean freefall uses it (the sources it seems). Haven't had a problem with it before now.
<matijja>btw: i can run reconfigure as normal user, but it gives me permission error
<thomassgn>matijja: that is expected, reconfigure changes your system - which only root should be allowed to. But you can build a system as user. I often do this, then I only need sudo when I actually reconfigure...
<thomassgn>try 'guix system build /etc/config.scm'
<matijja>build pass without errors
<thomassgn>nice
<thomassgn>if you try 'guix system reconfigure /etc/config.scm' now?
<thomassgn>same?
<thomassgn>also, do you use sudo or actually somehow login to the root account?
<matijja>same error
<thomassgn>(with su or similar I mean)
<matijja>i run it with sudo
<thomassgn>ok. Shouldn't make a difference, but maybe you can get a different kind of error if you first run 'sudo -s' and the try the reconfigure without sudo
<thomassgn>other things I've seen in the past that may interfere is if you're running something else than bash as your shell.
<matijja>i try with sudo -s in bash shell but result was the same
<soundtoxin>update: I used the --fallback option and ran updates again and they completed this time
<soundtoxin>oh and urxvt works now!
<pkill9>i'm getting this error in weechat when opening it: 17:38 =!= script: error downloading list of scripts: curl error 60 (server certificate verification failed. CAfile: none CRLfile: none) (URL: "http://weechat.org/files/plugins.xml.gz")
<pkill9>it's obviously unable to find certificates file
<pkill9>anyone else get this error?
<pkill9>sorry haven't checked the bug reports, should ahcecked
<thomassgn>You could also try 'strace -T -s 2000 -o /tmp/guix-reconf-strace.out -f guix system reconfigure /etc/config.scm' then inspect the output in the file /tmp/guix-reconf-strace.out
<thomassgn>maybe poste it to a pastebin or so if you like matijja ^^
<pkill9>yes it has been reported, nvm
<thomassgn>where is the cc program? it's not in gcc, which makes no sense to me...
<thomassgn>maybe it never was in gcc?
<rain1>I think cc usually points to gcc
<thomassgn>m, makes sense.
<pkill9>thomassgn: try the gcc-toolchain package
<pkill9>i *think* I had the same issue a while back
<thomassgn>ah, that's a package. Wonder what the difference is, I mean what is extra. Will be checked soon... :P
<pkill9>then again, i think for that particular issue i had to specify BUILDCC=gcc and/or CC=gcc as a buidl flag
<pkill9>build*
<thomassgn>right
<matijja>thomassgn: https://www.dropbox.com/s/6yow7pdp7ig90ix/guix-reconf-strace.out?dl=0
<matijja>better link: https://drive.google.com/open?id=1FfBIeU3PgQH8jJy2bXxiYY-QUvqe7_L5
<thomassgn>that's weird. if you run any other guix command as root, do you get a similar error? eg. 'guix build hello'?
<thomassgn>matijja: ^^ also, I think http://paste.lisp.org/ or https://paste.pound-python.org/ are much better.
<matijja>build hello works. Event system build /etc/config.scm works as root
<thomassgn>lol. wtf.
<thomassgn>idk. you are on guixsd?
<thomassgn>or are you on a different distribution?
<matijja>guixsd yes
<castilma>what's wrong with mirror.hydra?
<castilma>i get a lot 504s
<thomassgn>hmm
<thomassgn>castilma: same here, lots of fails from hydra. Add --fallback to your invocations and you should be good though.
<pkill9>i need to add berlin
<pkill9>i'm getting 504's from hydra as well
<thomassgn>mm.
<pkill9>lol, the source tarballs are also on the hydra server
<pkill9>and they're failing to download so *shrug*
<castilma>pkill9: i see the same. but what sources are hosted on hydra? shouldn't guix get those from upstream?
<pkill9>dunno, I just saw it trying to download a tar.xz file from hydra
<pkill9>one was eigen
<pkill9>hmm
<pkill9>the package definition points to bitbucket
<pkill9>that's the one i cancelled the building on
<pkill9>so maybe it would have tried upstream after
<efraim>freefall fixed
<efraim>spindle on i686 "fixed"
<thomassgn>efraim: Nice! (though I realized that I'm not on a hp laptop anymore so I don't need it...)
<nckx>Cool that freefall is still being used ^_^
<nckx>...oh.
<nckx>That'll teach me to go get coffee halfway through a sentence.
<marusich>pkill9, 'guix publish' can provide a content-addressed mirror for the source files referenced in 'origin' records. That's why you sometimes see source files being downloaded from Hydra instead of elsewhere.
<marusich>For details, see "Invoking guix publish" in the manual.
<pkill9>ah ok thanks