IRC channel logs

<mark_weaver>civodul: I just posted a patch to fix the single-package union problem.

<mark_weaver>civodul: guix-package.sh still fails for me. See http://www.netris.org/~mhw/test-suite.log

<civodul>mark_weaver: can you do a git pull and also test tests/pki.scm?

<mark_weaver>that's on the Loongson 3A

<mark_weaver>I see that you updated the submodule. is it important for me to run ./sync-with-upstream, etc? not sure if I should do that now since it would delay things.

<civodul>yeah, not now

<mark_weaver>okay

<mark_weaver>currently waiting for guix-package.sh to finish, then will test pki.scm.

<mark_weaver>looks like it's doing a bootstrap in the test store

<mark_weaver>it has certainly gotten farther than it did before

<mark_weaver>glib-2.40.0 is out. I wonder if we should update it.

<civodul>not now :-)

<mark_weaver>guix-package.sh passed. now running pki.scm

<civodul>perfect

<mark_weaver>pki.scm passed too :)

<civodul>wonderful!

<civodul>so feel free to push :-)

<mark_weaver>will do.

<civodul>now i can happily go to bed

<civodul>cheers!

<zacts>lo

<civodul>Hello Guix!

<phant0mas>good morning guix

<phant0mas>civodul: I am working to see why I am getting thelinking error. The workaround suggested by youpi is already present so I believe the problem is elsewhere

<civodul>ok

<phant0mas>if you can, check the attached file in my email, to see if there's something that seems wrong to you

<civodul>i'll look at it this afternoon

<civodul>i should really discuss with him

<civodul>we work in the same building, but i'm somewhat busy

<civodul>and i suppose he is too

<phant0mas>okay, thank you :-)

<phant0mas>civodul: you work at inria,right?

<civodul>yup!

<civodul>currently

<phant0mas>cool :-D

<atheia>'lo

<atheia>civodul: I've been thinking about the dmd work (readiness constructors)

<atheia>I think I need to spend more time on this then I can currently free up so I'm not sure I can come up with anything soon I'm afraid.

<civodul>hey atheia!

<civodul>ok, np

<civodul>we'll discuss that "in due time" :-)

<civodul>maybe i'll look into it after 0.6

<atheia>cheers!

<sriharsha>phant0mas, have you sent your GSoC application to the GNU project?

<phant0mas>sriharsha: of course, I sent it through melagne at 16-3

<sriharsha>OK, thanks

<bavier`>anyone have experience with perl modules in guix?

<bavier`>mostly, how do I ensure that the perl module an application needs are available after the application is installed?

<phant0mas>mark_weaver: what is the use of (snippet ..)?

<phant0mas>for example here http://paste.lisp.org/display/141912/raw

<phant0mas>how does it disable ldconfig?

<mark_weaver>phant0mas: in general, when posting excerpts, it's better to give me the relevant context.

<mark_weaver>where did that excerpt come from?

<phant0mas>glibc package in base.scm

<mark_weaver>within an 'origin' form, snippets are code segments that are executed after fetching the original upstream version, before repacking it up into the tarball that the rest of the recipe will see. sort of like 'patches', but more general.

<mark_weaver>so if you run "guix build -S glibc", you'll get a tarball that includes the patches and the results of the snippet.

<mark_weaver>one useful purpose for a snippet is to remove non-free code. we can't do that using patches, because we'd have to include the non-free code in our patch.

<mark_weaver>I'm not sure why civodul chose to put that subtitution in the 'origin' and not in a normal phase.

<phant0mas>I shall ask him about that, ok I think I understood, thank you :-)

<mark_weaver>you're welcome!

<phant0mas>civodul: I am starting to believe that I have the issue with the linking because it tries to use the native linker, not the cross one

<phant0mas>one question, in the glibc package in base.scm, why did you use this snippet

<phant0mas> http://paste.lisp.org/display/141919/raw

<civodul>phant0mas: it should be easy to see which linker it uses

<civodul>run "guix build ... -K"

<civodul>and then cd /tmp/nix-build-xxxx

<civodul>and "source environment-variables"

<civodul>and then "type -P ld" or so

<civodul>or even just look at LD in config.log

<civodul>the snippet is to make sure the Guix-built glibc doesn't use /etc/ld.so.cache or ldconfig when using Guix on some other distro

<phant0mas>it's using the same linker as the one for the native build on my system

<phant0mas>but for the rest of the process it's using the expected toolchain

<phant0mas>yeah I think i should do something about it

<civodul>can you post the complete build log?

<mark_weaver>civodul: what's the status of python 3 tests? does it work on your system now?

<phant0mas>of course

<civodul>mark_weaver: no, still blocked on the multiprocessing test

<civodul>Steap: any clue?

<civodul>we'll either disable this test, or all of them if we can't fix it

<mark_weaver>there might be one other test that fails on MIPS.

<civodul>oh

<Steap>civodul: the failing test in multiprocessing is a mystery to me

<Steap>I think we should disable it for now

<Steap>so we can release 0.6

<Steap>what should we do for systems with CONFIG_DEVPTS_MULTIPLE_INSTANCES=n ?

<civodul>nothing

<civodul>i don't know what to do with that

<civodul>but hopefully that's quite rare

<civodul>Steap: could you post a patch to skip this test?

<civodul>and perhaps report the problem to the Python folks?

<civodul>although it's annoying because we don't have many clues

<bavier`>civodul: regarding the licensing on the perl modules: the terms say "you can redistribute and/or modify it under the same terms as Perl itself"

<bavier`>(though both modules I proposed have some exceptions that should be noted)

<civodul>bavier`: ooh, interesting

<civodul>then your approach is correct

<civodul>perhaps add a line just above quoting this sentence?

<bavier`>sure, I can do that

<civodul>cool, thanks

<civodul>i see you're populating the HPC side of things ;-)

<bavier`>it's what I'm most comfortable with at the moment

<bavier`>and Guix offers a sane alternative to the environments I work in most of the time :)

<civodul>things like http://modules.sourceforge.net/ ? :-)

<civodul>i see people at work using that on a cluster

<civodul>it's a pain

<civodul>they're basically doing a distro by hand

<civodul>(well, like we do, but hey ;-))

<bavier`>yes, module is a nightmare

<bavier`>I'm not sure I could propose guix or nix as a full system replacement though, because many developers like to use compilers other than gcc, which gets difficult with ABI differences for libraries and such

<civodul>well, it could actually help with that, because Guix knows all the DAG

<civodul>however, in my experience, HPC people like to use non-free compilers

<civodul>that's the main problem

<bavier`>yes, intel and pgi compilers are quite common

<bavier`>I've been considering getting one of the "Run GCC" shirts that were just put in the GNU store ;)

<civodul>yeah, i noticed :-)

<civodul>we can get them from the /gnu/store, ahah ;-)

<bavier`>:P

<bavier`>wouldn't it be nice to hit 700 packages for the 0.6 release...

<civodul>yup!

<civodul>8 packages left :-)

<Steap>civodul: yeah, I'll try to send a patch

<phant0mas>my message to the mailist was rejected because the build.log was too big

<phant0mas>resending the message with the build.log in a url

IRC channel logs

2014-04-04.log