IRC channel logs


back to list of logs

<bavier>nice shoutout to the Guix team on lkcl's latest crowdsupply update:
<rekado>gobject-introspection fails to build on core-updates / rhel6
<rekado>something to do with unexpected types from glibc
<rekado>could also be a race condition due to parallel builds.
<rekado>yes, was a race condition.
<rekado>need to disable parallel building.
<rekado>mysql doesn’t build on core-updates / rhel6
<rekado>“catch (const boost::geometry::detail::self_get_turn_points::self_ip_exception &)”
<rekado>this looks like a code generation error. The ampersand is in the wrong place.
<civodul>is it? it's perfectly valid i think
<civodul>there was a boost update, no?
<rekado>yes, there was a boost update
<rekado>(I haven’t seen a trailing “&” before)
<civodul>it's a C++ "reference"
<rekado>ah, okay. I’m very ignorant when it comes to C++.
<rekado>I’ll see if we can upgrade mysql.
<rekado>no, this is the latest version.
<rekado>I wonder if we can build all packages that depend on mysql with mariadb instead.
<civodul>no idea
<rekado>built kentutils successfully with mariadb. It may be worth trying this for other packages.
<civodul>i think long ago we discussed a global move to mariadb, but i forgot the conclusions
<efraim>Mariadb might FTBFS on aarch64, but if it does we can worry about that later
<civodul>hey artyom-poptsov1
<civodul>artyom-poptsov1: when connecting with guile-ssh, i get this on some machines: no match for method kex algos: server [,diffie-hellman-group-exchange-sha256], client [diffie-hellman-group14-sha1,diffie-hellman-group1-sha1]
<civodul>it seems that guile-ssh uses sha1 as its default key exchange algorithms, which is going to cause more and more problems
<civodul>any ideas?
<rekado>the bootstrap phase fails for classpath-devel
<rekado>looks like it just need gettext
<rekado>oh, but it already *has* gettext-minimal among its inputs.
<rekado>(does the bootstrap phase happen too early?)
<rekado>The gnu build system misses a knob to tweak the bootstrap-scripts. That key is accepted by the bootstrap phase, but it cannot be passed as a build system argument.
***Cthulhux` is now known as Cthulhux
<efraim>mysql did fail on aarch64 on core-updates
<efraim>looks like we have a couple grafts still in core-updates
<mbakke>rekado: MySQL is rather picky about boost versions, that was the reason I never pushed boost 1.64 back in the day.
<efraim>Is it still looking for 1.59?
<mbakke>Not sure, there are probably patches floating around for newer Boost versions.
<mbakke>Or maybe we should switch to MariaDB as the canonical mysql package.
<rekado>mbakke: qtbase and qt are big packages using mysql. Someone should try building those with mariadb.
<thomassgn>how does privelege separation/setuid work with e.g. wireshark? I can't see a group as recommended by the wireshark docs...
<thomassgn>I see in the package definition there is no mention of setuid or similar. I'll look for examples of setuid and similar and see if I can get it working with privelege separation as the wireshark docs intend. :)
<rekado>thomassgn: setuid binaries must be whitelisted in GuixSD
<thomassgn>allright, whitelisted?
<thomassgn>aha, just found the setuid part of manual.
<thomassgn>rekado: Thanks for the hint :)
***fkz is now known as Guest86998
***krkini is now known as kini
<mbakke>rekado: I just realized, MySQL uses CMake, and Boost detection depends on the CMake version. Can you try giving it CMake 3.11?
<mbakke>My build host is busy bootstrapping the very latest core-updates and bisecting glibc 2.27 :P
<rejoice>Happy GNUvidad!
<mbakke>Nowadays I'm surprised when I don't already have a PGP signing key in my keyring.
<catonano>hello guixers
<catonano>I'm getting this:
<catonano>~$ guix pull
<catonano>Updating from Git repository at ''...
<catonano>guix pull: error: Git error: could not open '/home/catonano/.cache/guix/pull/pjmkglp4t7znuugeurpurzikxq3tnlaywmisyr27shj7apsnalwq/.git/logs/refs/remotes/origin/rhel6' for writing: Permesso negato
<catonano>rhel6 ?
<catonano>ah I hahd messed up the permissions on ~/.cache/guix
<mbakke>I've gotten a little further on glibc 2.27. The problem is limited to "glibc-final-with-bootstrap-bash" aka "glibc-intermediate".
<roptat>hi, I'm trying to build the website with haunt, but I get button-big: unbound variable
<roptat>I've entered an environment with: guix environment --ad-hoc git glibc-locales gnutls guile guile-json guile-syntax-highlight guix haunt
<roptat>and ran GUIX_WEB_SITE_LOCAL=yes haunt build as per README
<atw>The Agda package I made has a problem: Agda is trying to compile itself after it's installed, and this tried to write to the store. I got some help on the Agda mailing list:
<atw>I think the right approach would be to add a build phase that does a find -exec. Is there an existing package that does that?
<mbakke>atw: On core-updates, the "python" package has a phase that compiles all the bytecode, maybe you can use that for inspiration.
<dijong>hi, my guix installation seems to work ok but when trying to link any c++ programs with g++ -lstdc++, i get this error "test.cpp:(.text+0x7648): undefined reference to `operator delete(void*, unsigned long)'" at the end. i do have libstdc++:out installed, so what else could be the issue?
***loonquawl__ is now known as loonquawl
<dijong>i was able to get it to work by re-installing gcc:lib -- i just did "guix package -i gcc:lib" and it said "gcc:lib 7.3.0 → 7.3.0" and installed, then it worked.
<dijong>speaking of which, what's the best way to check if i've installed a package? doing guix package --show doesn't show that info, currently i just try guix package -i and it says e.g. "7.3.0 → 7.3.0" then i know it's already installed and hit ctrl+C
<bavier`>dijong: `guix package -I`
***loonquawl_ is now known as loonquawl
<mbakke>I haven't been able to reproduce the glibc 2.27 build failure when testing with a (package (inherit (@@ (gnu packages commencement) glibc-final-with-bootstrap-bash)). Nor with the actual "glibc/linux". Very peculiar.
<civodul>these should be the same things, but do they lead to different derivations?