<wdkrnls>Hmm... after updating my locale it not being set automagically in my profile despite glibc-locales being installed in my profile. After running "export GUIX_LOCPATH=$HOME/.guix-profile/lib/locale" everything works as expected. However, I don't want to have to run this every time. I'm sure this question has been asked before, I just can't find a better answer than running that command all the time.
<lfam>wdkrnls: You'll have to log out and in again
<malaclyps>is anyone else having problems using the emacs-guix package at the moment? I get a "service-sexps: unbound variable" error whenever I try to use it at themoment, even with an empty .emacs config.
<lle-bout>samplet: don't know if you heard that ppc64 is now bootstrapped
<pkill9>on a normal distro, updating it's dependencies while it's running swaps out file sunderneath it
<xelxebar>How did you install your Hebrew fonts, btw?
<lispmacs>xelxebar: I don't recall ever installing any special Hebrew fonts, as it wasn't necessary in the past
<lispmacs>xelxebar: do I need to add something to the locale?
<xelxebar>lispmacs: I think your locale looks fine. Unfortunately, I don't know much about gnome, but my first hypothesis would be that there is some difference between your system profile (from which gnome typically runs) is different than your user profile (which ends up running your terminal and thus emacs).
<lispmacs>hmm... maybe default gnome font does not have Hebrew characters finished...?
<lispmacs>I think that is it. I see default font (at least on my system is Cantarell
<lispmacs>but character map program shows that font is missing the Hebrew letters
<xelxebar>Hrm, I wouldn't guess that's the problem, since AFAIK characters can be pulled from appropriate fonts as needed, but it's certainly worth a shot.
<xelxebar>Try something in the DejaVu family. That tends to contain a really good swath of unicode.
<KE0VVT>lispmacs: What issues are we having with Hebrew letters? I used to type Hebrew in Emacs, until I had to handle vowel markings. Emacs doesn't display them well.
<lispmacs>KE0VVT: emacs displays them fine, but not gnome applications, at least on my system
<KE0VVT>lispmacs: Interesting. Can I see how your Emacs displays the following text? מײַן נאָמען איז כָּלֵבֿ
<lispmacs>xelxebar: I switch gnome to libertine default fonts, and am abble to insert the Hebrew characters now
<lispmacs>KE0VVT: it looks great, am trying to think how to send you a screenshot
<KE0VVT>lispmacs: Though you could just upload to Framapic and put the link here.
<lispmacs>KE0VVT: just sent email. Haven't heard of framapic before, but have bookmarked it now
<malaclyps>Hmm, I am putting together some patches for new guix package updates. I'm doing it one package at a time, and want to test/lint the package definitions as changes to the guix gnu/package files. This means that guix has to rebuild itself, which takes a *long* time. I remember some reference in the manual to speeding this up -- does anyone have a suggestion?
<xelxebar>lispmacs: Hrm. Glad it's working, but that sounds like a broken fontconfig issue.
<lispmacs>xelxebar: hmm. do I need to rerun some fontconfig database command?
<xelxebar>malaclyps: How are you building? running make should just build files with newer timestamps than the last build, not the entire distribution.
<TZander>what do people think about the idea that guix knows its installed on a host system with the result that scripts and profiles change variables like PATH to append guix specific things instead of prepend.
<TZander>(not sure why it would even ever need prepend, but I guess thats done for a reason)
<cbaines>TZander, I guess prepending is done to avoid unreproducible behaviour with some search paths. I can imagine some search paths are searched from the start, so you want to prepend to ensure locations from the environment at the end are considered last.
<cbaines>Do you have an example of when you've had an issue with search path ordering?
<rekado_>TZander: can you please be specific? What libraries are “very old” and the “opposite of always the latest”?
<rekado_>please report them to firstname.lastname@example.org if you can
<TZander>guix is not about having the latest version. It makes it easy to use several different versions of boost, for instance. This is what I mean where the design, the conceptual basis of the system is the opposite of most distos.
<TZander>So this is not about a bug specifically; even though Qt is old and KDE is old on guix.
<cbaines>Guix generally does aim as far as I'm aware to just have the latest versions of things
<cbaines>barring some exceptions, where the aim is to have the latest, plus some older versions (PostgreSQL, Guile, Ruby, ...)
<cbaines>of course, the degree to which that is true is mostly limited by the rate at which packages are updated
<TZander>If you go to a debian or archlinux you will find is very hard to download a binary package of any lib from 2 versions ago.
<TZander>It is completely supported on arch to roll back to a state as your machine was a year ago.
<TZander>Which is great, but also utterly incompatible with software from the host system.
<nixo_>mbakke: deluge-related good news and bad news: builds&run, but it depends on python-setproctitle which is currently broken (I had to work around the problem by breaking support for python 2 on it). Also, in order to pass some tests, it requires a newer python-twisted. The output for guix refresh -l -r python-twisted isn't promising (5793 packages would ensure 15616)
<cbaines>TZander, but Arch is the host system in this example. Did you mean to talk about roll backs with Guix or Arch?
<TZander>cbaines: ah, typo. Yes, arch => host system. Guix allows rollbacks of its profiles.
<cbaines>I've not used KDE in a long while, but I've had similar issues with Gnome. When mixing Guix and Debian, sometimes I couldn't login. I think it was possible to address in some cases though by changing package definitions not to propagate certain packages that could interfere with software from Debian.
<cbaines>As a place to start investigating this though, it probably needs narrowing down to the environment variable, or Guix package that is leading to the issue.
<mbakke>nixo_: don't use -r with 'guix refresh -l'
<TZander>cbaines: after trying to find out why (and it just always failing) I ended up just moving the guix.sh out of /etc/profile.d and removing the lines in by bash_profile and now having a 'startGuix' function to start doing guix stuff
<nixo_>mbakke: then it's only 33-74, I'll try this afternoon
<bricewge>nixo_: I have a working deluge 2.0 but the patchset isn't finished, do you want it?
<cbaines>TZander, yeah, I realise not being able to login can be difficult to debug. At least I've had problems in the past with the GDM.
<nixo_>bricewge: mine is working, too, does your pass all tests?
<bricewge>Unfortunately not, few days ago I had ~90 failing...
<nixo_>i think some of them must be disabled as they require an internet connection, but I can't understand how many as the output is really terrible
<bricewge>nixo_: To build python-setproctitle on python 2 and 3, just remove the modified check phase
<nixo_>bricewge: yes, that's what I did, but I think this will break python2 build
*jonsger wonders how often one has to rebase and update a patch series until it gets merged ^^
<nikita`>hm. pkgsrc hat on: how do you handle ctan apparently not handling permanent files and replacing them in place? this seems to be an issue in pkgsrc right now and we have to track down packages and push sources to our ftp
<xelxebar>nikita`: Isn't that just an issue on Master? I'd expect the files in the tag/* hierarchies to be static.
<rekado_>jonsger: in issues.guix.gnu.org we could also add the closed pill when “-done” is used. Currently I’m only matching on a Debbugs internal message that apparently gets added to the log when close is used.
<rekado_>who would like to adopt a forgotten bug/patch today?
<civodul>"The segfaulting behaviour is still a bug in my opinion" :-)
<mbakke>civodul: do you have a trick for mapping a PID to a build job? On Berlin, 63882 is using a lot of CPU time, occasionally uninterruptible, and I suspect that is what's holding the sqlite DB lock for more than an hour: https://ci.guix.gnu.org/jobset/guix-master
<mbakke>the process name is .guix-offload, so could be anything (presumably something huge)
<TZander>mbakke: cat /proc/63882/cmdline not helpful?
<cbaines>mbakke, do you see the PID in guix processes ?
***drakonis1 is now known as drakonis
<jfred[m]>I've also gotten Flatpak to work system-wide by running it as root
<jfred[m]>but yeah, it segfaulting is not super user-friendly heh
<jfred[m]>(probably using `--user` is the "right" way on guix, I didn't know about that option at the time)
<civodul>mbakke: dunno but i would do: sudo guix processes |recsel -e 'ClientPID ~ "63882"'
<civodul>and maybe strace the thing to get an idea of what it's doing
<rekado_>jfred[m]: is it segfaulting because of something we did to it or is it just an upstream bug? If it’s the latter, has the bug been reported upstream?
<civodul>rekado_: what should be our daily threshold? 2.4T?
<civodul>i think it should be, say, 5 times our daily consumption
<jfred[m]>rekado_: IIRC the last thing I figured out wrt flatpak was that it was trying to access (write I think) files in /var/lib/flatpak, was unable to due to file permissions, then segfaulted. I just tested the same `flatpak remote-add` as an unprivileged user on a Debian 10 system and seem to be getting a prompt to escalate privileges
<lprndn>accountsservice looks in XDG_DATA_DIRS for dbus interfaces. It's used at least by lightdm. Currently, it doesn't find anything as this variable is not set. Should I just wrap accountsservice or are there better ways?
<Gooberpatrol66>for example, if you do guix system disk-image --target=INVALIDTARGET /etc/config.scm, it prints guix/ui.scm:1936:12: In procedure run-guix-command:, even though the actual error is printed in the function glibc-dynamic-linker, in a different file
<pkill9>cbaines: how do you configure it to do that?
<cbaines>pkill9, I think the default is to use the HTTP challenge. Assuming you've got NGinx or whatever webserver up and running, serving files from the webroot certbot is configured with, I think the system reconfiguration prints out the script to run
<ngz>civodul: AFAIK, we don't have arphic (?) font. And I'm no specialist of CJK fonts. So I wonder what font could replace it. Moreover, some fonts packages I checked provide .ttc, not .ttf. I don't know if that's the same.
<lfam>Is there a way for C code to export environment variables besides setenv(3)?
<lfam>Maybe "export" is not the correct word. I mean, to set a variable in the context of the running program
<lfam>I am working with alsa-lib, testing a feature that would make it read and use an environment variable that would contain a list of paths. In most cases, it works as expected, but there is one codepath that fails, and the environment variable seems to truncated when fetched with getenv(3)
<lfam>So I'm wondering if the program is setting the variable somehow, but there aren't any setenv calls
<NieDzejkob>it could be modifying the strings returned by getenv or messing with environ calls
<lfam>I'm printing the value of the pointer returned by getenv immediately after the getenv call
<lfam>But it only shows the first entry in the variable, when it should contain two entries
<NieDzejkob>there's also the option of firing up the program in gdb, enabling reverse debugging, getting to the point where the variable is not what you expect, setting a watchpoint on it, and continuing in reverse
<lfam>I was afraid of that (don't have any experience with gdb)
<lfam>Would that work for environment variables? The C variable has a short lifetime and I'm printf-debugging it at every step
<lfam>I know what's wrong. I didn't realize that strtok_r would mutate the environment