<shtwzrd>I'm trying to build a uboot package but it keeps dying early, with a fail on include <asm/string.h> but I'm having trouble understanding why
<shtwzrd>it feels like my build environment is missing linux headers or something? But I'm pretty clueless, it doesn't seem like I needed to include anything like that before. And I have all the inputs that other similiar uboot packages have.
<shtwzrd>It's either an issue with the repo I'm using for uboot (unlikely, it's branched directly from official and laid out identically it seems) or it's a mismatch in behavior between git-fetch (what I'm doing) and url-fetch
<shtwzrd>In my build, the source goes in a "source" dir, and in the normal build it's in 'uboot-2020.01', and I can see that in the normal build, there's a lot of output where (I think) it's copying files. Maybe the dirname is important. Is there any 'smart' way to change the dirname git-fetch uses or am I just supposed to `mv` it in a snippet?
<bricewge>shtwzrd: Still trying to build the u-boot package for the pinebook pro?
<shtwzrd>yea :) It's taking me a while because I'm a scrub but at least it's a learning experience :)
<shtwzrd>It's the source being used in the AUR and i know it addresses the issue of the bootloader not bringing up all of the cores correctly
<shtwzrd>Though I haven't read the patch, maybe it includes that fix too
<bricewge>Then try using the git repo as your source.
<shtwzrd>bricewge: That's what I'm doing, but then it just completely fails to build because it can't resolve its includes anymore.
<shtwzrd>bricewge: that package definition is very cool, though. Mine is very stupid and ugly because I didn't realize I could apply a change to origin or patches directly on the deriving package, so I was re-defining things all way up to the level of the u-boot package itself.
<shtwzrd>also, yours actually works ;) So, perhaps I should just use your package definition, and if I really want that source's changes, I can just generate a .patch from their master and run with that.
<pkill9>is it possible to extract package requirements from CMakeLists/meson.build/configure files?
***daviid` is now known as daviid
<pkill9>i'm thinking about how it could be possible to create importers for these
<bricewge>shtwzrd: I'm having the same issue about the `#include <asm/string.h>`. Did you managed to build the AUR package before?
<valignatev>Hey guix! long time no see. I can't build 227aeeb64c825870433cfca1755b7d30005ae259, it segfaults on [ 56%] GUILEC gnu/packages/maths.go with the error "mmap(PROT_NONE) failed".
<shtwzrd>bricewge: no, I haven't built it before. I could hop onto an arch machine real quick tho and just verify that I'm not sending us on a wild goose chase
<bricewge>shtwzrd: There isn't a asm/string.h in linux-libre-headers that's probably why. « find $(guix build linux-libre-headers) -name 'string.h' »
<bricewge>Or it's cross-compilation issue and I don't yet understand how it works in Guix.
<shtwzrd>bricewge: I think it's a cross compile issue, because if I do the same, I do have asm/string.h (I'm running from the pinebook itself)
<valignatev>Also, I've prepared to finally submit Alacritty and rebased all my stuff on top of master but notices that it doesn't build on the newest guix version: for some reason guix vendors alacritty while trying to build alacritty e.g. it puts the very package I'm trying to build under guix-vendor dir of this very package
<valignatev>That didn't happened before I pulled latest commit yesterday
<cnmne>hi guix! I followed the "guix cookbook - direct checkout hacking" section to the T and it seems like my guix isn't picking up my changes to `/gnu/packages/ruby.scm'. It recognized it was changed and `newer than compiled', so I ran `make' according to the mailing list thread. Now it doesn't complain about being outdated, but the packages I added can't be searched or built!
<alextee[m]>would upstream welcome a change to the cairo package to allow for xml support?
<ngz>cnmne: Did you "guix environment guix" first?
<valignatev>efraim: It's 100% didn't happen before. Luckily my logs from the previous builds are still exist. They show that cargo-build-system definitely didn't vendored the package as a dependency of itself
<valignatev>So it must've been introduced during the last month or so
<wajsel>Hi all, I got a problem after updating my system with sudo guix system reconfigure
<wajsel>all worked great until the update. Problem looks to be that gnome-session-binary warns about "unable to find required component 'org.gnome.SettingsDaemon.XSettings' (and other components)
<wajsel>from /var/log/messages I see those lines and also "gdm: GdmDisplay: display latest 1.286123 seconds"
<wajsel>this happens when I try to login on the graphic login display, which I assume is GDM
<wajsel>does anyone have any idea where things could have gone wrong
<cnmne>ngz, NieDzejkob : thanks for the help, but it turns out my file was just written wrong yikes! I was using `guix import ... >> ruby.scm' and didn't wrap anything with `define-public'. sorry for the trouble!
<ngz>Isn't Guix data service supposed to apply patches and see if they build, or some such?
<oriansj>the important part is to add -t 01:00:00 so that it stops after the limit (in this case 1 hour)
<g_bor[m]><pkill9 "what can be done to increase the"> pkill9 (@pkill-9:matrix.org): i believe that we made some progress in this direction in the last few days, and have quite good plans for the future.
<bandali>so if you suggest, i'll add it as a propagated input
<ngz>bandali: I have no strong opinion about message-x, if you think that's optional, we can ignore it. However, we should remove sqlite as a propagated input.
***jonsger1 is now known as jonsger
<wajsel>drakonis1: then I only need to remove gdm. Why was gdm started and not slim if both of them were in the %desktop-services ?
<bandali>ngz, ha, i'm a bit torn honestly. i'd personally probably add message-x along with gnus-harvest if i were to use gnus-harvest, but idk if i'd want to impose that onto others. maybe let's leave it out for now, and if people complain, we'll add it
<bandali>ngz, re sqlite, absolutely. the v2 i'm sending in a few mins has that sorted out, thanks for pointing that out :-)
<nckx>drakonis1: Available in the current system? sudo herd status.
<g_bor[m]>nckx : actually it already helps a lot, for example you can immediately see if the package lints. I also believe it is possible to check if it builds.
<g_bor[m]>More checks should be added, which is the easy part. Also patchwork attaches a state based on what checks passed, but does not yet allow to categorize patches based on that, which would be nice.
<g_bor[m]>wajsel: I believe this should not be allowed. I don't currently have access to the code, but these services should be made conflicting, which can be done by providing the same symbol, for example login-manager would do
<wajsel>can I dump what is the full config? so that all %base-magic is expanded?
<nckx>g_bor[m]: Yes, I was pleasantly suprised by the Data Service :-) I look forward to playing with it once I get home.
*nckx had forgotten how god-awful moshless ssh is.
<ngz>g_bor[m]: Is there a Data Service for the dummies available somewhere? All my visits to data.guix.gnu.org were totally unhelpful.
<g_bor[m]>wajsel: I don't know if the system provisioning information stores it interpolated, or in its plain form...
***jonsger1 is now known as jonsger
<g_bor[m]>But I believe if you are in the same guix state you can just copy it to a repl and query it or pretty print it
<g_bor[m]>pkill9 (@pkill-9:matrix.org): how do you think that a display manager service would improve the situation?
<pkill9>you wouldn't need to delete gdm service, there wouldn't be conflicts, it would be neater to have the name "display-manager" with a field to specify which display manager. honestly it probably wouldn't , i was just bikeshedding
<g_bor[m]>I mean it is almost trivial to do it, but you would still have to modify it if we include it in desktop services
<pkill9>i think it's easier though as you could just specify the service and it would 'overwrite' the one in %desktop-services, no deleting require
<g_bor[m]>I would rather vote for providing the symbol from the services, and expose a procedure, for example provision, that takes two list of services, combining in a way that all services in the first list are guaranteed and as much of the second list as possible without conflicts. Does that make sense?
<apteryx>after some readings, it seems NFS is the way to go for a network file system between GNU/Linux machines, and Sambda (CIFS) is a more supported alternative (less performant for small files) to access from Android.
<shtwzrd>is there any way to get guix to output how it is resolving dependencies, and how it decides what to build first? I'd like to see how it's working its way through the tree when I pass it an operating-system form.
<jackhill>wdkrnls: oh cool, I didn't realize all the dependencies were packaged. I guess that wasn't true a few months ago when someone asked me. I don't have time to work on it right now, but that sounds promising
<wdkrnls>jackhill: no worries, thanks for listening :) I have posted my plea to the help-guix mailing list. I'm hopeful someone like zimoun or rekado might not have too much trouble with packaging it given their other impressive contributions to R-support within Guix.
<wdkrnls>There are some emacs packages which if added to Guix, would make it easier for me to work with all my .Rmd files without it. Some seemed easy enough to package, but then unfortunately they didn't work right when I tried to use them.
<wdkrnls>These are extensions to emacs-polymode which is packaged in Guix.